The Benefits and Costs of the
Clean Air Act, 1970 to 1990
Prepared for
U.S. Congress
by
U.S. Environmental Protection Agency
October 1997
-------
-------
Contents
Tables[[[ xi
Figures[[[ xv
Acronyms and Abbreviations xvii
Acknowledgments[[[ xxiii
Executive Summary [[[ ES-1
Purpose of the Study ES-1
Study Design ES-1
Study Review ES-1
Summary of Results ES-2
Direct Costs ES-2
Emissions ES-2
Air Quality ES-3
Physical Effects ES-5
Economic Valuation ES-7
Monetized Benefits and Costs ES-8
Alternative Results ES-9
Conclusions and Future Directions ES-9
Chapter 1: Introduction 1
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Chapter 3: Emissions 13
Sector-Specific Approach 15
Summary of Results 15
Uncertainty in the Emissions Estimates 17
Chapter 4: Air Quality[[[ 19
General Methodology 20
Sample Results 21
Carbon Monoxide 21
Sulfur Dioxide 22
Nitrogen Dioxide 22
Particulate Matter 23
Ozone 23
Urban Ozone 23
Rural Ozone 24
Acid Deposition 24
Visibility 25
Uncertainty in the Air Quality Estimates 25
Chapter 5: Physical Effects [[[ 29
Human Health and Welfare Effects Modeling Approach 29
Air Quality 29
Population 29
Health and Welfare Effects 29
Key Analytical Assumptions 30
Mapping Populations to Monitors 32
Choice of Study 33
Variance Within Studies 33
PM-Related Mortality 34
Short-Term Exposure Studies 34
Long-Term Exposure Studies 35
Health Effects Modeling Results 37
Avoided Premature Mortality Estimates 37
Non-Fatal Health Impacts 37
Other Physical Effects 38
Ecological Effects 38
Aquatic and Forest Effects 38
Quantified Agricultural Effects 39
Effects of Air Toxics 39
Uncertainty in the Physical Effects Estimates 41
Chapter 6: Economic Valuation 43
-------
Contents
Other Valuation Estimates 47
Changes in Children's IQ 47
Work Loss Days and Worker Productivity 48
Agricultural Benefits 48
Valuation Uncertainties 48
Mortity Risk Benefits Transfer 48
Chapter 7: Results and Uncertainty 51
Quantified Uncertainty in the Benefits Analysis 51
Aggregate Monetized Benefits 52
Comparison of Monetized Benefits and Costs 55
Major Sources of Uncertainty 56
Alternative Results 57
PM Mortality Valuation Based on Life-Years Lost 57
Alternative Discount Rates 58
Appendix A: Cost and Macroeconomic Modeling [[[ A-l
Introduction A-l
Macroeconomic Modeling A-l
Choice of Macroeconomic Model A-2
Overview of the Jorgenson-Wilcoxen Model A-2
Structure of the Jorgenson-Wilcoxen Model A-3
The Business Sector A-4
The Household Sector A-4
The Government Sector A-5
The Rest-of-the-World Sector A-5
Environmental Regulation, Investment, and Capital Formation A-5
The General Equilibrium A-5
Configuration of the No-control Scenario A-6
Elimination of Compliance Costs in the No-Control Case A-7
Capital Costs - Stationary Sources A-7
Operating and Maintenance Costs - Stationary Sources A-8
Capital Costs - Mobile Sources A-8
Operating and Maintenance - Mobile Sources A-8
Direct Compliance Expenditures Data A-8
Sources of Cost Data A-8
Cost of Clean Data A-8
EPA Data A-8
Commerce Data A-9
Stationary Source Cost Data A-9
Capital Expenditures Data A-9
Operation and Maintenance Expenditures Data A-10
Recovered Costs A-10
Mobile Source Cost Data A-l 1
Capital Expenditures Data A-11
Operation and Maintenance Expenditures Data A-11
Fuel Price Penalty A-11
Fuel Economy Penalty A-12
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Assessment Results A-14
Compliance Expenditures and Costs A-14
Annualization Method A-16
Discounting Costs and Expenditures A-19
Indirect Economic Effects of the CAA A-20
GNP and Personal Consumption A-20
Prices A-23
Sectoral Effects: Changes in Prices and Output by Industry A-23
Changes in Employment Across Industries A-25
Uncertainties in the Cost Analysis A-26
Potential Sources of Error in the Cost Data A-26
Mobile Source Costs A-28
Stationary Source Cost Estimate Revisions A-29
Endogenous Productivity Growth in the Macro Model A-29
Amortization Period for Stationary Source Plant and Equipment A-30
Cost and Macroeconomic Modeling References A-31
Appendix B: Emissions Modeling [[[ B-l
Introduction B-l
Comparison of Emissions Projections with Other EPA Data B-l
Control Scenario Projections Versus EPA Trends Projections B-l
No-Control Scenario Projections Versus Historical EPA Trends Data B-3
Industrial Boilers and Processes B-4
Overview of Approach B-4
Industrial Boilers B-4
Industrial Processes and In-Process Fuel Combustion B-4
Establishment of Control Scenario Emissions B-5
Control Scenario Boiler Emissions B-5
Control Scenario Industrial Process Emissions B-7
Development of Economic Driver Data
for the Control Scenario - Industrial Boilers and Processes B-7
Economic Driver Data for Industrial Boiler Approach B-7
Economic Driver Data for the Industrial Process Approach B-8
No-control Scenario Emissions B-8
Industrial Boiler Emissions of S02, NOx, and TSP B-8
Industrial Boiler Emissions of CO and VOC B-9
Industrial Process Emissions B-9
Lead Emissions B-9
Off-Highwav Vehicles B-10
Overview of Approach B-10
Development of Control Scenario B-l 1
No-control Scenario Emissions Estimates B-l I
National and State-Level Off-Highway Emission Estimates B-l 1
On-Highway B-12
Overview of Approach B-l 3
Personal Travel B-l3
Iterative Proportional Fitting (IPF) B-l3
Vehicle Ownership Projection (VOP) B-14
Projection of Vehicle Fleet Composition B-14
Activity/Energy Computation B-14
Goods Movement B-15
Other Transportation Activities B-15
Lead Emissions B-15
-------
Contents
Development of Highway Pollutant Estimates B-16
Control Scenario Emissions Calculation B-16
No-control Scenario Emissions B-21
Utilities B-24
Overview of Approach B-24
Establishment of Control Scenario Emissions B-24
Key Assumptions in the Development of the 1CF Analysis B-24
ARGUS Modeling Assumptions B-26
No-control Scenario Emissions B-27
ICF Estimates of SO2. TSP, and NOx Emissions in the No-control Scenario B-27
ARGUS No-control Scenario B-29
Estimation of Lead Emissions from Utilities B-29
CEUM Sensitivity Case B-30
Commercial/Residential B-30
Control Scenario Emissions B-31
Emissions Data B-32
Energy Data B-33
Economic/Demographic Data B-33
No-control Scenario Emissions B-34
Emissions Data B-34
Energy Data B-34
Economic/Demographic Data B-35
Emissions Modeling References B-39
Appendix C: Air Quality Modeling [[[ C-l
introduction C-l
Carbon Monoxide C-l
Control scenario carbon monoxide profiles C-l
No-control scenario carbon monoxide profiles C-2
Summary differences in carbon monoxide air quality C-4
Key caveats and uncertainties for carbon monoxide C-4
Sulfur Dioxide C-5
Control scenario sulfur dioxide profiles C-5
No-control scenario sulfur dioxide profiles C-5
Summary differences in sulfur dioxide air quality C-6
Key caveats and uncertainties for sulfur dioxide C-6
Nitrogen Oxides C-6
Control scenario nitrogen oxides profiles C-7
No-control scenario nitrogen oxides profiles C-8
Summary differences in nitrogen oxides air quality C-8
Key caveats and uncertainties for nitrogen oxides C-8
Acid Deposition C-8
Control scenario acid deposition profiles C-9
No-control scenario acid deposition profiles C-l 1
Summary differences in acid deposition C-l2
Key caveats and uncertainties for acid deposition C-12
Particulate Matter C-l3
Control scenario particulate matter profiles C-14
No-control scenario particulate matter profiles C-l5
Summary differences in particulate matter air quality C-l6
Key caveats and uncertainties for particulate matter C-l6
Ozone C-l 8
Control scenario ozone profiles C-21
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Visibility C-25
Control scenario visibility C-25
No-control scenario visibility C-26
Summary differences in visibility C-26
DeciView Haze Index C-26
Modeling Results C-28
Key caveats and uncertainties for visibility C-28
Air Quality Modeling References C-30
Appendix D: Human Health and Welfare Effects of Criteria Pollutants D-1
Introduction D-1
Principles for the Section 812 Benefits Analysis D-1
General Modeling Approach D-2
Quantifying Changes in Pollutant Exposures D-2
Air Quality D-2
Population Distribution D-3
Census Data D-3
Gridding U.S. Population D-4
Allocating Exposure Estimates to the Population D-4
Method One D-4
Method Two D-4
Estimating Human Health Effects of Exposure D-5
Types of Health Studies D-5
Epidemiological Studies D-6
Human Clinical Studies D-7
Issues in Selecting Studies To Estimate Health Effects D-9
Peer-Review of Research D-9
Confounding Factors D-9
Uncertainty D-10
Magnitude of Exposure D-11
Duration of Exposure D-11
Thresholds D-11
Target Population D-11
Statistical Significance of Exposure-Response Relationships D-12
Relative Risks D-12
Baseline Incidence Data D-12
Estimating Mortality Effects D-13
Using PM as an Indicator D-13
Estimating the Relationship Between PM and Premature Mortality D-13
Prematurity of Mortality: Life-Years Lost as a Unit of Measure D-16
Estimating Morbidity Effects D-19
Overlapping Health Effects D-19
Studies Requiring Adjustments D-19
Concentration-Response Functions: Health Effects D-19
Particulate Matter D-19
Ozone D-26
Nitrogen Oxides D-34
Carbon Monoxide D-3 6
Sulfur Dioxide D-38
Estimating Welfare Effects of Exposure D-40
Agricultural Effects D-40
Materials Damage D-41
Visibility D-41
Worker Productivity D-41
Ecological Effects D-41
Modeling Results D-44
Human Health and Welfare Effects References D-48
vi
-------
Contents
Appendix E: Ecological Effects of Criteria Pollutants E-l
Introduction E-l
Benefits From Avoidance of Damages to Aquatic Ecosystems E-l
Acid Deposition E-2
Background E-2
Current Impacts of Acid Deposition E-2
Effects on Water Chemistry E-2
Effects on Fish Habitat Quality E-4
Economic Damages to Recreational Fishing E-5
Benefits From Acid Deposition Avoidance Under the CAA E-5
Recreational Fishing E-5
Eutrophication E-6
Atmospheric Deposition and Eutrophication E-7
Valuing Potential Benefits from Eutrophication Avoidance Under the CAA E-7
Mercury E-8
Benefits from Avoided Damages to Wetland Ecosystems E-9
Introduction E-9
Effects of Acidification E-9
Effects of Nutrient Loading E-10
Summary of Wetland Ecosystem Effects E-l 1
Benefits from Avoided Damages to Forests E-l 1
Introduction E-l 1
Current Air Pollutant Effects on Forests E-l2
Acid Deposition Impacts E-l2
Ozone Impacts E-l2
Experimental Evidence E-l2
Observational Evidence E-l3
Endangered species E-l4
Valuation of Benefits From CAA-Avoided Damages to Forests E-14
Background E-14
Commercial Timber Harvesting E-l5
Non-marketed Forest Services E-l6
Ecosystem Effects References E-l8
Appendix F: Effects of Criteria Pollutants on Agriculture F-1
Introduction F-l
Ozone Concentration Data F-l
Control and No-control Scenario Ozone Concentration Data F-2
Calculation of the W126 Statistic F-2
Aggregating Ozone Data to the County Level F-3
Yield Change Estimates F-3
Exposure-Response Functions F-3
Minimum/Maximum Exposure-Response Functions F- l
Calculation of Ozone Indices F-4
Calculations of County Weights F-5
Calculation of Percent Change in Yield F-5
Economic Impact Estimates F-5
Agricultural Simulation .Model (AGSIM) F-5
Conclusions F-9
Agricultural Effects References F-10
Appendix G: Lead Benefits Analysis [[[ G-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Changes in IQ G-2
Quantifying the Relationship Between Blood Lead Levels and IQ G-2
Valuing Changes in Children's Intelligence G-3
Children with IQs Less Than 70 G-7
Quantifying the Number of Children with IQs Less than 70 G-7
Valuing the Reduction in Number of Children with IQs less than 70 G-8
Changes in Neonatal Mortality G-8
Quantifying the relationship between PbB levels and neonatal mortality G-8
Valuing changes in neonatal mortality G-8
Health Benefits to Men G-8
Hypertension G-9
Quantifying the relationship between PbB levels and hypertension G-9
Valuing reductions in hypertension G-9
Quantifying the relationship between blood lead and blood pressure G-9
Changes In Coronary Heart Disease G-10
Quantifying the relationship between blood pressure and coronary heart disease G-10
Valuing reductions in CHD events G-l 1
Changes in Initial Cerebrovascular Accidents
and Initial Atherothrombotic Brain Infarctions G-l 2
Quantifying the relationship between blood pressure and first-time stroke G-l2
Valuing reductions in strokes G-l2
Changes in Premature Mortality G-l3
Quantifying the relationship between blood pressure and premature mortality G- l 3
Valuing reductions in premature mortality G-l3
Health Benefits to Women G-l3
Changes in Coronary Heart Disease G- l 4
Quantifying the relationship between blood pressure and coronary heart disease G-l4
Valuing reductions in CHD events G-l4
Changes in Athe rothrombotic Brain Infarctions and Initial Cerebrovascular Accidents G-l 4
Quantifying the relationship between blood pressure and first-time stroke G-l4
Valuing reductions in strokes G-l 5
Changes in Premature Mortality G-l5
Quantifying the relationship between blood pressure and premature mortality G-15
Quantifying Uncertainty G-15
Characterizing Uncertainty Surrounding the Dose-Response Relationships G-15
Characterizing Uncertainty Surrounding the Valuation Estimates G-15
Industrial Processes and Boilers and Electric Utilities G-16
Methods Used to Determine Changes in Lead Emissions
from Industrial Processes from 1970 to 1990 G-16
TRI Data G-16
Derivation of Industrial Process Emissions Differentials 1970-1990 G-17
Data sources G-17
Estimates of industrial process emissions in the control scenario G-17
Estimates of industrial process emissions in the no-control scenario G-l 8
Matching TRI Data to Industrial Process Emissions Differentials G-l8
Methods Used to Determine Changes in Lead Emissions
from Industrial Boilers from 1970 to 1990 G-19
TRI Data G-19
Derivation of Industrial Combustion Emissions 1970-1990 G-20
Estimates of combustion emissions under the control scenario G-20
Estimates of combustion emissions under the no-control scenario G-20
Matching TRI Data to Industrial Combustion Emissions Data G-21
Methods Used to Determine Changes in Lead Emissions
from Electric Utilities from 1975 to 1990 G-21
Coal-Use Data G-21
The EPA Interim Emissions Inventory G-21
Matching the Coal-Use Data to the Interim Emissions Inventory G-22
Emissions Factors and Control Efficiencies G-22
vrii
-------
Use of Air Dispersion Modeling to Estimate Ambient Air Lead Levels G-23
Determination of Blood Lead Levels from Air Lead Concentrations G-23
Relationship Between Air Lead Concentrations and Blood Lead Levels G-23
Children G-25
Adults G-25
Individuals with initial blood lead levels of 30 pg/dL and greater G-26
Estimates of Initial Blood Lead Concentrations G-26
Combination of Air Concentration Estimates with Population Data G-27
Results G-28
Reduction in Health Effects Attributable to Gasoline Lead Reductions G-31
Estimating Changes in Amount of Lead in Gasoline from 1970 to 1990 G-31
Estimating the Change in Blood Lead Levels
from the Change in the Amount of Lead in Gasoline G-31
1970-Forward and 1990-Backward Approaches G-32
Relating Blood Lead Levels to Population Health Effects G-32
Changes in Leaded Gasoline Emissions
and Resulting Decreased Blood Lead Levels and Health Effects G-32
Lead Benefits Analysis References G-36
Appendix H: Air Toxics [[[ H-l
Introduction H-l
Limited Scope of this Assessment 11-1
History of Air Toxics Standards under the Clean Air Act of 1970 H-2
Quantifiable Stationary Source Air Toxics Benefits H-3
EPA Analyses of Cancer Risks from Selected Air Toxic Pollutants H-3
Cancer Risk Estimates from NHS HAP Risk Assessments H-4
Non-utilitv Stationary Source Cancer Incidence Reductions H-4
PES Study H-5
Methodology H-5
Findings 11-6
ICF Re-analvsis H-7
Methodology H-7
Findings H-8
Mobile Source HAP Exposure Reductions 11-9
Methodology H-l 0
Results H-10
Non-Cancer Health Effects H-l 1
Ecological Effects H-l I
Conclusions — Research Needs H-l2
Health Effects H-l2
Exposure Assessment H-l3
Ecosystem Effects H-l3
Economic Valuation H-l 3
Air Toxics References H-l4
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants 1-1
Methods Used to Value Health and Welfare Effects 1-1
Valuation of Specific Health Endpoints 1-3
Valuation of Premature Mortality Avoided 1-3
Valuation of Hospital Admissions Avoided 1-3
Valuation of Chronic Bronchitis Avoided 1-4
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Uncertainties 1-16
The Effect of Discount Rates 1-20
The Relative Importance of Different Components of Uncertainty 1-20
Economic Benefits Associated with Reducing Premature Mortality 1-21
Economic Valuation References 1-27
Appendix J: Future Directions[[[ J-1
Research Implications J-1
-------
Tables
Table ES-1 Criteria Pollutant Health Benefits - Distributions of 1990 Incidences of Avoided
Health Effects (In Thousands of Incidences Reduced) for 48 State Population ES-4
Table ES-2 Major Nonmonetized, Adverse Effects Reduced by the Clean Air Act ES-5
Table ES-3 Central Estimates of Economic Value per Unit of Avoided Effect
(In 1990 Dollars) ES-6
Table ES-4 Total Monetized Benefits by Endpoint Category for 48 State Population for 1970
to 1990 Period (In Billions of 1990 Dollars) ES-7
Table ES-5 Alternative Mortality Benefits Mean Estimates for 1970 to 1990 (In Trillions
of 1990 Dollars) Compared to Total 1970 to 1990 Compliance Costs ES-9
Table 1 Estimated Annual CAA Compliance Costs (SBillions) 8
Table 2 Compliance Cost G N P. and Consumption Impacts Discounted to 1990
($1990 Billions) 11
Table 3 Summary of Sector-Specific Emission Modeling Approaches 14
Table 4 Uncertainties Associated with Emissions Modeling 18
Table 5 Key Uncertainties Associated with Air Quality Modeling 26
Table 6 Human Health Effects of Criteria Pollutants 31
Table 7 Selected Welfare Effects of Criteria Pollutants 32
Table 8 Percent of Population (of the Continental US) Within 50km of a Monitor (Or in a
County with PM monitors), 1970-1990 33
Table 9 Criteria Pollutants Health Benefits — Distributions of 1990 Avoided Premature
Mortalities (Thousands of Cases Reduced) for 48 State Population 37
Table 10 Criteria Pollutants Health Benefits — Distributions of 1990 Non-Fatal Avoided
Incidence (Thousands of Cases Reduced) for 48 State Population 38
Table 11. Health and Welfare Effects of Hazardous Air Pollutants 40
Table 12 Uncertainties Associated with Physical Effects Modeling 42
Table 13 Health and Welfare Effects Unit Valuation (1990 Dollars) 44
Table 14 Summary of Mortality Valuation Estimates (Millions of $ 1990) 45
Table 15 Estimating Mortality Risk Based on Wage-Risk Studies: Potential Sources
and Likely Direction of Bias 50
Table 16 Present Value of 1970 to 1990 Monetized Benefits by Endpoint Category for 48
State Population (Billions of $1990, Discounted to 1990 at 5 Percent) 52
Table 17 Total Monetized Benefits for 48 State Population (Present Value in Billions of
1990 Dollars, Discounted to 1990 at 5 Percent) 53
Table 18 Quantified Uncertainty Ranges for Monetized Annual Benefits and Benefit/Cost
Ratios, 1970-1990 (In Billions of 1990-Value Dollars) 55
Table 19 Alternative Mortality' Benefits Mean Estimates for 1970 to 1990 (in Trillions of
1990 Dollars, Discounted at 5 percent) Compared to Total 1970 to 1990
Compliance Costs 57
Table 20 Effect of Alternative Discount Rates on Present Value of Total Monetized
Benefits/Costs for 1970 to 1990 (In Trillions of 1990 Dollars) 57
Table A-l Key Distinguishing Characteristics of the Jorgenson-Wilcoxen Model A-3
Table A-2 Definitions of Industries Within the J/W Model A-4
Table A-3 Estimated Capital and O&M Expenditures for Stationary Source Air Pollution
Control (Millions of Current Dollars) A-10
Table A-4 Estimated Recovered Costs for Stationary Source Air Pollution Control
(Millions of Current Dollars) A-l 1
Table A-5 Estimated Capital and Operation and Maintenance Expenditures for Mobile
Source Air Pollution Control (Millions of Current Dollars) A-12
xi
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Table A-6 O&M Costs and Credits (Millions of Current Dollars) A-12
Table A-7 Other Air Pollution Control Expenditures (Millions of Current Dollars) A-14
Table A-8 Summary of Expenditures and Conversion to 1990 Dollars
(Millions of Dollars) A-15
Table A-9 Annualized Costs, 1973-1990 (Millions of 1990 Dollars; Capital Expenditures
Annualized at 5 Percent) A-16
Table A-10 Amortization of Capital Expenditures for Stationary Sources
(Millions of 1990 Dollars) A-17
Table A-11 Amortization of Capital Expenditures for Mobile Sources
(Millions of 1990 Dollars) A-18
Table A-12 Compliance Expenditures and Annualized Costs, 1973 to 1990 ($1990 millions) A-19
Table A-13 Costs Discounted to 1990 ($1990 Millions) A-20
Table A-14 Differences in Gross National Product Between the Control and No-Control
Scenarios A-20
Table A -15 Difference in Personal Consumption Between the Control and No-Control
Scenarios A-21
Table A-16 GNP and Consumption Impacts Discounted to 1990 ($1990 Billions) A-21
Table A-17 Percentage Difference in Energy Prices Between the Control and No-Control
Scenarios A-23
Table A-18 Potential Sources of Error and Their Effect on Total Costs of Compliance A-26
Table A-19 Stationary Source O&M Expenditures as a Percentage of Capital Stock
(Millions of 1990 Dollars) A-27
Table A-20 Comparison of EPA and BEA Stationary Source Expenditure Estimates
(Millions of Current Dollars) A-28
Table A-21 BEA Estimates of Mobile Source Costs A-29
Table A-22 Annualized Costs Assuming 40-Year Stationary- Source Capital Amortization
Period, 1973 to 1990 ($1990 Millions) A-30
Table A-23 Effect of Amortization Periods on Annualized Costs Discounted to 1990
(Billions of $1990) A-30
Table B-l Correspondence Between Process Emissions Categories Used by MSCET,
Trends, and J/W Industrial Sectors and Identifier Codes B-6
Table B-2 Fuel Use Changes Between Control and No-control Scenarios B-9
Table B-3 Difference in Control and No-control Scenario Off-Highway Mobile Source
Emissions B-12
Table B-4 Sources of Data for Transportation Sector Control Scenario Activity Projection B-l 7
Table B-5 Distribution of Households by Demographic Attributes for Control Scenario B-18
Table B-6 Economic and Vehicle Usage Data for Vehicle Ownership Projection Control
Scenario B-l 9
Table B-7 Control Scenario Personal Characteristics B-20
Table B-8 Distribution of Households by Income Class for No-Control Scenario B-21
Table B-9 Economic and Vehicle Usage Data for Vehicle Ownership Projection
No-Control Scenario B-22
Table B-10 Percent Changes in Key Vehicle Characteristics Between the Control and
No-Control Scenarios B-23
Table B-l 1 J/W Estimates of Percentage Increases in National Electricity Generation
Under No-Control Scenario B-29
Table B-l2 Trends Source Categories and (1975 to 1985) Scaling Factors for TSP and CO B-33
Table B-l3 Percentage Change in Real Energy Demand by Households from Control to
No-Control Scenario B-34
Table B-l4 Percentage Change in Commercial Energy Demand from Control to
No-Control Scenario B-35
Table B-15 J/W Percent Differential in Economic Variables Used in CRESS B-35
Table B-16 TSP Emissions Under the Control and No-Control Scenarios by Target Year
(In Thousands of Short Tons) B-36
Table B-l 7 S02 Emissions Under the Control and No-Control Scenarios by Target Year
(In Thousands of Short Tons) B-36
xii
-------
Tables
Table B-l 8 NOx Emissions Under the Control and No-Control Scenarios by Target Year
(In Thousands of Short Tons) B-37
Table B-19 YOC Emissions Under the Control and No-Control Scenarios by Target Year
(In Thousands of Short Tons) B-37
Table B-20 CO Emissions Under the Control and No-Control Scenarios by Target Year
(In Thousands of Short Tons) B-38
Table B-21 Lead (Pb) Emissions Under the Control and No-Control Scenarios by Target
Year (In Thousands of Short Tons) B-38
Table C-l Summary of CO Monitoring Data C-2
Table C-2 Format of Air Quality Profile Databases C-3
Table C-3 Summary of S02 Monitoring Data C-5
Table C-4 Summary of N02 Monitoring Data C-7
Table C-5 Summary of NO Monitoring Data C-7
Table C-6 Summary of TSP Monitoring Data C-14
Table C-7 Summary of PM10 Monitoring Data C-15
Table C-8 Fine Particle (PM2.5) Chemical Composition by U.S. Region C-16
Table C-9 Coarse Particle (PM2.5 to PM 10) Chemical Composition by U.S. Region C-l7
Table C-10 PM Control Scenario Air Quality Profile Filenames C-l 7
Table C-l 1 PM' No-Control Scenario Air Quality Profile Filenames C-l 8
Table C-l2 Urban Areas Modeled with OZIPM4 C-l9
Table C-13 Summary of Ozone Monitoring Data C-21
Table C-14 Apportionment of Emissions Inventories for SAQM Runs C-22
Table C-15 1990 Control Scenario Visibility Conditions for 30 Southwestern U.S. Cities C-27
Table C-16 1990 No-control Scenario Visibility Conditions for 30 Southwestern U.S. Cities C-27
Table C-17 Summary of Relative Change in Visual Range and DeciView Between 1990
Control and No-Control Scenario Visibility Conditions for 30 Southwestern
U.S. Cities " C-29
Table D-l Criteria Air Pollutant Monitors in the U.S., 1.970 - 1990 D-3
Table D-2 Population Coverage in the "Within 50 km" Model Runs
(Percent of Continental U.S. Population) D-4
Table D-3 Population Coverage for "Extrapolated to All U.S." Model Runs (Percent of
Continental U.S. Population) D-5
Table D-4 Human Health Effects of Criteria Pollutants D-6
Table D-5 PM2.5/PM10 Ratios Used to Estimate PM2.5 Data Used With Pope et al. (1995)
Mortality Relationship D-l6
Table D-6 Summary of Concentration-Response Functions for Particulate Matter D-20
Table D-7 Summary of Concentration-Response Functions for Ozone D-27
Table D-8 Summary of Concentration-Response Functions for N02 D-35
Table D-9 Summary of Concentration-Response Functions for Carbon Monoxide D-3 7
Table D-10 Summary of Concentration-Response Functions for Sulfur Dioxide D-39
Table D-l 1 Selected Welfare Effects of Criteria Pollutants D-40
Table D-l 2 Summary of Functions Quantifying Welfare Benefits D-42
Table D-13 Criteria Pollutants Health Effects — Extrapolated to 48 State U.S. Population
(Cases Per Year-Mean Estimates) D-45
Table D-14 Mortality Distribution by Age: Proportion of PM- and Pb-related Premature
Mortalities and Associated Life Expectancies D-46
Table D-l5 Quantified Benefits Which Could Not Be Monetized — Extrapolated to the
Entire 48 State Population D-47
Table E-l Summary of Biological Changes with Surface Water Acidification E-3
Table E-2 Comparison of Population of Acidic National Surface Water Survey (NSWS) by
Chemical Category E-4
Table E-3 Results from Benefits Assessments of Aquatic Ecosystem Use Values from Acid
Deposition Avoidance E-6
xiii
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Table F-l Agriculture Exposure-Response Functions F-4
Table F-2 Relative No-Control to Control Percent Yield Change (harvested acres) for the
Minimum Scenario F-6
Table F-3 Relative No-Control to Control Percent Yield Change (harvested acres) for the
Maximum Scenario F-6
Table F-4 Change in Farm Program Payments, Net Crop Income, Consumer Surplus,
and Net Surplus Due to the CAA (Millions 1990 $) F-8
Table G-1 Quantified and Unquantified Health Effects of Lead G-1
Table G-2 Uncertainty Analysis: Distributions Associated With Dose-Response
Coefficients Used to Estimate Lead Health Effects G-16
Table G-3 Air Modeling Parameters G-24
Table G-4 Estimated Indirect Intake Slopes: Increment of Blood Lead Concentration
(in jog/dL) per Unit of Air Lead Concentration (|ig/m3) G-26
Table G-5 Estimated Lead Emissions from Electric Utilities, Industrial Processes,
and Industrial Combustion (in Tons) G-28
Table G-6 Yearly Differences in Number of Health Effects Between the Control and
No-Control Scenarios: Industrial Processes, Boilers, and Electric Utilities
(Holding Other Lead Sources at Constant 1970 Levels) G-29
Table G-7 Yearly Differences in Number of Health Effects Between the Controlled and
Uncontrolled Scenarios: Industrial Processes, Boilers, and Electric Utilities
(Holding Other Lead Sources at Constant 1990 Levels) G-30
Table G-8 Lead Burned in Gasoline (In Tons) G-33
Table G-9 Yearly Differences in Number of Health Effects Between the Control and
No-Control Scenarios: Lead in Gasoline only (Holding Other Lead Sources at
Constant 1970 Levels) G-34
Table G-10 Yearly Differences in Number of Health Effects Between the Control and
No-Control Scenarios: Lead in Gasoline only (Holding Other Lead Sources at
Constant 1990 Levels) G-35
Table H-l Health and Welfare Effects of Hazardous Air Pollutants H-2
Table H-2 Cancer Incidence Reductions and Monetized Benefits for NESHAPs H-5
Table I-l Summary of Mortality Valuation Estimates (Millions of 1990 Dollars) 1-3
Table 1-2 Unit Values Used for Economically Valuing Health and Welfare Endpoints 1-8
Table 1-3 Criteria Pollutants Health and Welfare Benefits — Extrapolated to Entire 48 State
Population Present Value (In 1990 Using 5% Discount Rate) of Benefits from
1970-1990 (In Billions of 1990 Dollars) 1-17
Table 1-4 Present Value of 1970 to 1990 Monetized Benefits by Endpoint Category for
48 State Population (Billions of 1990 Dollars, Discounted to 1990 at 5 Percent) I-l 8
Table 1-5 Monte Carlo Simulation Model Results for Target Years, Plus Present Value in
1990 Terms of Total Monetized Benefits for Entire 1970 to 1990 Period
(In Billions of 1990-Value Dollars) 1-18
Table 1-6 Comparison of 1990 (Single Year) Monetized Benefits by Endpoint for 48 State
Population and Monitored Areas (In Millions of 1990 Dollars) 1-19
Table 1-7 Effect of Alternative Discount Rates on Present Value of Total Monetized
Benefits for 1970 to 1990 (In Trillions of 1990 Dollars) 1-20
Table 1-8 Alternative Estimates of the Present Value of Mortality Associated With P\t
(Based on Pope et al., 1996, in Trillions of 1990 Dollars) 1-25
xiv
-------
Figures
Figure ES-1 Total Direct Compliance Costs of the CAA (in billions of inflation-adjusted
dollars.) ES-1
Figure ES-2 1990 Control and No-control Scenario Emissions (in millions of short tons) ES-2
Figure ES-3 Total Direct Costs and Monetized Direct Benefits of the Clean Air Act, 1970
to 1990 (in trillions of 1990 dollars) ES-8
Figure 1 Summary of Analytical Sequence and Modeled versus Historical Data Basis 4
Figure 2 Control and No-control Scenario Total SO, Emission Estimates 16
Figure 3 Control and No-control Scenario Total NO Emission Estimates 16
Figure 4 Control and No-control Scenario Total VOC Emission Estimates 16
Figure 5 Control and No-control Scenario Total CO Emission Estimates 16
Figure 6 Control and No-control Scenario Total TSP Emission Estimates 16
Figure 7 Control and No-control Scenario Total Pb Emission Estimates 16
Figure 8 Frequency Distribution of Estimated Ratios for 1990 Control to No-control
Scenario 95th Percentile 1-Hour Average CO Concentrations, by Monitor 21
Figure 9 Frequency Distribution of Estimated Ratios for 1990 Control to No-control
Scenario 95th Percentile 1-Hour Average SO . Concentrations, by Monitor 22
Figure 10 Frequency Distribution of Estimated Ratios for 1990 Control to No-control
Scenario 95th Percentile 1-Hour Average NO, Concentrations, by Monitor 23
Figure 11 Distribution of Estimated Ratios for 1990 Control to No-Control Annual Mean
TSP Concentrations, by Monitored County 23
Figure 12 Distribution of Estimated Ratios for 1990 Control to No-control OZIPM4
Simulated 1-Hour Peak Ozone Concentrations, by Urban Area 23
Figure 13 Distribution of Estimated Ratios for 1990 Control to No-control SAQM Simulated
Daytime Average Ozone Concentrations, by SAQM Monitor 24
Figure 14 Distribution of Estimated Ratios for 1990 Control to No-control RADM Simulated
Daytime Average Ozone Concentrations, by RADM Grid Cell 24
Figure 15 RADM-Predicted Percent Increase in Total Sulfur Deposition (Wet + Dry) Under the
No-control Scenario 24
Figure 16 RADM-Predicted Percent Increase in Total Nitrogen Deposition (Wet + Dry) Under
the No-control Scenario 25
Figure 17 RADM-Predicted Increase in Visibility Degradation, Expressed in DeciViews, for
Poor Visibility Conditions (90th Percentile) Under the No-control Scenario 25
Figure 18 Monte Carlo Simulation Model Results for Target Years (in billions of 1990 dollars) 54
Figure 19 Distribution of 1990 Monetized Benefits of CAA (in billions of 1990 dollars) 54
Figure 20 Uncertainty Ranges Deriving From Individual Uncertainty Factors 55
Figure A -1 Percent Difference in Real Investment Between Control and No-control Scenarios A-22
Figure A-2 Percent Difference in Price of Output by Sector Between Control and No-control
Scenario for 1990 A-22
Figure A-3 Percent Difference in Quantity of Output by Sector Between Control and
No-control Scenario for 1990 A-24
Figure A-4 Percent Difference in Employment by Sector Between Control and No-control
Scenario for 1990 A-24
Figure B-l Comparison of Control, No-control, and Trends SO, Emission Estimates B-2
Figure B-2 Comparison of Control, No-control, and Trends NO Emission Estimates B-2
Figure B-3 Comparison of Control, No-control, and Trends VOC Emission Estimates B-2
Figure B-4 Comparison of Control, No-control, and Trends CO Emission Estimates B-2
Figure B-5 Comparison of Control, No-Control, and Trends TSP Emission Estimates B-2
xv
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Figure C-l Frequency Distribution of Estimated Ratios for 1990 Control to No-Control
Scenario 95th Percentile 1.-Hour Average CO Concentrations, by Monitor C-4
Figure C-2 Frequency Distribution of Estimated Ratios for 1990 Control to No-control
Scenario 95th Percentile 1-Hour Average SO, Concentrations, by Monitor C-6
Figure C-3 Frequency Distribution of Estimated Ratios for 1990 Control to No-control
Scenario 95th Percentile 1-Hour Average NO, Concentrations, by Monitor C-8
Figure C-4 Location of the High Resolution RADM 20-km Grid Nested Inside the 80-km
RADM Domain C-9
Figure C-5 RADM-Predicted 1990 Total Sulfur Deposition (Wet + Dry; in kg/ha) Under the
Control Scenario C-10
Figure C-6 RADM-Predicted 1990 Total Nitrogen Deposition (Wet + Dry: in kg/ha) Under the
Control Scenario C-10
Figure C-7 RADM-Predicted 1990 Total Sulfur Deposition (Wet + Dry; in kg/ha) Under the
No-control Scenario C-l 1
Figure C-8 RADM-Predicted 1990 Total Nitrogen Deposition (Wet + Dry; in kg/ha) Under the
No-control Scenario C-l 1
Figure C-9 RADM-Predicted Percent Increase in Total Sulfur Deposition (Wet + Dry; in kg/ha)
Under the No-control Scenario C-l 2
Figure C-10 RADM-Predicted Percent Increase in Total Nitrogen Deposition (Wet + Dry;
in kg/ha) Under the No-control Scenario C-l 2
Figure C-l 1 Distribution of Estimated Ratios for 1990 Control to No-Control Annual Mean
TSP Concentrations, by Monitored Count}* C-l8
Figure C-12 RADM and SAQM Modeling Domains, with Rural Ozone Monitor Locations C-20
Figure C-l 3 Distribution of Estimated Ratios for 1990 Control to No-control OZFPM4
Simulated 1-Hour Peak Ozone Concentrations, by Urban Area C-23
Figure C-14 Distribution of Estimated Ratios for 1990 Control to No-control RADM-Simulated
Daytime Average Rural Ozone Concentrations, by RADM Grid Cell C-23
Figure C-l 5 Distribution of Estimated Ratios for 1990 Control to No-control SAQM-Simulated
Daytime Average Ozone Concentrations, by SAQM Monitor C-23
Figure C-16 RADM-Predicted Visibility Degradation, Expressed in Annual Average
DeciView, for Poor Visibility Conditions (90th Percentile) Under the Control
Scenario C-2 6
Figure C-l7 RADM-Predicted Visibility Degradation, Expressed in Annual Average
DeciView, for Poor Visibility Conditions (90th Percentile) Under the No-Control
Scenario C-2 6
Figure C-l8 RADM-Predicted Increase in Visibility Degradation, Expressed in Annual
Average DeciView, for Poor Visibility Conditions (90th Percentile) Under the
No-Control Scenario C-2 8
Figure 11-1 PES Estimated Reductions in HAP-Related Cancer Cases H-7
Figure H-2 ICF Estimated Reductions in Total HAP-Related Cancer Cases Using Upper
Bound Asbestos Incidence and Lower Bound Non-Asbestos H AP Incidence H-8
Figure H-3 ICF Estimated Reduction in Total 11A P-Related Cancer Cases Using Upper
Bound Incidence for All HAPs H-8
Figure H-4 National Annual Average Motor Vehicle HAP Exposures (|jg/m3) H-l I
Figure I-1 Monte Carlo Simulation Model Results for Target Years
(in billions of 1990 dollars) 1-19
Figure 1-2 Uncertainty Ranges Deriving From Individual Uncertainty Factors 1-21
-------
Acronyms and Abbreviations
^leq/L
microequivalents per liter
|ig/m3
micrograms per cubic meter
Mg
micrograms
(im
micrometers, also referred to as microns
ACCACAPERS
SAB Advisory Council on Clean Air Compliance Analysis Physical
Effects Review Subcommittee
agsim:
AGricultural Simulation Model
AIRS
EPA Aerometric Information Retrieval System
Al3+
aluminum
ANC
acid neutralizing capacity
ANL
Argonne National Laboratories
APPI
Argonne Power Plant Inventory
AQCR
Air Quality Control Region
ARGUS
Argonne Utility Simulation Model
ASI
Acid Stress Index
ATERIS
Air Toxic Exposure and Risk Information System
ATLAS
Aggregate Timberland Assessment System
AUSM
Advanced Utility Simulation Model
BEA
Bureau of Economic Analysis
K
total light extinction
BG/ED
Block Group / Enumeration District
Bi
atherothrombotic brain infarction
BID
Background Information Document
BP
blood pressure
BTU
British Thermal Unit
c.i.
confidence interval
CA
cerebrovascular accident
CAA
Clean Air Act
CAAA90
Clean Air Act Amendments of 1990
CAPMS
EPA's Criteria Air Pollutant Modeling System
CARB
California Air Resources Board
CASAC
SAB Clean Air Scientific Advisory Committee
CDC
Centers for Disease Control (now CDCP, Centers for Disease Control
and Prevention)
CTRL
EPA/ORD Corvallis Environmental Research Laboratory (old name; see
NERL)
CEUM
ICF Coal and Electric Utility Model
CHD
coronary heart disease
CIPP
changes in production processes
CO
carbon monoxide
CO,
carbon dioxide
COM
coefficient of haze
COHb
blood level of carboxyhemoglobin
COPD
chronic obstructive pulmonary disease
Council
SAB Advisory- Council 011 Clean Air Compliance Analysis
CPUE
catch per unit effort
x\ni
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
( R concentration-response
CRESS Commercial and Residential Simulation System model
CSTM Coal Supply and Transportation Model
CTG Control Techniques Guidelines
CV contingent valuation
CVM contingent valuation method
D C. District of Columbia
DBP diastolic blood pressure
DDE dichlorodiphenyldichloroethylene
DDT diclilorodiplicnyltrichloroethanc
DFEVj decrement of forced expiratory volume (in one second)
dL deciliter
DOC Department of Commerce
DOE Department of Energy
DOI Department of Interior
DRi Data Resources Incorporated
dV DeciView Haze Index
DVSAM Disaggregate Vehicle Stock Allocation Model
EC extinction coefficient
EDB ethylene dibromide
EDO ethylene dichloride
EFI Electronic Fuel Injection
EI Electronic Ignition
EIA Energy Information Administration
EKMA Empirical Kinetics Modeling Approach
ELI Environmental Law Institute
EOL cnd-of-linc
EPA Environmental Protection Agency
EPRI Electric Power Research Institute
ESEERCO Empire State Electric Energy Research Corporation
ESP electrostatic precipitator
FERC Federal Energy Regulatory Commission
FEY! forced expiratory volume (in one second)
FGD flue gas desulfurization
FHWA Federal Highway Administration
FIFRA Federal Insecticide, Fungicide, and Rodcnticide Act
FIP Federal Information Processing System
FR Federal Register
FRP Forest Response Program
GDP gross domestic product
GEMS Graphical Exposure Modeling System
GM geometric mean
GNP Gross National Product
GSD geometric standard deviation
H SO, sulfuric acid
2 4
ha hectares
HAP Hazardous Air Pollutant
HAPEM-MS Hazardous Air Pollutant Exposure Model - Mobile Source
UNO, nitric acid
hp horsepower
HTCM Hedonic Travel-Cost Model
ICARUS Investigation of Costs and Reliability in Utility Systems
xviii
-------
A crony ms and A bbrevi at ions
ICD-9
International Classification of Diseases, Ninth Version (1975 Revision)
ICE
Industrial Combustion Emissions model
lEc
Industrial Economics, Incorporated
FEUBK
EPA's Integrated Exposure Uptake Biokinetic model
IMS
Integrated Model Set
IFF
iterative proportional fitting
IQ
intelligence quotient
ISO T
Industrial Source Complex Long Term air quality model
J/W
Jorgenson / Wilcoxen
kg
kilograms
km
kilometers
lbs
pounds
LRI
lower respiratory illness
111/s
meters per second
m
meters
m3
cubic meters
Mm
megameters
MMBTU
million BTU
M0BILE5a
EPA's mobile source emission factor model
mpg
miles per gallon
MR AD
minor restricted activity day
MSCET
Month and State Current Emission Trends
MTD
metric tons per day
MVATS
EPA's Motor Vehicle-Related Air Toxics Study
MVMA
Motor Vehicle Manufacturers Association
Mwe
megawatt equivalent
N
nitrogen
NA
not available
NAAQS
National Ambient Air Quality Standard
NAPAP
National Acid Precipitation Assessment Program
NARSTO
North American Research Strategy for Tropospheric Ozone
NATICH
National Air Toxics Information Clearinghouse
NCLAN
National Crop Loss Assessment Network
NEA
National Energy Accounts
NERA
National Economic Research Associates
NERC
North American Electric Reliability Council
NERL
EPA ORD National Exposure Research Laboratory (new name for
CERL)
NESHAP
National Emission Standard for Hazardous Air Pollutants
NHANES
First National Health and Nutrition Examination Survey
NHANES II
Second National Health and Nutrition Examination Survey
N i PA
National Income and Product Accounts
NMOCs
nonmethane organic compounds
NO
nitric oxide
NO,
nitrogen dioxide
NO-
nitrate ion
NO
X
nitrogen oxides
NPTS
Nationwide Personal Transportation Survey
NSPS
New Source Performance Standards
NSWS
National Surface Water Survey
O&M
operating and maintenance
03
ozone
xix
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
OAQPS EPA/OAR Office of Air Quality Planning and Standards
OAR EPA Office of Air and Radiation
OMS EPA/OAR Office of Mobile Sources
OPAR EPA/OAR Office of Policy Analysis and Review
OPPE EPA Office of Policy Planning and Evaluation
ORD EPA Office of Research and Development
OZIPM4 Ozone Isopleth Plotting with Optional Mechanism-IV
PACE Pollution Abatement Costs and Expenditures survey
PAN peroxyacetyl nitrate
PAPE Pollution Abatement Plant and Equipment surv ey
Pb lead
PbB blood lead level
PCB pol\chlorinated biphenyl
PES Pacific Environmental Services
pH the logarithm of the reciprocal of hydrogen ion concentration, a measure
of acidity
PIC product of incomplete combustion
PM10 particulates less than or equal to 10 microns in aerometric diameter
PM ., particulates less than or equal to 2.5 microns in aerometric diameter
POP population
PopmiH exposed population of exercising mild asthmatics
Popmod exposed population of exercising moderate asthmatics
ppb parts per billion
PPH people per household
pphm parts per hundred million
ppm parts per million
PPRG Pooling Project Research Group
PRYL percentage relative yield loss
PURHAPS PURchased Heat And Power
PVC polyvinyl chloride
r2 statistical correlation coefficient, squared
RAD restricted activity day
RADM Regional Acid Deposition Model
RADM/EM RADM Engineering Model
RAMC Resource Allocation and Mine Costing model
RfD reference dose
RIA Regulatory Impact Analysis
ROM Regional Oxidant Model
RRAD respiratory restricted activity day
RUM Random Utility Model
s.e. standard error
SAB Science Advisor}' Board
SAI Systems Applications International
SAQM SARMAP Air Quality Model
SARA Superfund Amendment Reauthorization Act
SARMAP SJVAQS/AUSPEX Regional Modeling Adaptation Project
SCC Source Classification Code
SEDS State Energy Data System
SIC Standard Industrial Classification
SIP State Implementation Plan
SJVAQS San Joaquin Valley Air Quality Study
SMSA Standard Metropolitan Statistical Area
xx
-------
A crony ms and A bbrevi at ions
so.
sulfur dioxide
SOj'
sulfate ion
SOS/T
State of Science and Technology (refers to a series of NAPAP reports)
SRaw
Specific Airway Resistance
STAR
Stability Array weather database
TAMM90
Timber Assessment Market Model (revised version)
TEEMS
Transportation Energy and Emissions Modeling System
TIUS
Truck Inventory and Use Surveys
TR1
Toxic Release Inventory
TSP
total suspended particulate
U.S.
United States
UAM
Urban Airshed Model
URI
upper respiratory illness
USD A
United States Department of Agriculture
USEPA
United States Environmental Protection Agency
vc
vinyl chloride
VMT
vehicle miles traveled
YOC
volatile organic compounds
VOP
Vehicle Ownership Projection
VR
visual range
W126
index of peak weighted average of cumulative ozone concentrations
WLD
Work Loss Day-
WTP
willingness to pay
xxi
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
-------
Acknowledgments
This project is managed under the direction of Robert D. Brenner, Director of the U.S. EPA Office of Air
and Radiation/Office of Policy Analysis and Review and Richard D. Morgcnstern, Associate Assistant Admin-
istrator for Policy Planning and Evaluation, U.S. EPA (currently on leave as Visiting Scholar, Resources for the
Future). Hie principal project managers are Jim Do Mocker, EPA/OAR/OPAR; A1 McGartland, Director, EPA/
OPPE/OEE; and Tom Gillis, EPA/OPPE/OEE.
Many EPA staff contributed or reviewed portions of this draft document, including Joel Schwartz, Michael
Shapiro, Peter Preuss, Tracey Woodruff, Diane DeWitt. Dan Axelrad, Joel Scheraga, Anne Grambsch. Jenny
Weinberger, Allyson Siwik, Richard Scheffe, Vasu Kilaru, Amy Vasu. Kathy Kaufmann, Mary Ann Stewart,
Eric Smith, Dennis J. Kotchmar, Warren Freas, Tom Braverman, Bruce Polkowsky, David Mobley, Sharon
Nizich, David Meisenheimer, Fred Dimmick, Harvey Richmond, John Haines, John Bachmann, Ron Evans,
Tom Mc Mullen. Bill Vatavuk, Larry Sorrels, Dave McKee, Susan Stone, Melissa McCullough, Rosalina
Rodriguez, Vickie Boothe, Tom Walton, Michele McKeever, Vicki Atwell, Kelly Rimer, Bob Fegley, Aparna
Koppikar, Les Grant, Judy Graham, Robin Dennis, Dennis Leaf, Ann Watkins, Penny Carey, Joe Somers, Pam
Brodowicz, Byron B linger, Allen Basala, David Lee, Bill O'Neill, Susan Herrod, and Susan Stendebach. Allyson
Siwik of EPA/OAR/OAQPS and Bob Fegley of EPA/ORD/OSPRE played particularly important roles in coor-
dinating substantive and review contributions from their respective offices.
A number of contractors developed key elements of the analysis and supporting documents. These contrac-
tors include Bob Unsworth, Jim Neumann, Mike Hester, and Jon Discher of Industrial Economics, Incorporated
(lEc); Leland Deck, Ellen Post, Lisa Akeson, Brad Firlie, Susan Keane, Kathleen Cunningham, and John Voyzey
of Abt Associates; Bruce Braine, Patricia Kim, Sandeep Kohli, Anne Button, Bam? Galef. Cynde Sears, and
Tony Bansal of ICF Resources; John Langstaff. Michelle Woolfolk, Shelly Eberly, Chris Emery, Till Stoekenius,
and Andy Gray of ICF/Systems Applications International (ICF/SAI); Dale Jorgenson, Peter Wilcoxen, and
Richard Goettle of Jorgenson Associates; Jim Lockhart of the Environmental Law Institute (ELI); Beverly
Goodrich, Rehan Aziz, Noel Roberts, and Lucille Bender of Computer Sciences Corporation; Margaret Sexsmith
of Analytical Sciences, Incorporated; Ken Meardon of Pacific Environmental Services (PES); David South,
Gale Boyd, Mclanie Tomkins, and K. Guziel of Argonne National Laboratory (ANL); Don Garner; Rex Brown
and Jacob Ulvila of Decision Science Consortium; and Jim Wilson and Dianne P. Crocker of Pechan Associ-
ates. John Pitcher and H. Glenn Court of STRA managed the technical production of an earlier version of the
draft document. The SARMAP AQM runs were provided by Carol Bohnenkamp of EPA Region 9 and Saffet
Tanrikulu of the California Air Resources Board.
Science Advisory Board review of this report is supervised by Donald G. Barnes, Director of the SAB Staff.
SAB staff coordinating the reviews have included Jack Fowle, Jack Kooyoomjian, Sam Rondberg, Fred Talcott,
and Randall Bond. Diana Pozun provided administrative support.
The SAB Council was chaired by Richard Schmalensee of MIT throughout the development of the present
study. The Council is now chaired by Maureen Cropper of the World Bank as the Council's focus shifts to the
upcoming prospective studies. Members who have participated in the review of this draft report include Morton
Lippmann of New York University Medical Center, William Nordhaus of Yale University, Paul Portney of
Resources for the Future, Kip Viscusi of Harvard University, A. My rick Freeman of Bowdoin College, Maureen
Cropper, Ronald Cummings of Georgia State University, Daniel Dudek of the Environmental Defense Fund,
Robert Mendelsohn of Yale University-, Wayne Rachel of MELE Associates, William Cooper of Michigan
State University, Thomas Tietenberg of Colby College, Paul Lioy of the Robert Wood Johnson School of
Medicine, Roger McClellan of the Chemical Industry Institute of Toxicology, George T. Wolff of General
Motors, Richard Conway of Union Carbide Corporation, and Wallace Gates of the University of Maryland.
xxiii
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
The SAB Council Physical Effects Review Subcommittee was chaired by Morton Lippmann. Members
who have participated in the review include David V. Bates of the University of British Columbia, A. Myrick
Freeman of Bowdoin College, Gardner Brown, Jr. of the University of Washington, Timothy Larson of the
University of Washington, Lester Lave of Carnegie Mellon University, Joseph Meyer of the University of
Wyoming, Robert Rowe of Hagler Bailly, Incorporated, George Taylor of the University of Nevada, Bernard
Weiss of the University of Rochester Medical Center, and George Wolff of the General Motors Research Labo-
ratory.
The SAB Council Air Quality Subcommittee was chaired by George Wolff. Members who have partici-
pated in the review include Benjamin Liu of the University of Minnesota, Peter Mueller of the Electric Power
Research Institute, Warren White of Washington University, Joe Mauderly of the Lovelace Biomedical & En-
vironmental Research Institute, Philip Hopke of Clarkson University, Paulette Middleton of Science Policy
Associates, James H. Price, Jr. of the Texas Natural Resource Conservation Commission, and Harvey Jeffries of
the University of North Carolina, Chapel Hill.
This report could not have been produced without the support of key administrative support staff. The
project managers are grateful to Cat rice Jefferson, Nona Smoke, Carolyn Hicks, Eunice Javis. Gloria Booker,
Thelma Butler, Wanda Farrar, Ladonya Langston, Michelle Olawiiyi. and Eileen Pritchard for their timely and
tireless support on this project.
xxiv
-------
Executive Summary
Purpose of the Study
Throughout the history of the Clean Air Act, ques-
tions have been raised as to whether the health and
environmental benefits of air pollution control justify
the costs incurred by industry, taxpayers, and consum-
ers. For the most part, questions about the costs and
benefits of individual regulatory standards continue
to be addressed during the regulatory development
process through Regulatory Impact Analyses (RIAs)
and other analyses which evaluate regulatory costs,
benefits, and such issues as scope, stringency, and tim-
ing. There has never been, however, any comprehen-
sive, long-term, scientifically valid and reliable study
which answered the broader question:
"How do the overall health, welfare,
ecological, and economic benefits of Clean
Air Act programs compare to the costs of
these programs?"
To address this void, Congress added to the 1990
Clean Air Act Amendments a requirement under sec-
tion 812 that EPA conduct periodic, scientifically re-
viewed studies to assess the benefits and the costs of
the Clean Air Act. Congress further required EPA to
conduct the assessments to reflect central tendency,
or "best estimate," assumptions rather than the con-
servative assumptions sometimes deemed appropri-
ate for setting protective standards.
This report is the first in this ongoing series of
Reports to Congress. By examining the benefits and
costs of the 1970 and 1977 Amendments, this report
addresses the question of the overall value of
America's historical investment in cleaner air. The
first Prospective Study, now in progress, will evalu-
ate the benefits and costs of the 1990 Amendments.
Study Design
Estimates of the benefits and costs of the histori-
cal Clean Air Act are derived by examining the dif-
ferences in economic, human health, and environmen-
tal outcomes under two alternative scenarios: a "con-
trol scenario" and a "no-control scenario." The con-
trol scenario reflects actual historical implementation
of clean air programs and is based largely on histori-
cal data. The no-control scenario is a hypothetical sce-
nario which reflects the assumption that no air pollu-
tion controls were established beyond those in place
prior to enactment of the 1970 Amendments. Each of
the two scenarios is evaluated by a sequence of eco-
nomic, emissions, air quality, physical effect, eco-
nomic valuation, and uncertainty models to measure
the differences between the scenarios in economic,
human health, and environmental outcomes. Details
of this analytical sequence are presented in Chapter 1
and are summarized in Figure 1 of that chapter.
Study Review
EPA is required, under section 812, to consult both
a panel of outside experts and the Departments of
Labor and Commerce in designing and implementing
the study.
The expert panel was organized in 1991 as the
Advisory Council on Clean Air Act Compliance
Analysis (hereafter "Council") under the auspices of
EPA's Science Advisory Board (SAB). Organizing
the external panel under the auspices of the SAB en-
sured that the peer review of the study would be con-
ducted in a rigorous, objective, and publicly open
manner. Eminent scholars and practitioners with ex-
pertise in economics, human health sciences, envi-
ronmental sciences, and air quality modeling served
on the Council and its technical subcommittees, and
these reviewers met many times throughout the de-
sign and implementation phases of the study. During
this ongoing, in-depth review, the Council provided
valuable advice pertaining to the development and
selection of data, selection of models and assumptions,
evaluation and interpretation of the analytical find-
ings, and characterization of those findings in several
successive drafts of the Report to Congress. The
present report was vastly improved as a result of the
Council's rigorous and constructive review effort.
ES-1
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
With respect to the interagency review process,
EPA expanded the list of consulted agencies and con-
vened a series of meetings during the design and early
implementation phases from 1991 through late 1994.
In late 1994, to ensure that all interested parties and
the public received consistent information about re-
maining analytical issues and emerging results, EPA
decided to use the public SAB review process as the
primary forum for presenting and discussing issues
and results. The Interagency Review Group was there-
fore discontinued as a separate process in late 1994.
A final, brief interagency review, pursuant to Cir-
cular A-19, was organized in August 1997 by the Of-
fice of Management and Budget and conducted fol-
lowing the completion of the extensive expert panel
peer review by the SAB Council. During the course
of the final interagency discussions, it became clear
that several agencies held different views pertaining
to several key assumptions in this study as well as to
the best techniques to apply in the context of environ-
mental program benefit-cost analyses, including the
present study. The concerns include: (1) the extent to
which air quality would have deteriorated from 1970
to 1990 in the absence of the Clean Air Act, (2) the
methods used to estimate the number of premature
deaths and illnesses avoided due to the CAA, (3) the
methods used to estimate the value that individuals
place on avoiding those risks, and (4) the methods
used to value non-health related benefits. However,
due to the court deadline the resulting concerns were
not resolved during this final, brief interagency re-
view. Therefore, this report reflects the findings of
EPA and not necessarily other agencies in the Ad-
ministration. Interagency discussion of some of these
issues will continue in the context of the future pro-
spective section 812 studies and potential regulatory
actions.
Summary of Results
Direct Costs
To comply with the Clean Air Act, businesses,
consumers, and government entities all incurred higher
costs for many goods and services. The costs of pro-
viding goods and services to the economy were higher
primarily due to requirements to install, operate, and
maintain pollution abatement equipment. In addition,
costs were incurred to design and implement regula-
tions, monitor and report regulatory compliance, and
invest in research and development. Ultimately, these
higher costs of production were borne by stockhold-
ers, business owners, consumers, and taxpayers.
Figure ES-1 summarizes the historical data on
Clean Air Act compliance costs by year, adjusted both
for inflation and for the value of long-term invest-
ments in equipment. Further adjusting the direct costs
incurred each year to reflect their equivalent worth in
the year 1990, and then summing these annual results,
yields an estimate of approximately $523 billion for
the total value of 1970 to 1990 direct expenditures
(see Appendix A for calculations).
Emissions
Emissions were substantially lower by 1990 un-
der the control scenario than under the no-control sce-
nario, as shown in Figure ES-2. Sulfur dioxide (S02)
emissions were 40 percent lower, primarily due to
utilities installing scrubbers and/or switching to lower
sulfur fuels. Nitrogen oxides (NOx) emissions were
30 percent lower by 1990, mostly because of the in-
stallation of catalytic converters on highway vehicles.
Volatile organic compound (VOC) emissions were 45
percent lower and carbon monoxide (CO) emissions
were 50 percent lower, also primarily due to motor
vehicle controls.
For particulate matter, it is important to recog-
nize the distinction between reductions in directly
emitted particulate matter and reductions in ambient
concentrations of particulate matter in the atmosphere.
As discussed further in the next section, changes in
particulate matter air quality depend both on changes
in emissions of primary particles (i.e., air pollution
which is already in solid particle form) and on changes
in emissions of gaseous pollutants, such as sulfur di-
oxide and nitrogen oxides, which can be converted to
particulate matter through chemical transformation in
the atmosphere. Emissions of primary particulates
Figure ES-1. Total Estimated Direct Compliance Costs of
the CAA (in billions of inflation-adjusted dollars).
1975 1980 1985 1990
Year
ES-2
-------
Executive Summary
Figure ES-2. 1990 Control and No-control Scenario
Emissions (in millions of short tons).
inn
1 Emissions (millions of short tons)
| © o S ©
TSP S02 NOx VOC CO
Pollutant
m No-control
11 Control
were 75 percent lower under the control scenario by
1990 than under the no-control scenario. This sub-
stantial difference is primarily due to vigorous efforts
in the 1970s to reduce visible emissions from utility
and industrial smokestacks.
Lead (Pb) emissions for 1990 are reduced by about
99 percent from a no-control level of 237,000 tons to
about 3,000 tons under the control scenario.1 The vast
majority of the difference in lead emissions under the
two scenarios is attributable to reductions in the use
of leaded gasoline.
These reductions were achieved during a period
in which population grew by 22.3 percent and the na-
tional economy grew by 70 percent.
Air Quality
The substantial reductions in air pollutant emis-
sions achieved by the Clean Air Act translate into sig-
nificantly improved air quality throughout the U.S.
For sulfur dioxide, nitrogen oxides, and carbon mon-
oxide, the improvements in air quality under the con-
trol scenario are assumed to be proportional to the
estimated reduction in emissions. This is because, for
these pollutants, changes in ambient concentrations
in a particular area are strongly related to changes in
emissions in that area. While the differences in con-
trol and no-control scenario air quality for each of these
pollutants vary from place to place because of local
variability in emissions reductions, by 1990 the na-
tional average improvements in air quality for these
pollutants were: 40 percent reduction in sulfur diox-
ide, 30 percent reduction in nitrogen oxides, and 50
percent reduction in carbon monoxide.
Ground-level ozone is formed by the chemical re-
action of certain airborne pollutants in the presence
of sunlight. Reductions in ground-level ozone are
therefore achieved through reductions in emissions
of its precursor pollutants, particularly volatile organic
compounds (VOCs) and nitrogen oxides (NOx).2 The
differences in ambient ozone concentrations estimated
under the control scenario vary significantly from one
location to another, primarily because of local differ-
ences in the relative proportion of VOCs and NOx,
weather conditions, and specific precursor emissions
reductions. On a national average basis, ozone con-
centrations in 1990 are about 15 percent lower under
the control scenario. For several reasons, this overall
reduction in ozone is significantly less than the 30
percent reduction in precursor NOx and 45 percent
reduction in precursor VOCs. First, significant natu-
ral (i.e., biogenic) sources of VOCs limit the level of
ozone reduction achieved by reductions in man-made
(i.e., anthropogenic) VOCs. Second, current knowl-
edge of atmospheric photochemistry suggests that
ozone reductions will tend to be proportionally smaller
than reductions in precursor emissions. Finally, the
plume model system used to estimate changes in ur-
ban ozone for this study is incapable of handling long-
range transport of ozone from upwind areas and multi-
day pollution events in a realistic manner.
There are many pollutants which contribute to
ambient concentrations of particulate matter. The rela-
tive contributions of these individual pollutant spe-
cies to ambient particulate matter concentrations vary
from one region of the country to the next, and from
urban areas to rural areas. The most important par-
ticle species, from a human health standpoint, may be
the fine particles which can be respired deep into the
lungs. While some fine particles are directly emitted
by sources, the most important fine particle species
are formed in the atmosphere through chemical con-
version of gaseous pollutants. These species are re-
ferred to as secondary particles. The three most im-
portant secondary particles are (1) sulfates, which
derive primarily from sulfur dioxide emissions; (2)
nitrates, which derive primarily from nitrogen oxides
emissions; and (3) organic aerosols, which can be di-
rectly emitted or can form from volatile organic com-
1 Results for lead are not shown in Figure ES-2 because the absolute levels of lead emissions are measured in thousands, not
millions, of tons and will not be discernible on a graph of this scale.
2 Ambient NOx concentrations are driven by anthropogenic emissions whereas ambient VOCs result from both anthropogenic
and biogenic sources (e.g., terpenes emitted by trees).
ES-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
n The following additional human welfare effects were quantified directly in economic terms: household soiling
damage, visibility impairment, decreased worker productivity, and agricultural yield changes.
12 The 5th and 95th percentile outcomes represent the lower and upper bounds, respectively, of the 90 percent
credible interval for each effect as estimated by uncertainty modeling. The mean is the arithmetic
average of all estimates derived by the uncertainty modeling. See Chapter 7 and Appendix I for details.
/3 In this analysis, PM is used as a proxy pollutant for all non-Lead (Pb) criteria pollutants which may contribute
to premature mortality. See Chapter 5 and Appendix D for additional discussion.
Tabic ES-1. Criteria Pollutant Health Benefits — Estimated Distributions of 1990 Incidences
of Avoided Health Effects (in thousands of incidences reduced) for 48 State Population, i
Atlci'tl'd
Population
Annual Effects Avoided :
(thousands)
Kn d point
l'olliitiintts)
5th
%ile
Mean
95th
%ile
I in it
Premature Mortality
PM'3
30 and over
112
184
257
cases
Premature Mortality
Lead
all
7
22
54
cases
Chronic Bronchitis
PM
all
498
674
886
cases
Lost IQ Points
Lead
children
7,440
10,400
13,000
points
IQ less than 70
Lead
children
31
45
60
cases
Hypertension
Lead
men 20-74
9,740
12,600
15,600
cases
Coronary Heart Disease
Lead
40-74
0
22
64
cases
Atherothrombotic brain infarction
Lead
40-74
0
4
15
cases
Initial cerebrovascular accident
Lead
40-74
0
6
19
cases
Hospital Admissions
All Respiratory
PM &
Ozone
all
75
89
103
cases
Chronic Obstructive Pulmonary
PM &
Ozone
over 65
52
62
72
cases
Disease & Pneumonia
Ischemic Heart Disease
PM
over 65
7
19
31
cases
Congestive Heart Failure
PM &
CO
65 and over
28
39
50
cases
Other Respiratory-Related Ailments
Shortness of breath, days
PM
children
14,800
68,000
133,000
days
Acute Bronchitis
PM
children
0
8,700
21,600
cases
Upper & Lower Respiratory
PM
children
5,400
9,500
13,400
cases
Symptoms
Any of 19 Acute Symptoms
PM &
Ozone
18-65
15,400
130,000
244,000
cases
Asthma Attacks
PM &
Ozone
asthmatics
170
850
1,520
cases
Increase in Respiratory Illness
NC)2
all
4,840
9,800
14,000
cases
Any Symptom
S02
asthmatics
26
264
706
cases
Restricted Activity and Work Loss Days
Minor Restricted Activity Days
PM &
Ozone
18-65
107,000
125,000
143,000
days
Work Loss Days
PM
18-65
19,400
22,600
25,600
days
pound emissions. This highlights an important and
unique feature of particulate matter as an ambient pol-
lutant: more than any other pollutant, reductions in
particulate matter are actually achieved through re-
ductions in a wide variety of air pollutants. In other
words, controlling particulate matter means control-
ling "air pollution" in a very broad sense. In the present
analysis, reductions in sulfur dioxide, nitrogen oxides,
volatile organic compounds, and directly-emitted pri-
mary particles achieved by the Clean Air Act result in
a national average reduction in total suspended par-
ticulate matter of about 45 percent by 1990. For the
smaller particles which are of greater concern from a
health effects standpoint (i.e., PM10 and PM25), the
national average reductions were also about 45 per-
cent.
Reductions in sulfur dioxide and nitrogen oxides
also translate into reductions in formation, transport,
and deposition of secondarily formed acidic com-
pounds such as sulfate and nitric acid. These are the
principal pollutants responsible for acid precipitation,
or "acid rain." Under the control scenario, sulfur and
nitrogen deposition are significantly lower by 1990
than under the no-control scenario throughout the 31
eastern states covered by EPA's Regional Acid Depo-
sition Model (RADM). Percentage decreases in sul-
fur deposition range up to more than 40 percent in the
upper Great Lakes and Florida-Southeast Atlantic
Coast areas, primarily because the no-control scenario
projects significant increases in the use of high-sulfur
fuels by utilities in the upper Great Lakes and Gulf
ES-4
-------
Executive Summary
Coast states. Nitrogen deposition is also signifi-
cantly lower under the control scenario, with per-
centage decreases reaching levels of 25 percent or
higher along the Eastern Seaboard, primarily due
to higher projected emissions of motor vehicle ni-
trogen oxides under the no-control scenario.
Finally, decreases in ambient concentrations of
light-scattering pollutants, such as sulfates and ni-
trates, are estimated to lead to perceptible improve-
ments in visibility throughout the eastern states and
southwestern urban areas modeled for this study.
Physical Effects
The lower ambient concentrations of sulfur di-
oxide, nitrogen oxides, particulate matter, carbon
monoxide, ozone and lead under the control sce-
nario yield a substantial variety of human health,
welfare and ecological benefits. For a number of
these benefit categories, quantitative functions are
available from the scientific literature which allow
estimation of the reduction in incidence of adverse
effects. Examples of these categories include the
human mortality and morbidity effects of a num-
ber of pollutants, the neurobehavioral effects among
children caused by exposure to lead, visibility im-
pairment, and effects on yields for some agricul-
tural products.
A number of benefit categories, however, can
not be quantified and/or monetized for a variety of
reasons. In some cases, substantial scientific un-
certainties prevail regarding the existence and mag-
nitude of adverse effects (e.g., the contribution of
ozone to air pollution-related mortality). In other
cases, strong scientific evidence of an effect exists,
but data are still too limited to support quantitative
estimates of incidence reduction (e.g., changes in
lung function associated with long-term exposure
to ozone). Finally, there are effects for which there
is sufficient information to estimate incidence re-
duction, but for which there are no available eco-
nomic value measures; thus reductions in adverse
effects cannot be expressed in monetary terms. Ex-
amples of this last category include relatively small
pulmonary function decrements caused by acute
exposures to ozone and reduced time to onset of
angina pain caused by carbon monoxide exposure.
Table ES-1 provides a summary of the key dif-
ferences in quantified human health outcomes esti-
mated under the control and no-control scenarios.
Results are presented as thousands of cases avoided
in 1990 due to control of the pollutants listed in the
table and reflect reductions estimated for the entire
U.S. population living in the 48 continental states. Epi-
demiological research alone cannot prove whether a
cause-effect relationship exists between an individual
Table ES-2. Major Nonmonetized, Adverse Effects
Reduced by the Clean Air Act.
Pollutant
Particulate
Matter
Ozone
Carbon
Monoxide
Sulfur
Dioxide
Nitrogen
Oxides
Lead
Air Toxics
Nonmonetized Adverse Effects
Large Changes in Pulmonary Function
Other Chronic Respiratory Diseases
Inflammation ofthe Lung
Chronic Asthma and Bronchitis
Changes in Pulmonary Function
Increased Airway Responsiveness to Stimuli
Centroacinar Fibrosis
Inflammation ofthe Lung
Immunological Changes
Chronic Respiratory Diseases
Extrapulmonary Effects (i.e., other organ systems)
Forest and other Ecological Effects
Materials Damage
Decreased Time to Onset of Angina
Behavioral Effects
Other Cardiovascular Effects
Developmental Effects
Respiratory Symptoms in N on-Asthmatics
Hospital Admissions
Agricultural Effects
Materials Damage
Ecological Effects
Increased Airway Responsiveness to Stimuli
Decreased Pulmonary Function
Inflammation ofthe Lung
Immunological Changes
Eye Irritation
Materials Damage
Eutrophication (e.g., Chesapeake Bay)
Acid Deposition
Cardiovascular Diseases
Reproductive Effects in Women
Other N eurobehavioral, Physiological Effects in
Children
Developmental Effects from Maternal Exposure, inc
IQ Loss "
Ecological Effects
All Human Health Effects
Ecological Effects
" IQ loss from direct, as opposed to maternal, exposure is quantified and
monetized. See Tables ES-1 And ES-3.
ES-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
pollutant and an observed health effect. Although not
universally accepted, this study uses the epidemiologi-
cal findings about correlations between pollution and
observed health effects to estimate changes in the num-
ber of health effects that would occur if pollution lev-
els change. A range is presented along with the mean
estimate for each effect, reflecting uncertainties which
have been quantified in the underlying health effects
literature.
Adverse human health effects of the Clean Air
Act "criteria pollutants" sulfur dioxide,
nitrogen oxides, ozone, particulate mat-
ter, carbon monoxide, and lead dominate
the quantitative estimates in part be-
cause, although there are important re-
sidual uncertainties, evidence of physi-
cal consequences is greatest for these
pollutants. The Clean Air Act yielded
other benefits, however, which are im-
portant even though they are uncertain
and/or difficult to quantify. These other
benefit categories include (a) all benefits
accruing from reductions in hazardous
air pollutants (also referred to as air
toxics), (b) reductions in damage to cul-
tural resources, buildings, and other ma-
terials, (c) reductions in adverse effects
on wetland, forest, and aquatic ecosys-
tems, and (d) a variety of additional hu-
man health and welfare effects of crite-
ria pollutants. A more complete list of
these nonmonetized effects is presented
in Table ES-2.
In addition to controlling the six cri-
teria pollutants, the 1970 and 1977 Clean
Air Act Amendments led to reductions
in ambient concentrations of a small
number of hazardous air pollutants. Al-
though they are not fully quantified in
this report, control of these pollutants
resulted both from regulatory standards
set specifically to control hazardous air
pollutants and from incidental reductions
achieved through programs aimed at
controlling criteria pollutants.
Existing scientific research suggests
that reductions in both hazardous air
pollutants and criteria pollutants yielded
widespread improvements in the func-
tioning and quality of aquatic and ter-
restrial ecosystems. In addition to any intrinsic value
to be attributed to these ecological systems, human
welfare is enhanced through improvements in a vari-
ety of ecological services. For example, protection of
freshwater ecosystems achieved through reductions
in deposition of acidic air pollutants may improve com-
mercial and recreational fishing. Other potential eco-
logical benefits of reduced acid deposition include im-
proved wildlife viewing, maintenance of biodiversity,
and nutrient cycling. Increased growth and produc-
tivity of U.S. forests may have resulted from reduc-
TableES-3. Central Estimates of Economic Value per Unit of
Avoided Effect (in 1990 dollars).
TRinnrfl|h (r-hiiirnfl-
PrmUltrmltamnft
Waltrmaltivnitm (Tmmtffijmm twdt
Mortality
PM & Lead
$4,800,000
per case /I
Chronic Bronchitis
PM
$260,000
per case
IQ Changes
Lost IQ Points
Lead
$3,000
per IQ point
IQ less than 70
Lead
$42,00 0
per case
Hypertension
Lead
$68 0
per case
Strokes /2
Lead
$200,000
$150,000
per case-males3
per case-females3
Coronary Heart Disease
Lead
$52,000
per case
Hospital Admissions
Ischemic Heart Disease
PM
$10,300
per case
Congestive Heart Failure
PM
$8,300
per case
COPD
PM & Ozone
$8,100
per case
Pneumonia
PM & Ozone
$7,90 0
per case
All Respiratory
PM & Ozone
$6,100
per case
Respiratory Illness and Symptoms
Acute Bronchitis
PM
$45
per case
Acute Asthma
PM & Ozone
$32
per case
Acute Respiratory Symptoms
PM, Ozone, NO2,
S02
$1 8
per case
Upper Respiratory Symptoms
PM
$19
per case
Lower Respiratory Symptoms
PM
$12
per case
Shortness of Breath
PM
$5.30
per day
Work Loss Days
PM
$83
per day
Mild Restricted Activity Days
PM & Ozone
$3 8
per day
Welfare Benefits
Visibility
DeciV iew
$14
per unit change
in DeciView
Household Soiling
PM
$2.50
per household
per PM-10
ch ang e
Decreased Worker Productivity
Ozone
$1
/¦I
Agriculture (Net Surplus)
Ozone
Change in Economic Surplus
/I Alternative results, based on assigning a value of $293,000 for each life-year lost are
presented on pg. ES-9.
/2 Strokes are comprised of atherothrombotic brain infarctions and cerebrovascular
accidents; both are estimated to have the same monetary value.
/3 The different valuations for stroke cases reflect differences in lo st earnings between
males and females. See Appendix G for a more complete discussion ofvaluing
reductions in strokes.
/4 Decreased productivity valued as change in daily wages: $1 per worker per 10%
decrease in ozone.
ES-6
-------
Executive Summary
tions in ground-level ozone. More vigorous forest eco-
systems in turn yield a variety of benefits, including
increased timber production; improved forest aesthet-
ics for people enjoying outdoor activities such as hunt-
ing, fishing, and camping; and improvements in eco-
logical services such as nutrient cycling and tempo-
rary sequestration of global warming gases. These im-
provements in ecological structure and function have
not been quantified in this assessment.
Economic Valuation
Estimating the reduced incidence of physical ef-
fects provides a valuable measure of health benefits
for individual endpoints. However, to compare or ag-
gregate benefits across endpoints, the benefits must
be monetized. Assigning a monetary value to avoided
incidences of each effect permits a summation, in
terms of dollars, of monetized benefits realized as a
result of the Clean Air Act, and allows that summa-
tion to be compared to the cost of the Clean Air Act.
Before proceeding through this step, it is impor-
tant to recognize the substantial controversies and un-
certainties which pervade attempts to characterize ad-
verse human health and ecological effects of pollu-
tion in dollar terms. To many, dollar-based estimates
of the value of avoiding outcomes such as loss of hu-
man life, pain and suffering, or ecological degrada-
tion do not capture the full and true value to society as
a whole of avoiding or reducing these effects. Adher-
ents to this view tend to favor assessment procedures
which (a) adopt the most technically defensible dol-
lar-based valuation estimates for analytical purposes
but (b) leave the moral dimensions of policy evalua-
tion to those who must decide whether, and how, to
use cost-benefit results in making public policy deci-
sions. This is the paradigm adopted in the present
study. Given the Congressional mandate to perform a
cost-benefit study of the Clean Air Act, the Project
Team has endeavored to apply widely-recognized,
customary techniques of Applied Economics to per-
form this cost-benefit analysis. However, EPA be-
lieves there are social and personal values furthered
by the Clean Air Act which have not been effectively
captured by the dollar-based measures used in this
study. Therefore, EPA strongly encourages readers to
look beyond the dollar-based comparison of costs and
benefits of the Clean Air Act and consider the broader
value of the reductions in adverse health and environ-
mental effects which have been achieved as well as
any additional adverse consequences of regulation
which may not be reflected in the cost estimates re-
ported herein.
For this study, unit valuation estimates are derived
from the economics literature and reported in dollars
per case (or, in some cases, episode or symptom-day)
avoided for health effects and dollars per unit of
Tabic ES-4. Total Estimated Monetized Benefits by Endpoint Category lor 48 State Population
for 1970 to 1990 Period (in billions of 1990 dollars).
].nil point
l'ollu tiint(s)
Present Value
5th %ilc
Me a it
95th %ilc
Mortal ity
PM
$2,369
$16,632
$40,597
Mortality
Lead
$121
$1,339
$3,910
Chronic Bronchitis
PM
$409
$3,313
$10,401
IQ (Lost IQ Pts. + Children w
Lead
$271
$399
$551
IQ 70)
Hypertension
Lead
$77
$98
$120
Hospital Admissions
PM, Ozone, Lead, & CO
$27
$57
$120
Respiratory-Related
PM, Ozone, N02, & S02
$123
$182
$261
Symptoms, Restricted
Activity, & Decreased
Productivity
Soiling Damage
PM
$6
$74
$192
Visibility
particulates
$38
$54
$71
Agriculture (Net Surplus)
Ozone
$11
$23
S3 5
3 All of these summary results are present values of the 1970 to 1990 streams of benefits and costs, discounted at five percent.
ES-7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
avoided damage for human welfare effects. Similar to
estimates of physical effects provided by health stud-
ies, each of the monetary values of benefits applied in
this analysis can be expressed in terms of a mean value
and a range around the mean estimate. This range re-
flects the uncertainty in the economic valuation lit-
erature associated with a given effect. These value
ranges, and the approaches used to derive them, are
described in Chapter 6 and Appendix I for each of the
effects monetized in this study. The mean values of
these ranges are shown in Table ES-3.
Monetized Benefits and Costs
The total monetized economic benefit attributable
to the Clean Air Act is derived by applying the unit
values (or ranges of values) to the stream of
monetizable physical effects estimated for the 1970
to 1990 period. In developing these estimates, steps
are taken to avoid double-counting of benefits. In ad-
dition, a computer simulation model is used to esti-
mate ranges of plausible outcomes for the benefits
estimates reflecting uncertainties in the physical ef-
fects and economic valuation literature (see Chapter
7 and Appendix I for details).
The economic benefit estimation model then gen-
erated a range of economic values for the differences
in physical outcomes under the control and no-con-
trol scenarios for the target years of the benefits analy-
sis: 1975, 1980, 1985, and 1990. Linear interpolation
between these target years is used to estimate ben-
efits in intervening years. These yearly results are then
adjusted to their equivalent value in the year 1990 and
summed to yield a range and mean estimate for the
total monetized benefits of the Clean Air Act from
1970 to 1990. These results are summarized in Table
ES-4.
Combining these benefits results with the cost es-
timates presented earlier yields the following analyti-
cal outcomes.3
• The total monetized benefits of the Clean
Air Act realized during the period from
1970 to 1990 range from 5.6 to 49.4 trillion
dollars, with a central estimate of 22.2 tril-
lion dollars.
• By comparison, the value of direct compli-
ance expenditures over the same period
equals approximately 0.5 trillion dollars.
• Subtracting costs from benefits results in
net, direct, monetized benefits ranging
from 5.1 to 48.9 trillion dollars, with a cen-
tral estimate of 21.7 trillion dollars, for the
1970 to 1990 period.
• The lower bound of this range may go down
and the upper bound may go up if analyti-
cal uncertainties associated with compli-
ance costs, macroeconomic effects, emis-
sions projections, and air quality model-
ing could be quantified and incorporated
in the uncertainty analysis. While the range
already reflects many important uncertain-
ties in the physical effects and economic
valuation steps, the range might also
broaden further if additional uncertainties
in these two steps could be quantified.
• The central estimate of 22.2 trillion dollars
in benefits may be a significant underesti-
mate due to the exclusion of large numbers
of benefits from the monetized benefit es-
timate (e.g., all air toxics effects, ecosystem
effects, numerous human health effects).
Figure ES-3 provides a graphical representation
of the estimated range of total monetized benefits and
compares this range to estimated direct compliance
costs. Clearly, even the lower bound estimate of mon-
etized benefits substantially exceeds the costs of the
historical Clean Air Act. As shown by the yearly data
presented in Chapter 7, monetized benefits consis-
tently and substantially exceeded costs throughout the
1970 to 1990 period.
Figure ES-3. Total Estimated Direct Compliance Costs of
the CAA (in trillions of inflation-adjusted dollars).
50
« 40
"o
Q
o 30
CT>
CT>
° 20
-------
Executive Summary
Tabic ES-5. Alternative Mortality Benefits Mean
Estimates for 1970 to 1990 (in trillions of 1990
dollars) Compared to Total 1970 to 1990 Compliance
Costs.
Mortality
Benefits
(trillions of dollars)
RenefitFstimation Method
PM+Pb
Statistical life method ($4.8M/case)
16.6
18.0
Life-years lost method ($293,000/year
9.1
10.1
Total compliance cost
—
0.5
Alternative Results
The primary results of this analysis, including ag-
gregate cost and benefit estimates which reflect many
elements of the uncertainty associated with them, are
presented above. However, some additional analysis
is required to address an important issue raised by the
EPA Science Advisory Board Council on Clean Air
Act Compliance Analysis (a.k.a. Council) charged
with reviewing the present study. Specifically, the
Council believes it is appropriate to also display al-
ternative premature mortality results based on an ap-
proach which estimates, and assigns a value to, the
loss of life-years (i.e., the reduction in years of re-
maining life expectancy) resulting from the pollution
exposure. The Council's position is based on the con-
clusion that older individuals are more susceptible to
air pollution-induced mortality. EPA believes, how-
ever, that the simplifying assumptions which must be
adopted to implement a life-years lost approach ren-
der its results less reliable, even for the purposes of
economic efficiency analysis, than a value of statisti-
cal life approach. In addition, EPA is concerned about
any analytical methodology which may be interpreted
to justify conferring less environmental protection on
particular individuals or groups of individuals (e.g.,
the elderly and/or sick). EPA therefore prefers at this
time to continue with its current practice of assigning
the same economic value to incidences of premature
mortality regardless of the age and health status of
those affected, and the primary results presented above
reflect this view. Nevertheless, complete alternative
results based on a value of statistical life-years lost
(VSLY) approach are presented in Chapter 7 and Ap-
pendix I and are summarized below.
Table ES-5 summarizes and compares the results
of the mortality benefits estimates based on the value
of statistical life (VSL) and VSLY approaches. Esti-
mated 1970 to 1990 benefits from PM-related mor-
tality alone and total mortality (i.e., PM plus Lead)
benefits are reported, along with total compliance costs
for the same period. Adding the VSLY-based mortal-
ity benefits estimates to the non-mortality benefits
estimates from Table ES-4 yields the following re-
sults for the overall analysis.
• Alternate Result: The total monetized ben-
efits of the Clean Air Act realized during
the period from 1970 to 1990 range from
4.8 to 28.7 trillion dollars, with a central
estimate of 14.3 trillion dollars.
• Alternate Result: Subtracting costs from
benefits results in net, direct, monetized
benefits ranging from 4.3 to 28.2 trillion
dollars, with a central estimate of 13.7 tril-
lion dollars, for the 1970 to 1990 period.
The results indicate that the choice of valuation
methodology significantly affects the estimated mon-
etized value of historical reductions in air pollution-
related premature mortality. However, the downward
adjustment which would result from applying a VSLY
approach in lieu of a VSL approach does not change
the basic outcome of this study, viz. the estimated
monetized benefits of the historical Clean Air Act
substantially exceed the estimated historical costs of
compliance.
Conclusions and Future
Directions
First and foremost, these results indicate that the
benefits of the Clean Air Act and associated control
programs substantially exceeded costs. Even consid-
ering the large number of important uncertainties per-
meating each step of the analysis, it is extremely un-
likely that the converse could be true.
A second important implication of this study is
that a large proportion of the monetized benefits of
the historical Clean Air Act derive from reducing two
pollutants: lead and particulate matter4 (see Table ES-
4). Some may argue that, while programs to control
these two pollutants may have yielded measurable
4 Ambient particulate matter results from emissions of a wide array of precursor pollutants, including sulfur dioxide, nitrogen
oxides, and organic compounds.
ES-9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
benefits in excess of measurable costs, estimates of
measurable benefits of many other historical Clean
Air Act programs and standards considered in isola-
tion might not have exceeded measurable costs. While
this may or may not be true, this analysis provides no
evidence to support or reject such conjectures. On the
cost side, the historical expenditure data used in this
analysis are not structured in ways which allow attri-
bution of control costs to specific programs or stan-
dards. On the benefit side, most control programs
yielded a variety of benefits, many of which included
reductions in other pollutants such as ambient par-
ticulate matter. For example, new source performance
standards for sulfur dioxide emissions from coal-fired
utility plants yielded benefits beyond those associated
with reducing exposures to gaseous sulfur dioxide.
The reductions in sulfur dioxide emissions also led to
reductions in ambient fine particle sulfates, yielding
human health, ecological, and visibility benefits.
This retrospective study highlights important ar-
eas of uncertainty associated with many of the mon-
etized benefits included in the quantitative analysis
and lists benefit categories which could not be quan-
tified or monetized given the current state of the sci-
ence. Additional research in these areas may reduce
critical uncertainties and/or improve the comprehen-
siveness of future assessments. Particularly important
areas where further research might reduce critical
uncertainties include particulate matter-related mor-
tality incidence, valuation of premature mortality, and
valuation of particulate-related chronic bronchitis and
cardiovascular disease. Additional research on haz-
ardous air pollutants and on air pollution-related
changes in ecosystem structure and function might
help improve the comprehensiveness of future ben-
efit studies. (See Appendix J for further discussion.)
Finally, the results of this retrospective study pro-
vide useful lessons with respect to the value and the
limitations of cost-benefit analysis as a tool for evalu-
ating environmental programs. Cost-benefit analysis
can provide a valuable framework for organizing and
evaluating information on the effects of environmen-
tal programs. When used properly, cost-benefit analy-
sis can help illuminate important effects of changes
in policy and can help set priorities for closing infor-
mation gaps and reducing uncertainty. Such proper
use, however, requires that sufficient levels of time
and resources be provided to permit careful, thorough,
and technically and scientifically sound data-gather-
ing and analysis. When cost-benefit analyses are pre-
sented without effective characterization of the un-
certainties associated with the results, cost-benefit
studies can be used in highly misleading and damag-
ing ways. Given the substantial uncertainties which
permeate cost-benefit assessment of environmental
programs, as demonstrated by the broad range of esti-
mated benefits presented in this study, cost-benefit
analysis is best used to inform, but not dictate, deci-
sions related to environmental protection policies,
programs, and research.
ES-10
-------
1
Introduction
Background and Purpose
As part of the Clean Air Act Amendments of 1990,
Congress established a requirement under section 812
that EPA develop periodic Reports to Congress esti-
mating the benefits and costs of the Clean Air Act
itself. The first such report was to be a retrospective
analysis, with a series of prospective analyses to fol-
low every two years thereafter. This report represents
the retrospective study, covering the period beginning
with passage of the Clean Air Act Amendments of
1970, until 1990 when Congress enacted the most re-
cent comprehensive amendments to the Act.
Since the legislative history associated with sec-
tion 812 is sparse, there is considerable uncertainty
regarding Congressional intent behind the requirement
for periodic cost-benefit evaluations of the Clean Air
Act (CAA). However, EPA believes the principal goal
of these amendments was that EPA should develop,
and periodically exercise, the ability to provide Con-
gress and the public with up-to-date, comprehensive
information about the economic costs, economic ben-
efits, and health, welfare, and ecological effects of
CAA programs. The results of such analyses might
then provide useful information for refinement of CAA
programs during future reauthorizations of the Act.
The retrospective analysis presented in this Re-
port to Congress has been designed to provide an un-
precedented examination of the overall costs and ben-
efits of the historical Clean Air Act. Many other analy-
ses have attempted to identify the isolated effects of
individual standards or programs, but no analysis with
the present degree of validity, breadth and integration
has ever been successfully developed. Despite data
limitations, considerable scientific uncertainties, and
severe resource constraints; the EPA Project Team was
able to develop a broad assessment of the costs and
benefits associated with the major CAA programs of
the 1970 to 1990 period. Beyond the statutory goals
of section 812, EPA intends to use the results of this
study to help support decisions on future investments
in air pollution research. Finally, many of the meth-
odologies and modeling systems developed for the
retrospective study may be applied in the future to the
ongoing series of section 812 prospective studies.
Clean Air Act Requirements,
1970 to 1990
The Clean Air Act establishes a framework for
the attainment and maintenance of clean and health-
ful air quality levels. The Clean Air Act was enacted
in 1970 and amended twice — in 1977 and most re-
cently in 1990. The 1970 Clean Air Act contained a
number of key provisions. First, EPA was directed to
establish national ambient air quality standards for the
major criteria air pollutants. The states were required
to develop implementation plans describing how they
would control emission limits from individual sources
to meet and maintain the national standards. Second,
the 1970 CAA contained deadlines and strengthened
enforcement of emission limitations and state plans
with measures involving both the states and the fed-
eral government. Third, the 1970 Act forced new
sources to meet standards based on the best available
technology. Finally, the Clean Air Act of 1970 ad-
dressed hazardous pollutants and automobile exhausts.
The 1977 Clean Air Act Amendments also set new
requirements on clean areas already in attainment with
the national ambient air quality standards. In addition,
the 1977 Amendments set out provisions to help ar-
eas that failed to comply with deadlines for achieve-
ment of the national ambient air quality standards. For
example, permits for new major sources and modifi-
cations were required.
The 1990 Clean Air Act Amendments consider-
ably strengthened the earlier versions of the Act. With
respect to nonattainment, the Act set forth a detailed
and graduated program, reflecting the fact that prob-
lems in some areas are more difficult and complex
than others. The 1990 Act also established a list of
189 regulated hazardous air pollutants and a multi-
step program for controlling emissions of these toxic
air pollutants. Significant control programs were also
established for emissions of acid rain precursors and
stratospheric ozone-depleting chemicals. The biggest
regulatory procedural change in the Act is the new
permit program where all major sources are now re-
quired to obtain an operating permit. Finally, the
amendments considerably expanded the enforcement
provisions of the Clean Air Act, adding administra-
tive penalties and increasing potential civil penalties.
1
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Section 812 of the Clean Air Act
Amendments of 1990
Section 812 of the Clean Air Act Amendments of
1990 requires the EPA to perform a "retrospective"
analysis which assesses the costs and benefits to the
public health, economy and the environment of clean
air legislation enacted prior to the 1990 amendments.
Section 812 directs that EPA shall measure the effects
on "employment, productivity, cost of living, economic
growth, and the overall economy of the United States"
of the Clean Air Act. Section 812 also requires that
EPA consider all of the economic, public health, and
environmental benefits of efforts to comply with air
pollution standards. Finally, section 812 requires EPA
to evaluate the prospective costs and benefits of the
Clean Air Act every two years.
Analytical Design and Review
Target Variable
The retrospective analysis was designed to answer
the following question:
"How do the overall health, welfare,
ecological, and economic benefits of Clean
Air Act programs compare to the costs of
these programs?"
By examining the overall effects of the Clean Air
Act, this analysis complements the Regulatory Impact
Analyses (RIAs) developed by EPA over the years to
evaluate individual regulations. Resources were used
more efficiently by recognizing that these RIAs, and
other EPA analyses, provide complete information
about the costs and benefits of specific rules. Further-
more, in addition to the fact that the RIAs already pro-
vide rule-specific benefit and cost estimates, the broad-
scale approach adopted in the present study precludes
reliable re-estimation of the benefits and costs of in-
dividual standards or programs. On the cost side, this
study relies on aggregated compliance expenditure
data from existing surveys. Unfortunately, these data
do not support reliable allocation of total costs incurred
to specific emissions reductions for the various pol-
lutants emitted from individual facilities. Therefore,
it is infeasible in the context of this study to assign
costs to specific changes in emissions. Further com-
plications emerge on the benefit side. To estimate
benefits, this study calculates the change in incidences
of adverse effects implied by changes in ambient con-
centrations of air pollutants. However, reductions
achieved in emitted pollutants contribute to changes
in ambient concentrations of those, or secondarily
formed, pollutants in ways which are highly complex,
interactive, and often nonlinear. Therefore, even if
costs could be reliably matched to changes in emis-
sions, benefits cannot be reliably matched to changes
in emissions because of the complex, nonlinear rela-
tionships between emissions and the changes in am-
bient concentrations which are used to estimate ben-
efits.
Focusing on the broader taiget variables of "over-
all costs" and "overall benefits" of the Clean Air Act,
the EPA Project Team adopted an approach based on
construction and comparison of two distinct scenarios:
a "no-control scenario" and a "control scenario." The
no-control scenario essentially freezes federal, state,
and local air pollution controls at the levels of strin-
gency and effectiveness which prevailed in 1970. The
control scenario assumes that all federal, state, and
local rules promulgated pursuant to, or in support of,
the CAA during 1970 to 1990 were implemented. This
analysis then estimates the differences between the
economic and environmental outcomes associated
with these two scenarios. For more information on
the scenarios and their relationship to historical trends,
see Appendix B.
Key Assumptions
Two key assumptions were made during the sce-
nario design process to avoid miring the analytical
process in endless speculation. First, the "no-control"
scenario was defined to reflect the assumption that no
additional air pollution controls were imposed by any
level of government or voluntarily initiated by pri-
vate entities after 1970. Second, it is assumed that the
geographic distribution of population and economic
activity remains the same between the two scenarios.
The first assumption is an obvious oversimplifi-
cation. In the absence of the CAA, one would expect
to see some air pollution abatement activity, either
voluntary or due to state or local regulations. It is con-
ceivable that state and local regulation would have
required air pollution abatement equal to—or even
greater than—that required by the CAA; particularly
since some states, most notably California, have done
so. If one were to assume that state and local regula-
tions would have been equivalent to CAA standards,
then a cost-benefit analysis of the CAA would be a
meaningless exercise since both costs and benefits
would equal zero. Any attempt to predict how state
and local regulations would have differed from the
CAA would be too speculative to support the cred-
ibility of the ensuing analysis. Instead, the no-control
scenario has been structured to reflect the assumption
that states and localities would not have invested fur-
ther in air pollution control programs after 1970 in
the absence of the federal CAA. That is, this analysis
accounts for the costs and benefits of all air pollution
2
-------
control from 1970 to 1990. Speculation about the pre-
cise fraction of costs and benefits attributable exclu-
sively to the federal CAA is left to others. Neverthe-
less, it is important to note that state and local govern-
ments and private initiatives are responsible for a sig-
nificant portion of these total costs and total benefits.
At the same time, it must also be acknowledged that
the federal CAA played an essential role in achieving
these results by helping minimize the advent of pollu-
tion havens1, establishing greater incentives for pol-
lution control research and development than indi-
vidual state or local rules could provide; organizing
and promoting health and environmental research,
technology transfer and other information management
and dissemination services; addressing critical inter-
state air pollution problems, including the regional fine
particle pollution which is responsible for much of
the estimated monetary benefit of historical air pollu-
tion control; providing financial resources to state and
local government programs; and many other services.
In the end, however, the benefits of historical air pol-
lution controls were achieved through partnerships
among all levels of government and with the active
participation and cooperation of private entities and
individuals.
The second assumption concerns changing demo-
graphic patterns in response to air pollution. In the
hypothetical no-control world, air quality is worse than
that in the historical "control" world particularly in
urban industrial areas. It is possible that in the no-
control case more people, relative to the control case,
would move away from the most heavily polluted ar-
eas. Rather than speculate on the scale of population
movement, the analysis assumes no differences in
demographic patterns between the two scenarios. Simi-
larly, the analysis assumes no changes in the spatial
pattern of economic activity. For example: if, in the
no-control case, an industry is expected to produce
greater output than it did in the control case, that in-
creased output is produced by actual historical plants,
avoiding the need to speculate about the location or
other characteristics of new plants providing additional
productive capacity.
Analytic Sequence
The analysis was designed and implemented in a
sequential manner following seven basic steps which
are summarized below and described in detail later in
this report. The seven major steps were:
Chapter 1: Introduction
direct cost estimation
• macroeconomic modeling
emissions modeling
air quality modeling
health and environmental effects estimation
economic valuation
results aggregation and uncertainty character-
ization
By necessity, these components had to be com-
pleted sequentially. The emissions modeling effort had
to be completed entirely before the air quality models
could be configured and run; the air quality modeling
results had to be completed before the health and en-
vironmental consequences of air quality changes could
be derived; and so on. The analytical sequence, and
the modeled versus actual data basis for each analyti-
cal component, are summarized in Figure 1 and de-
scribed in the remainder of this section.
The first step of the analysis was to estimate the
total direct costs incurred by public and private enti-
ties to comply with post-1970 CAA requirements.
These data were obtained directly from Census Bu-
reau and Bureau of Economic Analysis (BEA) data
on compliance expenditures reported by sources, and
from EPA analyses. These direct cost data were then
adopted as inputs to the macroeconomic model used
to project economic conditions-such as production
levels, prices, employment patterns, and other eco-
nomic indicators-under the two scenarios. To ensure
a consistent basis for scenario comparison, the analy-
sis applied the same macroeconomic modeling sys-
tem to estimate control and no-control scenario eco-
nomic conditions.2 First, a control scenario was con-
structed by running the macroeconomic model using
actual historical data for input factors such as eco-
nomic growth rates during the 1970 to 1990 period.
The model was then re-run for the no-control scenario
by, in essence, returning all post-1970 CAA compli-
ance expenditures to the economy. With these addi-
tional resources available for capital formation, per-
sonal consumption, and other purposes, overall eco-
nomic conditions under the no-control scenario dif-
fered from those of the control scenario. In addition
to providing estimates of the difference in overall eco-
nomic growth and other outcomes under the two sce-
narios, these first two analytical steps were used to
define specific economic conditions used as inputs to
the emissions modeling effort, the first step in the es-
timation of CAA benefits.3
1 "Pollution havens" is a term used to identify individual states or localities which permit comparatively high levels of pollution in
order to attract and hold polluting industries and other activities.
2 Using modeled economic conditions for both scenarios has both advantages and disadvantages. The principal disadvantage is that
historical economic conditions "predicted" by a macroeconomic model will not precisely duplicate actual historical events and condi-
tions. However, this disadvantage is outweighed by the avoidance of distortions and biases which would result from comparing a
modeled no-control scenario with actual historical conditions. By using the same macroeconomic model for both scenarios, model errors
and biases essentially cancel out, yielding more robust estimates of scenario differences, which are what this analysis seeks to evaluate.
3 For example, the macroeconomic model projected different electricity sales levels under the two scenarios, and these sales levels
were used as key input assumptions by the utility sector emissions model.
3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Figure 1. Summary of Analytical Sequence and Modeled versus Historical Data Basis.
Control Scenario
No-Control Scenario
Compile historical compliance expenditure
data
Develop modeled macroeconomic scenario
based on actual historical economic data
Develop modeled macroeconomic scenario
by rerunning control scenario with
compliance expenditures added back to the
economy
Project emissions by year, pollutant, and
sector using control scenario
macroeconomic projection as input to
sector-specific emissions models
Re-run sector-specific emissions models
using no-control scenario macroeconomic
projection
Develop statistical profiles of historical air
quality for each pollutant based on
historical monitoring data (plus
extrapolations to cover unmonitored areas)
Derive no-control air quality profiles by
adjusting control scenario profiles based on
differences in air quality modeling of
control scenario and no-control scenario
emissions inventories
Estimate physical effects based on
application of concentration-response
functions to historical air quality profiles
Estimate physical effects based on
application of concentration-response
functions to no-control scenario air quality
profiles
Calculate differences in physical outcomes
between control and no-control scenario
Estimate economic value of differences in
physical outcomes between the two
scenarios*
Compare historical, direct compliance costs
with estimated economic value of
monetized benefits, considering additional
benefits which could not be quantified
and/or monetized
* In some cases, economic value is derived directly from physical effects modeling (e.g., agricultural yield loss).
-------
Chapter 1: Introduction
Using appropriate economic indicators from the
macroeconomic model results as inputs, a variety of
emissions models were run to estimate emissions lev-
els under the two scenarios. These emissions models
provided estimates of emissions of six major pollut-
ants4 from each of six key emitting sectors: utilities,
industrial processes, industrial combustion, on-high-
way vehicles, off-highway vehicles, and commercial/
residential sources. The resulting emissions profiles
reflect state-wide total emissions from each pollut-
ant-sector combination for the years 1975,1980, 1985,
and 1990.5
The next step toward estimation of benefits in-
volved translating these emissions inventories into
estimates of air quality conditions under each scenario.
Given the complexity, data requirements, and operat-
ing costs of state-of-the-art air quality models-and the
afore-mentioned resource constraints-the EPA Project
Team adopted simplified, linear scaling approaches
for a number of pollutants. However, for ozone and
other pollutants or air quality conditions which involve
substantial non-linear formation effects and/or long-
range atmospheric transport and transformation, the
EPA Project Team invested the time and resources
needed to use more sophisticated modeling systems.
For example, urban area-specific ozone modeling was
conducted for 147 urban areas throughout the 48 con-
tiguous states.
Up to this point of the analysis, both the control
and no-control scenario were based on modeled con-
ditions and outcomes. However, at the air quality
modeling step, the analysis returned to a foundation
based on actual historical conditions and data. Spe-
cifically, actual historical air quality monitoring data
from 1970 to 1990 were used to define the control
scenario. Air quality conditions under the no-control
scenario were then derived by scaling the historical
data adopted for the control scenario by the ratio of
the modeled control and no-control scenario air qual-
ity. This approach took advantage of the richness of
the historical data on air quality, provided a realistic
grounding for the benefit measures, and yet retained
the analytical consistency conferred by using the same
modeling approach for both scenarios. The outputs of
this step of the analysis were statistical profiles for
each pollutant characterizing air quality conditions at
each monitoring site in the lower 48 states.6
The control and no-control scenario air quality
profiles were then used as inputs to a modeling sys-
tem which translates air quality to physical outcomes
-such as mortality, emergency room visits, or crop
yield losses- through the use of concentration-re-
sponse functions. These concentration-response func-
tions were in turn derived from studies found in the
scientific literature on the health and ecological ef-
fects of air pollutants. At this point, estimates were
derived of the differences between the two scenarios
in terms of incidence rates for a broad range of human
health and other effects of air pollution by year, by
pollutant, and by monitor.7
In the next step, economic valuation models or
coefficients were used to estimate the economic value
of the reduction in incidence of those adverse effects
which were amenable to such monetization. For ex-
ample, a distribution of unit values derived from the
economic literature was used to estimate the value of
reductions in mortality risk associated with exposure
to particulate matter. In addition, benefits which could
not be expressed in economic terms were compiled
and are presented herein. In some cases, quantitative
estimates of scenario differences in the incidence of a
nonmonetized effect were calculated.8 In many other
cases, available data and techniques were insufficient
to support anything more than a qualitative character-
ization of the change in effects.
Finally, the costs and monetized benefits were
combined to provide a range of estimates for the par-
tial, net economic benefit of the CAA with the range
reflecting quantified uncertainties associated with the
physical effects and economic valuation steps.9 The
term "partial" is emphasized because only a subset of
the total potential benefits of the CAA could be rep-
resented in economic terms due to limitations in ancal
4 These six pollutants are total suspended particulates (TSP), sulfur dioxide (S02), nitrogen oxides (NO ), carbon monoxide (CO),
volatile organic compounds (VOCs), and lead (Pb). The other CAA criteria pollutant, ozone (03), is formed in the atmosphere through
the interaction of sunlight and ozone precursor pollutants such as NOx and VOCs.
5 By definition, 1970 emissions under the two scenarios are identical.
6 The one exception is particulate matter (PM). For PM, air quality profiles for both Total Suspended Particulates (TSP) and
particulates less than or equal to 10 microns in diameter (PM ) were constructed at the county level rather than the individual monitor
level.
7 Or, for PM, by county.
8 For example, changes in forced expiratory volume in one second (FEVj) as a result of exposure to ozone were quantified but
could not be expressed in terms of economic value.
9 Although considerable uncertainties surround the direct cost, macroeconomic modeling, emissions modeling,, and air quality
modeling steps, the ranges of aggregate costs and benefits presented in this analysis do not reflect these uncertainties. While the
uncertainties in these components were assessed qualitatively, and in some cases quantitatively, resource limitations precluded the
multiple macroeconomic model, emissions model, and air quality model runs which would have been required to propagate these
uncertainties through the entire analytical sequence. As a result, complete quantitative measures of the aggregate uncertainty in the cost
and benefit estimates could not be derived. However, the ranges presented do reflect quantitative measures of the uncertainties in the
two most uncertain analytical steps: physical effects estimation and economic valuation.
5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
cal resources, available data and models, and the state
of the science.10 Of paramount concern to the EPA
Project Team was the paucity of concentration-re-
sponse functions needed to translate air quality
changes into measures of ecological effect. In addi-
tion, significant scientific evidence exists linking air
pollution to a number of adverse human health ef-
fects which could not be effectively quantified and/or
monetized.11
Review Process
The CAA requires EPA to consult with an out-
side panel of experts-referred to statutorily as the
Advisory Council on Clean Air Act Compliance
Analysis (the Council)-in developing the section 812
analyses. In addition, EPA is required to consult with
the Department of Labor and the Department of Com-
merce.
The Council was organized in 1991 under the aus-
pices and procedures of EPA's Science Advisory
Board (SAB). Organizing the review committee un-
der the SAB ensured that review of the section 812
studies would be conducted by highly qualified ex-
perts in an objective, rigorous, and publicly open
manner. The Council has met many times during the
development of the retrospective study to review meth-
odologies and interim results. While the full Council
retains overall review responsibility for the section
812 studies, some specific issues concerning physical
effects and air quality modeling have been referred to
subcommittees comprised of both Council members
and members of other SAB committees. The Council's
Physical Effects Review Subcommittee met several
times and provided its own review findings to the full
Council. Similarly, the Council's Air Quality Subcom-
mittee, comprised of members and consultants of the
SAB Clean Air Scientific Advisory Committee
(CASAC), held several teleconference meetings to
review methodology proposals and modeling results.
With respect to the interagency review process,
EPA expanded the list of consulted agencies and con-
vened a series of meetings during the design and early
implementation phases from 1991 through late 1994.
In late 1994, to ensure that all interested parties and
the public received consistent information about re-
maining analytical issues and emerging results, EPA
decided to use the public SAB review process as the
primary forum for presenting and discussing issues
and results. The Interagency Review Group was there-
fore discontinued as a separate process in late 1994.
A final, brief interagency review, pursuant to Cir-
cular A-19, was organized in August 1997 by the Of-
fice of Management and Budget and conducted fol-
lowing the completion of the extensive expert panel
peer review by the SAB Council. During the course
of the final interagency discussions, it became clear
that several agencies held different views pertaining
to several key assumptions in this study as well as to
the best techniques to apply in the context of environ-
mental program benefit-cost analyses, including the
present study. The concerns include: (1) the extent to
which air quality would have deteriorated from 1970
to 1990 in the absence of the Clean Air Act, (2) the
methods used to estimate the number of premature
deaths and illnesses avoided due to the CAA, (3) the
methods used to estimate the value that individuals
place on avoiding those risks, and (4) the methods
used to value non-health related benefits. However,
due to the court deadline the resulting concerns were
not resolved during this final, brief interagency re-
view. Therefore, this report reflects the findings of
EPA and not necessarily other agencies in the Ad-
ministration. Interagency discussion of some of these
issues will continue in the context of the future pro-
spective section 812 studies and potential regulatory
actions.
Report Organization
The remainder of the main text of this report sum-
marizes the key methodologies and findings of retro-
spective study. The direct cost estimation and macro-
economic modeling steps are presented in Chapter 2.
The emissions modeling is summarized in Chapter 3.
Chapter 4 presents the air quality modeling method-
ology and sample results. Chapter 5 describes the ap-
proaches used and principal results obtained through
the physical effects estimation process. Economic
valuation methodologies are described in Chapter 6.
Chapter 7 presents the aggregated results of the cost
and benefit estimates and describes and evaluates
important uncertainties in the results.
Additional details regarding the methodologies
and results are presented in the appendices and in the
referenced supporting documents. Appendix A cov-
ers the direct cost and macroeconomic modeling. Ap-
pendix B provides additional detail on the sector-spe-
cific emissions modeling effort. Details of the air qual-
ity models used and results obtained are presented or
referenced in Appendix C. The effects of the CAA on
human health and visibility; aquatic, wetland, and for-
est ecosystems; and agriculture are presented in Ap-
pendices D, E, and F, respectively. Appendix G pre-
sents details of the lead (Pb) benefits analysis. Air
toxics reduction benefits are discussed in Appendix
H. The methods and assumptions used to value quan-
tified effects of the CAA in economic terms are de-
scribed in Appendix I. Appendix J describes some ar-
eas of research which may increase comprehensive-
ness and reduce uncertainties in effect estimates for
future assessments, and describes plans for future sec-
tion 812 analyses.
10 It should be noted that there is some uncertainty associated with the estimates of economic costs as well and that some omitted
components of adverse economic consequences of pollution control programs may be significant. For example, some economists
argue that the economic costs of the CAA reported herein may be significantly underestimated to the extent potential adverse effects
of regulation on technological innovation are not captured. Nevertheless, it is clear that the geographic, population, and categorical
coverage of monetary cost effects is significantly greater than coverage of monetized benefits in this analysis.
11 For example, while there is strong evidence of a link between exposure to carbon monoxide and reduced time of onset of
angina attack, there are no valuation functions available to estimate the economic loss associated with this effect.
-------
2
Cost and Macroeconomic Effects
The costs of complying with Clean Air Act (CAA)
requirements through the 1970 to 1990 period affected
patterns of industrial production, capital investment,
productivity, consumption, employment, and overall
economic growth. The purpose of the analyses sum-
marized in this chapter was to estimate those direct
costs and the magnitude and significance of resulting
changes to the overall economy. This was accom-
plished by comparing economic indicators under two
alternative scenarios: a control scenario serving as the
historical benchmark, including the historical CAA
as implemented; and a no-control scenario which as-
sumes historical CAA programs did not exist. The
estimated economic consequences of the historical
CAA were taken as the difference between these two
scenarios.
Data used as inputs to the cost analysis can be
classified into two somewhat overlapping categories
based on the information source: survey-based infor-
mation (generally gathered by the Census Bureau) and
information derived from various EPA analyses. For
the most part, cost estimates for stationary air pollu-
tion sources (e.g., factory smokestacks) are based on
surveys of private businesses that attempt to elicit in-
formation on annual pollution control outlays by those
businesses. Estimates of pollution control costs for
mobile sources (e.g., automobiles) are largely based
on EPA analyses, rather than on direct observation
and measurement of compliance expenditures. For
example, to determine one component of the cost of
reducing lead emissions from mobile sources, the
Project Team used an oil refinery production cost
model to calculate the incremental cost required to
produce unleaded (or less-leaded, as appropriate)
rather than leaded gasoline, while maintaining the
octane level produced by leaded gasoline.
As is the case with many policy analyses, a sig-
nificant uncertainty arises in the cost analysis as a
consequence of constructing a hypothetical scenario.
With this retrospective analysis covering almost
twenty years, difficulties arise in projecting alterna-
tive technological development paths. In some cases,
the analytical assumptions used to project the alterna-
tive scenario are not immediately apparent. For ex-
ample, the surveys covering stationary source com-
pliance expenditures require respondents to report
pollution abatement expenditures—implicitly asking
them to determine by how much the company's costs
would decline if there were no CAA compliance re-
quirements. While a response might be relatively
straightforward in the few years following passage of
the CAA, a meaningful response becomes more diffi-
cult after many years of technical change and invest-
ment in less-polluting plant and equipment make it
difficult to determine the degree to which total costs
would differ under a "no CAA" scenario. In cases such
as this, assumptions concerning the alternative hypo-
thetical scenario are made by thousands of individual
survey respondents. Where cost data are derived from
EPA analyses, the hypothetical scenario assumptions
are, at least in theory, more apparent. For example,
when determining the incremental cost caused by pol-
lution-control requirements, one needs to make as-
sumptions (at least implicitly) about what an auto
would look like absent pollution control requirements.
In either case, the need to project hypothetical tech-
nology change for two decades introduces uncertainty
into the assessment results, and this uncertainty may
be difficult to quantify.
The remainder of this chapter summarizes the
basic methods and results of the direct compliance
cost and macroeconomic analyses. Further details re-
garding the modeling methods and assumptions em-
ployed, as well as additional analytical results, are
presented in Appendix A.
Direct Compliance Costs
Compliance with the CAA imposed direct costs
on businesses, consumers, and governmental units; and
triggered other expenditures such as governmental
regulation and monitoring costs and expenditures for
7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Table 1. Estimated Annual CAA
Compliance Costs ($bil lions).
Expenditures Annualized Costs
$1990 at:
Year Scurrent SI990
3%
5%
7%
1973
7.2
19.6
11.0
11.0
11.1
1974
8.5
21.4
13.2
13.4
13.7
1975
10.6
24.4
13.3
13.6
14.0
1976
11.2
24.1
14.1
14.6
15.1
1977
11.9
24.1
15.3
15.9
16.6
1978
12.0
22.6
15.0
15.8
16.7
1979
14.4
24.8
17.3
18.3
19.3
1980
16.3
25.7
19.7
20.8
22.0
1981
17.0
24.4
19.6
20.9
22.3
1982
16.0
21.6
18.6
20.1
21.7
1983
15.5
20.1
19.1
20.7
22.5
1984
17.3
21.6
20.1
21.9
23.8
1985
19.1
22.9
22.5
24.4
26.5
1986
17.8
20.8
21.1
23.2
25.4
1987
18.2
20.6
22.1
24.2
26.6
1988
18.2
19.8
22.0
24.3
26.7
1989
19.0
19.8
22.9
25.3
27.8
1990
190
190
23,6
26,1
28,7
research and development by both government and
industry. Although expenditures unadjusted for infla-
tion — that is, expenditures denominated in "current
dollars"— increased steadily from $7 billion to $19
billion per year over the 1973 to 1990 period,12 an-
nual CAA compliance expenditures adjusted for in-
flation were relatively stable, averaging near $25 bil-
lion (in 1990 dollars) during the 1970s and close to
$20 billion during most of the 1980s (see Table 1).
Aggregate compliance expenditures were somewhat
less than one half of one percent of total domestic
output during that period, with the percentage falling
from two thirds of one percent of total output in 1975
to one third of one percent in 1990.
Although useful for many purposes, a summary
of direct annual expenditures may not the best cost
measure to use when comparing costs to benefits.
Capital expenditures are investments, generating a
stream of benefits and opportunity cost13 over the life
of the investment. The appropriate accounting tech-
nique to use for capital expenditures in a cost/benefit
analysis is to annualize the expenditure. This tech-
nique, analogous to calculating the monthly payment
associated with a home mortgage, involves spreading
the cost of the capital equipment over the useful life
of the equipment using a discount rate to account for
the time value of money.
For this cost/benefit analysis, "annualized" costs
reported for any given year are equal to O&M expen-
ditures — including R&D and other similarly recur-
ring expenditures — plus amortized capital costs (i.e.,
depreciation plus interest costs associated with the
existing capital stock) for that year. Stationary source
air pollution control capital costs were amortized over
20 years; mobile source air pollution control costs were
amortized over 10 years.14 All capital expenditures
were annualized using a five percent, inflation-ad-
justed rate of interest. Additionally, annualized costs
were calculated using discount rates of three and seven
percent to determine the sensitivity of the cost results
to changes in the discount rate. Table 1 summarizes
costs annualized at three, five, and seven percent, as
well as annual expenditures.
Total expenditures over the 1973-1990 period,
discounted to 1990 using a five percent (net of infla-
tion) discount rate, amount to 628 billion dollars (in
1990 dollars). Discounting the annualized cost stream
to 1990 (with both annualization and discounting pro-
cedures using a five percent rate) gives total costs of
523 billion dollars (in 1990 dollars). Aggregate annu-
alized costs are less than expenditures because the
annualization procedure spreads some of the capital
cost beyond 1990.15
12 Due to data limitations, the cost analysis for this CAA retrospective starts in 1973, missing costs incurred in 1970-72. This
limitation is not likely to be significant, however, because relatively little in the way of compliance with the "new" provisions of the
1970 CAA was required in the first two years following passage.
13 In this context, "opportunity cost" is defined as the value of alternative investments or other uses of funds foregone as a result of
the investment.
14 Although complete data are available only for the period 1973-1990, EPA's Cost of Clean report includes capital expenditures
for 1972 (see Appendix A for more details and complete citation). Those capital expenditure data have been used here. Therefore,
amortized costs arising from 1972 capital investments are included in the 1973-1990 annualized costs, even though 1972 costs are not
otherwise included in the analysis. Conversely, some capital expenditures incurred in the 1973-1990 period are not reflected in the
1973-1990 annualized costs — those costs are spread through the following two decades, thus falling outside of the scope of this study
(e.g., only one year of depreciation and interest expense is included for 1989 capital expenditures). Similarly, benefits arising from
emission reductions realized after 1990 as a result of capital investments made during the 1970 to 1990 period of this analysis are not
included in the estimates of benefits included in this report.
15 This adjustment is required because many 1970 to 1990 investments in control equipment continue to yield benefits beyond
1990. Annualization of costs beyond 1990 ensures that the costs and benefits of any particular investment are properly scaled and
matched over the lifetime of the investment.
8
-------
Chapter 2: Cost and Macro economic Effect
Indirect Effects of the CAA
Through changing production costs, CAA imple-
mentation induced changes in consumer good prices,
and thus in the size and composition of economic out-
put. The Project Team used a general equilibrium
macroeconomic model to assess the extent of such
second-order effects. This type of model is useful be-
cause it can capture the feedback effects of an action.
In the section 812 macroeconomic modeling exercise,
the feedback effects arising from expenditure changes
were captured, but the analogous effects arising from
improvements in human health were not captured by
the model. For example, the macroeconomic model
results do not reflect the indirect economic effects of
worker productivity improvements and medical ex-
penditure savings caused by the CAA. Consequently,
the macroeconomic modeling exercise provides lim-
ited and incomplete information on the type and po-
tential scale of indirect economic effects.
The effects estimated by the macroeconomic
model can be grouped into two broad classes: sectoral
impacts (i.e., changes in the composition of economic
output), and aggregate effects (i.e., changes in the
degree of output or of some measure of human wel-
fare). The predicted sectoral effects were used as in-
puts to the emissions models as discussed in Chapter
3. In general, the estimated second-order macroeco-
nomic effects were small relative to the size of the
U.S. economy. See Appendix A for more detail on
data sources, analytical methods, and results for the
macroeconomic modeling performed for this assess-
ment.
Sectoral Impacts
The CAA had variable compliance impacts across
economic sectors. The greatest effects were on the
largest eneigy producers and consumers, particularly
those sectors which relied most heavily on consump-
tion of fossil fuels (or energy generated from fossil
fuels). In addition, production costs increased more
for capital-intensive industries than for less capital-
intensive industries under the control scenario due to
a projected increase in interest rates. The interest rate
increase, which resulted in an increase in the cost of
capital, occurred under the control scenario because
CAA-mandated investment in pollution abatement
reduced the level of resources available for other uses,
including capital formation.
Generally, the estimated diiference in cost impacts
under the control and no-control scenarios for a par-
ticular economic sector was a function of the relative
eneigy-intensity and capital-intensity of that sector.
Increased production costs in eneigy- and capital-in-
tensive sectors under the control scenario were re-
flected in higher consumer prices, which resulted in
reductions in the quantity of consumer purchases of
goods and services produced by those sectors. This
reduction in consumer demand under the control sce-
nario led, ultimately, to reductions in output and em-
ployment in those sectors. The sectors most affected
by the CAA were motor vehicles, petroleum refining,
and electricity generation. The electricity generation
sector, for example, incurred a two to four percent
increase in consumer prices by 1990, resulting in a
three to five and a half percent reduction in output.
Many other manufacturing sectors saw an output ef-
fect in the one percent range.
Some other sectors, however, were projected to
increase output under the control scenario. Apart from
the pollution control equipment industry, which was
not separately identified and captured in the macro-
economic modeling performed for this study, two ex-
ample sectors for which output was higher and prices
were lower under the control scenario are food and
furniture. These two sectors showed production cost
and consumer price reductions of one to two percent
relative to other industries under the control scenario,
resulting in output and employment increases of simi-
lar magnitudes.
Aggregate Effects
As noted above, the control and no-control sce-
narios yield different estimated mixes of investment.
In particular, the control scenario was associated with
more pollution control capital expenditure and less
consumer commodity capital expenditure. As a result,
the growth pattern of the economy under the control
scenario differed from the no-control scenario. Under
the control scenario, the macroeconomic model pro-
jected a rate of long-run GNP growth about one twen-
tieth of one percent per year lower than under the no-
control scenario. Aggregating these slower growth
effects of the control scenario over the entire 1970 to
1990 period of this study results, by 1990, in a level
of GNP one percent (or approximately $55 billion)
lower than that projected under the no-control sce-
nario.
9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Although small relative to the economy as a whole,
the estimated changes in GNP imply that the poten-
tial impact of the CAA on the economy by 1990 was
greater than that implied by expenditures ($19 billion
in 1990) or annualized costs ($26 billion in 1990, an-
nualized at five percent). Discounting the stream of
1973-1990 GNP effects to 1990 gives an aggregate
impact on production of 1,005 billion dollars (in 1990
dollars discounted at five percent). Of that total, $569
billion represent reductions in household consump-
tion, and another $200 billion represent government
consumption, for an aggregate effect on U.S. consump-
tion of goods and services equal to 769 billion dol-
lars. Both the aggregate GNP effects and aggregate
consumption effects exceed total 1973-1990 expen-
ditures ($628 billion) and annualized costs ($523 bil-
lion, with all dollar quantities in $1990, discounted at
five percent).
Changes in GNP (or, even, changes in the national
product account category "consumption") do not nec-
essarily provide a good indication of changes in so-
cial welfare. Social welfare is not improved, for ex-
ample, by major oil tanker spills even though mea-
sured GNP is increased by the "production" associ-
ated with clean-up activities. Nevertheless, the effects
of the CAA on long-term economic growth would be
expected to have had some effect on economic wel-
fare. One of the characteristics of the macroeconomic
model used by the Project Team is its ability to esti-
mate a measure of social welfare change which is su-
perior to GNP changes. This social welfare measure
estimates the monetary compensation which would be
required to offset the losses in consumption (broadly
defined) associated with a given policy change. The
model reports a range of results, with the range sensi-
tive to assumptions regarding how cost impacts are
distributed through society. For the CAA, the model
reports an aggregate welfare effect of 493 billion to
621 billion dollars (in 1990 dollars), depending on
the distributional assumptions used. This range does
not differ greatly from the range of results represented
by 1973-1990 expenditures, compliance costs, and
consumption changes.
Uncertainties and Sensitivities in
the Cost and Macroeconomic
Analysis
The cost and macroeconomic analyses for the
present assessment relied upon survey responses, EPA
analyses, and a macroeconomic simulation model.
Although the Project Team believes that the results of
the cost and macroeconomic analyses are reasonably
reliable, it recognizes that every analytical step is sub-
ject to uncertainty. As noted at the beginning of this
chapter, explicit and implicit assumptions regarding
hypothetical technology development paths are cru-
cial to framing the question of the cost impact of the
CAA. In addition, there is no way to verify the accu-
racy of the survey results used;16 alternative, plausible
cost analyses exist that arrive at results that differ from
some of the results derived from EPA analyses; and it
is not clear how the use of a general equilibrium mac-
roeconomic model affects the accuracy of macroeco-
nomic projections in a macroeconomy characterized
by disequilibrium. For many factors engendering un-
certainty, the degree or even the direction of bias is
unknown. In several areas, nevertheless, uncertainties
and/or sensitivities can be identified that may bias the
results of the analysis.
Productivity and Technical Change
An important component of the macroeconomic
model used by the Project Team is its treatment of
technical change and productivity growth. Three fac-
tors associated with productivity and technical change
have been identified which may bias the results of the
macroeconomic simulation: (1) the long-run effects
of reducing the "stock" of technology, (2) the pos-
sible "chilling" effect of regulations on innovation and
technical change, and (3) the role of endogenous pro-
ductivity growth within the macroeconomic model.
The macroeconomic model projected a decrease
in the growth of GNP as a result of CAA compliance.
Decreased growth was due not only to decreased capi-
tal investment, but also to decreased factor productiv-
ity. The annual decrement in productivity can be
thought of as a reduction of the stock of available tech-
nology. That reduction in stock could be expected to
affect macroeconomic activity after 1990, as well as
16 For an example of the difficulties one encounters in assessing the veracity of survey results, see the discussion in Appendix A
on the apparently anomalous growth in stationary source O&M expenditures in relation to the size of the stationary source air
pollution control capital stock.
10
-------
Chapter 2: Cost and Macro economic Effect
during the 1973-1990 period studied by the Project
Team. Thus, to the extent that this effect exists, the
Project Team has underestimated the macroeconomic
impact ofthe CAA by disregarding the effect of 1973-
1990 productivity change decrements on post-1990
GNP
Some economists contend that regulations have a
"chilling" effect on technological innovation and,
hence, on productivity growth. Two recent studies by
Gray and Shadbegian,17 which are sometimes cited in
support of this contention, suggest that pollution abate-
ment regulations may decrease productivity levels in
some manufacturing industries. The macroeconomic
model allowed policy-induced productivity change
through the mechanism of price changes and result-
ant factor share changes. To the extent that additional
policy-induced effects on productivity growth exist,
the Project Team has underestimated the impact of
the CAA on productivity growth during the 1973-1990
period, and, thus, has underestimated macroeconomic
impacts during the 1973-1990 period and beyond.
The macroeconomic model allowed productivity
growth to vary with changes in prices generated by
the model. This use of "endogenous" productivity
growth is not universal in the economic growth litera-
ture — that is, many similar macroeconomic models
do not employ analogous forms of productivity growth.
The Project Team tested the sensitivity of the model
results to the use of endogenous productivity growth.
If the model is run without endogenous productivity
growth, then the predicted macroeconomic impacts
(GNP, personal consumption, etc.) of the CAA are
reduced by approximately 20 percent. That is, to the
extent that use of endogenous productivity growth in
the macroeconomic model is an inaccurate simulation
technique, then the Project Team has overestimated
the macroeconomic impact of the CAA.
Discount Rates
There is a broad range of opinion in the econom-
ics profession regarding the appropriate discount rate
to use in analyses such as the current assessment. Some
economists believe that the appropriate rate is one that
approximates the social rate of time preference — that
is, the rate of return at which individuals are willing
to defer consumption to the future. A three percent
rate would approximate the social rate of time prefer-
ence (all rates used here are "real", i.e., net of price
inflation impacts). Others believe that a rate that ap-
proximates the opportunity cost of capital (e.g., seven
percent or greater) should be used.18 A third school of
thought holds that some combination ofthe social rate
of time preference and the opportunity cost of capital
is appropriate, with the combination effected either
by use of an intermediate rate or by use of a multiple-
step procedure employing the social rate of time pref-
erence as the "discount rate," but still accounting for
the opportunity cost of capital.
The Project Team elected to use an intermediate
rate (five percent), but recognizes that analytical re-
sults aggregated across the study period are sensitive
to the discount rate used. Consequently, all cost mea-
sures are presented at three and seven percent, as well
as the base case five percent. Table 2 summarizes
major cost and macroeconomic impact measures ex-
pressed in constant 1990 dollars, and discounted to
1990 at rates of three, five, and seven percent.
Tabic 2. Compliance Cost. GNP. and
Consumption Impacts Discounted to 1990
($1990 billions)
3%
S»/n
7%
Expenditures
$52
628
761
Annualized Costs
417
523
657
GNP
880
1005
1151
Household Consumption
500
569
653
HH and Gov't Consnmntion
676
769
881
17 Gray, Wayne B., and Ronald J. Shadbegian, "Environmental Regulation and Manufacturing Productivity at the Plant Level,"
Center for Economic Studies Discussion Paper, CES 93-6, March 1993. Gray, Wayne B., and Ronald J. Shadbegian, "Pollution
Abatement Costs, Regulation, and Plant-Level Productivity," National Bureau of Economic Research, Inc., Working Paper Series,
Working Paper No. 4994, January 1995.
18 Some would argue that use of the opportunity cost of capital approach would be inappropriate in the current assessment if the
results of the macroeconomic modeling (such as GNP) were used as the definition of "cost," since the macro model already accounts
for the opportunity cost of capital. The appropriate rate would then be the social rate of time preference.
11
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Exclusion of Health Benefits from the
Macroeconomic Model
The macroeconomic modeling exercise was de-
signed to capture the second-order macroeconomic
effects arising from CAA compliance expenditures.
Those predicted second-order effects are among the
factors used to drive the emissions estimates and, ul-
timately, the benefits modeled for this assessment. The
benefits of the CAA, however, would also be expected
to induce second-order macroeconomic effects. For
example, increased longevity and decreased incidence
of non-fatal heart attacks and strokes would be ex-
pected to improve macroeconomic performance mea-
sures. The structure of the overall analysis, however,
necessitated that these impacts be excluded from the
macroeconomic simulation.
The first-order CAA beneficial effects have been
included in the benefits analysis for this study, includ-
ing measures that approximate production changes
(e.g., income loss due to illness, or lost or restricted
work days; income loss due to impaired cognitive abil-
ity; and income loss due to reduced worker produc-
tion in certain economic sectors). These measures are
analogous to compliance expenditures in the cost
analysis. The second-order benefits impacts, which
would result from price changes induced by CAA-
related benefits, have not been estimated. It is likely
that the estimated adverse second-order macroeco-
nomic impacts would have been reduced had the im-
pact of CAA benefits been included in the macroeco-
nomic modeling exercise; however, the magnitude of
this potential upward bias in the estimate of adverse
macroeconomic impact was not quantitatively as-
sessed.
12
-------
3
Emissions
This chapter presents estimates of emissions re-
ductions due to the Clean Air Act (CAA) for six crite-
ria air pollutants. Reductions are calculated by esti-
mating, on a sector-by-sector basis, the differences in
emissions between the control and no-control sce-
narios. While the relevant years in this analysis are
1970 through 1990, full reporting of emissions was
only made for the 1975 to 1990 period since 1970
emission levels are, by assumption, identical for the
two scenarios. The criteria pollutants for which emis-
sions are reported in this analysis are: total suspended
particulates (TSP),19 carbon monoxide (CO), volatile
organic compounds (VOC), sulfur dioxide (S02), ni-
trogen oxides (NOx), and Lead (Pb).
The purpose of the present study is to estimate
the differences in economic and environmental con-
ditions between a scenario reflecting implementation
of historical CAA controls and a scenario which as-
sumes that no additional CAA-related control pro-
grams were introduced after 1970. Because of the fo-
cus on differences in -rather than absolute levels of-
emissions between the scenarios, the various sector-
specific emission models were used to estimate both
the control and no-control scenario emission invento-
ries. This approach ensures that differences between
the scenarios are not distorted by differences between
modeled and actual historical emission estimates.20
Despite the use of models to estimate control sce-
nario emission inventories, the models used were con-
figured and/or calibrated using historical emissions
estimates. The control scenario utility emissions esti-
mates, for example, were based on the ICF CEUM
model which was calibrated using historical emissions
inventory data.21 In other cases, such as the EPA Emis-
sions Trends Report (Trends) methodology22 used to
estimate industrial process emissions, historical data
were used as the basis for control scenario emissions
with little or no subsequent modification. Neverthe-
less, differences in model selection, model configura-
tion, and macroeconomic input data23 result in un-
avoidable, but in this case justifiable, differences be-
tween national total historical emission estimates and
national total control scenario emission estimates for
each pollutant. Comparisons between no-control, con-
trol, and official EPA Trends Report historical emis-
sions inventories are presented in Appendix B.24
19 In 1987, EPA replaced the earlier TSP standard with a standard for particulate matter of 10 microns or smaller (PM10).
20 By necessity, emission models must be used to estimate the hypothetical no-CAA scenario. If actual historical emissions data
were adopted for the control scenario, differences between the monitoring data and/or models used to develop historical emission
inventories and the models used to develop no-control scenario emission estimates would bias the estimates of the differences between
the scenarios.
21 See ICF Resources, Inc., "Results of Retrospective Electric Utility Clean Air Act Analysis - 1980, 1985 and 1990," September
30, 1992, Appendix C.
22 EPA, 1994a: U.S. Environmental Protection Agency, "National Air Pollutant Emission Trends, 1900-1993," EPA-454/R-94-
027, Office of Air Quality Planning and Standards, Research Triangle Park, NC, October 1994.
23 The Jorgenson/Wilcoxen macroeconomic model outputs were used to configure both the control and no-control scenario
emission model runs. While this satisfies the primary objective of avoiding "across model" bias between the scenarios, the macroeco-
nomic conditions associated with the control scenario would not be expected to match actual historical economic events and condi-
tions. To the extent actual historical economic conditions are used to estimate official historical emission inventories, conformity
between these historical emissions estimates and control scenario emission estimates would be further reduced.
24 In general, these comparisons show close correspondence between control scenario and Trends estimates with the largest
differences occurring for VOC and CO emissions. The Trends report VOC estimates are generally higher than the control scenario
estimates due primarily to the inclusion of Waste Disposal and Recycling as a VOC source in the Trends report. This inconsistency is
of no consequence since Waste Disposal and Recycling sources were essentially uncontrolled by the historical CAA and therefore do
not appear as a difference between the control and no-control scenarios. The higher CO emission estimates in the Trends Report are
primarily associated with higher off-highway vehicle emissions estimates. Again, since off-highway emissions do not change between
the control and no-control scenario in the present analysis, this inconsistency is of no consequence.
13
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
To estimate no-control scenario emissions, sec-
tor-specific historical emissions are adjusted based on
changes in the following two factors: (1) growth by
sector predicted to occur under the no-control scenario;
and (2) the exclusion of controls attributable to spe-
cific provisions of the CAA.
To adjust emissions for economic changes under
the no-control scenario, activity levels that affect emis-
sions from each sector were identified. These activity
levels include, for example, fuel use, industrial activ-
ity, and vehicle miles traveled (VMT). The Jorgenson-
Wilcoxen (J/W) general equilibrium model was used
to estimate changes in general economic conditions,
as well as sector-specific economic outcomes used as
inputs to the individual sector emission models.25
"able 3. Summary of Scctor-Spccific Emission Modeling Approaches.
Sector
Modeling Approach
On-Highway Vehicles
Modeled using ANL's TEEMS; adjusted automobile emission estimates by
changes in personal travel and economic activity in the without CAA case.
Truck VMT was obtained from the Federal Highway Administration (FHWA).
MOBILE5 a emission factors were used to calculate emissions.
Lead emission changes from gasoline were estimated by Abt Associates based
on historical gasoline sales and the lead content of leaded gasoline in each
target year.
Off-Highway Vehicles
ELI analysis based on Trends methods. Recalculated historical emissions
using 1 970 control efficiencies from Trends. No adjustment was made to
activity levels in the without the CAA case.
Electric Utilities
ICF's Coal and Electric Utility Model (CEUM)usedto assess SO2, NO*, and
TSP emission changes. Electricity sales levels were adjusted with results of
the J/W model.
The Argonne Utility SimulationModel(ARGUS)provided CO and VOC
results. Changes in activity levels were adjusted with results of the J/W model.
Lead emissions were calculated based on energy consumption data and Trends
emission factors and control efficiencies.
Industrial Combustion
ANL industrial boiler analysis for S02, NO*, and TSP using the Industrial
Combustion Emissions (ICE) model.
VOC and CO emissions from industrial boilers were calculated based on
Trends methods; recalculated using 1970 control efficiencies.
Lead emissions calculated for boilers and processes based on Trends fuel
consumption data, emission factors, and 1970 control efficiencies.
Industrial Processes
ELI analyzed industrial process emissions based on Trends methods. Adjusted
historical emissions with J/W sectoral changes in output, and 1970 control
efficiencies from Trends.
Lead emissions calculated for industrial processes and processes based on
Trends fuel consumption data, emission factors, and 1970 control efficiencies.
Commercial / Residential
ANL's Commercial and Residential Simulation System (CRESS) model was
used.
25 For example, the change in distribution of households by income class predicted by the J/W model was used as input to the
transportation sector model system. Changes in household income resulted in changes in vehicle ownership and usage patterns which,
in turn, influence VMT and emissions. (See Pechan, 1995, p. 43).
14
-------
Chapter 3: Emissions
The specific outputs from the J/W model used in
this analysis are the percentage changes in gross na-
tional product (GNP), personal consumption, and out-
put for various economic sectors under the control and
no-control scenario for the years 1975, 1980, 1985,
and 1990.26 The sectors for which the results of the J/
W model are used include: industrial processes, elec-
tric utilities, highway vehicles, industrial boilers, and
the commercial/residential sector. For the off-highway
sector, economic growth was not taken into account
as there was no direct correspondence between J/W
sectors and the off-highway vehicle source category
activity.
In addition to adjusting for economic activity
changes, any CAA-related control efficiencies that
were applied to calculate control scenario emissions
were removed for the no-control scenario. In most
instances, emissions were recalculated based on 1970
control levels.
Uncertainty associated with several key model-
ing inputs and processes may contribute to potential
errors in the emission estimates presented herein. Al-
though the potential errors are likely to contribute in
only a minor way to overall uncertainty in the esti-
mated monetary benefits of the Clean Air Act, the most
significant emission modeling uncertainties are de-
scribed at the end of this chapter.
Sector-Specific Approach
The approaches used to calculate emissions for
each sector vary based on the complexity of estimat-
ing emissions in the absence of CAA controls, taking
economic activity levels and CAA regulations into
account. For the off-highway vehicle and industrial
process sectors, a relatively simple methodology was
developed. The approaches used for the highway ve-
hicles, electric utilities, industrial boilers, and com-
mercial/residential sectors were more complex be-
cause the J/W model does not address all of the deter-
minants of economic activity in these sectors that
might have changed in the absence of regulation. The
approaches by sector used to estimate emissions for
the two scenarios are summarized in Table 3, and are
described in more detail in Appendix B.
Summary of Results
Figure 2 compares the total estimated sulfur di-
oxide emission from all sectors under the control and
no-control scenarios over the period from 1975 to
1990. Figures 3, 4, 5, 6, and 7 provide similar com-
parisons forNOx, VOCs, CO, TSP, and Lead (Pb) re-
spectively.
Additional tables presented in Appendix B pro-
vide further breakdown of the emissions estimates by
individual sector. The essential results are character-
ized below. For most sectors, emission levels under
the control scenario were substantially lower than lev-
els projected under the no-control scenario. For some
pollutants, for example NOx, most of the reductions
achieved under the control scenario offset the growth
in emissions which would have occurred under the
no-control case as a result of increases in population
and economic activity. For other pollutants, particu-
larly lead, most of the difference in 1990 emissions
projected under the two scenarios reflects significant
improvement relative to 1970 emission levels. Ap-
pendix B also assesses the consistency of the control
and no-control scenario estimates for 1970 to 1990
with pre-1970 historical emissions trends data.
The CAA controls that affected S02 emitting
sources had the greatest proportional effect on indus-
trial process emissions, which were 60 percent lower
in 1990 than they would have been under the
no-control scenario. S02 emissions from electric utili-
ties and industrial boilers were each nearly 40 percent
lower in 1990 as a result of the controls. In terms of
absolute tons of emission reductions, controls on elec-
tric utilities account for over 10 million of the total 16
million ton difference between the 1990 control and
no-control scenario S02 emission estimates.
CAA regulation of the highway vehicles sector
led to the greatest percent reductions in VOC andNOx.
Control scenario emissions of these pollutants in 1990
were 66 percent and 47 percent lower, respectively,
than the levels estimated under the no-control scenario.
In absolute terms, highway vehicle VOC controls ac-
count for over 15 million of the roughly 17 million
ton difference in control and no-control scenario emis-
sions.
Differences between control and no-control sce-
nario CO emissions are also most significant for high-
way vehicles. In percentage terms, highway vehicle
CO emissions were 56 percent lower in 1990 under
the control scenario than under the no-control scenario.
Industrial process CO emission estimates under the
control scenario were about half the levels projected
under the no-control scenario. Of the roughly 89 mil-
26 For details regarding the data linkages between the J/W model and the various emission sector models, see Pechan (1995).
15
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Figure 2. Control and No-control Scenario Total S02
Emission Estimates.
f2
c
2.
00
W
40
30
¦S 20
10
I Control
. No-Contro
1975
1980 1985
Year
1990
Figure 5. Control and No-control Scenario Total CO
Emission Estimates.
e
o
H
C
o
.a
c
200
150
a 100
a s
o
I
50
I Control
„ No-Control
1975
1980 1985
Year
1990
Figure 3. Control and No-control Scenario Total NOx
Emission Estimates.
40
s2
G
o
on
c
W
30
20
10
^ Control
. No-Control
1975
1980 1985
Year
1990
Figure 6. Control and No-control Scenario Total TSP
Emission Estimates.
40
C
o
H
ts
o
JS
CO
W
30
c
•2 20
0 L_l_
j Control
. No-Control
1975
1980
1985
1990
Year
Figure 4. Control and No-control Scenario Total VOC
Emission Estimates.
40
c
e2
ts
o
J3
cn
c
w
30
20
10
I Control
. No-Control
1975
1980
1985
1990
Year
Figure 7. Control and No-control Scenario Total Pb
Emission Estimates.
C
o
H
t:
o
4=
00
G
C
200
150
5 100
50
l Control
. No-Contro
1975
1980
1985
1990
Year
16
-------
Chapter 3: Emissions
lion ton difference in CO emissions between the two
scenarios, 84 million tons are attributable to highway
vehicle controls and the rest is associated with reduc-
tions from industrial process emissions.
For TSP, the highest level of reductions on a per-
centage basis was achieved in the electric utilities sec-
tor. TSP emissions from electric utilities were 93 per-
cent lower in 1990 under the control scenario than
projected under the no-control scenario. TSP emis-
sions from industrial processes were also significantly
lower on a percentage basis under the control scenario,
with the differential reaching 76 percent by 1990.
This is not an unexpected result as air pollution
control regulations in the 1970's focused on solving
the visible particulate problems from laige fuel com-
bustors. In terms of absolute tons, electric utilities
account for over 5 million of the 16 million ton differ-
ence between the two scenarios and industrial pro-
cesses account for almost 10 million tons.
The vast majority of the difference in lead emis-
sions under the two scenarios is attributable to reduc-
tions in burning of leaded gasoline. By 1990, reduc-
tions in highway vehicle emissions account for 221
thousand of the total 234 thousand ton difference in
lead emissions. As shown in more detail in Appendix
B, airborne lead emissions from all sectors were vir-
tually eliminated by 1990.
As described in the following chapter and in Ap-
pendix C, these emissions inventories were used as
inputs to a series of air quality models. These air qual-
ity models were used to estimate air quality condi-
tions under the control and no-control scenarios.
Uncertainty in the Emissions
Estimates
The emissions inventories developed for the con-
trol and no-control scenarios reflect at least two ma-
jor sources of uncertainty. First, potential errors in the
macroeconomic scenarios used to configure the sec-
tor-specific emissions model contribute to uncertain-
ties in the emissions model outputs. Second, the emis-
sions models themselves rely on emission factors,
source allocation, source location, and other param-
eters which may be erroneous.
An important specific source of potential error
manifest in the present study relates to hypothetical
emission rates from various sources under the no-con-
trol scenario. Emission rates from motor vehicles, for
example, would have been expected to change during
the 1970 to 1990 period due to technological changes
not directly related to implementation of the Clean
Air Act (e.g., advent of electronic fuel injection, or
EFI). However, the lack of emissions data from ve-
hicles with EFI but without catalytic converters com-
pelled the Project Team to use 1970 emission factors
throughout the 1970 to 1990 period for the no-control
scenario. Although this creates a potential bias in the
emissions inventories, the potential errors from this
and other uncertainties in the emissions inventories
are considered unlikely to contribute significantly to
overall uncertainty in the monetary estimates of Clean
Air Act benefits. This conclusion is based on the de-
monstrably greater influence on the monetary benefit
estimates of uncertainties in other analytical compo-
nents (e.g., concentration-response functions). A list
of the most significant potential errors in the emis-
sions modeling, and their significance relative to over-
all uncertainty in the monetary benefit estimate, is
presented in Table 4.
17
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic 4. Uncertainties Associated with Emissions Modeling.
Potential Source of Erro r
Direction of Potential
Bias in Estimate of
Emission Reduction
Benefits
Significance Relative to Key
Uncertainties in Overall Monetary
Benefit Estimate
Use of 1970 motor vehicle emission factors
forno-control scenario without adjustment
for advent of Electronic Fuel Injection
(EFI) and Electronic Ignition (EI).
Overestimate.
Unknown, but likely to be minor due
to overwhelming significance of
catalysts in determining emission
rales.
Use of ARGUS for utility CO and VOC
rather than CEUM.
Unknown.
Negligible.
Use of historical fuel consumption to
estimate 1975 SO2, NO*, TSP utility
emissions.
Unknown.
Negligible.
Adoption of assumption that utility unit
inventories remain fixed between the
control and no-control scenarios.
Overestimate.
Unknown, but likely to be small
since the CAA had virtually no effect
on costs of new coal-fired plants
built priorto 1975 and these plants
comprise a large majority of total
coal-fired capacity operating in the
1970 to 1990period. (SeelCF
CEUM Report, p. 7).
18
-------
4
Air Quality
Air quality modeling is the crucial analytical step
which links emissions to changes in atmospheric con-
centrations of pollutants which affect human health
and the environment. It is also one of the more com-
plex and resource-intensive steps, and contributes sig-
nificantly to overall uncertainty in the bottom-line
estimate of net benefits of air pollution control pro-
grams. The assumptions required to estimate hypo-
thetical no-control scenario air quality conditions are
particularly significant sources of uncertainty in the
estimates of air quality change, especially for those
pollutants which are not linearly related to changes in
associated emissions. Specific uncertainties are de-
scribed in detail at the end of this chapter.
The key challenges faced by air quality modelers
attempting to translate emission inventories into air
quality measures involve modeling of pollutant dis-
persion and atmospheric transport, and modeling of
atmospheric chemistry and pollutant transformation.
These challenges are particularly acute for those pol-
lutants which, rather than being directly emitted, are
formed through secondary formation processes. Ozone
is the paramount example since it is formed in the
atmosphere through complex, nonlinear chemical in-
teractions of precursor pollutants, particularly vola-
tile organic compounds (VOCs) and nitrogen oxides
(NOx). In addition, atmospheric transport and trans-
formation of gaseous sulfur dioxide and nitrogen ox-
ides to particulate sulfates and nitrates, respectively,
contributes significantly to ambient concentrations of
fine particulate matter. In addition to managing the
complex atmospheric chemistry relevant for some
pollutants, air quality modelers also must deal with
uncertainties associated with variable meteorology and
the spatial and temporal distribution of emissions.
Given its comprehensive nature, the present analy-
sis entails all of the aforementioned challenges, and
involves additional complications as well. For many
pollutants which cause a variety of human health and
environmental effects, the concentration-response
functions which have been developed to estimate those
effects require, as inputs, different air quality indica-
tors. For example, adverse human health effects of
particulate matter are primarily associated with the
respirable particle fraction;27 whereas household soil-
ing is a function of total suspended particulates, espe-
cially coarse particles. It is not enough, therefore, to
simply provide a single measure of particulate matter
air quality. Even for pollutants for which particle size
and other characteristics are not an issue, different air
quality indicators are needed which reflect different
periods of cumulative exposure (i.e., "averaging peri-
ods"). For example, 3-month growing season averages
are needed to estimate effects of ozone on yields of
some agricultural crops, whereas adverse human health
effect estimates require ozone concentration profiles
based on a variety of short-term averaging periods.28
Fortunately, in responding to the need for scien-
tifically valid and reliable estimation of air quality
changes, air quality modelers and researchers have
developed a number of highly sophisticated atmo-
spheric dispersion and transformation models. These
models have been employed for years supporting the
development of overall federal clean air programs,
national assessment studies, State Implementation
Plans (SIPs), and individual air toxic source risk as-
sessments. Some of these models, however, require
massive amounts of computing power. For example,
completing the 160 runs of the Regional Acid Depo-
sition Model (RADM) required for the present study
required approximately 1,080 hours of CPU time on a
Cray-YMP supercomputer at EPA's Bay City
Supercomputing Center.
Given the resource-intensity of many state-of-the-
art models, the Project Team was forced to make dif-
ficult choices regarding where to invest the limited
27 Particles with an aerometric diameter of less than or equal to 10 microns.
28 For example, ozone concentration-response data exists for effects associated with 1-hour, 2.5-hour, and 6.6-hour exposures.
19
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
resources available for air quality modeling. With a
mandate to analyze all of the key pollutants affected
by historical Clean Air Act programs, to estimate all
of the significant endpoints associated with those pol-
lutants, and to do so for a 20 year period covering the
entire continental U.S., it was necessary to use sim-
plified approaches for most of the pollutants to be
analyzed. In several cases related to primary emissions
—particularly sulfur dioxide (S02), nitrogen oxides
(NOx), and carbon monoxide (CO)— simple "roll-up
model" strategies were adopted based on the expecta-
tion that changes in emissions of these pollutants
would be highly correlated with subsequent changes
in air quality.29 Significant pollutants involving sec-
ondary atmospheric formation, nonlinear formation
mechanisms, and/or long-range transport were ana-
lyzed using the best air quality model which was af-
fordable given time and resource limitations. These
models, discussed in detail in Appendix C, included
the Ozone Isopleth Plotting with Optional Mechanism-
IV (OZIPM4) model for urban ozone; various forms
of the above-referenced RADM model for background
ozone, acid deposition, sulfate, nitrate, and visibility
effects in the eastern U.S.; and the SJVAQS/AUSPEX
Regional Modeling Adaptation Project (SARMAP)
Air Quality Model (SAQM) for rural ozone in Cali-
fornia agricultural areas. In addition, a linear scaling
approach was developed and implemented to estimate
visibility changes in large southwestern U.S. urban
areas.
By adopting simplified approaches for some pol-
lutants, the air quality modeling step adds to the over-
all uncertainties and limitations of the present analy-
sis. The limited expanse and density of the U.S. air
quality monitoring network and the limited coverage
by available air quality models of major geographic
areas30 further constrain the achievable scope of the
present study. Under these circumstances, it is impor-
tant to remember the extent and significance of gaps
in geographic coverage for key pollutants when con-
sidering the overall results of this analysis. Key un-
certainties are summarized at the end of this chapter
in Table 5. More extensive discussion of the caveats
and uncertainties associated with the air quality model-
ing step is presented in Appendix C. In addition, in-
formation regarding the specific air quality models
used, the characteristics of the historical monitoring
data used as the basis for the control scenario pro-
files, pollutant-specific modeling strategies and as-
sumptions, references to key supporting documents,
and important caveats and uncertainties are also pre-
sented in Appendix C.
General Methodology
The general methodological approach taken in this
analysis starts with the assumption that actual histori-
cal air quality will be taken to represent the control
scenario. This may seem somewhat inconsistent with
the approach taken in earlier steps of the analysis,
which used modeled macroeconomic conditions as the
basis for estimating macroeconomic effects and emis-
sions. However, the central focus of the overall analy-
sis is to estimate the difference in cost and benefit
outcomes between the control and no-control sce-
narios. It is consistent with this central paradigm to
use actual historical air quality data as the basis for
estimating how air quality might have changed in the
absence of the Clean Air Act.
The initial step, then, for each of the five non-
lead (Pb) criteria pollutants31 was to compile com-
prehensive air quality profiles covering the entire ana-
lytical period from 1970 to 1990. The source for these
data was EPA's Aerometric Information Retrieval
System (AIRS), which is a publicly accessible data-
base of historical air quality data. The vast number of
air quality observations occurring over this twenty year
period from the thousands of monitors in the U.S. in-
dicates the need to represent these observations by
statistical distributions. As documented in detail in
the supporting documents covering S02, NOx, CO, and
ozone,32 both lognormal and gamma distributional
forms were tested against actual data to determine the
29 It is important to emphasize that the correlation expected is between changes in emissions and changes in air quality. Direct
correlations between the absolute emissions estimates and empirical air quality measurements used in the present analysis may not be
strong due to expected inconsistencies between the geographically local, monitor-proximate emissions densities affecting air quality
data.
30 For example, the regional oxidant models available for the present study do not cover some key Midwestern states, where
human health, agricultural crop, and other effects from ozone may be significant.
31 Lead (Pb), the sixth criteria pollutant, is analyzed separately. The ability to correlate emissions directly with blood lead levels
obviates the need for using air quality modeling as an intermediate step toward estimation of exposure.
32 See SAI S02, NOx, and CO Report (1994) and SAI Ozone Report (1995).
20
-------
Chapter 4: Air Quality
form which provided the best fit to the historical data.33
Based on these tests, one or the other statistical distri-
bution was adopted for the air quality profiles devel-
oped for each pollutant. In addition to reducing the
air quality data to a manageable form, this approach
facilitated transformations of air quality profiles from
one averaging period basis to another.
Once the control scenario profiles based on his-
torical data were developed, no-control scenarios were
derived based on the results of the various air quality
modeling efforts. Again, the focus of the overall analy-
sis is to isolate the difference in outcomes between
the control and no-control scenarios. The no-control
scenario air quality profiles were therefore derived by
adjusting the control scenario profiles upward (or
downward) based on an appropriate measure of the
difference in modeled air quality outcomes. To illus-
trate this approach, consider a simplified example
where the modeled concentration of Pollutant A un-
der the no-control scenario is 0.12 ppm, compared to
a modeled concentration under the control scenario
of 0.10 ppm. An appropriate measure of the differ-
ence between these outcomes, whether it is the 0.02
ppm difference in concentration or the 20 percent per-
centage differential, is then used to ratchet up the con-
trol case profile to derive the no-control case profile.
Generally, the modeled differential is applied across
the entire control case profile to derive the no-control
case profile. As described below in the individual sec-
tions covering particulate matter and ozone, however,
more refined approaches are used where necessary to
take account of differential outcomes for component
species (i.e., particulate matter), long-range transport,
and background levels of pollutants.
Sample Results
The results of the air quality modeling effort in-
clude a vast array of monitor-specific air quality pro-
files for particulate matter (PM10 and TSP),34 S02,
N02, NO, CO, and ozone; RADM grid cell-based esti-
mates of sulfur and nitrogen deposition; and estimates
of visibility degradation for eastern U.S. RADM grid
cells and southwestern U.S. urban areas. All of these
data were transferred to the effects modelers for use in
configuring the human health, welfare, and ecosystem
physical effects models. Given the massive quantity
and intermediate nature of the air quality data, they
are not exhaustively reported herein.35 To provide the
reader with some sense of the magnitude of the differ-
ence in modeled air quality conditions under the con-
trol and no-control scenarios, some illustrative results
for 1990 are presented in this chapter and in Appen-
dix C. In addition, maps depicting absolute levels of
control and no-control scenario acid deposition and
visibility are presented to avoid potential confusion
which might arise through examination of percent
change maps alone.36
Carbon Monoxide
Figure 8 provides an illustrative comparison of
1990 control versus no-control scenario CO concen-
trations, expressed as a frequency distribution of the
ratios of 1990 control to no-control scenario 95th per-
centile 1-hour average concentrations at individual CO
monitors. Consistent with the emission changes un-
derlying these air quality results, CO concentrations
under the control scenario tend to be about half those
projected under the no-control scenario, with most
individual monitor ratios ranging from about 0.40 to
0.60 percent, and a few with ratios in the 0.60 to 0.80
range.
Figure 8. Frequency Distribution of Estimated Ratios for
1990 Control to No-control Scenario 95th Percentile 1-
Hour Average CO Concentrations, by Monitor.
300
200
"I 100
3
£
I I l l
_J I I L.
0.05 0.25 0.45 0.65 0.85 1.05 1.25
Ratio of CAA:No-CAA 95th Percentile 1-Hour Average
21
33 The statistical tests used to determine goodness of fit are descnbed m the SAI reports.
34 PM data are reported as county-wide values for counties with PM monitors and a sufficient number of monitor observations.
35 The actual air quality profiles, however, are available on disk from EPA. See Appendix C for further information.
36 Large percentage changes can result from even modest absolute changes when they occur in areas with good initial (e.g.,
control scenario) air quality. Considering percentage changes alone might create false impressions regarding absolute changes in air
quality in some areas. For example, Appendix C discusses in detail two such cases: the Upper Great Lakes and Florida-Southeast
Atlantic Coast areas, which show high percentage changes in sulfur deposition and visibility.
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
In considering these results, it is important to note
that CO is essentially a "hot spot" pollutant, meaning
that higher concentrations tend to be observed in lo-
calized areas of relatively high emissions. Examples
of such areas include major highways, major inter-
sections, and tunnels. Since CO monitors tend to be
located in order to monitor the high CO concentra-
tions observed in such locations, one might suspect
that using state-wide emissions changes to scale air
quality concentration estimates at strategically located
monitors might create some bias in the estimates.
However, the vast majority of ambient CO is contrib-
uted from on-highway vehicles. In addition, the vast
majority of the change in CO emissions between the
control and no-control scenario occurs due to catalyst
controls on highway vehicles. Since CO hot spots re-
sult primarily from highway vehicles emissions, con-
trolling such vehicles would mean CO concentrations
would be commensurately lowered at CO monitors.
While variability in monitor location relative to ac-
tual hot spots and other factors raise legitimate con-
cerns about assuming ambient concentrations are cor-
related with emission changes at any given monitor,
the Project Team believes that the results observed
provide a reasonable characterization of the aggregate
change in ambient CO concentrations between the two
scenarios.
Sulfur Dioxide
As for CO, no-control scenario S02 concentra-
tions were derived by scaling control scenario air qual-
ity profiles based on the difference in emissions pre-
dicted under the two scenarios. Unlike CO, S02 is
predominantly emitted from industrial and utility
sources. This means that emissions, and the changes
in emissions predicted under the two scenarios, will
tend to be concentrated in the vicinity of major point
sources. Again, while state-wide emissions changes
are used to scale S02 concentrations between the sce-
narios, these state-wide emission changes reflect the
controls placed on these individual point sources.
Therefore, the Project Team again considers the dis-
tribution of control to no-control ratios to be a rea-
sonable characterization of the aggregate results de-
spite the uncertainties associated with estimation of
changes at individual monitors.
Figure 9 provides a histogram of the predicted
control to no-control ratios for S02 which is similar
to the one presented for CO. The results indicate that,
on an overall basis, S02 concentrations were reduced
by about one-third. The histogram also shows a much
wider distribution of control to no-control ratios for
individual monitors than was projected for CO. This
result reflects the greater state to state variability in
S02 emission changes projected in this analysis. This
greater state to state variability in turn is a function of
the variable responses of S02 point sources to histori-
cal C control requirements.37 This source-specific vari-
ability was not observed for CO because controls were
applied relatively uniformly on highway vehicles.
Figure 9. Frequency Distribution of Estimated Ratios for
1990 Control to No-control Scenario 95th Percentile 1-
Hour Average S02 Concentrations, by Monitor.
300
100
0
0.05
0.25
0.45
0.65
0.85
1.05
1.25
Ratio of CAA:No-CAA 95th Percentile 1-Hour Average
Nitrogen Dioxide
Results forN02 are presented in Figure 10. These
results are similar to the results observed for CO, and
for a similar reason: the vast majority of change in
N02 emissions between the two scenarios is related
to control of highway vehicle emissions. While
baseline emissions ofN02 from stationary sources may
be significant, these sources were subject to minimal
controls during the historical period of this analysis.
On an aggregated basis, overall N02 concentrations
are estimated to be roughly one-third lower under the
control scenario than under the no-control scenario.
37 Figure 9 indicates that six monitors were projected to have higher S02 concentrations for 1990 under the control scenario than
under the no-control scenario. All six of these monitors are located in Georgia, a state for which higher 1990 utility SO^ emissions are
projected in the control scenario due to increased use of higher-sulfur coal. The projected increase in overall Georgia utility consump-
tion of higher sulfur coal under the control case is a result of increased competition for the low-sulfur southern Appalachian coal
projected to occur under the control scenario.
22
-------
Chapter 4: Air Quality
Figure 10. Frequency Distribution of Estimated Ratios for
1990 Control to No-control Scenario 95th Percentile 1-
Hour Average N02 Concentrations, by Monitor.
300
200
*2
0 ioo
S3
J I I ' L.
-J I I l_
0.05 0.25 0.45 0.65 0.85 1.05 1.25
Ratio of CAA:No-CAA 95th Percentile 1-Hour Average
Particulate Matter
An indication of the difference in outcomes for
particulate matter between the two scenarios is pro-
vided by Figure 11. This graph shows the distribution
of control to no-control ratios for annual mean TSP in
1990 for those counties which both had particulate
monitors and a sufficient number of observations from
those monitors.38 While the distribution of results is
relatively wide, reflecting significant county to county
variability in ambient concentration, on a national
aggregate basis particulate matter concentrations un-
Figure 11. Frequency Distribution of Estimated Ratios for
1990 Control to No-control Annual Mean TSP Concentra-
tions, by Monitored County.
0.00 0.20 0.40 0.60 0.80 1.00
Ratio of CAA:No-CAA Annual Mean TSP (interval midpoint)
der the control scenario were just over half the level
projected under the no-control scenario. The signifi-
cant county to county variability observed in this case
reflects point source-specific controls on particulate
matter precursors, especially S02, and the effects of
long-range transport and transformation.
Ozone
Urban Ozone
Figure 12 presents a summary of the results of the
1990 OZIPM4 ozone results for all 147 of the mod-
eled urban areas. In this case, the graph depicts the
distribution of ratios of peak ozone concentrations
estimated for the control and no-control scenarios.
While the vast majority of simulated peak ozone con-
centration ratios fall below 1.00, eight urban areas
show lower simulated peak ozone for the no-control
scenario than for the control scenario. For these eight
urban areas, emissions of precursors were higher un-
der the no-control scenario; however, the high pro-
portion of ambient NOx compared to ambient non-
methane oiganic compounds (NMOCs) in these areas
results in a decrease in net ozone production in the
vicinity of the monitor when N O emissions increase.39
Figure 12. Distribution of Estimated Ratios for 1990
Control to No-control OZIPM4 Simulated 1-Hour Peak
Ozone Concentrations, by Urban Area.
30
20
£ 10
_a a_
IBl I
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Ratio ofCA A :No-CAA Peak Ozone (intervalmidpoint)
38 Given the relative importance of particulate matter changes to the bottom line estimate of CAA benefits, and the fact that a
substantial portion of the population lives in unmonitored counties, a methodology was developed to allow estimation of particulate
matter benefits for these unmonitored counties. This methodology was based on the use of regional air quality modeling to interpolate
between monitored counties. It is summarized in Appendix C and described in detail in the SAI PM Interpolation Report (1996).
39 Over an unbounded geographic area, NOx reductions generally decrease net ozone production.
23
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Rural Ozone
Acid Deposition
Figures 13 and 14 present frequency distributions
for control to no-control ratios of average ozone-sea-
son daytime ozone concentrations at rural monitors
as simulated by SAQM and RADM, respectively.
Figure 13. Distribution of Estimated Ratios for 1990
Control to No-control SAQM Simulated Daytime Average
Ozone Concentrations, by SAQM Monitor.
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Ratio ofCAA :No-CAA Ozone-Season Daytime Average Ozone (interval midpoint)
Both the RADM and SAQM results indicate rela-
tively little overall change in rural ozone concentra-
tions. This is primarily because reductions in ozone
precursor emissions were concentrated in populated
areas.
Figure 14. Distribution of Estimated Ratios for 1990
Control to No-control RADM Simulated Daytime Average
Ozone Concentrations, by RADM Grid Cell.
200
u
•G 150
O
(§
I
100
50
.1
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Ratio ofCA A :No-CAA Ozone-Season Daytime Average Ozone (intervalmidpoint)
Figure 15 is a contour map showing the estimated
percent increase in sulfur deposition under the no-con-
trol scenario relative to the control scenario for 1990.
Figure 16 provides comparable information for nitro-
gen deposition.
Figure 15. RADM-Predicted Percent Increase in Total
Sulfur Deposition (Wet + Dry) Under the No-control
Scenario.
jiggiiB
r
f i/
.v
130 - 35
> 40
These results show that acid deposition rates in-
crease significantly under the no-control scenario,
particularly in the Atlantic Coast area and in the vi-
cinity of states for which relatively large increases in
emissions are projected under the no-control scenario
(i.e., Kentucky, Florida, Michigan, Mississippi, Con-
necticut, and Florida).
In the areas associated with laige increases in sul-
fur dioxide emissions, rates of sulfur deposition in-
crease to greater than or equal to 40 percent. The high
proportional increase in these areas reflects both the
significant increase in acid deposition precursor emis-
sions in upwind areas and the relatively low deposi-
tion rates observed under the control scenario.40
Along the Atlantic Coast, 1990 nitrogen deposi-
tion rates increase by greater than or equal to 25 per-
cent under the no-control scenario. This is primarily
due to the significant increase in mobile source nitro-
gen oxide emissions along the major urban corridors
of the eastern seaboard.
40 Even small changes in absolute deposition can yield large percentage changes when initial absolute deposition is low. See
Appendix C for further discussion of this issue.
24
-------
Chapter 4: Air Quality
Figure 16. RADM-Predicted Percent Increase in Total
Nitrogen Deposition (Wet + Dry) Under the No-
control Scenario.
1
lEuEimD:
:: 0-10
mo - 15
115 - 20
120-25
> 25
Visibility
The difference in modeled 1990 control and
no-control scenario visibility conditions projected by
the RADM/EM for the eastern U.S. is depicted by the
contour map presented in Figure 17. This figure shows
the increase in modeled annual average visibility deg-
radation, in DeciView41 terms, for 1990 when mov-
Figure 17. RADM-Predicted Percent Increase in
Visibility Degradation, Expressed in DeciViews, for
Poor Visibility Conditions (90th Percentile) Under the
No-control Scenario.
ing from the control to the no-control scenario. Since
the DeciView metric is based on perceptible changes
in visibility, these results indicate noticeable deterio-
ration of visibility in the eastern U.S. underthe no-
control scenario.
Visibility changes in 30 southwestern U.S. urban
areas were also estimated using emissions scaling tech-
niques. This analysis also found significant, percep-
tible changes in visibility between the two scenarios.
Details of this analysis, including the specific out-
comes for the 30 individual urban areas, are presented
in Appendix C.
Uncertainty in the Air Quality
Estimates
Uncertainty prevades the projected changes in air
quality presented in this study. These uncertainties
arise due to potential inaccuracies in the emissions
inventories used as air quality modeling inputs and
due to potential errors in the structure and parameter-
ization of the air quality models themselves. In addi-
tion, an important limitation of the present study is
the lack of available data and/or modeling results for
some pollutants in some regions of the country (e.g.,
visibility changes in western U.S. Class I areas such
as the Grand Canyon). The inability to provide com-
prehensive estimates of changes in air quality due to
the Clean Air Act creates a downward bias in the
monetary benefit estimates.
The most important specific sources of uncertainty
are presented in Table 5, and are described further in
Appendix C. While the list of potential errors pre-
sented in Table 5 is not exhaustive, it incorporates the
uncertainties with the greatest potential for contribut-
ing to error in the monetary benefit estimates. Over-
all, the uncertainties in the estimated change in air
quality are considered small relative to uncertainties
contributed by other components of the analysis.
41 The DeciView Haze Index (dV) is a relatively new visibility indicator aimed at measuring visibility changes in terms of human
perception. It is described in detail in Appendix C.
25
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic 5. Key Uncertainties Associated with Air Quality Modeling.
Potential Source of Error
Direction of
Potential Bias
in Estimate of
Air Quality
Benefits
Significance Relative to Key
Uncertainties in Overall Monetary
Benefit Estimate
Use of OZIPM4 model, which does not
capture long-range and night-time transport of
ozone. Use of a regional oxidant model, such
as UAM-V, would mitigate errors associated
with neglecting transport.
Underestimate.
Significant, but probably not major.
Overall average ozone response of 15% to
NOx and VOC reductions of
approximately 30% and 45%,
respectively. Even if ozone response
doubled to 30%, estimate of monetized
benefits of CAA will not change very
much. Significant benefits of ozone
reduction, however, could not be
monetized.
Use of early biogenic emission estimates in
RADM to estimate rural ozone changes in the
eastern 31 states.
Underestimate.
Probably minor. Errors are estimated to
be within -15% to +25 % of the ozone
predictions.
Use of proxy pollutants to scale up some
particulate species in some areas. Uncertainty
is created to the extent species of concern are
not perfectly correlated with the proxy
pollutants.
Unknown.
Potentially significant. Given the relative
importance ofthe estimated changes in
fine particle concentrations to the
monetized benefit estimate, any
uncertainty associated with fine particles
is potentially significant. However, the
potential error is mitigated to some extent
since proxy pollutant measures are applied
to both scenarios.
Use of state-wide average emission reductions
to configure air quality models. In some
cases, control programs may have been
targeted to problem areas, so using state-wide
averages would miss relatively large
reductions in populated areas.
Underestimate.
Probably minor.
Exclusion of visibility benefits in Class I
areas in the Southwestern U.S.
Underestimate.
Probably minor. No sensitivity analysis
has been performed; however, monetized
benefits of reduced visibility impairment
in the Southwest would probably not
significantly alter the estimate of
monetized benefits.
26
-------
Chapter 4: Air Quality
Tabic 5 (con"l). Key Uncertainties Associated with Air Quality Modeling.
Potential Source of Error
Direction of
Potential Bias
in Estim ate of
Air Quality
Benefits
Significance Relative to Key
Uncertainties in Overall Monetary
Benefit Estimate
Lack of model coverage in western 17 states
for acid deposition.
Underestimate.
Probably minor. No sensitivity analysis
has been performed; however, monetized
benefits of reduced acid deposition in the
17 western states would probably not
significantly alter the estimate of
monetized benefits.
Use of spatially and geographically
aggregated emissions data to configure
RADM. Lack of available day-specific
meteorological data results in inability to
account for temperature effects on VOCs and
effect of localized meteorology around major
point sources.
Unknown.
Potentially significant. Any effect which
might influence the direction of long-
range transport of fine particulates such as
sulfates and nitrates could significantly
influence the estimates of total monetized
benefits of the CAA.
Use of constant concentration for organic
aerosols between the two scenarios. Holding
organic aerosol concentrations fixed omits the
effect of changes in this constituent of fine
particulate matter.
Underestimate.
Probably minor, because (a) nitrates were
also held fixed and nitrates and organic
aerosols move in opposite directions so
the exclusion of both mitigates the effect
of omitting either, (b) sulfates are by far
the dominant species in the eastern U.S.,
and (c) larger errors would be introduced
by using emissions scaling to estimate
changes in organic aerosols since a
significant fraction of organic aerosols are
caused by biogenic gas-phase VOC
emissions which do not change between
the scenarios.
Unavailability of ozone models for rural areas
outside the RADM and SAQM domains.
Underestimate.
Probably minor. Misses potential human
health, welfare, and ecological benefits of
reducing rural ozone in agricultural and
other rural areas; however, ozone changes
are likely to be small given limited
precursor reductions in rural areas.
RADM control:no-control ratios are in
fact, relatively small.
Use of peak episode changes to estimate
changes in annual distribution of ozone
concentration.
Unknown.
Probably minor, particularly since relative
changes in ozone concentration between
the scenarios were small.
27
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
28
-------
5
Physical Effects
Human Health and Welfare
Effects Modeling Approach
This chapter identifies and, where possible, esti-
mates the principal health and welfare benefits en-
joyed by Americans due to improved air quality re-
sulting from the CAA. Health benefits have resulted
from avoidance of air pollution-related health effects,
such as premature mortality, respiratory illness, and
heart disease. Welfare benefits accrued where im-
proved air quality averted damage to measurable re-
sources, including agricultural production and visibil-
ity. The analysis of physical effects required a combi-
nation of three components: air quality, population,
and health or welfare effects. As structured in this
study, the 3-step process involved (1) estimating
changes in air quality between the control and no-con-
trol scenarios, (2) estimating the human populations
and natural resources exposed to these changed air
quality conditions, and (3) applying a series of con-
centration-response equations which translated
changes in air quality to changes in physical health
and welfare outcomes for the affected populations.
Air Quality
The Project Team first estimated changes in con-
centrations of criteria air pollutants between the con-
trol scenario, which at this step was based on histori-
cal air quality, and the no-control scenario. Air qual-
ity improvements resulting from the Act were evalu-
ated in terms of both their temporal distribution from
1970 to 1990 and their spatial distribution across the
48 conterminous United States. Generally, air pollu-
tion monitoring data provided baseline ambient air
quality levels for the control scenario. Air quality
modeling was used to generate estimated ambient con-
centrations for the no-control scenario. A variety of
modeling techniques was applied, depending on the
pollutant modeled. These modeling approaches and
results are summarized in Chapter 4 and presented in
detail in Appendix C.
Population
Health and some welfare benefits resulting from
air quality improvements were distributed to individu-
als in proportion to the reduction in exposure. Pre-
dicting individual exposures, then, was a necessary
step in estimating health effects. Evaluating exposure
changes for the present analysis required not only an
understanding of where air quality improved as a re-
sult of the CAA, but also how many individuals were
affected by varying levels of air quality improvements.
Thus, a critical component of the benefits analysis
required that the distribution of the U.S. population
nationwide be established.
Three years of U.S. Census data were used to rep-
resent the geographical distribution of U.S. residents:
1970, 1980, and 1990. Population data was supplied
at the census block group level, with approximately
290,000 block groups nationwide. Allocating air qual-
ity improvements to the population for the other tar-
get years ofthis study - 1975 and 1985 - necessitated
interpolation of the three years of population data.
Linear interpolation was accomplished for each block
group in order to maintain the variability in growth
rates throughout the country.
Health and Welfare Effects
Benefits attributable to the CAA were measured
in terms of the avoided incidence of physical health
effects and measured welfare effects. To quantify such
benefits, it was necessary to identify concentration-
response relationships for each effect being consid-
ered. As detailed in Appendix D, such relationships
were derived from the published science literature. In
the case of health effects, concentration-response func-
tions combined the air quality improvement and popu-
lation distribution data with estimates of the number
of fewer individuals that suffer an adverse health ef-
fect per unit change in air quality. By evaluating each
concentration-response function for every monitored
location throughout the country, and aggregating the
29
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
resulting incidence estimates, it was possible to gen-
erate national estimates of incidence under the con-
trol and no-control scenarios.
In performing this step of the analysis, the Project
Team discovered that it was impossible to estimate
all of the health and welfare benefits which have re-
sulted from the Clean Air Act. While scientific infor-
mation was available to support estimation of some
effects, many other important health and welfare ef-
fects could not be estimated. Furthermore, even though
some physical effects could be quantified, the state of
the science did not support assessment of the economic
value of all of these effects. Table 6 shows the health
effects for which quantitative analysis was prepared,
as well as some of the health effects which could not
be quantified in the analysis. Table 7 provides similar
information for selected welfare effects.
While the 3-step analytical process described
above was applied for most pollutants, health effects
for lead were evaluated using a different methodol-
ogy. Gasoline as a source of lead exposure was ad-
dressed separately from conventional point sources.
Instead of using ambient concentrations of lead re-
sulting from use of leaded gasoline, the concentra-
tion-response functions linked changes in lead releases
directly to changes in the population's mean blood
lead level. The amount of leaded gasoline used each
year was directly related to mean blood lead levels
using a relationship described in the 1985 Lead Regu-
latory Impact Analysis (U.S. EPA, 1985). Health ef-
fects resulting from exposure to point sources of at-
mospheric lead, such as industrial facilities, were con-
sidered using the air concentration distributions mod-
eled around these point sources. Concentration-re-
sponse functions were then used to estimate changes
in blood lead levels in nearby populations.
Most welfare effects were analyzed using the same
basic 3-step process used to analyze health effects,
with one major difference in the concentration-re-
sponse functions used. Instead of quantifying the re-
lationship between a given air quality change and the
number of cases of a physical outcome, welfare ef-
fects were measured in terms of the avoided resource
losses. An example is the reduction in agricultural crop
losses resulting from lower ambient ozone concentra-
tions under the control scenario. These agricultural
benefits were measured in terms of net economic sur-
plus.
Another important welfare effect is the benefit
accruing from improvements in visibility under the
control scenario. Again, a slightly different method-
ological approach was used to evaluate visibility im-
provements. Visibility changes were a direct output
of the models used to estimate changes in air qual-
ity.42 The models provided estimates of changes in
light extinction, which were then translated mathemati-
cally into various specific measures of perceived vis-
ibility change.43 These visibility change measures were
then combined with population data to estimate the
economic value of the visibility changes. Other wel-
fare effects quantified in terms of avoided resource
losses include household soiling damage by PM10 and
decreased worker productivity due to ozone exposure.
The results of the welfare effects analysis are found
in Chapter 6 and in Appendices D and F.
Because of a lack of available concentration-re-
sponse functions (or a lack of information concerning
affected populations), ecological effects were not
quantified for this analysis. However, Appendix E
provides discussion of many of the important ecologi-
cal benefits which may have accrued due to historical
implementation of the CAA.
Key Analytical Assumptions
Several important analytical assumptions affect
the confidence which can be placed in the results of
the physical effects analysis. The most important of
these assumptions relate to (a) mapping of potentially
exposed populations to the ambient air quality moni-
toring network, (b) choosing among competing scien-
tific studies in developing quantitative estimates of
physical effects, (c) quantifying the contribution to
analytical uncertainty of within-study variances in
effects estimates and, perhaps most important in the
context of the present study, (d) estimating particu-
late matter-related mortality based on the currently
available scientific literature.
Because these resultant uncertainties were caused
by the inadequacy of currently available scientific in-
formation, there is no compelling reason to believe
42 These models, and the specific visibility changes estimated by these models, are described in summary fashion in the previous
chapter and are discussed in detail in Appendix C.
43 These visibility measures are described in Appendix C.
30
-------
Chapter 5: Physical Effects
Tabic 6. Human Health Effects of Criteria Pollutants.
Pollutant
Quantified Hen It h Effects
llnqiiiintified Health Effects
Other Possible Effects
Ozon e
Respiratory symptoms
Minor restricted activity days
Respiratory restricted activity
days
Hospital admissions
Emergency room visits
Asthma attacks
Changes in pulmonary function
Chronic Sinusitis & Hay Fever
Increased airway
responsiveness to stimuli
Centroacinar fibrosis
Inflammation in the lung
Immunologic changes
Chronic respiratory diseases
Extrapulmonary effects (e.g.,
changes in structure,
function of other organs)
Reduced UV-B exposure
attenuation
Psi rticu kite Mutter/
I SP/ Sulfates
Mortality*
Bronchitis - Chronic and Acute
Hospital admissions
Lower resp irat ory il lnes s
Upper respiratory illness
Chest illness
Respiratory symptoms
Minor restricted activity days
All restricted activity days
Days of work loss
Moderate or worse asthma
status (asthmatics)
Changes in pulmonary function
Chronic respiratory diseases
other than chronic
bronchitis
Inflammation in the lung
(jirbon Monoxide
Hospital Admissions -
congestive heart failure
Decreased time to onset of
angina
Behavioral effects
Other hospital admissions
Other cardiovascular effects
Developmental effects
Nitrogen Oxides
Respiratory illness
Increased airway
responsiveness
Decreased pulmonary function
Inflammation in the lung
Immunological changes
Sulfii i Dioxide
In exercising asthmatics:
Changes in pulmonary function
Respiratory symptoms
Combined responses of
respiratory symptoms and
pulmonary function
changes
Respiratory symptoms in non-
asthmatics
Hospital admissions
Ia;i d
Mortality
Hypertension
Non-fatal coronary heart
disease
Non-fatal strokes
IQ loss effect on lifetime
earnings
IQ loss effects on special
education needs
Health effects for individuals in
age ranges other than those
studied
Neurobehavioral function
Other cardiovascular diseases
Reproductive effects
Fetal effects from maternal
exposure
Delinquent and anti-social
behavior in children
* This analysis estimates excess mortality using PM as an indicator ofthepollutant mix to which individuals were exposed.
31
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic 7. Selected Welfare Effects of Criteria Pollutants.
Pollutii lit
Quantified Welfare Effects
I nqu;intitled Welfsire Effects
O/onc
Changes in crop yields (for 7 crops)
Decreased worker productivity
Changes in other crop yields
Materials damage
Effects on forests
Effects on wildlife
l\irtk ul;iU' Mutter/
TS1V Sulfates
Household soiling
Visibility
Other materials damage
Effects on wildlife
Nitrogen Oxides
Visibility
Crop losses due to acid deposition
Materials damage due to acid deposition
Effects on fisheries due to acidic
deposition
Effects on forests
Sulfur Dioxide
Visibility
Crop losses due to acid deposition
Materials damage due to acid deposition
Effects on fisheries due to acidic
deposition
Effects on forests
that the results of the present analysis are biased in a
particular direction. Some significant uncertainties,
however, may have arisen from interpretation of model
results, underlying data, and supporting scientific stud-
ies. These assumptions and uncertainties are charac-
terized in this report to allow the reader to understand
the degree of uncertainty and the potential for mises-
timation of results. In addition, the overall results are
presented in ranges to reflect the aggregate effect of
uncertainty in key variables. A quantitative assessment
of some of the uncertainties in the present study is
presented in Chapter 7. In addition, the key uncertain-
ties in the physical effects modeling step of this analy-
sis are summarized in Table 12 at the end of this chap-
ter. The remainder of this section discusses each of
the four critical modeling procedures and associated
assumptions.
Mapping Populations to Monitors
The Project Team's method of calculating ben-
efits of air pollution reductions required a correlation
of air quality data changes to exposed populations.
For pollutants with monitor-level data (i.e., S02, 03,
N02, CO), it was assumed that all individuals were
exposed to air quality changes estimated at the near-
est monitor. For PM historical air quality data were
available at the county level. All individuals residing
in a county were assumed to be exposed to that
county's PM10 air quality.44
Many counties did not contain particulate matter
air quality monitors or did not have a sufficient num-
ber of monitor observations to provide reliable esti-
mates of air quality. For those counties, the Project
Team conducted additional analyses to estimate PM10
air quality changes during the study period. For coun-
ties in the eastern 31 states, the grid cell-specific sul-
fate particle concentrations predicted by the RADM
model were used to provide a scaled interpolation
between monitored counties.45 For counties outside
the RADM domain, an alternative method based on
state-wide average concentrations was used. With this
supplemental analysis, estimates were developed of
the health effects of the CAA on almost the entire
continental U.S. population.46 Compliance costs in-
44 In some counties and in the early years of the study period, particulate matter was monitored as TSP rather than as PM10 In these
cases, PM10 was estimated by applying TSP:PM10 ratios derived from historical data. This methodology is described in Appendix C.
45 The specific methodology is described in detail in Appendix C.
46 While this modeling approach captures the vast majority of the U.S. population, it does not model exposure for everyone. To
improve computational efficiency, those grid cells with populations less than 500 were not modeled; thus, the analysis covered
somewhat more than 97 percent of the population.
32
-------
Chapter 5: Physical Effects
curred in Alaska and Hawaii were included in this
study, but the benefits of historical air pollution re-
ductions were not. In addition, the CAA yielded ben-
efits to Mexico and Canada that were not captured in
this study.
Air quality monitors are more likely to be found
in high pollution areas rather than low-pollution ar-
eas. Consequently, mapping population to the nearest
monitor regardless of the distance to that monitor al-
most certainly results in an overstatement of health
impacts due to air quality changes for those popula-
tions. The Project Team conducted a sensitivity analy-
sis to illustrate the importance of the "mapping to near-
est monitor" assumption. For comparison to the base
case, which modeled exposure for the 48 state popu-
lation, Table 8 presents the percentage of the total 48-
state population covered in the "50 km" sensitivity
scenario. For most pollutants in most years, 25 per-
cent or more of the population resided more than 50
km from an air quality monitor (or in a county with-
out PMJ0 monitors). Estimated health benefits are ap-
proximately linear to population covered — that is, if
the population modeled for a pollutant in a given year
in the sensitivity analysis is 25 percent smaller than
the corresponding population modeled in the base case,
then estimated health benefits are reduced by roughly
25 percent in the sensitivity case. This sensitivity
analysis demonstrates that limiting the benefits analy-
sis to reflect only those living within 50 km of a moni-
tor or within a PM-monitored county would lead to a
substantial underestimate of the historical benefits of
the CAA. Since these alternative results may have led
to severely misleading comparisons of the costs and
benefits of the Act, the Project Team decided to adopt
the full 48-state population estimate as the central case
for this analysis despite the greater uncertainties and
potential biases associated with estimating exposures
from distant monitoring sites.
Tabic 8. Percent of Population (of the Continental
US) within 50km of a monitor (or in a County
with PM monitors). 1970-1990.
Pollutant
Year
EM»
Ds
N£b
S£k
m
1975
79%
56%
53%
65%
67%
1980
80%
71%
59%
73%
68%
1985
75%
72%
61%
73%
68%
1990
68%
74%
62%
71%
70%
Choice of Study
The Project Team relied on the most recent avail-
able, published scientific literature to ascertain the
relationship between air pollution and human health
and welfare effects. The choice of studies, and the
uncertainties underlying those studies, also created
uncertainties in the results. For example, to the extent
the published literature may collectively overstate the
effects of pollution, EPA's analysis will overstate the
effects of the CAA. Such outcomes may occur be-
cause scientific research which fails to find signifi-
cant relationships is less likely to be published than
research with positive results. On the other hand, his-
tory has shown that it is highly likely that scientific
understanding of the effects of air pollution will im-
prove in the future, resulting in discovery of previ-
ously unknown effects. Important examples of this
phenomenon are the substantial expected health and
welfare benefits of reductions in lead and ambient
particulate matter, both of which have been shown in
recent studies to impose more severe effects than sci-
entists previously believed. To the extent the present
analysis misses effects of air pollution that have not
yet been subject to adequate scientific inquiry, the
analysis may understate the effects of the CAA.
For some health endpoints, the peer-reviewed sci-
entific literature provides multiple, significantly dif-
fering alternative CR functions. In fact, it is not un-
usual for two equally-reputable studies to differ by a
factor of three or four in implied health impact. The
difference in implied health effects across studies can
be considered an indication of the degree of scientific
uncertainty associated with measurement of that health
effect. Where more than one acceptable study was
available, the Project Team used CR functions from
all relevant studies to infer health effects. That is, the
health effect implied by each study is reported (see
Appendix D), and a range of reported results for a
particular health endpoint can be interpreted as a mea-
sure of the uncertainty of the estimate.
Variance Within Studies
Even where only one CR function was available
for use, the uncertainty associated with application of
that function to estimate physical outcomes can be
evaluated quantitatively. Health effects studies pro-
vided "best estimates" of the relationship between air
quality changes and health effects, and a measure of
the statistical uncertainty of the relationship. In this
analysis, the Project Team used simulation modeling
33
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
techniques to evaluate the overall uncertainty of the
results given uncertainties within individual studies,
across studies examining a given endpoint, and in the
economic valuation coefficients applied to each end-
point. The analysis estimating aggregate quantitative
uncertainty is presented in Chapter 7.
PM-Related Mortality
The most serious human health impact of air pol-
lution is an increase in incidences of premature mor-
tality. In the present study, excess premature mortal-
ity is principally related to increased exposure to lead
(Pb)47 and to particulate matter (PM) and associated
non-Pb criteria pollutants.48 With respect to PM, a
substantial body of published health science literature
recognizes a correlation between elevated PM con-
centrations and increased mortality rates. However,
there is a diversity of opinion among scientific ex-
perts regarding the reasonableness of applying these
studies to derive quantitative estimates of premature
mortality associated with exposure to PM. While 19
of 21 members of the Science Advisory Board Clean
Air Act Scientific Advisory Committee agree that
present evidence warrants concern and implementa-
tion of a fine particle (PM2 ) standard to supplement
the PM10 standard, they also point out that the causal
mechanism has not been clearly established.
For the purposes of the present study, the Project
Team has concluded that the well-established corre-
lation between exposure to elevated PM and prema-
ture mortality is sufficiently compelling to warrant an
assumption of a causal relationship and derivation of
quantitative estimates of a PM-related premature mor-
tality effect. In addition to the assumption of causal-
ity, a number of other factors contribute to uncertainty
in the quantitative estimates of PM-related mortality.49
First, although there is uncertainty regarding the shape
of the CR functions derived from the epidemiological
studies, the present analysis assumes the relationship
to be linear throughout the relevant range of expo-
sures. Second, there is significant variability among
the underlying studies which may reflect, at least in
part, location-specific differences in CR functions.
Transferring CR functions derived from one or more
specific locations to all other locations may contrib-
ute significantly to uncertainty in the effect estimate.
Third, a number of potentially significant biases and
uncertainties specifically associated with each of the
two types of PM-related mortality study further con-
tribute to uncertainty. The remainder of this section
discusses these two groups of studies and their atten-
dant uncertainties and potential biases. (See Appen-
dix D for a more complete discussion of these studies
and their associated uncertainties.)
Short-Term Exposure Studies
Many of the studies examining the relationship
between PM exposure and mortality evaluate changes
in mortality rates several days after a period of el-
evated PM concentrations. In general, significant cor-
relations have been found. These "short-term expo-
sure" or "episodic" studies are unable to address two
important issues: (1) the degree to which the observed
excess mortalities are "premature," and (2) the degree
to which daily mortality rates are correlated with long-
term exposure to elevated PM concentrations (i.e.,
exposures over many years rather than a few days).
Because the episodic mortality studies evaluate
the mortality rate impact only a few days after a high-
pollution event, it is likely that many of the "excess
mortality" cases represented individuals who were
already suffering impaired health, and for whom the
high-pollution event represented an exacerbation of
an already serious condition. Based on the episodic
studies only, however, it is unknown how many of the
victims would have otherwise lived only a few more
days or weeks, or how many would have recovered to
enjoy many years of a healthy life in the absence of
the high-pollution event. For the purpose of cost-ben-
efit analysis, it can be important to determine whether
a pollution event reduces the average lifespan by sev-
eral days or by many years. Although the episodic
mortality studies do not provide an estimate of the
expected life years lost (nor do they address the health
status of victims), some have evaluated the age of the
excess premature mortality cases, and have estimated
that 80 to 85 percent of the victims are age 65 or older.
In addition to causing short-term health problems,
air pollution (measured by elevated annual PM con-
47 Detailed information on methods, sources, and results of the Pb mortality analysis are presented in Appendix G.
48 PM concentrations are highly correlated with concentrations of other criteria pollutants. It is difficult to determine which
pollutant is the causative factor in elevated mortality rates. In this study, the Project Team has used PM as a surrogate for a mix of
criteria pollutants.
49 It should also be noted that some of the morbidity studies, most notably the PM/chronic bronchitis epidemiological studies,
involve many of the same uncertainties.
34
-------
Chapter 5: Physical Effects
centrations) can cause longer-term health problems
that may lead to premature mortality. Such long-term
changes in susceptibility to premature mortality in the
future will be missed by efforts to correlate prema-
ture mortalities with near-term episodes of elevated
pollution concentrations. Consequently, excess pre-
mature mortality estimates based on the results of the
"episodic" mortality studies will underestimate the
effect of long-term elevated pollution concentrations
on mortality rates.
Long-Term Exposure Studies
The other type of PM-related mortality study in-
volves examination of the potential relationship be-
tween long-term exposure to PM and annual mortal-
ity rates. These studies are able to avoid some of the
weaknesses of the episodic studies. In particular, by
investigating changes in annual (rather than daily)
mortality rates, the long-term studies do not predict
most cases of excess premature mortality where mor-
tality is deferred for only a few days; also, the long-
term studies are able to discern changes in mortality
rates due to long-term exposure to elevated air pollu-
tion concentrations. Additionally, the long-term ex-
posure studies are not limited to measuring mortali-
ties that occur within a few days of a high-pollution
event. Consequently, use of the results of the long-
term studies is likely to result in a more complete as-
sessment of the effect of air pollution on mortality
risk.
The long-term exposure studies, however, have
some significant limitations and potential biases. Al-
though studies that are well-executed attempt to con-
trol for those factors that may confound the results of
the study, there is always the possibility of insuffi-
cient or inappropriate adjustment for those factors that
affect long-term mortality rates and may be con-
founded with the factor of interest (e.g., PM concen-
trations). Prospective cohort studies have an advan-
tage over ecologic, or population-based, studies in that
they gather individual-specific information on such
important risk factors as smoking. It is always pos-
sible, however, that a relevant, individual-specific risk
factor may not have been controlled for or that some
factor that is not individual-specific (e.g., climate) was
not adequately controlled for. It is therefore possible
that differences in mortality rates that have been as-
cribed to differences in average PM levels may be due,
in part, to some other factor or factors (e.g., differ-
ences among communities in diet, exercise, ethnicity,
climate, industrial effluents, etc.) that have not been
adequately controlled for.
Another source of uncertainty surrounding the
prospective cohort studies concerns possible histori-
cal trends in PM concentrations and the relevant pe-
riod of exposure, which is as yet unknown. TSP con-
centrations were substantially higher in many loca-
tions for several years prior to the cohort studies and
had declined substantially by the time these studies
were conducted. If this is also true for PM2 5 and PM10,
it is possible that the laiger PM coefficients reported
by the long-term exposure studies (as opposed to the
short-term exposure studies) reflect an upward bias.
If the relevant exposure period extends over a decade
or more, then a coefficient based on PM concentra-
tions at the beginning of the study or in those years
immediately prior to the study could be biased up-
ward if pollution levels had been decreasing mark-
edly for a decade or longer prior to the study.
On the other hand, if a downward trend in PM
concentrations continued throughout the period of the
study, and if a much shorter exposure period is rel-
evant (e.g., contained within the study period itself),
then characterizing PM levels throughout the study
by those levels just prior to the study would tend to
bias the PM coefficient downward. Suppose, for ex-
ample, that PM levels were converging across the dif-
ferent study locations overtime, and in particular, into
the study period. (That is, suppose PM levels were
decreasing overtime, but decreasing faster in the high-
PM locations than in the low-PM locations, so that at
the beginning of the study period the interlocational
differences in PM concentrations were smaller than
they were a decade earlier.) Suppose also that the rel-
evant exposure period is about one year, rather than
many years. The Pope study characterizes the long-
term PM concentration in each of the study locations
by the median PM concentration in the location dur-
ing the five year period 1979-1983. Study subjects
were followed, however, from 1982 through 1989. If
the difference in median PM concentrations across the
50 study locations during the period 1979-1983 was
greater than the difference during the period 1983-
1988, and if it is PM levels during the period 1983-
1988 that most affect premature mortality during the
study period (rather than PM levels during the period
1979-1983), then the study would have attributed
interlocational differences in mortality to larger
interlocational differences in PM concentrations than
were actually relevant. This would result in a down-
ward bias of the PM coefficient estimated in the study.
35
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
The relevant exposure period is one of a cluster
of characteristics of the mortality-PM relationship that
are as yet unknown and potentially important. It is
also unknown whether there is a time lag in the PM
effect. Finally, it is unknown whether there may be
cumulative effects of chronic exposure — that is,
whether the relative risk of mortality actually increases
as the period of exposure increases.
Three recent studies have examined the relation-
ship between mortality and long-term exposure to PM:
Pope et al. (1995), Dockery et al. (1993), and Abbey
et al. (1991). The Pope et al. study is considered a
better choice of long-term exposure study than either
of the other two studies. Pope et al. examined a much
larger population and many more locations than ei-
ther the Dockery study or the Abbey study. The
Dockery study covered only six cities. The Abbey
study covered a cohort of only 6,000 people in Cali-
fornia. In particular, the cohort in the Abbey study
was considered substantially too small and too young
to enable the detection of small increases in mortality
risk. The study was therefore omitted from consider-
ation in this analysis. Even though Pope et al. (1995)
reports a smaller premature mortality response to el-
evated PM than Dockery et al. (1993), the results of
the Pope study are nevertheless consistent with those
of the Dockery study.
Pope et al., (1995) is also unique in that it fol-
lowed a laigely white and middle class population,
decreasing the likelihood that interlocational differ-
ences in premature mortality were attributable to dif-
ferences in socioeconomic status or related factors.
Furthermore, the generally lower mortality rates and
possibly lower exposures to pollution among this
group, in comparison to poorer minority populations,
would tend to bias the PM coefficient from this study
downward, counteracting a possible upward bias as-
sociated with historical air quality trends discussed
above.
Another source of downward bias in the PM co-
efficient in Pope et al., (1995) is that intercity move-
ment of cohort members was not considered. Migra-
tion across study cities would result in exposures of
cohort members being more similar than would be
indicated by assigning city-specific annual average
pollution levels to each member of the cohort. The
more intercity migration there is, the more exposure
will tend toward an intercity mean. If this is ignored,
differences in exposure levels, proxied by differences
in city-specific annual average PM levels, will be ex-
aggerated, resulting in a downward bias of the PM
coefficient. This is because a given difference in mor-
tality rates is being associated with a laiger difference
in PM levels than is actually the case.
An additional source of uncertainty in the Pope et
al., study arises from the PM indicator used in the
study. The Pope et al. study examined the health ef-
fects associated with two indices of PM exposure;
sulfate particles and fine particles (PM25). The PM25
relationship is used in this analysis because it is more
consistent with the air quality data selected for this
analysis (PM|(). Because we use a PM2 mortality re-
lationship, air quality profiles were developed from
the PM10 profiles generated for the entire 20 year pe-
riod. The same regional information about the PM10
components (sulfate, nitrate, oiganic particulate and
primary particulate) used to develop the PM10 profiles
was used to develop regional PM25/PM10 ratios. Al-
though both urban and rural ratios are available, for
computational simplicity, only the regional urban ra-
tios were used to estimate the PM25 profiles from the
PMJ0 profiles used in the analysis. This reflects the
exposure of the majority of the modeled population
(i.e., the urban population), while introducing some
error in the exposure changes for the rural popula-
tion. In the east and west, where the rural ratio is laiger
than the urban ratio, the change in PM2 exposure will
be underestimated for the rural population. In the cen-
tral region the PM25 change will be overestimated.
These ratios were used in each year during 1970-1990,
introducing another source of uncertainty in the analy-
sis.
After considering the relative advantages and dis-
advantages of the various alternative studies available
in the peer-reviewed literature, the Project Team de-
cided that the long-term exposure studies were pref-
erable for the purposes of the present study, primarily
because the long-term exposure studies appear to pro-
vide a more comprehensive estimate of the premature
mortality incidences attributable to PM exposure.
Among the long-term exposure studies, the Pope et
al., (1995) study appears more likely to mitigate a key
source of potential confounding. For these reasons,
the CR function estimated in Pope et al., (1995) is
considered the most reasonable choice for this analy-
sis and is utilized in spite of the several important re-
sidual uncertainties and potential biases which are sub-
sequently reflected in the PM-related mortality effect
estimate.
36
-------
Chapter 5: Physical Effects
Health Effects Modeling Results
This section provides a summary of the differences
in health effects estimated under the control and no-
control scenarios. Because the differences in air qual-
ity between the two scenarios generally increased from
1970 to 1990, and the affected population grew laiger
during that period, the beneficial health effects of the
CAA increased steadily during the 1970 to 1990 pe-
riod. More detailed results are presented in Appendix
D.
Avoided Premature Mortality Estimates
The Project Team determined that, despite their
limitations, the long-term particulate matter exposure
studies provided the superior basis for estimating
mortality effects for the purpose of benefit-cost analy-
sis. Three prospective cohort studies were identified
(Pope et al. (1995), Dockery et al. (1993), and Abbey
et al. (1991)), although the Abbey study was omitted
from consideration because the cohort in that study
was considered insufficient to allow the detection of
small increases in mortality risk. Exposure-response
relationships inferred from the Pope et al. study were
used in the health benefits model to estimate avoided
mortality impacts of the CAA. The Pope et al. study
was selected because it is based on a much larger popu-
lation and a greater number of communities (50) than
is the six-city Dockery et al. Study. The results of the
Pope et al. are consistent with those of the other study,
and are consistent with earlier ecological population
mortality studies. See Appendix D for additional dis-
cussion of the selection of mortality effects studies.
Table 9 presents estimated avoided excess pre-
mature mortalities for 1990 only, with the mean esti-
mate and 90 percent confidence interval. See Appen-
dix D for more detail on results implied by individual
epidemiological studies, and on the temporal pattern
of impacts.50 The model reports a range of results for
each health endpoint. Here, the fifth percentile, mean,
and ninety-fifth percentile estimates are used to char-
acterize the distribution. The total number of avoided
cases of premature mortality due to reduced exposure
to lead (Pb) and particulate matter are presented. Ad-
ditionally, avoided mortality cases are listed by age
cohort of those who have avoided premature mortal-
ity in 1990, along with the expected remaining lifespan
(in years) for the average person in each age cohort.
The average expected remaining lifespan across all
age groups is also indicated. These averages might be
higher if data were available for PM-related mortality
in the under 30 age group and for Pb-related mortality
in the 5-39 age group.
Tabic 9. Criteria Pollutants Health Benefits ~
Distributions of 1990 Avoided Premature Mortalities
(thousands of cases reduced) lor 48 State Population.
* Averages calculated from proportions of premature mortalities by age
group, from Table D-14.
Non-Fatal Health Impacts
The health benefits model reports non-fatal health
effects estimates similarly to estimates of premature
mortalities: as a range of estimates for each quanti-
fied health endpoint, with the range dependent on the
quantified uncertainties in the underlying concentra-
tion-response functions. The range of results for 1990
only is characterized in Table 10 with fifth percentile,
mean, and ninety-fifth percentile estimates. All esti-
mates are expressed as thousands of new cases avoided
in 1990. "Lost IQ Points" representthe aggregate num-
ber of points (in thousands) across the population af-
fected by lead concentrations in 1990. All "Hospital
Admissions" estimates are in thousands of admissions,
regardless of the length of time spent in the hospital.
"Shortness of breath" is expressed as thousands of
Remaining
]Lii&
Annual Casts Avoided
(thousands;)
IPyUhuttaunft
Age group
Kxpeetanev
(yrs)
altn
%Oe
Meam
951 h
I'M,,
30 and over
112
184
257
30-34
48
9
3
5
35-44
38
5
8
ii
45-54
29
7
11
15
55-64
21
14
23
33
65-74
14
26
43
62
75-84
9
"K)
54
76
>84
6
24
41
59
Avg.: 14*
Lead
all ages
7
??
54
infants
75
5
5
5
40-44
38
0
o
13
45-54
29
0
4
20
55-64
21
0
6
18
65-74
14
0
4
15
Avg.: 38*
TOTAL
166
205
252
50 Earlier years are estimated to have had fewer excess premature mortalities.
37
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic 10. Criteria Pollutants Health Benefits — Distributions of 1990 Non-Fatal Avoided
ncidcncc (thousands of eases reduced) for 48 State Population.
Endpoint
Pollutant(s)
Affected
Population
(age group)
Annual Effects Avoided
(thousands)
Unit
5th
%ile
Mean
95 th
%ile
Chronic Bronchitis
PM
all
493
674
886
cases
Lo st IQ Points
Lead
children
7,440
10,400
13,000
points
IQ 70
Lead
children
31
45
60
cases
Hypertension
Lead
men 20-74
9,740
12,600
15,600
cases
Chronic Heart Disease
Lead
40-74
0
22
64
cases
Atherothrombotic brain infarction
Lead
40-74
0
4
15
cases
Initial cerebrovascular accident
Lead
40-47
0
6
19
cases
Hospital Admissions
All Respiratory
PM&Q3
all
75
89
103
cases
COPD + Pneumonia
PM &
03
over 65
52
62
72
cases
Ischem ic Heart Disease
PM
over 65
7
19
31
cases
Congestive Heart Failure
PM & CO
65 and over
28
39
50
cases
Other Respiratory-Related Ailments
Shortness of breath, days
PM
children
14,800
68,800
133,000
days
Acute Bronchitis
PM
children
0
8,700
21,600
cases
Upper & Lower Resp. Symptoms
PM
children
5,400
9,500
13,400
cases
Any of 19 Acute Symptoms
PM &
03
18-65
15,400
130,000
244,000
cases
Asthm a Attacks
PM &
03
asthmatics
170
850
1,520
cases
Increase in Respiratory Illness
NO 2
all
4,840
9,800
14,000
cases
Any Symptom
S02
asthmatics
26
264
706
cases
Restricted Activity and WorkLoss Days
MRAD
PM &
03
18-65
107,000
125,000
143,000
da vs
WorkLoss Days (WLD)
PM
18-65
19,400
22,60 0
25,6 0 0
da vs
The following additional welfare benefits were quantified directly in economicterms: household soiling
dam age, visibility, decreased worker productivity, and agricultural benefits (m easured in terms of net
surplus).
days: that is, one "case" represents one child experi-
encing shortness of breath for one day. Likewise, "Re-
stricted Activity Days" and "Work Loss Days" are
expressed in person-days.
Other Physical Effects
Human health impacts of criteria pollutants domi-
nate quantitative analyses of the effects of the CAA,
in part because the scientific bases for quantifying air
quality and physical effect relationships are most ad-
vanced for health effects. The CAA yielded other ben-
efits, however, which are important even though they
were sometimes difficult or impossible to quantify
fully given currently available scientific and applied
economic information.
Ecological Effects
The CAA yielded important benefits in the form
of healthier ecological resources, including: stream,
river, lake and estuarine ecosystems; forest and wet-
land ecosystems; and agricultural ecosystems. These
benefits are important because of both the intrinsic
value of these ecological resources and the intimate
linkage between human health and the health and vi-
tality of our sustaining ecosystems. Given the com-
plexity of natural and agricultural ecosystems and the
laige spatial and temporal dimensions involved, it has
been difficult or impossible to quantify benefits fully
given currently available scientific and applied eco-
nomic information.
Aquatic and Forest Effects
Beyond the intrinsic value of preserving natural
aquatic (i.e., lakes, streams, rivers, and estuaries), ter-
restrial (i.e., forest and grassland), and wetland eco-
systems and the life they support, protection of eco-
systems from the adverse effects of air pollution can
yield significant benefits to human welfare. The his-
torical reductions in air pollution achieved under the
CAA probably led to significant improvements in the
38
-------
Chapter 5: Physical Effects
health of ecosystems and the myriad ecological ser-
vices they provide. Reductions in acid deposition (SOx
and NOx) and mercury may have reduced adverse ef-
fects on aquatic ecosystems, including finfish, shell-
fish, and amphibian mortality and morbidity, reduced
acidification of poorly buffered systems, and reduced
eutrophication of estuarine systems. Ecological pro-
tection, in turn, can enhance human welfare through
improvements in commercial and recreational fishing,
wildlife viewing, maintenance of biodiversity, im-
provements in drinking water quality, and improve-
ments in visibility.
Wetlands ecosystems are broadly characterized as
transitional areas between terrestrial and aquatic sys-
tems in which the water table is at or near the surface
or the land is periodically covered by shallow water.
Valuable products and services of wetlands include:
flood control, water quality protection and improve-
ment, fish and wildlife habitat, and landscape and bio-
logical diversity. High levels of air pollutants have
the potential to adversely impact wetlands. Reductions
of these pollutants due to compliance with the CAA
have reduced the adverse effects of acidification and
eutrophication of wetlands, which in turn has protected
habitat and drinking water quality.
Forest ecosystems, which cover 33 percent of the
land in the United States, provide an extensive array
of products and services to humans. Products include
lumber, plywood, paper, fiielwood, mulch, wildlife
(game), water (quality), seeds, edible products (e.g.,
nuts, syrup), drugs, and pesticides. Forest services
include recreation, biological and landscape diversity,
amenity functions (e.g., urban forest), reduced runoff
and erosion, increased soil and nutrient conservation,
pollutant sequestration (e.g., C02, heavy metals) and
pollutant detoxification (e.g., oiganochlorines). The
greatest adverse effect on forest systems are imposed
by ozone. No studies have attempted to quantify the
economic benefits associated with all product and ser-
vice functions from any U.S. forest. Some studies have
attempted to estimate the net economic damage from
forest exposure to air pollutants by calculating hypo-
thetical or assumed reductions in growth rates of com-
mercial species. While quantification of forest dam-
ages remains incomplete, available evidence suggests
that recreational, service, and non-use benefits may
be substantial.
For a more comprehensive discussion of the pos-
sible ecological effects of the CAA, see Appendix E.
Quantified Agricultural Effects
Quantification of the effects of the CAA on agri-
culture was limited to the major agronomic crop spe-
cies including barley, corn, soybeans, peanuts, cotton,
wheat, and sorghum. These species account for 70
percent of all cropland in the U.S., and 73 percent of
the nation's agricultural receipts. Ozone is the primary
pollutant affecting agricultural production. Nationwide
crop damages were estimated under the control and
no-control scenarios. Net changes in economic sur-
plus (in 1990 dollars) annually and as a cumulative
present value (discounted at 5%) over the period 1976-
1990 were estimated. Positive surpluses were exhib-
ited in almost all years and were the result of the in-
crease in yields associated with decreased ozone con-
centrations under the control scenario. The present
value (in 1990) of the estimated agricultural benefits
of the CAA ranges from $7.8 billion in the minimum
response case to approximately $37 billion in the
maximum response case51 (note that discounting 1976-
1990 benefits to 1990 amounts to a compounding of
benefits). Exposure-response relationships and culti-
var mix reflect historical patterns and do not account
for possible substitution of more ozone-resistant cul-
tivars in the no-control scenario. Thus, the upper end
of the range of benefit calculations may overestimate
the actual agricultural benefits of the CAA with re-
spect to these crops. Because numerous crops are ex-
cluded from the analysis, including high value crops
that may be sensitive to ozone, the lower end of the
range is not likely to fully capture the agricultural
benefits of reductions in ozone.
Effects of Air Toxics
In addition to control of criteria pollutants, the
Clean Air Act resulted in control of some air toxics
— defined as non-criteria pollutants which can cause
adverse effects to human health and to ecological re-
sources. Control of these pollutants resulted both from
incidental control due to criteria pollutant programs
and specific controls taigeted at air toxics through the
National Emission Standards for Hazardous Air Pol-
lutants (NESHAPs) under Section 112 of the Act.
Air toxics are capable of producing a wide vari-
ety of effects. Table 11 presents the range of potential
human health and ecological effects which can occur
due to air toxics exposure. For several years, the pri-
mary focus of risk assessments and control programs
designed to reduce air toxics has been cancer. Accord-
51 Ranges reflect usage of alternate exposure-response functions.
39
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic 11. Health and Welfare Effects of Hazardous Air Pollutants.
Effect Category
Quantified Effects
Unquantificd Effects
Other Possible Effects
Human Health
Cancer Mortality
- nonutility stationary
source
-mobile source
Cancer Mortality
- utility source
- area source
Noncancer effects
- neurological
- respiratory
- reproductive
- hematopoietic
- developmental
- immunological
- organ toxicity
Human Welfare
Decreased income and
recreation opportunities
due to fish advisories
Odors
Decreased income
resulting from decreased
physical performance
Ecological
Effects on wildlife
Effects on plants
Ecosystem effects
Loss of bio logical
diversity
Effects on global climate
Other Welfare
Visibility
Building Deterioration
Loss ofbiological diversity
ing to present EPA criteria, there are over 100 known
or suspected carcinogens. EPA's 1990 Cancer Risk
study indicated that as many as 1,000 to 3,000 can-
cers annually may be attributable to the air toxics for
which assessments were available (virtually all of this
estimate came from assessments of about a dozen well-
studied pollutants).52
In addition to cancer, these pollutants can cause a
wide variety of health effects, ranging from respira-
tory problems to reproductive and developmental ef-
fects. There has been considerably less work done to
assess the magnitude of non-cancer effects from air
toxics, but one survey study has shown that some pol-
lutants are present in the atmosphere at reference lev-
els that have caused adverse effects in animals.53
Emissions of air toxics can also cause adverse
health effects via non-inhalation exposure routes. Per-
sistent bioaccumulating pollutants, such as mercury
and dioxins, can be deposited into water or soil and
subsequently taken up by living organisms. The pol-
lutants can biomagnify through the food chain and
exist in high concentrations when consumed by hu-
mans in foods such as fish or beef. The resulting ex-
posures can cause adverse effects in humans, and can
also disrupt ecosystems by affecting top food chain
species.
Finally, there are a host of other potential eco-
logical and welfare effects associated with air toxics,
for which very little exists in the way of quantitative
analysis. Toxic effects of these pollutants have the
potential to disrupt both terrestrial and aquatic eco-
systems and contribute to adverse welfare effects such
as fish consumption advisories in the Great Lakes.54
52 U.S. EPA, Cancer Risk from Outdoor Exposure to Air Toxics. EPA-450/l-90-004f. Prepared by EPA/OAR/OAQPS.
53 U.S. EPA, "Toxic Air Pollutants and Noncancer Risks: Screening Studies," External Review Draft, September, 1990.
54 U.S. EPA, Office of Air Quality Planning and Standards. "Deposition of Air Pollutants to the Great Waters, First Report to
Congress," May 1994. EPA-453/R-93-055.
40
-------
Chapter 5: Physical Effects
Unfortunately, the effects of air toxics emissions
reductions could not be quantified for the present
study. Unlike criteria pollutants, there was relatively
little monitoring data available for air toxics, and that
which exists covered only a handful of pollutants.
Emissions inventories were very limited and incon-
sistent, and air quality modeling has only been done
for a few source categories. In addition, the scientific
literature on the effects of air toxics was generally
much weaker than that available for criteria pollut-
ants.
Limitations in the underlying data and analyses
of air toxics led the Project Team to exclude the avail-
able quantitative results from the primary analysis of
CAA costs and benefits. The estimates of cancer inci-
dence benefits of CAA air toxics control which were
developed, but ultimately rejected, are presented in
Appendix H. Also found in Appendix H is a list of
research needs identified by the Project Team which,
if met, would enable at least a partial assessment of
air toxics benefits in future section 812 studies.
Uncertainty In The Physical Effects
Estimates
As discussed above, and in greater detail in Ap-
pendix D, a number of important assumptions and
uncertainties in the physical effects analysis may in-
fluence the estimate of monetary benefits presented
in this study. Several of these key uncertainties, their
potential directional bias, and the potential signifi-
cance of this uncertainty for the overall results of the
analysis are summarized in Table 12.
41
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Table 12. Uncertainties Associated with Physical Effects Modeling.
Potential Source of Krror
Direction of Potentiul Bins
in Plivsicnl KITects Kstiniiilc
Significance Kcl 21 live to Key Iincerhiinlies in
Ovcnill Monetiirv Benefit Ksliniiite
Estimation of PM25 from modeled PM]0
and TSP data (to support mortality
estimation)
Unknown
Significant. Estimated PM25 profiles are used
to calculate most of the premature mortality.
There is significant uncertainty about how the
fine particle share of overall PM levels varies
temporally and spatially throughout the 20 year
period.
Extrapolation of health effects to
populations distantfrommonitors (or
monitored counties in the case of PM).
Probable overestimate.
Probably minor. In addition, this adjustment
avoids the underestimation which would result
by estimating effects for onlythose people
living near monitors. Potential overestimate
may result to the extent air quality in areas
distant from monitors is significantly better than
in monitored areas. This disparity should be
quite minor for regional pollutants, such as
ozone andfine particulates.
Estimation of degree of life-shortening
associated with PM-related premature
mortality.
Unknown.
Unknown, possibly significant when using a
value of life-years approach. Varyingthe
estimate of degree of prematurity has no effect
on the aggregate benefit estimate when a value
of statistical life approach is used since all
incidences of premature mortality are valued
equally. Under the alternative approach based
on valuing individual life-years, the influence
of alternative values for numbers of average
life-years lost may be significant.
Assumption ofzero lagbetween
exposure and incidence of PM-related
premature mortality.
Overestimate.
Probably minor. The short-term mortality
studies indicate that a significant portion of the
premature mortality associated with exposure to
elevated PM concentrations is very short-term
(i.e., a matter of a few days). In addition, the
available epidemiological studies do not
provide evidence of a significant lag between
exposure andincidence. The lag is therefore
likely to be a few years at most and application
of reasonable discount rates over a few years
would not alter the monetized benefit estimate
significantly.
Choice of CR function (i.e., "across-
study" uncertainties)
Unknown.
Significant. The differences in implied physical
outcomes estimated by different underlying
studies are large.
Uncertainty associated withCR
functions derived from each individual
study (i.e., "within study" uncertainty)
Unknown.
Probably minor.
Exclusion of potential UV-B attenuation
benefits associated withhigher
concentrations of tropospheric ozone
undertheno-control case.
Overestimate.
Insignificant.In addition to the incomplete
scientific evidence that there is a UV-B
exposure disbenefit associated specifically with
tropospheric ozone reductions, the potential
contribution toward total ozone column
attenuation from the tropospheric layer is
probably very small.
Exclusion of potential substitution of
ozone-iesistant cultivars in agriculture
analysis.
Overestimate.
Insignificant, given small relative contribution
of quantified agricultural effects to overall
quantified benefit estimate.
Exclusion of other agricultural effects
(crops, pollutants)
Underestimate.
Unknown, possibly significant.
Exclusion of effects on terrestrial,
wetland, and aquatic ecosystems, and
forests.
Underestimate.
Unknown, possibly significant.
No quantification of materials damage
Underestimate
Unknown, possibly significant.
42
-------
6
Economic Valuation
Estimating the reduced incidence of physical ef-
fects represents a valuable measure of health benefits
for individual endpoints; however, to compare or ag-
gregate benefits across endpoints, the benefits must
be monetized. Assigning a monetary value to avoided
incidences of each effect permits a summation, in terms
of dollars, of monetized benefits realized as a result
of the CAA, and allows that summation to be com-
pared to the cost of the CAA.
For the present analysis of health and welfare ben-
efits, valuation estimates were obtained from the eco-
nomic literature, and are reported in dollars per case
reduced for health effects and dollars per unit of
avoided damage for welfare effects.55 Similar to esti-
mates of physical effects provided by health studies,
each of the monetary values of benefits applied in this
analysis is reported in terms of a mean value and a
probability distribution around the mean estimate. The
statistical form of the probability distribution used for
the valuation measures varies by endpoint. For ex-
ample, while the estimate of the dollar value of an
avoided premature mortality is described by the
Weibull distribution, the estimate for the value of a
reduced case of acute bronchitis is assumed to be uni-
formly distributed between a minimum and maximum
value.
Methods for Valuation of Health
and Welfare Effects
In environmental benefit-cost analysis, the dollar
value of an environmental benefit (e.g., a health-re-
lated improvement in environmental quality) conferred
on a person is the dollar amount such that the person
would be indifferent between having the environmen-
tal benefit and having the money. In some cases, this
value is measured by studies which estimate the dol-
lar amount required to compensate a person for new
or additional exposure to an adverse effect. Estimates
derived in this manner are referred to as "willingness-
to-accept" (WTA) estimates. In other cases, the value
of a welfare change is measured by estimating the
amount of money a person is willing to pay to elimi-
nate or reduce a current hazard. This welfare change
concept is referred to as "willingness-to-pay" (WTP).
For small changes in risk, WTP and WTA are virtu-
ally identical, primarily because the budget constraints
normally associated with expressions of WTP are not
significant enough to drive a wedge between the esti-
mates. For larger risk changes, however, the WTP and
WTA values may diverge, with WTP normally being
less than WTA because of the budget constraint ef-
fect. While the underlying economic valuation litera-
ture is based on studies which elicited expressions of
WTP and/or WTA, the remainder of this report refers
to all valuation coefficients as WTP estimates. In some
cases (e.g., stroke-related hospital admissions), nei-
ther WTA nor WTP estimates are available and WTP
is approximated by cost of illness (COI) estimates, a
clear underestimate of the true welfare change since
important value components (e.g., pain and suffering
associated with the stroke) are not reflected in the out-
of-pocket costs for the hospital stay.
For most goods, WTP can be observed by exam-
ining actual market transactions. For example, if a
gallon of bottled drinking water sells for one dollar, it
can be observed that at least some persons are willing
to pay one dollar for such water. For goods that are
not exchanged in the market, such as most environ-
mental "goods," valuation is not so straightforward.
Nevertheless, value may be inferred from observed
behavior, such as through estimation of the WTP for
mortality risk reductions based on observed sales and
prices of safety devices such as smoke detectors. Al-
ternatively, surveys may be used in an attempt to elicit
directly WTP for an environmental improvement.
Wherever possible, this analysis uses estimates
of the mean WTP of the U.S. population to avoid an
environmental effect as the value of avoiding that ef-
fect. In some cases, such estimates are not available,
and the cost of mitigating or avoiding the effect is
used as a rough estimate of the value of avoiding the
effect. For example, if an effect results in hospitaliza-
tion, the avoided medical costs were considered as a
possible estimate of the value of avoiding the effect.
Finally, where even the "avoided cost" estimate is not
available, the analysis relies on other available meth-
ods to provide a rough approximation of WTP. As
noted above, this analysis uses a range of values for
most environmental effects, or endpoints. Table 13
55 The literature reviews and valuation estimate development process is described in detail in Appendix I and in the referenced
supporting reports.
43
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic 13. Health and Welfare Effects Unit Valuation
(1990 dollars).
FiTfit<
Agriculture (Net Surplus)
g3
Estimated Change In
Economic Surplus
* Strokes are comprised of atherothrombotic brain infarctions and cerebrovascular
accidents; both are estimated to have the same monetary value.
** Decreased productivity valued as change in daily wages: $1 per worker per 10%
decrease in 03.
provides a summary of the mean unit value estimates
used in the analysis. The full range of values can be
found in Appendix I.
Mortality
Some forms of air pollution increase the probabil-
ity that individuals will die prematurely. The concen-
tration-response functions for mortality used in this
analysis express this increase in mortality risk as cases
of "excess premature mortality" per time pe-
riod (e.g., per year).
The benefit, however, is the avoidance
of small increases in the risk of mortality. If
individuals' WTP to avoid small increases in
risk is summed over enough individuals, the
value of a statistical premature death avoided
can be inferred.56 For expository purposes,
this valuation is expressed as "dollars per
mortality avoided," or "value of a statistical
life" (VSL), even though the actual valuation
is of small changes in mortality risk.
The mortality risk valuation estimate
used in this study is based on an analysis of
26 policy-relevant value-of-life studies (see
Table 14). Five of the 26 studies are contin-
gent valuation (CV) studies, which directly
solicit WTP information from subjects; the
rest are wage-risk studies, which base WTP
estimates on estimates of the additional com-
pensation demanded in the labor market for
riskier jobs. The Project Team used the best
estimate from each of the 26 studies to con-
struct a distribution of mortality risk valua-
tion estimates for the section 812 study. A
Weibull distribution, with a mean of $4.8 mil-
lion and standard deviation of $3.24 million,
provided the best fit to the 26 estimates. There
is considerable uncertainty associated with
this approach, however, which is discussed
in detail later in this chapter and in Appen-
dix I.
In addition, the Project Team developed
alternative calculations based on a life-years
lost approach. To employ the value of statis-
tical life-year (VSLY) approach, the Project
Team had to first estimate the age distribu-
tion of those lives which would be saved by
reducing air pollution. Based on life expect-
ancy tables, the life-years saved from each statistical
life saved within each age and sex cohort were calcu-
lated. To value these statistical life-years, a concep-
tual model was hypothesized which depicted the rela-
tionship between the value of life and the value of
life-years. As noted earlier in Table 9, the average
number of life-years saved across all age groups
for which data were available are 14 for PM-
related mortality and 38 for Pb-related mortality. The
56 Because people are valuing small decreases in the risk of premature mortality, it is expected deaths that are inferred. For
example, suppose that a given reduction in pollution confers on each exposed individual a decrease in mortal risk of 1/100,000. Then
among 100,000 such individuals, one fewer individual can be expected to die prematurely . If each individual's WTP for that risk
reduction is $50, then the implied value of a statistical premature death avoided is $50 x 100,000 = $5 million.
44
-------
Chapter 6: Economic Valuation
Tabic 14. Summary of Mortality Valuation Estimates
(millions of $ 1990)
average for PM, in particular, differs from the 3 5-year
expected remaining lifespan derived from existing
wage-risk studies.57
Using the same distribution of value of life esti-
mates used above (i.e. the Weibull distribution with a
mean estimate of $4.8 million), a distribution for the
value of a life-year was then estimated and combined
with the total number of estimated life-years lost. The
details of these calculations are presented in Appen-
dix I.
Survey-Based Values
Willingness-to pay for environmental improve-
ment is often elicited through survey methods (such
as the "contingent valuation" method). Use of such
methods in this context is controversial within the
economics profession. In general, economists prefer
to infer WTP from observed behavior. There are times
when such inferences are impossible, however, and
some type of survey technique may be the only means
of eliciting WTP. Economists' beliefs regarding the
reliability of such survey-based data cover a broad
spectrum, from unqualified acceptances of the results
of properly-conducted surveys to outright rejections
of all survey-based valuations.
In this analysis, unit valuations which rely exclu-
sively on the contingent valuation method are chronic
bronchitis, respiratory-related ailments, minor re-
stricted activity days, and visibility. As indicated
above, the value derived for excess premature mortal-
ity stems from 26 studies, of which five use the con-
tingent valuation method. These five studies are within
the range of the remaining 21 labor market studies.
All five report mortality valuations lower than the
central estimate used in this analysis. Excluding the
contingent valuation studies from the mortality valu-
ation estimate would yield a central estimate approxi-
mately ten percent higher than the 4.8 million dollar
value reported above. The endpoints with unit valua-
tions based exclusively on contingent valuation ac-
count for approximately 30 percent of the present value
of total monetized benefits. Most of the CV-based
benefits are attributable to avoided cases of chronic
bronchitis.
Chronic Bronchitis
The best available estimate of WTP to avoid a
case of chronic bronchitis (CB) comes from Viscusi
et al.( 1991). The case of CB described to the respon-
dents in the Viscusi study, however, was described by
the authors as a severe case. The Project Team em-
ployed an estimate of WTP to avoid a pollution-re-
lated case of CB that was based on adjusting the WTP
to avoid a severe case, estimated by Viscusi et al.
(1991), to account for the likelihood that an average
case of pollution-related CB is not as severe as the
case described in the Viscusi study.
The central tendency estimate of WTP to avoid a
pollution-related case of chronic bronchitis (CB) used
in this analysis is the mean of a distribution of WTP
estimates. This distribution incorporates the uncer-
tainty from three sources: (1) the WTP to avoid a case
of severe CB, as described by Viscusi et al., 1991; (2)
the severity level of an average pollution-related case
Slttaadbf
Tvpc of
Waflmmttibmi
(ftm iilllHiflwmHR
1990S)
Kneisner and Leeth (1991) (US)
Labor Market
0.6
Smith and Gilbert (1984)
Labor Market
0.7
Dillingham (1985)
Labor Market
0.9
Butler (1983)
Labor Market
1.1
Miller and Guria (1991)
Cont. Value
1.2
Moore and Viscusi (1988a)
Labor Market
2.5
Viscusi, Magat, aid Huber (1991b)
Cont. Value
2.7
Gegax et al. (1985)
Cont. Value
3.3
Marin and Rsacharopoulos (1982)
Labor Market
2.8
Kneisner and Leeth (1991)
(Australia)
Labor Market
3.3
Gerking, de Ha an, and Schulze
(1988)
Cont. Value
3.4
Cousineau, Lacroix, and Girard
(1988)
Labor Market
3.6
Jones-Lee (1989)
Cont. Value
3.8
Dillingham (1985)
Labor Market
3.9
Viscusi (1978, 1979)
Labor Market
4.1
R.S. Smith (1976)
Labor Market
4.6
V.K. Smith (1976)
Labor Market
4.7
Olson (1981)
Labor Market
5.2
Viscusi (1981)
Labor Market
6.5
R.S. Smith (1974)
Labor Market
7.2
Moore and Viscusi (1988a)
Labor Market
7.3
Kneisner and Leeth (1991) (Japan)
Labor Market
7.6
Herzog and Schlottman (1987)
Labor Market
9.1
Leigh and Folson (1984)
Labor Market
9.7
Leigh (1987)
Labor Market
10.4
Gaten (1988)
Labor Market
13.5
SOURCE: Viscusi, 1992
57 See, for example, Moore and Viscusi (1988) or Viscusi (1992).
45
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
of CB (relative to that of the case described by Viscusi
et al.(1991); and (3) the elasticity of WTP with re-
spect to severity of the illness. Based on assumptions
about the distributions of each of these three uncer-
tain components, a distribution of WTP to avoid a
pollution-related case of CB was derived by Monte
Carlo methods. The mean of this distribution, which
was about $260,000, is taken as the central tendency
estimate of WTP to avoid a pollution-related case of
CB. The three underlying distributions, and the gen-
eration of the resulting distribution of WTP, are de-
scribed in Appendix I.
Respiratory-Related Ailments
In general, the valuations assigned to the respira-
tory-related ailments listed in Table 14 represent a
combination of willingness to pay estimates for indi-
vidual symptoms which comprise each ailment. For
example, a willingness to pay estimate to avoid the
combination of specific upper respiratory symptoms
defined in the concentration-response relationship
measured by Pope et al. (1991) is not available. How-
ever, while that study defined upper respiratory symp-
toms as one suite of ailments (runny or stuffy nose;
wet cough; and burning, aching, or red eyes), the valu-
ation literature reported individual WTP estimates for
three closely matching symptoms (head/sinus conges-
tion, cough, and eye irritation). The available WTP
estimates were therefore used as a surrogate to the
values for the precise symptoms defined in the con-
centration-response study.
To capture the uncertainty associated with the
valuation of respiratory-related ailments, this analy-
sis incorporated a range of values reflecting the fact
that an ailment, as defined in the concentration-re-
sponse relationship, could be comprised of just one
symptom or several. At the high end of the range, the
valuation represents an aggregate of WTP estimates
for several individual symptoms. The low end repre-
sents the value of avoiding a single mild symptom.
Minor Restricted Activity Days
An individual suffering from a single severe or a
combination of pollution-related symptoms may ex-
perience a Minor Restricted Activity Day (MRAD).
Krupnick and Kopp (1988) aigue that mild symptoms
will not be sufficient to result in a MRAD, so that
WTP to avoid a MRAD should exceed WTP to avoid
any single mild symptom. On the other hand, WTP to
avoid a MRAD should not exceed the WTP to avoid a
work loss day (which results when the individual ex-
periences more severe symptoms). No studies are re-
ported to have estimated WTP to avoid a day of mi-
nor restricted activity. Instead, this analysis uses an
estimate derived from WTP estimates for avoiding
combinations of symptoms which may result in a day
of minor restricted activity ($38 per day). The uncer-
tainty range associated with this value extends from
the highest value for a single symptom to the value
for a work loss day. Furthermore, the distribution ac-
knowledges that the actual value is likely to be closer
to the central estimate than either extreme.
Visibility
The value of avoided visibility impairment was
derived from existing contingent valuation studies of
the household WTP to improve visibility, as reported
in the economics literature. These studies were used
to define a single, consistent basis for the valuation of
visibility benefits nationwide. The central tendency
of the benefits estimate is based on an annual WTP of
$14 per household per unit improvement in the
DeciView index, with upper and lower bounds of $21
and $8, respectively, on the uncertainty range of the
estimate.
Avoided Cost Estimates
For some health effects, WTP estimates are not
available, and the Project Team instead used "costs
avoided" as a substitute for WTP. Avoided costs were
used to value the following endpoints: hypertension,
hospital admissions, and household soiling.
Hypertension and Hospital Admissions
Avoided medical costs and the avoided cost of lost
work time were used to value hypertension (high blood
pressure) and hospital admissions (this includes hos-
pital admissions for respiratory ailments as well as
heart disease, heart attacks, and strokes) .
For those hospital admissions which were speci-
fied to be the initial hospital admission (in particular,
hospital admissions for coronary heart disease (CHD)
events and stroke), avoided cost estimates should con-
sist of the present discounted value of the stream of
medical expenditures related to the illness, as well as
the present discounted value of the stream of lost earn-
ings related to the illness. While an estimate of present
discounted value of both medical expenditures and
lost earnings was available for stroke ($200,000 for
46
-------
Chapter 6: Economic Valuation
males and $150,000 for females), the best available
estimate for CHD ($52,000) did not include lost earn-
ings. Although no published estimates of the value of
lost earnings due to CHD events are available, one
unpublished study suggests that this value could be
substantial, possibly exceeding the value of medical
expenditures. The estimate of $52,000 for CHD may
therefore be a substantial underestimate. The deriva-
tions of the avoided cost estimates for CHD and stroke
are discussed in Appendix G.
In those cases for which it is unspecified whether
the hospital admission is the initial one or not (that is,
for all hospital admissions endpoints other than CHD
and stroke), it is unclear what portion of medical ex-
penditures and lost earnings after hospital discharge
can reasonably be attributed to pollution exposure and
what portion might have resulted from an individual's
pre-existing condition even in the absence of a par-
ticular pollution-related hospital admission. In such
cases, the estimates of avoided cost include only those
costs associated with the hospital stay, including the
hospital charge, the associated physician charge, and
the lost earnings while in the hospital ($6,100 to
$10,300, depending on the ailment for which hospi-
talization is required).
The estimate of avoided cost for hypertension in-
cluded physician chaiges, medication costs, and hos-
pitalization costs, as well as the cost of lost work time,
valued at the rate estimated for a work loss day (see
discussion below). Based on this approach, the value
per year of avoiding a case of hypertension is taken to
equal the sum of medical costs per year plus work
loss costs per year; the resulting value is $680 per case
per year.
Presumably, willingness-to-pay to avoid the ef-
fects (and treatment) of hypertension would reflect
the value of avoiding any associated pain and suffer-
ing, and the value placed on dietary changes, etc. Like-
wise, the value of avoiding a health effect that would
require hospitalization or doctor's care would include
the value of avoiding the pain and suffering caused
by the health effect as well as lost leisure time, in ad-
dition to medical costs and lost work time. Conse-
quently, the valuations for these endpoints used in this
analysis likely represent lower-bound estimates of the
true social values for avoiding such health effects.
Household Soiling
This analysis values benefits for this welfare ef-
fect by considering the avoided costs of cleaning
houses due to particulate matter soiling. The Project
Team's estimate reflects the average household's an-
nual cost of cleaning per jig/m3 particulate matter
($2.50). Considered in this valuation are issues such
as the nature of the particulate matter, and the propor-
tion of households likely to do the cleaning themselves.
Since the avoided costs of cleaning used herein do
not reflect the loss of leisure time (and perhaps work
time) incurred by those who do their own cleaning,
the valuation function likely underestimates true WTP
to avoid additional soiling.
Other Valuation Estimates
Changes in Children's IQ
One of the major effects of lead exposure is per-
manently impaired cognitive development in children.
No ready estimates of society's WTP for improved
cognitive ability are currently available. Two effects
of IQ decrements can be monetized, however: reduc-
tions in expected lifetime income, and increases in
societal expenditures for compensatory education.
These two effects almost certainly understate the WTP
to avoid impaired cognitive development in children,
and probably should be considered lower bound esti-
mates. In the absence of better estimates, however,
the Project Team has assumed that the two monetized
effects represent a useful approximation of WTP.
The effect of IQ on expected lifetime income com-
prises a direct and an indirect effect. The direct effect
is drawn from studies that estimate, all else being
equal, the effect of IQ on income. The indirect effect
occurs as a result of the influence of IQ on educa-
tional attainment: higher IQ leads to more years of
education, and more education leads in turn to higher
expected future income. However, this indirect ben-
efit is mitigated, but not eliminated, by the added costs
of the additional education and by the potential earn-
ings foigone by the student while enrolled in school.58
Combining the direct and indirect influences, the net
effect of higher IQ on expected lifetime income (dis-
58 Theoretically, the indirect effect should be small relative to the direct effect of IQ on future earnings. The empirical research
used to derive values for this analysis, however, implies that the indirect effect is roughly equal in magnitude to the direct effect. One
can infer from this information that there is a market distortion of some sort present (such as imperfect knowledge of the returns to
education), or, perhaps, that individuals make their education "investments" for purposes other than (or in addition to) "maximizing
lifetime income." See Appendix G for further discussion of this issue.
47
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
counted to the present at five percent) is estimated to
be $3,000 per additional IQ point.
In this analysis, it is assumed that part-time com-
pensatory education is required for all children with
IQ less than 70. The Project Team assumed that the
WTP to avoid cases of children with IQ less than 70
can be approximated by the cost ($42,000 per child)
of part-time special education in regular classrooms
from grades one through twelve (as opposed to inde-
pendent special education programs), discounted to
the present at five percent. See Appendix G for more
detail on valuation methods and data sources for IQ
effects and other lead-related health impacts.
Work Loss Days and Worker
Productivity
For this analysis, it was assumed that the median
daily 1990 wage income of 83 dollars was a reason-
able approximation of WTP to avoid a day of lost
work. Although a work loss day may or may not af-
fect the income of the worker, depending on the terms
of employment, it does affect economic output and is
thus a cost to society. Conversely, avoiding the work
loss day is a benefit.
A decline in worker productivity has been mea-
sured in outdoor workers exposed to ozone. Reduced
productivity is measured in terms of the reduction in
daily income of the average worker engaged in strenu-
ous outdoor labor, estimated at $1 per 10 percent in-
crease in ozone concentration.
Agricultural Benefits
Similar to the other welfare effects, the agricul-
tural benefits analysis estimated benefits in dollars per
unit of avoided damage, based on estimated changes
in crop yields predicted by an agricultural sector
model. This model incorporated agricultural price,
farm policy, and other data for each year. Based on
expected yields, the model estimated the production
levels for each crop, and the economic benefits to con-
sumers, and to producers, associated with these pro-
duction levels. To the extent that alternative exposure-
response relationships were available, a range of po-
tential benefits was calculated (see Appendix F).
Valuation Uncertainties
The Project Team attempted to handle most valu-
ation uncertainties explicitly and quantitatively by
expressing values as distributions (see Appendix I for
a complete description of distributions employed),
using a Monte-Carlo simulation technique to apply
the valuations to physical effects (see Chapter 7) with
the mean of each valuation distribution equal to the
"best estimate" valuation. This approach does not, of
course, guarantee that all uncertainties have been ad-
equately characterized, nor that the valuation estimates
are unbiased. It is possible that the actual WTP to avoid
an air pollution-related impact is outside of the range
of estimates used in this analysis. Nevertheless, the
Project Team believes that the distributions employed
are reasonable approximations of the ranges of uncer-
tainty, and that there is no compelling reason to be-
lieve that the mean values employed are systemati-
cally biased (except for the IQ-related and avoided
cost-based values, both of which probably underesti-
mate WTP).
One particularly important area of uncertainty is
valuation of mortality risk reduction. As noted in Chap-
ter 7, changes in mortality risk are a very important
component of aggregate benefits, and mortality risk
valuation is an extremely laige component of the quan-
tified uncertainty. Consequently, any uncertainty con-
cerning mortality risk valuation beyond that addressed
by the quantitative uncertainty assessment (i.e., that
related to the Weibull distribution with a mean value
of $4.8 million) deserves note. One issue merits spe-
cial attention: uncertainties and possible biases related
to the "benefits transfer" from the 26 valuation source
studies to valuation of reductions in PM-related mor-
tality rates.
Mortality Risk Benefits Transfer
Although each of the mortality risk valuation
source studies (see Table 14) estimated the average
WTP for a given reduction in mortality risk, the de-
gree of reduction in risk being valued varied across
studies and is not necessarily the same as the degree
of mortality risk reduction estimated in this analysis.
The transferability of estimates of the value of a sta-
tistical life from the 26 studies to the section 812 ben-
efit analysis rests on the assumption that, within a rea-
sonable range, WTP for reductions in mortality risk
is linear in risk reduction. For example, suppose a study
48
-------
Chapter 6: Economic Valuation
estimates that the average WTP for a reduction in
mortality risk of 1/100,000 is 50 dollars, but that the
actual mortality risk reduction resulting from a given
pollutant reduction is 1/10,000. IfWTP for reductions
in mortality risk is linear in risk reduction, then a WTP
of 50 dollars for a reduction of 1/100,000 implies a
WTP of 500 dollars for a risk reduction of 1/10,000
(which is ten times the risk reduction valued in the
study). Underthe assumption of linearity, the estimate
of the value of a statistical life does not depend on the
particular amount of risk reduction being valued.
Although the particular amount of mortality risk
reduction being valued in a study may not affect the
transferability of the WTP estimate from the study to
the benefit analysis, the characteristics of the study
subjects and the nature of the mortality risk being val-
ued in the study could be important. Certain charac-
teristics of both the population affected and the mor-
tality risk facing that population are believed to affect
the average WTP to reduce risk. The appropriateness
of the mean of the WTP estimates from the 26 studies
for valuing the mortality-related benefits of reductions
in pollutant concentrations therefore depends not only
on the quality of the studies (i.e., how well they mea-
sure what they are trying to measure), but also on (1)
the extent to which the subjects in the studies are simi-
lar to the population affected by changes in air pollu-
tion and (2) the extent to which the risks being valued
are similar.
The substantial majority of the 26 studies relied
upon are wage-risk (or labor market) studies. Com-
pared with the subjects in these wage-risk studies, the
population most affected by air pollution-related mor-
tality risk changes is likely to be, on average, older
and probably more risk averse. Some evidence sug-
gests that approximately 85 percent of those identi-
fied in short-term ("episodic") studies who die pre-
maturely from PM-related causes are over 65,59 The
average age of subjects in wage-risk studies, in con-
trast, would be well under 65.
The direction of bias resulting from the age dif-
ference is unclear. It could be argued that, because an
older person has fewer expected years left to lose, his
or her WTP to reduce mortality risk would be less
than that of a younger person. This hypothesis is sup-
ported by one empirical study, Jones-Lee et al. (1985),
which found WTP to avoid mortality risk at age 65 to
be about 90 percent of what it is at age 40. On the
other hand, there is reason to believe that those over
65 are, in general, more risk averse than the general
population, while workers in wage-risk studies are
likely to be less risk averse than the general popula-
tion. Although the list of 26 studies used here excludes
studies that consider only much-higher-than-average
occupational risks, there is nevertheless likely to be
some selection bias in the remaining studies—that is,
these studies are likely to be based on samples of
workers who are, on average, more risk-loving than
the general population. In contrast, older people as a
group exhibit more risk-averse behavior.
There is substantial evidence that the income elas-
ticity of WTP for health risk reductions is positive
(although there is uncertainty about the exact value of
this elasticity). Individuals with higher incomes (or
greater wealth) should, then, be willing to pay more
to reduce risk, all else equal, than individuals with
lower incomes or wealth. The comparison between
the (actual and potential) income or wealth of the
workers in the wage-risk studies versus that of the
population of individuals most likely to be affected
by changes in pollution concentrations, however, is
unclear. One could aigue that because the elderly are
relatively wealthy, the affected population is also
wealthier, on average, than are the wage-risk study
subjects, who tend to be middle-aged (on average)
blue-collar workers. On the other hand, the workers
in the wage-risk studies will have potentially more
years remaining in which to acquire streams of in-
come from future earnings. In addition, it is possible
that among the elderly it is laigely the poor elderly
who are most vulnerable to air pollution-related mor-
tality risk (e.g., because of generally poorer health
care). On net, the potential income comparison is un-
clear.
Although there may be several ways in which job-
related mortality risks differ from air pollution-related
mortality risks, the most important difference may be
that job-related risks are incurred voluntarily whereas
air pollution-related risks are incurred involuntarily.
There is some evidence60 that people will pay more to
reduce involuntarily incurred risks than risks incurred
voluntarily. If this is the case, WTP estimates based
on wage-risk studies may be downward biased esti-
mates of WTP to reduce involuntarily incurred air
pollution-related mortality risks.
59 See Schwartz and Dockery (1992), Ostro et al. (1995), and Chestnut (1995).
60See, for example, Violette and Chestnut, 1983.
49
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Finally, another important difference related to the
nature of the risk may be that some workplace mortal-
ity risks tend to involve sudden, catastrophic events,
whereas air pollution-related risks tend to involve
longer periods of disease and suffering prior to death.
Some evidence suggests that WTP to avoid a risk of a
protracted death involving prolonged suffering and
loss of dignity and personal control is greater than the
WTP to avoid a risk (of identical magnitude) of sud-
den death. To the extent that the mortality risks ad-
dressed in this assessment are associated with longer
periods of illness or greater pain and suffering than
are the risks addressed in the valuation literature, the
WTP measurements employed in the present analysis
would reflect a downward bias.
The potential sources of bias introduced by rely-
ing on wage-risk studies to derive an estimate of the
WTP to reduce air pollution-related mortality risk are
summarized in Table 15. Among these potential bi-
ases, it is disparities in age and income between the
subjects of the wage-risk studies and those affected
by air pollution which have thus far motivated spe-
cific suggestions for quantitative adjustment61; how-
ever, the appropriateness and the proper magnitude of
such potential adjustments remain unclear given pres-
ently available information. These uncertainties are
particularly acute given the possibility that age and
income biases might offset each other in the case of
pollution-related mortality risk aversion. Furthermore,
the other potential biases discussed above, and sum-
marized in Table 16, add additional uncertainty re-
garding the transferability of WTP estimates from
wage-risk studies to environmental policy and pro-
gram assessments.
Tabic 15. Estimating Mortality Risk Based on Wage-
Risk Studies: Potential Sources and Likely Direction of
Bias.
Factor
Likdv Direction ofBias in W1 P
Estimate
Age
Uncertain, perhaps upward
Degree of Risk Aversion
Downward
Income
Uncertain
Voluntary vs.
Involuntary
Downward
Catastrophic vs.
Protracted Death
Uncertain, perhaps downward
61 Chestnut, 1995; IEc, 1992.
50
-------
7
Results and Uncertainty
This chapter presents a summary of the monetized
benefits of the CAA from 1970 to 1990, compares
these with the corresponding costs, explores some of
the major sources of uncertainty in the benefits esti-
mates, and presents alternative results reflecting di-
verging viewpoints on two key variables: PM-related
mortality valuation and the discount rate.
Monetized economic benefits for the 1970 to 1990
period were derived by applying the unit valuations
discussed in Chapter 6 to the stream of physical ef-
fects estimated by the method documented in Chapter
5. The range of estimates for monetized benefits is
based on the quantified uncertainty associated with
the health and welfare effects estimates and the quan-
tified uncertainty associated with the unit valuations
applied to them. Quantitative estimates of uncertain-
ties in earlier steps of the analysis (i.e., estimation of
compliance costs,62 emissions changes, and air qual-
ity changes) could not be adequately developed and
are therefore not applied in the present study. As a
result, the range of estimates for monetized benefits
presented in this chapter is narrower than would be
expected with a complete accounting of the uncertain-
ties in all analytical components. However, the uncer-
tainties in the estimates of physical effects and unit
values are considered to be large relative to these ear-
lier components. The characterization of the uncer-
tainty surrounding unit valuations is discussed in de-
tail in Appendix I. The characterization of the uncer-
tainty surrounding health and welfare effects estimates,
as well as the characterization of overall uncertainty
surrounding monetized benefits, is discussed below.
Quantified Uncertainty in the
Benefits Analysis
Alternative studies published in the scientific lit-
erature which examine the health or welfare conse-
quences of exposure to a given pollutant often obtain
different estimates of the concentration-response (CR)
relationship between the pollutant and the effect. In
some instances the differences among CR functions
estimated by, or derived from, the various studies are
substantial. In addition to sampling error, these dif-
ferences may reflect actual variability of the concen-
tration-response relationship across locations. Instead
of a single CR coefficient characterizing the relation-
ship between an endpoint and a pollutant in the CR
function, there could be a distribution of CR coeffi-
cients which reflect geographic differences.63 Because
it is not feasible to estimate the CR coefficient for a
given endpoint-pollutant combination in each county
in the nation, however, the national benefits analysis
applies the mean of the distribution of CR coefficients
to each county. This mean is estimated based on the
estimates of CR coefficients reported in the available
studies and the information about the uncertainty of
these estimates, also reported in the studies.
Based on the assumption that for each endpoint-
pollutant combination there is a distribution of CR
coefficients, the Project team used a Monte Carlo ap-
proach to estimate the mean of each distribution and
to characterize the uncertainty surrounding each esti-
mate. For most health and welfare effects, only a single
study is considered. In this case, the best estimate of
the mean of the distribution of CR coefficients is the
reported estimate in the study. The uncertainty sur-
rounding the estimate of the mean CR coefficient is
62 Although compliance cost estimation is primarily of concern to the cost side of this analysis, uncertainty in the estimates for
compliance costs does influence the uncertainty in the benefit estimates because compliance cost changes were used to estimate
changes in macroeconomic conditions which, in turn, influenced the estimated changes in emissions, air quality, and physical effects.
63 Geographic variability may result from differences in lifestyle (e.g., time spent indoors vs outdoors), deposition rates, or other
localized factors which influence exposure of the population to a given atmospheric concentration of the pollutant.
51
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
best characterized by the standard error of the reported
estimate. This yields a normal distribution, centered
at the reported estimate of the mean. If two or more
studies are considered for a given endpoint-pollutant
combination, a normal distribution is derived for each
study, centered at the mean estimate reported in the
study. On each iteration of a Monte Carlo procedure,
a CR coefficient is randomly selected from each of
the normal distributions, and the selected values are
averaged. This yields an estimate of the mean CR co-
efficient for that endpoint-pollutant combination. It-
erating this procedure many times results in a distri-
bution of estimates of the mean CR coefficient.
Each estimate randomly selected from this distri-
bution was evaluated for each county in the nation,
and the results were aggregated into an estimate of
the national incidence of the health or welfare effect.
Through repeated sampling from the distribution of
mean CR coefficients, a distribution of the estimated
change in effect outcomes due to the change in air
quality between the control and no-control scenarios
was generated.
Once a distribution of estimated outcomes was
generated for each health and welfare effect, Monte
Carlo methods were used again to characterize the
overall uncertainty surrounding monetized benefits.
For each health and welfare effect in a set of non-
overlapping effects, an estimated incidence was ran-
domly selected from the distribution of estimated in-
cidences for that endpoint, and a unit value was ran-
domly selected from the corresponding distribution
of unit values, on each iteration of the Monte Carlo
procedure. The estimated monetized benefit for that
endpoint produced on that iteration is the product of
these two factors. Repeating the process many times
generated a distribution of estimated monetized ben-
efits by endpoint. Combining the results for the indi-
vidual endpoints using the Monte Carlo procedure
yielded a distribution of total estimated monetized
benefits for each taiget year (1975, 1980, 1985 and
1990). This technique enabled a representation of
uncertainty in current scientific and economic opin-
ion in these benefits estimates.
Aggregate Monetized Benefits
For each of the taiget years of the analysis, the
monetized benefits associated with the different health
and welfare effects for that year must be aggregated.
These aggregate benefits by taiget year must then be
aggregated across the entire 1970 to 1990 period of
the study to yield a present discounted value of aggre-
gate benefits for the period. The issues involved in
each stage of aggregation, as well as the results of
aggregation, are presented in this section. (The de-
tailed results for the taiget years are presented in Ap-
pendix I.)
Present Value
Endpoint
Pollutant(s)
5th %ilc
Mean
95 th %ile
Mortality
PM
$2,369
$16,632
$40,597
Mortality
Pb
$121
$1,339
$3,910
Chronic Bronchitis
PM
$409
$3,313
$10,401
IQ (LostlQ Pts. + Children w/IQ<70)
Pb
$271
$399
$551
Hypertension
Pb
$77
$98
$120
Ho spital Admissions
PM,03,Pb,& CO
$27
$57
$120
Respiratory-Related Symptoms, Restricted PM, 03, N02, & S02
$123
$182
$261
Activity, & Decreased Productivity
Soiling Damage
PM
$6
$74
$192
Visibility
particulates
$38
$54
$71
Agriculture (Net Surplus)
03
$11
$23
$35
52
Tabic 16. Present Value of 1970 to 1990 Monetized Benefits by Endpoint Category lor 48 State
Population (billions of $1990. discounted to 1990 at 5 percent).
-------
Chapter 7: Results and Uncertainty
Table 16 presents monetized benefits for each
quantified and monetized health and welfare endpoint
(orgroupofendpoints), aggregated from 1970to 1990.
The mean estimate resulting from the Monte Carlo
simulation is presented, along with the measured cred-
ible range (upper and lower fifth percentiles of the
distribution). Aggregating the stream of monetized
benefits across years involved compounding the stream
of monetized benefits estimated for each year to the
1990 present value (using a five percent discount rate).
Since the present value estimates combine streams
of benefits from 1970 to 1990, the calculation required
monetized estimates for each year. However, Monte
Carlo modeling was carried out only for the four tar-
get years (1975, 1980, 1985 and 1990). In the inter-
vening years, only a central estimate of benefits was
estimated for each health and welfare endpoint (by
multiplying the central incidence estimate for the given
year by the central estimate of the unit valuation). The
resulting annual benefit estimates provided a tempo-
ral trend of monetized benefits across the period re-
sulting from the annual changes in air quality. They
Table 16 offers a comparison of benefits by health
or welfare endpoint. The effect categories listed in
the table are mutually exclusive, allowing the mon-
etized benefits associated with them to be added. It
should be noted, however, that the listed categories
combine estimates that are not mutually exclusive. To
avoid double counting, care was taken to treat the ben-
efits associated with overlapping effects as alterna-
tive estimates. For example, the "Hospital Admis-
sions" category includes admissions for specific ail-
ments (Pneumonia and COPD) as well as the broader
classification of "all respiratory" ailments. Clearly,
benefits accruing from the first two represent a subset
of the last and adding all three together would result
in an overestimate of total monetized benefits. To avoid
this, the sum of benefits from Pneumonia and COPD
was treated as an alternative to the benefits estimated
for all respiratory ailments (the sum of the first two
was averaged with the third). This issue of double-
counting also arose for two other cases of overlap-
ping health effects, both of which have been combined
into the "Respiratory-Related Symptoms, Restricted
Activity, & Decreased Productivity" category in Table
Tabic 17. Total Monetized Benefits for 48 State Population (Present Value in billions of 1990$.
discounted to 1990 at 5 percent).
Present Value
5th %ile
Mean
95th %ile
TOTAL (Billions of 1990-value dollars)
$5,600
$22,200
$49,400
did not, however, characterize the uncertainty associ-
ated with the yearly estimates for intervening years.
In an attempt to capture uncertainty associated with
these estimates, the Project Team relied on the ratios
of the 5th percentile to the mean and the 95th percen-
tile to the mean in the target years. In general, these
ratios were fairly constant across the target years, for
a given endpoint. The ratios were interpolated between
the target years, yielding ratios for the intervening
years. Multiplying the ratios for each intervening year
by the central estimate generated for that year pro-
vided estimates of the 5th and 95th percentiles, which
were used to characterize uncertainty about the cen-
tral estimate. Thus, the present value of the stream of
benefits, including the credible range estimates, could
be computed.
16. First, acute bronchitis was treated as an alterna-
tive (i.e., averaged with) the combination of upper and
lower respiratory symptoms, since their definitions of
symptoms overlap. Second, various estimates of re-
stricted activity, with different degrees of severity,
were combined into a single benefit category.
Table 17 reports the estimated total national mon-
etized benefits attributed in this analysis to the CAA
from 1970 to 1990. The benefits, valued in 1990 dol-
lars, range from $5.6 to $49.4 trillion with a central
estimate of $22.2 trillion. The Monte Carlo technique
was used to aggregate monetized benefits across end-
points. For each of several thousand iterations, a ran-
dom draw of the monetized benefits for each endpoint
was selected from the distributions summarized in
53
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Table 16 and the individual endpoint estimates were
then summed. This resulted in the distribution of total
national monetized benefits reported above.64
The temporal pattern of benefits during the 1970
to 1990 period is related to the difference in emis-
sions between the control and no-control scenarios and
is magnified by population growth during that period.
As illustrated by Figure 18, quantified annual ben-
efits increased steadily during the study period, with
the greatest increases occurring during the late 1970s.
The mean estimate of quantified annual benefits grew
from 355 billion dollars in 1975 (expressed as infla-
tion-adjusted 1990 dollars) to 930 billion dollars in
1980, 1,155 billion dollars in 1985, and 1,248 billion
dollars in 1990.
Figure 19 depicts the distribution of monetized
benefits for 1990 (similar distributions were gener-
ated for other years in the analysis period). The solid
vertical bars in the figure represent the relative fre-
quency of a given result in the 1990 Monte Carlo
analysis. The largest bar, located above the "<$ 1,000",
indicates that more Monte Carlo iterations generated
monetized benefits of $900 billion to $ 1 trillion than
in any other $100 billion range bin, making this the
modal bin. The expected value of the estimate for to-
tal monetized benefit for 1990 (i.e., the mean of the
distribution) is $ 1.25 trillion. The ninety percent con-
fidence interval, a summary description of the spread
of a distribution, is also noted in the figure.
Figure 18. Monte Carlo Simulation Model Results for
Taiget Years (in billions of 1990 dollars).
$3,000
§ $2,500--
H $2,000
is $1,500
&
$1,000"
$500-
$0
EE
Mean
5th%
^ Mean
^ 5th%
^ Mean
^ 5th%
^ 5th%
1975
1980
1985
1990
Figure 19. Distribution of 1990 Monetized Benefits of
CAA (in billions of 1990 dollars).
Jl
Distribution Summary (SBillions)
5th percentile = $329
mean = $1,250
95th percentile = $2,760
CN CN
CN CN
f 5th percentile
95th percentile^
Total Monetary Benefits ($ Billions)
On initial inspection, the estimated $ 1.25 trillion
value for monetized benefits in 1990 may seem im-
plausibly laige, even though 1990 is the year in which
the differences between outcomes under the control
and no-control scenarios are at their most extreme.
The plausibility of this estimate may seem particu-
larly questionable to some if one considers that the
$1.25 trillion value for 1990 is over five percent of
the estimated $22.8 trillion value for total 1990 assets
of households and nonprofit oiganizations. Consid-
ered from this perspective, $ 1.25 trillion may seem to
represent a large share of total wealth, and some might
question whether Americans would really be willing
to pay this much money for the reductions in risk
achieved by the Clean Air Act and related programs,
even if the risk in question involves premature death.
However, in the end it is clear that such comparisons
are overly simplistic and uninformative because they
ignore the magnitude and nature of the welfare change
being measured.
First, with respect to the magnitude of the differ-
ence in estimated social welfare under the two sce-
narios, it is important to recognize how severe air qual-
ity conditions and health risks would be under the
hypothetical no-control scenario. Focusing on ambi-
ent particulate matter, the pollutant responsible for the
vast majority of the estimated monetary benefits, a
comparison of the estimated annual mean concentra-
tions of total suspended particulates (TSP) projected
in the U.S. under the no-control scenario with esti-
64 Comparing Tables 16 and 17, it can be seen that the sum of benefits across endpoints at a given percentile level does not result
in the total monetized benefits estimate at the same percentile level in Table 17. For example, if the fifth percentile benefits of the
endpoints shown in Table 16 were added, the resulting total would be substantially less than $5.6 trillion, the fifth percentile value of
the distribution of aggregate monetized benefits reported in Table 17. This is because the various health and welfare effects are treated
as stochastically independent, so that the probability that the aggregate monetized benefit is less than or equal to the sum of the
separate five percentile values is substantially less than five percent.
54
-------
Chapter 7: Results and Uncertainty
mated annual mean TSP concentrations in other parts
of the world65 indicates that in 1990—
60 metropolitan areas in the U.S. would have
had higher TSP concentrations than Moscow,
Russia
7 metropolitan areas would be worse than
Bangkok, Thailand
6 metropolitan areas would be worse than
Bombay, India
2 metropolitan areas would be worse than Ma-
nila, Philippines
One metropolitan area would be worse than
Delhi, India (one of the most polluted cities
in the world)
Under the control scenario, TSP levels in only 3
metropolitan areas were projected to exceed those in
Moscow, and none exceeded levels found in the other
foreign cities listed above. The principal reason air
quality conditions are so poor under the no-control
scenario is that air pollution control requirements re-
main fixed at their 1970 levels of scope and stringency
while total economic activity, including polluting ac-
tivity, grows by 70 percent and population grows by
22.3 percent between 1970 and 1990. Under the se-
vere air quality conditions projected throughout the
U.S. in 1990 under the no-control case, an additional
205,000 people would be projected to die prematurely
due to the effects of particulate matter, lead, and other
criteria pollutants. This represents a very laige increase
in the risk of premature mortality. Since the estimate
that the average loss of life for those who actually
succumb to PM exposure related health effects is ap-
proximately 14 years, and life-shortening due to lead
exposure is even greater, it is no longer surprising that
the estimated value of avoiding these severe condi-
tions is so high.
Second, with respect to the nature of the welfare
change reflected in the monetized benefit estimate,
the concern about the effects of limited budgets con-
straining Americans' collective ability to pay to avoid
these severe no-control scenario conditions is mis-
placed. In reality, what society actually had to pay to
avoid these conditions is measured on the cost side of
the analysis, which sums up the total expenditures
made by manufacturers and others to achieve these
air pollution reductions. The most reasonable estimate
of the value Americans place on avoiding those se-
vere no-control scenario conditions, however, is pro-
vided by measuring the amount of compensation
Americans would have demanded from polluting com-
panies and others to accept, willingly, all of that extra
pollution and its associated risks of premature death.
Under this concept of welfare change measurement,
there is no inherent limit on the amount of money citi-
zens would demand from companies to accept their
pollution and so individual personal wealth does not
constrain this value.
The monetized benefit estimate presented in this
study, therefore, does not necessarily represent an at-
tempt to mirror what Americans would pay out of their
own pockets to reduce air pollution from levels they
never experienced; rather, it provides an estimate of
the value Americans place on the protection they re-
ceived against the dire air pollution conditions which
might have prevailed in the absence of the 1970 and
1977 Clean Air Acts and related programs. Viewed
from this perspective, the estimated monetized ben-
efits presented herein appear entirely plausible.
Comparison of Monetized
Benefits and Costs
Table 18 presents summary quantitative results for
the retrospective assessment. Annual results are pre-
sented for four individual years, with all dollar fig-
ures expressed as inflation-adjusted 1990 dollars. The
final column sums the stream of costs and benefits
from 1970 to 1990, discounted (i.e., compounded) to
1990 at five percent. "Monetized benefits" indicate
both the mean of the Monte Carlo analysis and the
credible range. "Net Benefits" are mean monetized
benefits less annualized costs for each year. The table
also notes the benefit/cost ratios implied by the ben-
efit ranges. The distribution of benefits changes little
(except in scale) from year to year: The mean esti-
mate is somewhat greater than twice the fifth percen-
tile estimate, and the ninety-fifth percentile estimate
is somewhat less than twice the mean estimate. The
distribution shape changes little across years because
the sources of uncertainty (i.e., CRfunctions and eco-
nomic valuations) and their characterizations are un-
changed from year to year. Some variability is induced
by changes in relative pollutant concentrations over
time, which then change the relative impact of indi-
vidual CRfunctions.
Several measures of "cost" are available for use
in this analysis (see Chapter 2). The Project Team
65 "Urban Air Pollution in Megacities of the World," UNEP/WHO, 1992a, Published by the World Health Organization and
United Nations Environment Program, Blackwell Publishers, Oxford, England, 1992. "City Air Quality Trends," UNEP/WHO, 1992b,
Published by the United Nations Environment Program, Nairobi, Kenya, 1992.
55
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic 18. Quantified Uncertainty Ranges lor Monetized
Annual Benefits and Benefit/Cost Ratios. 1970-1990 (in
millions of 1990-valuc dollars).
1975
1980
1985
1990
1>V
Monetized Benefits
5th percentile
87
235
293
329
5.600
Mean estimate
355
930
1,155
1.248
22.200
95th percentile
799
2,063
2,569
2.762
49.400
Annualized Costs (5%)
14
21
25
26
523
Net Benefits
Mean benefits - Costs
341
909
1,130
1.220
21.700
Benefit/Cost ratio
5th percentile
6/1
11/1
12/1
13/1
1 1/1
Mean estimate
25/1
44/1
46/1
48/1
42/1
95th Dercentile
57/1
98/1
103/1
106/1
94/1
Notes- PV=1990 present value reflecting compounding of costs and benefits
from 1971 to 1990 at 5 percent.
employs "annualized cost" as the primary cost mea-
sure because it measures cost in a fashion most analo-
gous to the benefits estimation method. An alternative
measure, "compliance expenditure," is a reasonable
cost measure. Some capital expenditures, however,
generate a benefit stream beyond the period of the
analysis (i.e., beyond 1990). Those post-1990 benefits
are not, in general, included in the benefit estimates
presented above. The annualization procedure reduces
the bias introduced by the use of capital expenditures
by spreading the cost of the capital investment over its
expected life, then counting as a "cost" only those costs
incurred in the 1970 to 1990 period.
The macroeconomic analysis employed for this
analysis (see Chapter 2) indicates that compliance
expenditures induce significant second-order ef-
fects, and it can be aigued that those effects should
be included in a comprehensive cost analysis. Ben-
efits resulting from compliance expenditures
should also induce second-order macroeconomic
effects (which would, one would expect, partly or
completely offset the estimated second-order ad-
verse effects induced by compliance expenditures).
Due to the sequencing of the analytical steps in
this assessment, it was not practical to estimate
the second-order cost and benefit impacts induced
by the estimated health and welfare benefits. Be-
cause second-order impacts of benefits are not
estimated, the Project Team refrained from choos-
ing as the primary cost measure one that included
second-order impacts, and instead employed "an-
nualized costs" as the primary cost measure.
Major Sources of Uncertainty
The methods used to aggregate monetized ben-
efits and characterize the uncertainty surrounding es-
timates of these benefits have been discussed above,
and the resulting estimates of aggregate benefits have
been compared to the corresponding estimates of cost.
Additional insights into key assumptions and findings
can, however, be obtained by further analysis of po-
tentially important variables.
For some factors in the present analysis, both the
degree of uncertainty and the direction of any associ-
ated bias are unknown; for some other factors, no
employable quantitative estimates could be used even
though available evidence suggests a positive and
potentially substantial value. An example of the latter
deficiency is the lack of quantitative estimates for some
human health effects, some human welfare effects, and
all ecological effects. Despite the exclusion of poten-
tially important variables, it is worthwhile to evaluate
the relative contribution of included variables to quan-
tifiable uncertainty in the net benefit estimate. One of
these variables, premature mortality valuation, is also
given special attention in the subsequent section on
alternative results.
The estimated uncertainty ranges for each end-
point category summarized in Table 16 reflect the mea-
sured uncertainty associated with both avoided inci-
dence and economic valuation. The Project Team con-
ducted a sensitivity analysis to determine the variables
with the greatest contribution to the quantified uncer-
tainty range. The results of this sensitivity analysis
are illustrated in Figure 20.
Figure 20. Uncertainty Ranges Deriving From Individual
Uncertainty Factors
72 $50 "
| $45 -
¦e
$40-
4} $35 -
a
$30-
$25-
$20-
$15-
$10-
$5 -
95th %ile
Mean
5th %ile
£ o =
ro '¦£
¦f to
O -2
s 5
Z-
a;
W
c
>•
Oi
w oi
c
c
^ £
o
c
o
c
2
c
m —
CL
o
o
56
-------
Chapter 7: Results and Uncertainty
In this sensitivity analysis, all the inputs to the
Monte Carlo uncertainty analysis are held constant
(at their mean values), allowing only one variable ~
for example, the economic valuation of mortality ~
to vary across the range of that variable's uncertainty.
The sensitivity analysis then isolates how this single
source of uncertainty contributes to the total measured
uncertainty in estimated aggregate benefits. The first
uncertainty bar represents the credible range associ-
ated with the total monetized benefits of the Clean
Air Act, as reported above. This captures the multiple
uncertainties in the quantified benefits estimation. The
rest of the uncertainty bars represent the quantified
uncertainty ranges generated by single variables. As
shown in Figure 20, the most important contributors
to aggregate quantified uncertainty are mortality valu-
ation and incidence, followed by chronic bronchitis
valuation and incidence.
Alternative Results
The primary results of this analysis, including
aggregate cost and benefit estimates and the uncer-
tainty associated with them, are presented and dis-
cussed above. However, although the range of net
benefit estimates presented reflects uncertainty in
many important elements of the analysis, there are
two key variables which require further discussion and
analysis: PM-related mortality valuation and the dis-
count rate. This additional treatment is necessary be-
cause reasonable people may disagree with the Project
Team's methodological choices for these two vari-
ables, and these choices might be considered ex ante
to significantly influence the results of the study. The
purpose of this section, therefore, is to present alter-
native quantitative results which reflect, separately,
(1) an alternative approach to valuation of premature
mortality associated with particulate matter exposure,
and (2) alternative values for the discount rate used to
adjust the monetary values of effects occurring in vari-
ous years to a particular reference year (i.e., 1990).
PM Mortality Valuation Based on Life-
Years Lost
The primary analytical results presented earlier
in this chapter assign the same economic value to in-
cidences of premature mortality regardless of the age
and health status of those affected. Although this has
been the traditional practice for benefit-cost studies
conducted within the Agency, this may not be the most
appropriate method for valuation of premature mor-
tality caused by PM exposure. Some short-term PM
exposure studies suggest that a significantly dispro-
portionate share of PM-related premature mortality
occurs among persons 65 years of age or older. Com-
bining standard life expectancy tables with the lim-
ited available data on age-specific incidence allows
crude approximations of the number of life-years lost
by those who die prematurely as a result of exposure
to PM or, alternatively, the changes in age-specific
life expectancy of those who are exposed to PM.
The ability to estimate, however crudely, changes
in age-specific life expectancy raises the issue of
whether available measures of the economic value of
mortality risk reduction can, and should, be adapted
to measure the value of specific numbers of life-years
saved.66 Although the Agency has on occasion per-
formed sensitivity calculations which adjust mortal-
ity values for those over age 65, the Agency is skepti-
cal that the current state of knowledge and available
analytical tools support using a life-years lost approach
or any other approach which assigns different risk re-
duction values to people of different ages or circum-
stances. This skepticism is mirrored in the OMB guid-
ance on implementing Executive Order 12866 per-
taining to economic analysis methods, which states
on page 31:
While there are theoretical advantages to
using a value of statistical life-year-extended
approach, current research does not provide
a definitive way of developing estimates of
VSLY that are sensitive to such factors as
current age, latency of effect, life years
remaining, and social valuation of different
risk reductions. In lieu of such information,
there are several options for deriving the
value of a life-year savedfrom an estimate of
the value of life, but each of these methods
has drawbacks. One approach is to use results
from the wage compensation literature (which
focuses on the effect of age on WTP to avoid
risk of occupationalfatality). However, these
results may not be appropriate for other types
of risks. Another approach is to annualize the
VSL using an appropriate rate of discount and
the average life years remaining. This
approach does not provide an independent
estimate of VSLY; it simply rescales the VSL
estimate. Agencies should consider providing
estimates of both VSL and VSLY, while
recognizing the developing state of know ledge
in this area.
While the Agency continues to prefer an approach
which makes no valuation distinctions based on age
or other characteristics of the affected population, al-
ternative results based on a VSLY approach are pre-
66 This issue was extensively discussed during the Science Advisory Board Council review of drafts of the present study. The
Council suggested it would be reasonable and appropriate to show PM mortality benefit estimates based on value of statistical life-
years (VSLY) saved as well as the value of statistical life (VSL) approach traditionally applied by the Agency to all incidences of
premature mortality.
57
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
sented below. The method used to develop VSLY es-
timates is described briefly in Chapter 6 and in more
detail in Appendix I.
Table 19 summarizes and compares the results of
the VSL and VSLY approaches. Estimated 1970 to
1990 benefits from PM-related mortality alone and
total assessment benefits are reported, along with to-
tal compliance costs for the same period, in 1990 dol-
lars discounted to 1990 at five percent. The results
indicate that the choice of valuation methodology sig-
nificantly affects the estimated monetized value of
historical reductions in air pollution-related prema-
ture mortality. However, the downward adjustment
which would result from applying a VSLY approach
in lieu of a VSL approach does not change the basic
outcome of this study, viz. the estimated monetized
benefits of the historical CAA substantially exceed
the historical costs of compliance.
1970 toward 1990 (see Table 18 above), benefit cost
ratios decline as the discount rate increases (because
earlier periods are given greater weight). Overall, the
results of the benefit-cost assessment appear to be
generally insensitive to the choice of discount rate.
Tabic 20. Eflcci of Alternative Discount Rates on
Present Value of Total Monetized Benefits/Costs
lor 1970 to 1990 (in trillions of 1990 dollars).
Discount rate
2%
m.
2%.
Mean Estimated Benefits
19.2
22.2
25.8
Annualized Costs
0.4
0.5
0.7
Net Benefits
18.8
21.7
25.1
Benefit/Cost ratio
48/1
42/1
37/1
Tabic 19. Alternative Mortality Benefits Mean
Estimates lor 1970 to 1990 (in trillions of 1990
dollars, discounted at 5 percent) Compared to
Total 1970 to 1990 Compliance Costs.
Rmiefits
Renefit Estimation Method PM Tot.
Statistical life method ($4.8M/case) 16.6 18.0
Life-years lost method ($293,GOO/year) 9.1 10.1
Total compliance cost --- 0.5
Alternative Discount Rates
In some instances, the choice of discount rate can
have an important effect on the results of a benefit-
cost analysis; particularly for those analyses with rela-
tively long time horizons for costs and/or benefits. In
this assessment, the discount rate affects only four
factors: IQ-related benefits estimates (especially esti-
mates of changes in discounted lifetime income), life-
time income losses due to other health effects (e.g.,
stroke), annualized costs (i.e., amortized capital ex-
penditures), and compounding of all costs and ben-
efits to 1990. Table 20 summarizes the effect of alter-
native discount rates on the "best estimate" results of
this analysis. Because monetized benefits exceed costs
for all years in the analysis period, net benefits in-
crease as the discount rate increases. Because the an-
nual benefit/cost ratio increases as one moves from
58
-------
Chapter 7: Results and Uncertainty
59
-------
Appendix A: Cost and Macroeconomic Modeling
Introduction
The purpose of this appendix is to describe in de-
tail the estimation of direct compliance costs associ-
ated with the CAA and the effect of those expendi-
tures on U.S. economic conditions from 1970 to 1990.
The first section of this appendix describes the dy-
namic, general equilibrium macroeconomic model
used to examine economy-wide effects. Two broad
categories of models were considered for use in the
assessment: Macroeconomic forecasting models (e.g.,
the Data Resources Inc. model of the U.S. economy),
and general equilibrium models (e.g., Hazilla and
Kopp [1990], and Jorgenson and Wilcoxen [1990a]).
The project team selected the Jorgenson-Wilcoxen (J/
W) general equilibrium model of the United States
for this analysis (Jorgenson and Wilcoxen [1990a]).
There are two main reasons for choosing a dynamic
general equilibrium approach: To capture both the
direct and indirect economic effects of environmen-
tal regulation, and to capture the long-run dynamics
of the adjustment of the economy. The general equi-
librium framework enabled the project team to assess
shifts in economic activity between industries, includ-
ing changes in distributions of labor, capital, and other
production factors within the economy, and changes
in the distribution of goods and services.
The second section describes the data sources for
direct compliance expenditures and presents estimates
of historical air pollution control expenditures. These
estimates are derived primarily from EPA's 1990 re-
port entitled "Environmental Investments: The Cost
of a Clean Environment"1 (hereafter referred to as Cost
of Clean). Specific adjustments to the Cost of Clean
stationary source and mobile source O&M data needed
to adapt these data for use in the present study are
also described. These adjusted expenditure estimates
represent the compliance cost data used as inputs to
the J/W model to determine macroeconomic effects.
The final section presents a summary of the di-
rect expenditure data, presents direct costs in a form
that can be compared to the benefits estimates found
elsewhere in the study, and discusses indirect effects
arising from compliance expenditures estimated by
the macroeconomic model. The indirect effects re-
ported by the model are sectoral impacts and changes
in aggregate measures of economic activity such as
household consumption and gross national product.
These indirect effects are second-order impacts of
compliance expenditures — a parallel modeling ex-
ercise to estimate second-order economic impacts aris-
ing from the benefits of compliance (e.g., increased
output as a result of improved longevity or fewer
workdays lost as a result of non-fatal heart attacks)
has not been attempted.
Macroeconomic Modeling
EPA analyses of the costs of environmental regu-
lations typically quantify the direct costs of pollution
abatement equipment and related operating and main-
tenance expenses. However, this approach does not
fully account for all of the broader economic conse-
quences of reallocating resources to the production
and use of pollution abatement equipment. A general
equilibrium, macroeconomic model could, in theory,
capture the complex interactions between sectors in
the economy and assess the full economic cost of air
pollution control. This would be particularly useful
for assessing regulations that may produce significant
interaction effects between markets. Another advan-
tage of a general equilibrium, macroeconomic frame-
work is that it is internally consistent. The consistency
of sectoral forecasts with realistic projections of U.S.
economic growth is ensured since they are estimated
within the context of a single model.2 This contrasts
1 Environmental Investments: The Cost of a Clean Environment, Report of the Administrator of the Environmental Protection
Agency to the Congress of the United States, EPA-230-11-90-083, November 1990.
2 In the present study, both benefits and costs are driven by of the same macroeconomic projections from the Jorgenson/
Wilcoxen model, to ensure that the estimates are based on a consistent set of economic assumptions.
A-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
with typical EPA analyses that compile cost estimates
from disparate sectoral and partial equilibrium mod-
els.
The economic effects of the CAA may be over-
or underestimated, if general equilibrium effects are
ignored, to the extent that sectors not directly regu-
lated are affected. For example, it is well known that
the CAA imposed significant direct costs on the en-
ergy industry. Economic sectors not directly regulated
will nonetheless be affected by changes in energy
prices. However, an examination of the broader ef-
fects of the CAA on the entire economy might reveal
that the CAA also led to more rapid technological
development and market penetration of environmen-
tally "clean" renewable sources of energy (e.g., pho-
tovoltaics). These effects would partially offset ad-
verse effects on the energy industry, and lead to a dif-
ferent estimate of the total economic cost to society
of the CAA.
The significance of general equilibrium effects in
the context of any particular analysis is an empirical
question. Kokoski and Smith (1987) used a comput-
able general equilibrium model to demonstrate that
partial-equilibrium welfare measures can offer rea-
sonable approximations of the true welfare changes
for large exogenous changes. In contrast, the results
of Jorgenson and Wilcoxen (1990a) and Hazilla and
Kopp (1990) suggest that total pollution abatement in
the U.S. has been a major claimant on productive re-
sources, and the effect on long-run economic growth
may be significant. Again, such conclusions must be
considered in light of the limitations of general equi-
librium models.
Choice of Macroeconomic Model
The adequacy of any model or modeling approach
must be judged in light of the policy questions being
asked. One goal of the present study is to assess the
effects of clean air regulations on macroeconomic
activity. Two broad categories of macroeconomic
models were considered for use in the assessment:
short run, Keynesian models and long-run, general
equilibrium models.
Recognizing that structural differences exist be-
tween the models, one needs to focus in on the par-
ticular questions that should be answered with any
particular model. The Congressional Budget Office
(1990) noted:
"Both the [Data Resources Incorporated] DRI
and the IPCAEO models show relatively
limited possibilities for increasing energy
efficiency and substituting other goods for
energy in the short run... Both models focus
primarily on short-term responses to higher
energy prices, and neither is very good at
examining how the structure of the economy
could change in response to changing energy
prices. The [Jorgenson-Wilcoxen] model
completes this part of the picture..."3
One strategy for assessing the macroeconomic
effects of the CAA would be to use a DRI-type model
in conjunction with the Jorgenson-Wilcoxen model
to assess both the long-term effects and the short-run
transitions, in much the same way that the Congres-
sional Budget Office used these models to assess the
effects of carbon taxes. However, because of signifi-
cant difficulties in trying to implement the DRI model
in a meaningful way, the project team chose to focus
on the long-run effects of the CAA. Structural changes
(e.g., changes in employment in the coal sector due to
the CAA) can be identified with the Jorgenson-
Wilcoxen model.
Overview of the Jorgenson-
Wilcoxen Model
The discussion below focuses on those character-
istics of the Jorgenson-Wilcoxen model that have
important implications for its use in the assessment
of environmental regulations (see Table A-l). The J/
W model is a detailed dynamic general equilibrium
model of the U.S. economy designed for medium run
analysis of regulatory and tax policy (Jorgenson and
Wilcoxen [1990a]). It provides projections of key
macroeconomic variables, such as GNP and aggre-
gate consumption, as well as energy flows between
economic sectors. As a result, the model is particu-
larly useful for examining how the structure of the
economy could change in response to changes in re-
3 The Congressional Budget Office report (1990) refers to an older (1981) version of the Jorgenson model, not the current
(1988) version. The approach to long-run dynamics differs between the two models. The newer Jorgenson-Wilcoxen model contains
both the capital accumulation equation and the capital asset pricing equation. The 1981 version of the model contained only the
capital accumulation equation.
A-2
-------
Appendix A: Cost andMacroeconomic Modeling
Tabic A-l. Key Distinguishing Characteristics of
the Jorgcnson-Wilcoxcn Model.
¦ Dynamic, general equilibrium,
macroeconomic model of the U.S. economy.
¦ Econometrically estimated using historic
data.
¦ Free mobility of a single type of capital and
labor between industries.
¦ Detailed treatment of pro duction and
consumption.
¦ Rigorous representation of savings and
investment.
¦ Endogenous model of technical change.
¦ Does not capture unemployment,
underemployment, or the costs of moving
capital from one industry to another.
source prices. For the purpose of this study, it has five
key features: a detailed treatment of production and
consumption, parameters estimated econometrically
from historical data, an endogenous model of techni-
cal change, a rigorous representation of saving and
investment, and free mobility of labor and capital be-
tween industries.
The first two features, industry and consumer de-
tail and econometric estimation, allow the model to
capture the effects of the CAA at each point in time
for given levels of technology and the size of the
economy's capital stock. A detailed treatment of pro-
duction and consumption is important because the
principal effects of the Clean Air Act fell most heavily
on a handful of industries. The J/W model divides
total U.S. production into 35 industries which allows
the primary economic effects of the CAA to be cap-
tured. Econometric estimation is equally important
because it ensures that the behavior of households and
firms in the model is consistent with the historical
record.
The model's second two features —its represen-
tations of technical change and capital accumulation—
complement the model's intratemporal features by
providing specific information on how the Act affected
technical change and the accumulation of capital.
Many analyses of environmental regulations overlook
or ignore intertemporal effects but these effects can
be very important. Jorgenson and Wilcoxen (1990a)
suggests that the largest cost of all U.S. environmen-
tal regulations together was that the regulations re-
duced the rate of capital accumulation.
The model's last feature, free mobility of a single
type of capital and a single type of labor, is important
because it limits the model's ability to measure the
short run costs of changes in policy. J/W is a full-
employment model that describes the long-run dynam-
ics of transitions from one equilibrium to another.
Capital and labor are both assumed to be freely mo-
bile between sectors (that is, they can be moved from
one industry to another at zero cost) and to be fully
used at all times. Over the medium to long run, this is
a reasonable assumption, but in the short run it is too
optimistic. In particular, the model will understate the
short run costs of a change in policy because it does
not capture unemployment, underemployment, or the
costs of moving capital from one industry to another.
A single rate of return on capital exists that efficiently
allocates the capital in each period among sectors.
Similarly, a single equilibrium wage rate allocates
labor throughout the economy.
Structure of the Jorgenson-Wilcoxen
Model
The J/W model assesses a broad array of economic
effects of environmental regulations. Direct costs are
captured as increased expenditures on factors of pro-
duction —capital, labor, energy and materials— that
the various industries must make to comply with the
regulations, as well as additional out-of-pocket ex-
penditures that consumers must make. Indirect costs
are captured as general equilibrium effects that occur
throughout the economy as the prices of factors of
production change (e.g., energy prices). Also, the rate
of technological change can respond to changes in the
prices of factors of production, causing changes in
productivity (Jorgenson and Fraumeni, 1981).
The model is divided into four major sectors: the
business, household, government, and rest-of-the-
world sectors. The business sector is further subdi-
vided into 35 industries (see Table A-2)4 Each sector
produces a primary product, and some produce sec-
ondary products. These outputs serve as inputs to the
production processes of the other industries, are used
for investment, satisfy final demands by the house-
hold and government sectors, and are exported. The
model also allows for imports from the rest of the
world.
4 The 35 industries roughly correspond to a two-digit SIC code classification scheme.
A-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic A-2. Definitions of Industries Within
the J/W Model.
Industry
Nil m ho r
Description
1
Agriculture, forestry, and
fisheries
2
Metal mining
3
Coalmining
4
Crude petroleum and natural gas
5
Nonmetallic mineral mining
6
Construction
7
Food and kindred products
8
Tobacco manufacturers
9
Textile mill products
10
Apparel and other textile
products
II
Lumber and wood products
12
Furniture and fixtures
13
Paper and allied products
14
Printing and publishing
15
Chemicals and allied products
16
Petroleum refining
17
Rubber and plastic products
18
Leather and leather products
19
Stone, clay, and glass products
20
Primary metals
21
Fabricated metal products
22
Machinery, except electrical
23
Electrical machinery
24
Motor vehicles
25
Other transportation equipment
26
Instruments
27
Miscellaneous manufacturing
28
Transportation and warehousing
29
Communication
30
Electric utilities
31
Gas utilities
32
Trade
33
Finance, insurance, and real
estate
34
Other services
35
Government enterprises
The Business Sector
The model of producer behavior allocates the
value of output of each industry among the inputs of
the 35 commodity groups, capital services, labor ser-
vices, and noncompeting imports. Output supply and
factor demands of each sector are modeled as the re-
sults of choices made by wealth maximizing, price
taking firms which are subject to technological con-
straints. Firms have perfect foresight of all future
prices and interest rates. Production technologies are
represented by econometrically estimated cost func-
tions that fully capture factor substitution possibili-
ties and industry-level biased technological change.
Capital and energy are specified separately in the
factor demand functions of each industry. The ability
of the model to estimate the degree of substitutability
between factor inputs facilitates the assessment of the
effect of environmental regulations. A high degree of
substitutability between inputs implies that the cost
of environmental regulation is low, while a low de-
gree of substitutability implies high costs of environ-
mental regulation. Also, different types of regulations
lead to different responses on the part of producers.
Some regulations require the use of specific types of
equipment. Others regulations restrict the use of par-
ticular factor inputs; for example, through restrictions
on the combustion of certain types of fuels. Both of
these effects can change the rate of productivity growth
in an industry through changes in factor prices.
The Household Sector
In the model of consumer behavior, consumer
choices between labor and leisure and between con-
sumption and saving are determined. A system of in-
dividual, demographically defined household demand
functions are also econometrically estimated. House-
hold consumption is modeled as a three stage optimi-
zation process. In the first stage households allocate
lifetime wealth to full consumption in current and fu-
ture time periods to maximize intertemporal utility.
Lifetime wealth includes financial wealth, discounted
labor income, and the imputed value of leisure. House-
holds have perfect foresight of future prices and in-
terest rates. In the second stage, for each time period
full consumption is allocated between goods and ser-
vices and leisure to maximize intratemporal utility.
This yields an allocation of a household's time en-
dowment between the labor market (giving rise to la-
bor supply and labor income) and leisure time and
demands for goods and services. In the third stage,
personal consumption expenditures are allocated
among capital, labor, noncompeting imports and the
outputs of the 35 production sectors to maximize a
subutility function for goods consumption. As with
the business sector, substitution possibilities exist in
consumption decisions. The model's flexibility en-
ables it to capture the substitution of nonpolluting
products for polluting ones that may be induced by
environmental regulations. Towards this end, pur-
chases of energy and capital services by households
are specified separately within the consumer demand
functions for individual commodities.
A-4
-------
Appendix A: Cost andMacroeconomic Modeling
It is important to be clear regarding the notions of
labor supply and demand within the J/W model, and
what is meant by "employment" throughout this re-
port. Labor demands and supplies are represented as
quality-adjusted hours denominated in constant dol-
lars. The labor market clears in each period; the quan-
tity of labor services offered by households is absorbed
fully by the economy's producing sectors. However,
inferences regarding the number of persons employed
require information on labor quality and work-hours
per person over time and across simulations. Neither
of these are explicitly modeled.
The Government Sector
The behavior of government is constrained by
exogenously specified budget deficits. Government
tax revenues are determined by exogenously speci-
fied tax rates applied to appropriate transactions in
the business and household sectors. Levels of eco-
nomic activity in these sectors are endogenously de-
termined. Capital income from government enterprises
(determined endogenously), and nontax receipts
(given exogenously), are added to tax revenues to
obtain total government revenues. Government expen-
ditures adjust to satisfy the exogenous budget deficit
constraint.
The Rest-of-the-World Sector
The current account balance is exogenous, limit-
ing the usefulness of the model to assess trade com-
petitiveness effects. Imports are treated as imperfect
substitutes for similar domestic commodities and com-
pete on price. Export demands are functions of for-
eign incomes and ratios of commodity prices in U.S.
currency to the exchange rate. Import prices, foreign
incomes, and tariff policies are exogenously speci-
fied. Foreign prices of U.S. exports are determined
endogenously by domestic prices and the exchange
rate. The exchange rate adjusts to satisfy the exog-
enous constraint on net exports.
Environmental Regulation, Investment,
and Capital Formation
Environmental regulations have several important
effects on capital formation. At the most obvious level,
regulations often require investment in specific pieces
of pollution abatement equipment. If the economy's
pool of savings were essentially fixed, the need to in-
vest in abatement equipment would reduce, or crowd
out, investment in other kinds of capital on a dollar
for dollar basis. On the other hand, if the supply of
savings were very elastic then abatement investments
might not crowd out other investment at all. In the J/
W model, both the current account and government
budget deficits are fixed exogenously so any change
in the supply of funds for domestic investment must
come from a change in domestic savings. Because
households choose consumption, and hence savings,
to maximize a lifetime utility function, domestic sav-
ings will be somewhat elastic. Thus, abatement in-
vestment will crowd out other investment, although
not on a dollar for dollar basis.
The J/W assumption that the current account does
not change as a result of environmental regulation is
probably unrealistic, but it is not at all clear that this
biases the crowding out effects in any particular di-
rection. By itself, the need to invest in abatement capi-
tal would tend to raise U.S. interest rates and draw in
foreign savings. To the extent this occurred, crowd-
ing out would be reduced. At the same time, how-
ever, regulation reduces the profitability of domestic
firms. This effect would tend to lower the return on
domestic assets, leading to a reduced supply of for-
eign savings which would exacerbate crowding out.
Which effect dominates is an empirical question be-
yond the scope of this study.
In additional to crowding out ordinary investment,
environmental regulation also has a more subtle ef-
fect on the rate of capital formation. Regulations raise
the prices of intermediate goods used to produce new
capital. This leads to a reduction in the number of capi-
tal goods which can be purchased with a given pool
of savings. This is not crowding out in the usual sense
of the term, but it is an important means by which
regulation reduces capital formation.5
The General Equilibrium
The J/W framework contains intertemporal and
intratemporal models (Jorgenson and Wilcoxen
[1990c]). In any particular time period, all markets
clear. This market clearing process occurs in response
to any changes in the levels of variables that are speci-
5 Wilcoxen (1988) suggests that environmental regulation may actually lead to a "crowding in" phenomenon. Wilcoxen
examined the effects of regulation at the firm level, and introduced costs into the model related to the installation of capital. He found
that when firms shut down their plants to install environmental capital, they take account of the adjustment costs and often concur-
rently replace other older capital equipment. This effect, however, is not captured in the current version of the Jorgenson-Wilcoxen
model.
A-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
fied exogenously to the model. The interactions among
sectors determine, for each period, aggregate domes-
tic output, capital accumulation, employment, the
composition of output, the allocation of output across
different household types, and other variables.
The model also produces an intertemporal equi-
librium path from the initial conditions at the start of
the simulation to the stationary state. (A stationary
solution for the model is obtained by merging the
intertemporal and intratemporal models.) The dynam-
ics of the J/W model have two elements: An accumu-
lation equation for capital, and a capital asset pricing
equation. Changes in exogenous variables cause sev-
eral adjustments to occur within the model. First, the
single stock of capital is efficiently allocated among
all sectors, including the household sector. Capital is
assumed to be perfectly malleable and mobile among
sectors, so that the price of capital services in each
sector is proportional to a single capital service price
for the economy as a whole. The value of capital ser-
vices is equal to capital income. The supply of capital
available in each period is the result of past invest-
ment, i.e., capital at the end of each period is a func-
tion of investment during the period and capital at the
beginning of the period. This capital accumulation
equation is backward-looking and captures the effect
of investments in all past periods on the capital avail-
able in the current period.
The capital asset pricing equation specifies the
price of capital services in terms of the price of in-
vestment goods at the beginning and end of each pe-
riod, the rate of return to capital for the economy as a
whole, the rate of depreciation, and variables describ-
ing the tax structure for income from capital. The cur-
rent price of investment goods incorporates an assump-
tion of perfect foresight or rational expectations. Un-
der this assumption, the price of investment goods in
every period is based on expectations of future capi-
tal service prices and discount rates that are fulfilled
by the solution of the model. This equation for the
investment goods price in each time period is forward-
looking.6
One way to characterize the J/W model —or any
other neoclassical growth model— is that the short-
run supply of capital is perfectly inelastic, since it is
completely determined by past investment. However,
the supply of capital is perfectly elastic in the long
run. The capital stock adjusts to the time endowment,
while the rate of return depends only on the
intertemporal preferences of the household sector.
A predetermined amount of technical progress
also takes place that serves to lower the cost of sectoral
production. Finally, the quality of labor is enhanced,
giving rise to higher productivity and lower costs of
production.
Given all of these changes, the model solves for a
new price vector and attains a new general equilib-
rium. Across all time periods, the model solves for
the time paths of the capital stock, household con-
sumption, and prices. The outcomes represent a gen-
eral equilibrium in all time periods and in all markets
covered by the J/W model.
Configuration of the No-control
Scenario
One of the difficulties in describing the no-con-
trol scenario is ascertaining how much environmen-
tal regulation would have been initiated by state and
local governments in the absence of a federal program.
It may reasonably be argued that many state and local
governments would have initiated their own control
programs in the absence of a federal role. This view
is further supported by the fact that many states and
localities have, in fact, issued rules and ordinances
which are significantly more stringent and encompass-
ing than federal minimum requirements. However, it
may also be argued that the federal CAA has moti-
vated a substantial number of stringent state and local
control programs.
Specifying the range and stringency of state and
local programs that would have occurred in the ab-
sence of the federal CAA would be almost entirely
speculative. For example, factors which would com-
plicate developing assumptions about stringency and
scope of unilateral state and local programs include:
(i) the significance of federal funding to support state
and local program development; (ii) the influence of
more severe air pollution episodes which might be
expected in the absence of federally-mandated con-
trols; (iii) the potential emergence of pollution havens,
as well as anti-pollution havens, motivated by local
6 The price of capital assets is also equal to the cost of production, so that changes in the rate of capital accumulation result in an
increase in the cost of producing investment goods. This has to be equilibrated with the discounted value of future rentals in order to
produce an intertemporal equilibrium. The rising cost of producing investment is a cost of adjusting to a new intertemporal equilib-
rium path.
A-6
-------
Appendix A: Cost andMacroeconomic Modeling
political and economic conditions; (iv) the influence
of federally-sponsored research on the development
of pollution effects information and control technolo-
gies; and (v) the need to make specific assumptions
about individual state and local control levels for in-
dividual pollutants to allow estimation of incremen-
tal reductions attributable to federal control programs.
Another complication associated with the no-con-
trol scenario is the treatment of air pollution control
requirements among the major trading partners of the
U.S. Real-world manifestation of ano-control scenario
would imply that public health and environmental
goals were not deemed sufficiently compelling by U.S.
policy makers. Under these conditions, major trading
partners ofthe U.S. in Japan, Europe, and Canada may
well reach similar policy conclusions. Simply put, if
the U.S. saw no need for air pollution controls, there
is little reason to assume other developed industrial
countries would have either. In this case, some of the
estimated economic benefits of reducing or eliminat-
ing air pollution controls in the U.S. would not mate-
rialize because U.S. manufacturers would not neces-
sarily gain a production cost advantage over foreign
competitors. However, like the question of state and
local programs in the absence of a federal program,
foreign government policies under a no-control sce-
nario would be highly speculative.
Given the severity of these confounding factors,
the only analytically feasible assumptions with respect
to the no-control scenario are that (a) no new control
programs would have been initiated after 1970 by the
states or local governments in the absence of a fed-
eral role, and (b) environmental policies of U.S. trad-
ing partners remain constant regardless of U.S. policy.
Elimination of Compliance Costs in the
No-Control Case
Industries that are affected by environmental regu-
lations can generally respond in three ways: (i) with
process changes (e.g., fluidized bed combustion); (ii)
through input substitution (e.g., switching from high
sulfur coal to low sulfur coal); and (iii) end-of-pipe
abatement (e.g., the use of electrostatic precipitation
to reduce the emissions of particulates by combus-
tion equipment).7 Clean air regulations have typically
led to the latter two responses, especially in the short
run. End-of-pipe abatement is usually the method of
choice for existing facilities, since modifying exist-
ing production processes can be costly. This approach
is also encouraged by EPA's setting of standards based
on the notion of "best available technology" (Free-
man, 1978).
All three possible responses may lead to: (i) un-
anticipated losses to equity owners; (ii) changes in
current output; and (iii) changes in long-run profit-
ability. If firms were initially maximizing profits, then
any of the above three responses will increase its costs.
Fixed costs of investment will be capitalized imme-
diately. This will result in a loss to owners of equity
when regulations are introduced. As far as firms are
concerned, this is just like a lump sum tax on sunk
capital. Such effects will not affect growth or effi-
ciency. However, regulations could also change mar-
ginal costs and therefore current output. In addition,
they could change profits (i.e., the earnings of capi-
tal), and thus affect investment. Both of these effects
will reduce the measured output of the economy.
On the consumption side, environmental regula-
tions change consumers' expectations of their lifetime
wealth. In the no-control scenario of this assessment,
lifetime wealth increases. This causes an increase in
consumption. In fact, with perfect foresight, consump-
tion rises more in earlier time periods. This also re-
sults in a change in savings.
Capital Costs - Stationary Sources
To appropriately model investment in pollution
control requires a recognition that the CAA had two
different effects on capital markets. First, CAA regu-
lations led to the retrofitting of existing capital stock
in order to meet environmental standards. In the no-
control scenario, these expenditures do not occur. In-
stead, the resources that were invested in pollution
abatement equipment to retrofit existing sources are
available to go to other competing investments. Thus,
at each point in time, these resources might go to in-
vestments in capital in the regulated industry, or may
go into investments in other industries, depending
upon relative rates of return on those investments. This
will affect the processes of capital formation and deep-
ening.
Second, the CAA placed restrictions on new
sources of emissions. When making investment deci-
sions, firms take into account the additional cost of
pollution abatement equipment. Effectively, the
7 Regulation may also affect the rate of investment, and change the rate of capital accumulation.
A-7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
"price" of investment goods is higher because more
units of capital are required to produce the same
amount of output. In the no-control scenario, there
are no restrictions on new sources and hence no re-
quirements for pollution control expenditures. Effec-
tively, the "price" of investment goods is lower. Thus,
at each point in time, investors are faced with a lower
price of investment goods. This results in a different
profile for investment over time.
Operating and Maintenance Costs - Stationary
Sources
In addition to purchasing pollution abatement
equipment, firms incurred costs to run and maintain
the pollution abatement equipment. In the no-control
scenario, resources used to pay for these operating
and maintenance (O&M) costs are freed up for other
uses. The model assumes that the resources required
to run and maintain pollution control equipment are
in the same proportions as the factor inputs used in
the underlying production technology. For example,
if 1 unit of labor and 2 units of materials are used to
produce 1 unit of output, then one-third of pollution
control O&M costs are allocated to labor and two-
thirds are allocated to materials. These adjustments
were introduced at the sector level. O&M expendi-
tures are exclusive of depreciation charges and offset
by any recovered costs.
Capital Costs - Mobile Sources
Capital costs associated with pollution control
equipment were represented by changing costs for
motor vehicles (sector 24) and other transportation
equipment (sector 26). Prices (unit costs) were reduced
in proportion to the value of the pollution control de-
vices contained in cars, trucks, motorcycles, and air-
craft.
Operating and Maintenance - Mobile Sources
Prices for refined petroleum products (sector 16)
were changed to reflect the resource costs associated
with producing unleaded and reduced lead gasoline
(fuel price penalty), the change in fuel economy for
vehicles equipped with pollution control devices (fuel
economy penalty), and the change in fuel economy
due to the increased fuel density of lower leaded and
no lead gasoline (fuel economy credit). Third, inspec-
tion and maintenance costs and a maintenance credit
associated with the use of unleaded and lower leaded
(i.e., unleaded and lower leaded gasoline is less cor-
rosive, and therefore results in fewer muffler replace-
ments, less spark plug corrosion, and less degrada-
tion of engine oil) were represented as changes in
prices for other services (sector 34).
Direct Compliance Expenditures
Data
Sources of Cost Data
Cost data for this study are derived primarily from
the 1990 Cost of Clean report. EPA publishes cost
data in response to requirements of the Clean Air and
Clean Water Acts. The following subsections describe
Cost of Clean data in detail, as well as adjustments
made to the data and data from other sources.
Cost of Clean Data
EPA is required to compile and publish public
and private costs resulting from enactment of the Clean
Air Act and the Clean Water Act. The 1990 Cost of
Clean report presents estimates of historical pollution
control expenditures for the years 1972 through 1988
and projected future costs for the years 1989 through
2000. This includes federal, state, and local govern-
ments as well as the private sector. Estimates of capi-
tal costs, operation and maintenance (O&M) costs,
and total annualized costs for five categories of envi-
ronmental media, including air, water, land, chemi-
cal, and multi-media, are presented. It should be noted
that these estimates represent direct regulatory imple-
mentation and compliance costs rather than social
costs. The Cost of Clean relied on data from two gov-
ernmental sources, the EPA and the U.S. Department
of Commerce (Commerce).
EPA Data
EPA expenditures were estimated from EPA bud-
get justification documents.8 Estimates of capital and
operating costs resulting from new and forthcoming
regulations were derived from EPA's Regulatory Im-
pact Analyses (RIAs). RIAs have been prepared prior
to the issuance of all major regulations since 1981.
Finally, special analyses conducted by EPA program
offices or contractors were used when other data
sources did not provide adequate or reliable data.
8 The main source of data for EPA expenditures is the Justification of Appropriation Estimates for Committee on Appropriations.
A-8
-------
Appendix A: Cost andMacroeconomic Modeling
Commerce Data
Data collected by Commerce were used exten-
sively in the Cost of Clean for estimates of historical
pollution control expenditures made by government
agencies other than EPA and by the private sector.
Two Commerce agencies, the Bureau of Economic
Analysis (BEA) and the Bureau of the Census (Cen-
sus), have collected capital and operating costs for
compliance with environmental regulations since the
early 1970's. Commerce is, in fact, the primary source
of original survey data for environmental regulation
compliance costs. Commerce publishes a number of
documents that report responses to surveys and com-
prise most of the current domain of known pollution
abatement and control costs in the United States, in-
cluding:
A series of articles entitled "Pollution Abate-
ment and Control Expenditures" published
annually in the Survey of Current Business
by BEA (BEA articles);
A series of documents entitled "Pollution
Abatement Costs and Expenditures" pub-
lished annually in the Current Industrial Re-
ports by Census (PACE reports); and
• A series of documents entitled Government
Finances published annually by Census (Gov-
ernment Finances).
BEA articles contain data derived from a number
of sources, including two key agency surveys —the
"Pollution Abatement Costs and Expenditures Sur-
vey" (PACE Survey) and the "Pollution Abatement
Plant and Equipment Survey" (PAPE Survey)—
which are conducted annually by Census for BEA.
Data have been reported for 1972 through 1987.9
PACE reports have been published annually since
1973 with the exception of 1987. Figures for 1987
were estimated on the basis of historical shares within
total manufacturing. These reports contain expendi-
ture estimates derived from surveys of about 20,000
manufacturing establishments. Pollution abatement
expenditures for air, water and solid waste are reported
by state and Standard Industrial Code (SIC) at the four-
digit level. According to Census, surveys conducted
since 1976 have not included establishments with
fewer than 20 employees because early surveys
showed that they contributed only about 2 percent to
the pollution estimates while constituting more than
10 percent of the sample size.
Each year Census conducts a survey of state, lo-
cal, and county governments; and survey results are
published in Government Finances. Census asks gov-
ernment units to report revenue and expenditures, in-
cluding expenditures for pollution control and abate-
ment.
Non-EPA Federal expenditures were estimated
from surveys completed by federal agencies detailing
their pollution control expenditures, which are sub-
mitted to BEA. Private sector air pollution control
expenditures, as well as state and local government
air pollution expenditures, were taken from BEA ar-
ticles.
Stationary Source Cost Data
Capital Expenditures Data
Capital expenditures for stationary air pollution
control are made by factories and electric utilities for
plant and equipment that abate pollutants through end-
of-line (EOL) techniques or that reduce or eliminate
the generation of pollutants through changes in pro-
duction processes (CIPP). For the purposes of this
report EOL and CIPP expenditures are aggregated.10
Table A-3 summarizes capital expenditures for sta-
tionary air pollution control, categorized as "nonfarm
business" or "government enterprise" expenditures.
Nonfarm business capital expenditures consist of
plant and equipment expenditures made by 1) manu-
facturing companies, 2) privately and cooperatively
owned electric utilities, and 3) other nonmanufacturing
companies. "Government enterprise" is, according to
BEA, an agency of the government whose operating
costs, to a substantial extent, are covered by the sale
of goods and services. Here, government enterprise
means specifically government enterprise electric
9 The most recent BEA article used as a source for air pollution control costs in the Cost of Clean was "Pollution Abatement and
Control Expenditures, 1984-87" in Survey of Current Business, June 1989.
10 Survey respondents to the Census annual Pollution Abatement Surveys report the difference between expenditures for CIPP
and what they would have spent for comparable plant and equipment without pollution abatement features. Disaggregated capital
expenditures by private manufacturing establishments can be found in annual issues of Census reports.
A-9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic A-3. Estimated Capital and O&M
Expenditures for Stationary Source Air
Pollution Control (millions of current dollars).
Nont'a im
Government
Business
Enterprise
Year
Cap a
(WMb
Cap c
Otf.M"
1972
2,172
63
1973
2,968
1,407
82
29
1974
3,328
1,839
104
56
1975
3,914
2,195
102
45
1976
3,798
2,607
156
58
1977
3,811
3,163
197
60
1978
3,977
3,652
205
72
1979
4,613
4,499
285
106
1980
5,051
5,420
398
148
1981
5,135
5,988
451
135
1982
5,086
5,674
508
141
1983
4,155
6,149
422
143
1984
4,282
6,690
416
147
1985
4,141
6,997
328
189
1986
4,090
7,116
312
140
1987
4,179
7,469
277
130
1988
4,267
7,313
243
161
1989
4,760
7,743
235
173
1990
4,169
8,688
226
154
Sources:
a. Non-farm capital expenditures for 1972-87 are from Cost
of Clean T able B-l, line 2.
b. Non-farm O&M expenditures for 1973-85 are from Cost
of Clean, Table B-l, line 8.
c. Government enterprise capital expenditures for 1972-87
are from Cost of Clean,J able B-9, line 1.
d. Government enterprise O&M expenditures for 1973-85
are from Cost of Clean, T able B-9, line 5.
All other reported expenditures are EPA estimates.
utilities. Government enterprise capital expenditures
are pollution abatement expenditures made by pub-
licly owned electric utilities.11
Operation and Maintenance Expenditures Data
Stationary source O&M expenditures are made
by manufacturing establishments, private and public
electric utilities, and other nonmanufacturing busi-
nesses to operate air pollution abatement equipment.
O&M expenditures for electric utilities are made up
of two parts: 1) expenditures for operating air pollu-
tion equipment and 2) the additional expenditures as-
sociated with switching to alternative fuels that have
lower sulfur content (fuel differential). Expenditures
to operate air pollution abatement equipment are for
the collection and disposal of flyash, bottom ash, sul-
fur and sulfur products, and other products from flue
gases.12 O&M expenditures are net of depreciation
and payments to governmental units, and are summa-
rized in Table A-3. O&M data were disaggregated to
the two digit SIC level for use in the macroeconomic
model.
For both capital and O&M expenditures, histori-
cal survey data were not available for each year
through 1990 prior to publication of Cost of Clean.
For the purpose of the section 812 analysis, EPA pro-
jected 1988-1990 capital expenditures and 1986-1990
O&M expenditures. Those projections were used in
the macroeconomic simulation, and have been retained
as cost estimates to ensure consistency between the
macroeconomic results and the direct cost estimates.
Since completion of the macroeconomic modeling,
however, BEA has published expenditure estimates
through 1990. A comparison of more recent BEA es-
timates with the EPA projections used in the section
812 analysis can be found in the "Uncertainties in the
Cost Analysis" section, below.
Recovered Costs
"Recovered costs" are costs recovered (i.e., rev-
enues realized) by private manufacturing establish-
ments through abatement activities. According to in-
structions provided to survey participants by Census,
recovered costs consist of 1) the value of materials or
energy reclaimed through abatement activities that
were reused in production and 2) revenue that was
obtained from the sale of materials or energy reclaimed
through abatement activities. Estimates of recovered
costs were obtained from the PACE reports and are
summarized in Table A-4. In this analysis, recovered
costs were removed from total stationary source air
pollution control O&M costs — that is, net O&M cost
in any year would be O&M expenditures (see Table
A-3) less recovered costs. Recovered cost data were
disaggregated to the two digit SIC level for use in the
macroeconomic model.
11 BEA calculates these expenditures using numbers obtained from Energy Information Agency (EIA) Form 767 on steam-
electric plant air quality control.
12 Farber, Kit D. and Gary L. Rutledge, "Pollution Abatement and Control Expenditures: Methods and Sources for Current-
Dollar Estimates," Unpublished paper, Bureau of Economic Analysis, U.S. Department of Commerce, October 1989.
A-10
-------
Appendix A: Cost andMacroeconomic Modeling
Tabic A-4. Estimated Recovered Costs lor
Stationary Source Air Pollution Control
(millions of current dollars).
Vi'iir
PACF*
F.sti m nti'il
1972
248
1973
199
1974
296
1975
389
1976
496
1977
557
1978
617
1979
750
750
1980
862
862
1981
1,000
997
1982
858
857
1983
822
822
1984
866
870
1985
767
768
1986
860
867
1987
987
1988
1,103
1,107
1989
1,122
1990
1,256
* Air cost recovered as reported in PACE
Source: "Pollution Abatement Costs and
Expenditures" published annually in the Current
Industrial Reports by Census.
Mobile Source Cost Data
Costs of controlling pollution emissions from
motor vehicles were estimated by calculating the pur-
chase price and O&M cost premiums associated with
vehicles equipped with pollution abatement controls
over the costs for vehicles not equipped with such
controls. These costs were derived using EPA analy-
ses, including EPA RIAs, the Cost of Clean, and other
EPA reports.13 This Appendix summarizes the sec-
tion 812 mobile source compliance cost estimates and
provides references to published data sources where
possible. Further information on specific methods,
analytical steps, and assumptions can be found in
McConnell etal. (1995),14 which provides a detailed
description of the section 812 mobile source cost es-
timation exercise and compares the method and re-
sults to other similar analyses (including Cost of Clean
(1990)).
Capital Expenditures Data
Capital expenditures for mobile source emission
control are associated primarily with pollution abate-
ment equipment on passenger cars, which comprise
the bulk of all mobile sources of pollution. These capi-
tal costs reflect increasingly stringent regulatory re-
quirements and improvements in pollution control
technologies overtime. Each of the following devices
have been used at one time or another dating back to
the Clean Air Act Amendments of 1965: air pumps,
exhaust-gas recirculation valves, high altitude con-
trols, evaporative emissions controls, and catalysts.
The cost estimates for each component were computed
on a per-vehicle basis by engineering cost analyses
commissioned by EPA. The resulting per-vehicle capi-
tal costs were multiplied by vehicle production esti-
mates to determine annual capital costs. Table A-5
summarizes mobile source capital costs.
Operation and Maintenance Expenditures Data
Costs for operation and maintenance of emission
abatement devices include the costs of maintaining
pollution control equipment plus the cost of vehicle
inspection/maintenance programs. Operating costs per
vehicle were multiplied by total vehicles in use to
determine annual cost. Mobile source O&M costs are
made up of three factors: 1) fuel price penalty, 2) fuel
economy penalty, and 3) inspection and maintenance
program costs as described below. These costs are
mitigated by cost savings in the form of maintenance
economy and fuel density economy. Table A-6 sum-
marizes mobile source O&M expenditures and cost
savings by categories, with net O&M costs summa-
rized above in Table A-5. The following sections de-
scribe the components of the mobile source O&M cost
estimates.
Fuel Price Penalty
Historically, the price of unleaded fuel has been
several cents per gallon higher than the price of leaded
fuel. CAA costs were calculated as the difference be-
13 A complete listing of sources used in calculating mobile source capital and operating expenditures can be found in Environ-
mental Investments: The Cost of a Clean Environment, Report of the Administrator of the Environmental Protection Agency to the
Congress of the United State, EPA-230-11-90-083, November 1990.
14 Evaluating the Cost of Compliance with Mobile Source Emission Control Requirements: Retrospective Analysis, Resources
for the Future Discussion Paper, 1995. Note that McConnell et al. refer to the section 812 estimates as: CostofClean (1993,unpub-
lished).
A-ll
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic A-5. Estimated Capital and
Operation and Maintenance Expenditures
lor Mobile Source Air Pollution Control
(millions of current dollars).
Year
Capital"
O&M"1
1973
276
1,765
1974
242
2,351
1975
1,570
2,282
1976
1,961
2,060
1977
2,248
1,786
1978
2,513
908
1979
2,941
1,229
1980
2,949
1,790
1981
3,534
1,389
1982
3,551
555
1983
4,331
-155
1984
5,679
-326
1985
6,387
337
1986
6,886
-1,394
1987
6,851
-1,302
1988
7,206
-1,575
1989
7,053
-1,636
1990
7,299
-1,816
Sources:
a. Capital exp.: Cost ofClean^ Tables C-2 to C-9, line 3
on each; Tables C-2Ato C-9A, line 10 on each; converted
from $198 6 to current dollars.
b. O&M exp.: EPA analyses based on sources and
methods in: Costs and Benefits of Reducing Lead in
Gasoline: Final Regulatory Impact Analysis^ U.S.
EnvironmentalProtection Agency, Office of Fblicy
Analysis, EPA-230-05-85-006, February 1985; and Cost
of Clean.
tween the cost of making unleaded gasoline and leaded
gasoline with lower lead levels and the cost of mak-
ing only leaded gasoline with a lead content set at
pre-regulatory levels. These cost estimates were de-
veloped using a linear programming model of the re-
finery industry. Prices of crude oil and other unfin-
ished oils, along with the prices of refinery outputs,
were adjusted annually according to price indices for
imported crude oil over the period of analysis. The
relative shares of leaded and unleaded gasoline and
the average lead content in leaded gasoline also were
adjusted annually according to the historical record.
These estimates may tend to understate costs due
to a number of biases inherent in the analysis process.
For example, the refinery model was allowed to opti-
mize process capacities in each year. This procedure
is likely to understate costs because regulatory require-
ments and market developments cannot be perfectly
anticipated overtime. This procedure resulted in esti-
mates that are about ten percent less than estimates in
other EPA reports.15 However, new process technolo-
gies that were developed in the mid-1980s were not
reflected in either the base case or regulatory case runs.
It is reasonable to expect that regulatory requirements
would have encouraged development of technologies
at a faster rate than would have occurred otherwise.
Fuel Economy Penalty
The fuel economy penalty benefit is the cost as-
sociated with the increased/decreased amount of fuel
used by automobiles with air pollution control devices
(all else being equal). An assumption that can be made
is that the addition of devices, such as catalytic con-
Table A-6. O&M Costs and Credits (millions
of current dollars).
Fu el
Fuel Price
Fcon.
Net
Totii 1
Veil r
Pen nltv
Penultv
\& M*
fusts
1973
91
1700
-26
1765
1974
244
2205
-98
2351
1975
358
2213
-289
2282
1976
468
2106
-514
2060
1977
568
1956
-738
1786
1978
766
1669
-1527
908
1979
1187
1868
-1826
1229
1980
1912
1998
-2120
1790
1981
2181
1594
-2386
1389
1982
2071
1026
-2542
555
1983
1956
628
-2739
-155
1984
2012
313
-2651
-326
1985
3057
118
-2838
337
1986
2505
-40
-3859
-1394
1987
2982
-158
-4126
-1302
1988
3127
-210
-4492
-1575
1989
3476
-318
-4794
-1636
1990
3754
-481
-5089
-1816
* Inspection and maintenance costs less fuel density savings
and maintenance savings.
Sources: All results are presented in Jorgenson et al. (1993),
pg. A 17. FPP results are based on a petroleum refinery cost
model run for the retrospective analysis. FEP and Net I&M
are based on data and methods from Costs and Benefits of
Reducing Lead in Gasoline: Final Regulatory Impact
Analysis, U.S. Environmental Protection Agency, Office of
Policy Analysis, EPA-230-05-85-006, February 1985; and
Cost of Clean (1990). Specific mi alytic procedures are
summarized in McConnell et al. (1995).
15 Costs and Benefits of Reducing Lead in Gasoline: Final Regulatory Impact Analysis, U.S. Environmental Protection Agency,
Office of Policy Analysis, EPA-230-05-85-006, February 1985.
A-12
-------
Appendix A: Cost andMacroeconomic Modeling
verters, decrease automobile fuel efficiency.16 If this
assumption is true, air pollution control devices in-
crease the total fuel cost to consumers. An alternative
assumption is that the use of catalytic converters has
increased fuel economy. This increase has been at-
tributed in large measure to the feedback mechanism
built into three-way catalytic converters.17 Under this
assumption, the decrease in total fuel cost to consum-
ers is considered a benefit of the program.
For the purposes of this study, sensitivity analy-
ses were performed using data presented in the Cost
of Clean report. These analyses were conducted to
evaluate the significance of assumptions about the
relationship between mile per gallon (MPG) values
for controlled automobiles and MPG values for un-
controlled cars. Based on results of these and other
analyses, fuel economy was assumed to be equal for
controlled and uncontrolled vehicles from 1976 on-
ward. This may bias the cost estimates although in an
unknown direction.
Inspection and Maintenance Programs
Inspection and maintenance programs are admin-
istered by a number of states. Although these programs
are required by the Clean Air Act, the details of ad-
ministration were left to the discretion of state or lo-
cal officials. The primary purpose of inspection and
maintenance programs is to identify cars that require
maintenance —including cars that 1) have had poor
maintenance, 2) have been deliberately tampered with
or had pollution control devices removed, or 3) have
used leaded gasoline when unleaded is required— and
force the owners of those cars to make necessary re-
pairs or adjustments.18 Expenditures for inspection and
maintenance were taken from the Cost of Clean.
Beneficial effects of the mobile source control
program associated with maintenance and fuel den-
sity were also identified. These cost savings were in-
cluded in this study as credits to be attributed to the
mobile source control program. Credits were estimated
based on an EPA study,19 where more detailed expla-
nations may be found.
Maintenance Credits
Catalytic converters require the use of unleaded
fuel, which is less corrosive than leaded gasoline. On
the basis of fleet trials, the use of unleaded or lower
leaded gasoline results in fewer muffler replacements,
less spark plug corrosion, and less degradation of en-
gine oil, thus reducing maintenance costs. Mainte-
nance credits account for the majority of the direct
(non-health) economic benefits of reducing the lead
concentration in gasoline.
Fuel Density Credits
The process of refining unleaded gasoline in-
creases its density. The result is a gasoline that has
higher energy content. Furthermore, unleaded gaso-
line generates more deposits in engine combustion
chambers, resulting in slightly increased compression
and engine efficiency. Higher energy content of un-
leaded gasoline and increased engine efficiency from
the used of unleaded gasoline yield greater fuel
economy and therefore savings in refining, distribu-
tion, and retailing costs.
Other Direct Cost Data
The Cost of Clean report includes several other
categories of cost that are not easily classified as ei-
ther stationary source or mobile source expenditures.
Federal and state governments incur air pollution
abatement costs; additionally, federal and state gov-
ernments incur costs to develop and enforce CAA
regulations. Research and development expenditures
by the federal government, state and local govern-
ments, and (especially) the private sector can be at-
tributed to the CAA. These data are summarized by
year in Table A-7.
Unlike the other private sector expenditure data
used for this analysis, the survey data used as a source
for private sector R&D expenditures cannot be disag-
gregated into industry-specific expenditure totals.
Consequently, private sector R&D expenditures are
16 Memo from Joel Schwartz (EPA/OPPE) to Joe Somers and Jim DeMocker dated December 12, 1991, and entitled "Fuel
Economy Benefits." Schwartz states that since this analysis is relative to a no Clean Air Act baseline, not a 1973 baseline, fuel
economy benefits are not relevant. In the absence of regulation, tuning of engines for maximum economy would presumably be
optimal in the base case as well.
17 Memo from Joseph H. Somers, EPA Office of Mobile Sources, to Anne Grambsch (EPA/OPPE) and Joel Schwartz (EPA/
OPPE) entitled "Fuel Economy Penalties for section 812 Report," December 23, 1991.
18 Walsh, Michael P., "Motor Vehicles and Fuels: The Problem," EPA Journal, Vol. 17, No. 1, January/February 1991, p. 12.
19 Schwartz, J., etal. Costs and Benefits of Reducing Lead in Gasoline: Final Regulatory Impact Analysis, U. S. Environmental
Protection Agency, Economic Analysis Division, Office of Policy Analysis, February 1985.
A-13
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic A-7. Other Air Pollution Control Expenditures (millions of
current dollars).
Regulations
Research
Year
Aim foment
anil Monitoring
and Development
Total
State &
State &
State &
Fed.'
T,oca1b
Fed.0
Focal11
Private6
Fed.f Focal8
1973
47
0
50
115
492
126
6
836
1974
56
0
52
131
520
100
7
866
1975
88
1
66
139
487
108
8
897
1976
105
1
69
135
562
131
6
1.009
1977
106
1
80
161
675
144
7
1.174
1978
90
0
93
183
805
146
8
1.325
1979
103
0
100
200
933
105
7
1.448
1980
95
0
122
207
851
130
5
1.410
1981
85
0
108
226
798
131
0
1.348
1982
87
0
93
230
761
126
2
1.229
1983
136
4
88
239
691
133
6
1.297
1984
115
14
101
250
665
165
4
1.314
1985
98
12
103
250
775
247
3
1.488
1986
67
14
106
307
833
217
4
1.548
1987
80
15
110
300
887
200
2
1.594
1988
65
10
120
320
934
220
1
1.670
1989
70
12
130
360
984
230
2
1.788
1990
71
13
133
343
749
231
2
1,542
Sources:
a. Federal government abatement expenditures: 1973-82, "Pollution Abatement and Control
Expenditures", Survey of Current Business fBF.A^ July 1986 Table 9 line 13; 1983-87, BEA
June 1989 Table 71ine 13; 1988-90, BEAMay 1995 Table 7 line 13.
b. State and local abatement expenditures: 1973-87, Cost of Clean, Table B-9 line 2; 1988-90,
BEAMay 1995 Table7 line 14.
c. Federal government "regs/monitoring" expenditures: 1973-82, BEA July 1986, Table 9 line
17; 1983-87, BEA June 1989 Table 6 line 17; 1988-90, BEAMay 1995Table 7 line 17.
d. State and local government "regs/monitoring" expenditures: 1973-87, Cost of Clean, T able
B-9 line 3; 1988-90, BEA May 1995 Table 7 line 18.
e. Private sector R&D expenditures: 1973-86, BEA May 1994 Table 4 (no line #) [total R&D
expenditures in $1987 are converted to current dollars using the GDP price deflator series found
elsewhere in this Appendix -- netting outpublic sector R&D leaves private sector expenditures];
1987-90, BEAMay 1995 Table 7 line 20.
f Federal governmentR&D expenditures: 1973-82, BEA July 1986 Table 9 line 21; 1983-87,
BEA June 1989Table 6 line 21; 1988-90, BEAMay 1995, Table 7 line 21.
g. State and local governmentR&D expenditures: 1973-87, Cost of Clean, Table B-9 line 4;
1988-90, BEAMay 1995 Table 7 line 22.
from more recent issues of the Survey of
Current Business (BEA). Federal govern-
ment expenditures are from BEA (various
issues). Private R&D expenditures were
reported in Cost of Clean. Since publica-
tion of Cost of Clean, however, BEA has
revised its private sector R&D expenditure
series (BEA, 1994 and 1995). Since private
R&D expenditures were not included in the
macroeconomic modeling exercise, the re-
vised series can be (and has been) used
without causing inconsistency with other
portions of the section 812 analysis.
Assessment Results
Compliance Expenditures and
Costs
Compliance with the CAA imposed
direct costs on businesses, consumers, and
governmental units, and triggered other
expenditures such as governmental regula-
tion and monitoring costs and expenditures
for research and development by both gov-
ernment and industry. As shown in Table
A-8, annual CAA compliance expenditures
- including R&D, etc.- over the period
from 1973 to 1990 were remarkably
stable20, ranging from about $20 billion to
$25 billion in inflation-adjusted 1990 dol-
lars (expenditures are adjusted to 1990 dol-
lars through application of the GDP Implicit
Price Deflator). This is equal to approxi-
mately one third of one percent of total
domestic output during that period, with the
percentage falling from one half of one per-
cent of total output in 1973 to one quarter
of one percent in 1990.
omitted from the macroeconomic modeling exercise
(the macro model is industry-specific). The R&D ex-
penditures are, however, included in aggregate cost
totals used in the benefit-cost analysis.
The Cost of Clean and the series of articles "Pol-
lution Abatement and Control Expenditures" in the
Survey of Current Business (various issues) are the
data sources for "Other Air Pollution Control Expen-
ditures." State and local expenditures through 1987
are found in Cost of Clean; 1988-90 expenditures are
Although useful for many purposes, a summary
of direct annual expenditures is not the best cost mea-
sure to use when comparing costs to benefits. Capital
expenditures are investments, generating a stream of
benefits (and opportunity cost) over the life of the in-
vestment. The appropriate accounting technique to use
for capital expenditures in a cost/benefit analysis is to
annualize the expenditure — i.e., to spread the capi-
tal cost over the useful life of the investment, apply-
ing a discount rate to account for the time value of
money.
20 While total expenditures remained relatively constant over the period, the sector-specific data presented in Tables A-3 and A-5
above indicate that capital expenditures for stationary sources fell significantly throughout the period but that this decline was offset
by significant increases in mobile source capital expenditures.
AA4
-------
Tabic A-8. Summary of Expenditures and Conversion to 1990 Dollars (millions of dollars).
CURRENT YEAR DOLLARS
1990 DOLLARS
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
Stationary
K
2,235
3,050
3,432
4,016
3,954
4,008
4,182
4,898
5,449
5,586
5,594
4,577
4,698
4,469
4,402
4,456
4,510
4,995
4,395
O&M
na
1,436
1,895
2,240
2,665
3,223
3,724
4,605
5,568
6,123
5,815
6,292
6,837
7,186
7,256
7,599
7,474
7,916
8,842
Rec.
Costs
na
199
296
389
496
557
617
750
862
997
857
822
870
768
867
987
1,107
1,122
1,256
Mobile
K
na
276
242
1,570
1,961
2,248
2,513
2,941
2,949
3,534
3,551
4,331
5,679
6,387
6,886
6,851
7,206
7,053
7,312
Source
O&M
na
1,765
2,351
2,282
2,060
1,786
908
1,229
1,790
1,389
555
(155)
(326)
337
(1,394)
(1,302)
(1,575)
(1,636)
(1,816)
Other
na
836
866
897
1,009
1,174
1,325
1,448
1,410
1,348
1,299
1,297
1,314
1,488
1,548
1,594
1,670
1,788
1,542
TOTAL
EXP
na
7,164
8,490
10,616
11,153
11,882
12,035
14,371
16,304
16,983
15,957
15,520
17,332
19,099
17,831
18,211
18,178
18,994
19,019
GDP
price
defl.
38.8
41.3
44.9
49.2
52.3
55.9
60.3
65.5
71.7
78.9
83.8
87.2
91
94.4
96.9
100
103.9
108.5
113.2
Stationary
Rec.
Mobile Source
TOTA
K
O&M
Costs
K
O&M
Other
EXP
6,521
8,360
3,936
545
756
4,838
2,290
19,635
8,653
4,778
746
610
5,927
2,184
21,405
9,240
5,154
895
3,612
5,250
2,063
24,425
8,558
5,768
1,074
4,244
4,459
2,183
24,139
8,116
6,527
1,128
4,552
3,617
2,378
24,062
7,851
6,991
1,158
4,718
1,705
2,487
22,593
8,465
7,959
1,296
5,083
2,124
2,503
24,837
8,603
8,791
1,361
4,656
2,826
2,226
25,741
8,014
8,785
1,430
5,070
1,993
1,935
24,367
7,557
7,855
1,158
4,797
750
1,755
21,555
5,942
8,168
1,067
5,622
(201)
1,684
20,148
5,844
8,505
1,082
7,064
(406)
1,634
21,560
5,359
8,617
921
7,659
404
1,785
22,903
5,142
8,477
1,013
8,044
(1,628)
1,809
20,831
5,044
8,602
1,117
7,755
(1,474)
1,804
20,615
4,914
8,143
1,206
7,851
(1,716)
1,819
19,805
5,211
8,259
1,171
7,359
(1,707)
1,865
19,817
4,395
8,842
1,256
7,312
(1,816)
1,542
19,019
K = Capital expenditures: O&M = Operation and Maintenance expenditures.
Ree. Costs = recovered costs. Total expenditures arc the sum of stationary source, mobile source, and "other" expenditures, less recovered costs.
Stationary source expenditures arc the sum of"Nonrarin Business" and "Government Enterprise" expenditures (from Table A-3).
To calculate expenditures in 1990 dollars, current year expenditures are multiplied bv the ratio of the 1990 price deflator to the current year deflator. For example.
1989 expenditures arc multiplied by (l 13.2/108.5).
Source for price deflator scries: Economic Report of the President. February 1995. Table B-3.
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Annualization Method
For this cost/benefit analysis, all capital expendi-
tures have been annualized at 3 percent, 5 percent,
and 7 percent (real) rates of interest. Therefore, "an-
nualized" costs reported for any given year are equal
to O&M expenditures (plus R&D, etc., expenditures,
minus recovered costs) plus amortized capital costs
(i.e., depreciation plus interest costs associated with
the pre-existing capital stock) for that year. Station-
ary source air pollution control capital costs are am-
ortized over twenty years; mobile source air pollution
control costs are amortized over ten years. Capital
expenditures are amortized using the formula for an
annuity [that is, r/(l-(l+r)_t) , where r is the rate of
interest and t is the amortization period].21 Multiply-
ing the expenditure by the appropriate annuity factor
gives a constant annual cost to be incurred fort years,
the present value of which is equal to the expenditure.
Due to data limitations, the cost analysis for this
CAA retrospective starts in 1973, missing costs in-
curred in 1970-72. Cost of Clean, however, includes
stationary source capital expenditures for 1972. In this
analysis, amortized costs arising from 1972
capital investments are included in the 1973 -
1990 annualized costs, even though 1972
costs are not otherwise included in the analy-
sis. Conversely, only a portion of the (e.g.)
1989 capital expenditures are reflected in the
1990 annualized costs — the remainder of
the costs are spread through the following
two decades, which fall outside of the scope
of this study (similarly, benefits arising from
emission reductions in, e.g., 1995 caused by
1990 capital investments are not captured
by the benefits analysis). Table A-9 presents
CAA compliance costs from 1973 to 1990,
in 1990 dollars, with capital expenditures
amortized at a five percent real interest rate.
"Total" costs are the sum of stationary
source, mobile source, and "other" costs,
minus recovered costs.
Tables A-10 and A-ll provide details
of the amortization calculation (using a five
percent interest rate) for stationary sources
and mobile sources, respectively. Similar
calculations were performed to derive the
annualized cost results using discount rates
of three percent and seven percent.
The Stationary Source table reports a capital ex-
penditure of $6,521 million for 1972 (in 1990 dol-
lars). The cost is spread over the following twenty
years (which is the assumed useful life of the invest-
ment) using a discount rate of five percent; thus, the
amortization factor to be used is f(20)=0.0802. Mul-
tiplying $6,521 million by 0.0802 gives an annuity of
$523 million. That annuity is noted on the first data
row of the table, signifying that the 1972 expenditure
of $6,521 million implies an annual cost of $523 mil-
lion for the entire twenty-year period of 1973 to 1992
(the years following 1990 are not included on the
tables, since costs incurred in those years are not in-
cluded in this retrospective assessment). The first sum-
mary row near the bottom of the table (labeled "SUM")
reports aggregate annualized capital costs: for 1973
(the first data column), capital costs are $523 million.
Capital expenditures in 1973 amounted to $8,360
million. Using the amortization technique explained
above, one can compute an annualized cost of $671
million, incurred for the twenty-year period of 1974
to 1993. Aggregate annualized capital costs for 1974
include cost flows arising from 1972 and 1973 invest-
Tablc A-9. Annualized Costs. 1973-1990 (millions of 1990
dollars: capital expenditures annualized at 5 percent).
Stationary
rec.
Mobile Source
K
O&M
rosts
K
O&M
other
Total
1973
523
3,936
545
0
4,838
2,290
11,042
1974
1,194
4,778
746
98
5,927
2,184
13,435
1975
1,888
5,154
895
177
5,250
2,063
13,638
1976
2,630
5,768
1,074
645
4,459
2,183
14,611
1977
3,317
6,527
1,128
1,194
3,617
2,378
15,904
1978
3,968
6,991
1,158
1,784
1,705
2,487
15,776
1979
4,598
7,959
1,296
2,395
2,124
2,503
18,282
1980
5,277
8,791
1,361
3,053
2,826
2,226
20,812
1981
5,967
8,785
1,430
3,656
1,993
1,935
20,905
1982
6,610
7,855
1,158
4,313
750
1,755
20,125
1983
7,217
8,168
1,067
4,934
(201)
1,684
20,734
1984
7,694
8,505
1,082
5,564
(406)
1,634
21,909
1985
8,163
8,617
921
6,400
404
1,785
24,447
1986
8,593
8,477
1,013
6,924
(1,628)
1,809
23,161
1987
9,005
8,602
1,117
7,416
(1,474)
1,804
24,237
1988
9,410
8,143
1,206
7,831
(1,716)
1,819
24,281
1989
9,804
8,259
1,171
8,237
(1,707)
1,865
25,288
1990
10,222
8,842
1,256
8,531
(1,816)
1,542
26,066
Source: Stationary source capital costs and mobile source capital costs are from
Tables A-10 and A-ll, respectively. All other costs and offsetsare from Table
A-8.
21 Using an interest rate of five percent, the factor for a twenty year amortization period is 0.0802; that for a ten year amortiza-
tion period is 0.1295.
A-16
-------
Tabic A-10. Amortization of Capital Expenditures for Stationary Sources (millions of 1990 dollars).
EXPEND
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1972
6,521
523
523
523
523
523
523
523
523
523
523
523
523
523
523
523
523
523
523
1973
8,360
671
671
671
671
671
671
671
671
671
671
671
671
671
671
671
671
671
1974
8,653
694
694
694
694
694
694
694
694
694
694
694
694
694
694
694
694
1975
9,240
741
741
741
741
741
741
741
741
741
741
741
741
741
741
741
1976
8,558
687
687
687
687
687
687
687
687
687
687
687
687
687
687
1977
8,116
651
651
651
651
651
651
651
651
651
651
651
651
651
1978
7,851
630
630
630
630
630
630
630
630
630
630
630
630
1979
8,465
679
679
679
679
679
679
679
679
679
679
679
1980
8,603
690
690
690
690
690
690
690
690
690
690
1981
8,014
643
643
643
643
643
643
643
643
643
1982
7,557
606
606
606
606
606
606
606
606
1983
5,942
477
477
477
477
477
477
477
1984
5,844
469
469
469
469
469
469
1985
5,359
430
430
430
430
430
1986
5,142
413
413
413
413
1987
5,044
405
405
405
1988
4,914
394
394
1989
5,211
418
1990
4,395
SUM
523
1,194
1,888
2,630
3,317
3,968
4,598
5,277
5,967
6,610
7,217
7,694
8,163
8,593
9,005
9,410
9,804
10,222
Expenditures
8,360
8,653
9,240
8,558
8,116
7,851
8,465
8,603
8,014
7,557
5,942
5,844
5,359
5,142
5,044
4,914
5,211
4,395
K stock
6,521
14,880
23,533
32,773
41,331
49,448
57,299
65,763
74,366
82,381
89,937
95,879
101,723
107,082
112,225
117,269
122,182
127,394
K stock net depr.
6,521
14,684
22,876
31,372
38,869
45,612
51,776
58,232
64,469
69,740
74,173
76,606
78,587
79,713
80,249
80,300
79,819
79,217
Int
326
734
1,144
1,569
1,943
2,281
2,589
2,912
3,223
3,487
3,709
3,830
3,929
3,986
4,012
4,015
3,991
3,961
Depr
197
460
745
1,061
1,373
1,687
2,009
2,365
2,744
3,123
3,508
3,863
4,233
4,607
4,993
5,395
5,813
6,262
Capital expenditures for each year arc found in the "EXPEND" column. Expenditures arc amortized over 20 years (i.e.. years (t+1) to (t+20)) using a 5% real interest rate
to derive a constant cost per year for the entire amortization period. The present value (in year t) of the cost flow is equal to the expenditure in year I. Annualized C A A
compliance capital cost for each year (displayed in row "SUM") is the sum of the annuities calculated for capital expenditures from previous years. The capital slock ("K
slock") in place at the start of each year is equal to the sum of expenditures from previous years. Subtracting depreciation from the capital slock leaves "K slock net depr."
Annual interest expense is 5% of net capital slock. Annual interest expense plus depreciation equals annualized compliance cost (row "SUM").
-------
Tabic A-l 1.
Amortization of Capital Expenditures
for Mobile Sources (millions of 1990 dollars).
EXPEND
1974 1975 1976 1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1972
0
1973
756
CD
00
CD
00
CD
00
CD
00
98
98
98
98
98
98
1974
610
79 79 79
79
79
79
79
79
79
79
1975
3,612
468 468
468
468
468
468
468
468
468
468
1976
4,244
550
550
550
550
550
550
550
550
550
550
1977
4,552
590
590
590
590
590
590
590
590
590
590
1978
4,718
611
611
611
611
611
611
611
611
611
611
1979
5,083
658
658
658
658
658
658
658
658
658
658
1980
4,656
603
603
603
603
603
603
603
603
603
1981
5,070
657
657
657
657
657
657
657
657
1982
4,797
621
621
621
621
621
621
621
1983
5,622
728
728
728
728
728
728
1984
7,064
915
915
915
915
915
1985
7,659
992
992
992
992
1986
8,044
1,042
1,042
1,042
1987
7,755
1,004
1,004
1988
7,851
1,017
1989
7,359
1990
7,312
1990
603
657
621
728
915
992
1,042
1,004
1,017
953
SUM
98
177
645
1,194
1,784
2,395
3,053
3,656
4,313
4,934
5,564
6,400
6,924
7,416
7,831
8,237
8,531
Expenditures
610
3,612
4,244
4,552
4,718
5,083
4,656
5,070
4,797
5,622
7,064
7,659
8,044
7,755
7,851
7,359
7,312
K stock
756
1,367
4,979
9,223
13,776
18,493
23,576
28,232
33,302
38,099
42,965
49,419
53,466
57,266
60,469
63,602
65,878
K stock net depr.
756
1,306
4,807
8,647
12,437
15,993
19,480
22,057
24,574
26,287
28,289
31,204
34,023
36,845
39,026
40,997
42,169
Int
38
65
240
432
622
800
974
1,103
1,229
1,314
1,414
1,560
1,701
1,842
1,951
2,050
2,108
Depr
60
112
404
762
1,162
1,595
2,079
2,553
3,084
3,620
4,150
4,840
5,223
5,574
5,880
6,187
6,423
Capital expenditures for each year arc found in the "EXPEND" column. Expenditures arc amortized over 10 years (i.e.. years (1+1) to (1+1 ())) using a 5% real interest rate
to derive a constant cost per year for the entire amortization period. The present value (in year t) of the cost flow is equal to the expenditure in year t. Annualized CAA
compliance capital cost for each year (displayed in row "SUM") is the sum of the annuities calculated for capital expenditures from previous years. The capital stock ("K
stock") in place at the start of each year is equal to the sum of expenditures from the previous ten years. The sum of all previous expenditures less depreciation leaves "K
stock net depr." Annual interest expense is 5% of net capital slock. Annual interest expense plus depreciation equals annualized compliance cost (row "SUM").
-------
Appendix A: Cost andMacroeconomic Modeling
ments: that is, $523 million plus $671 million, or
$1,194 million (see the "SUM" row). Similar calcu-
lations are conducted for every year through 1990, to
derive aggregate annualized capital costs that increase
monotonically from 1973 to 1990, even though capi-
tal expenditures decline after 1975.22
An alternative calculation technique is available
that is procedurally simpler but analytically identical
to that outlined above. Instead of calculating an annu-
ity for each capital expenditure (by multiplying the
expenditure by the annuity factor /). then summing
the annuities associated with all expenditures in pre-
vious years, one can sum all previous expenditures
and multiply the sum (i..e., the capital stock at the
start of the year) by f. The third summary row (la-
beled "K stock") near the bottom of the amortization
summary tables give the pollution control capital stock
at the start of each year. For example, the stationary
sources capital stock in place at the start of 1975 was
$23,533 million (this is the sum of 1972, 1973, and
1974 capital expenditures). Multiplying the capital
stock by the annuity factor 0.0802 gives $1,888 mil-
lion, which is the aggregate annualized stationary
source capital cost for 1975.
One can perform further calculations to decom-
pose the annualized capital costs into "interest" and
"financial depreciation" components.23 For example,
at the start of 1973, the stationary source capital stock
was $6,521 million. A five percent interest rate im-
plies an "interest expense" for 1973 of $326 million.
Given a 1973 annualized cost of $523 million, this
implies a "depreciation expense" for that year of ($523
million minus $326 million =) $197 million. For 1974,
the existing capital stock net of "financial deprecia-
tion" was $14,684 million (that is, the $6,521 million
in place at the start of 1973, plus the investment of
$8,360 million during 1973, minus the depreciation
of $197 million during 1973); five percent of $14,684
million is the interest expense of $734 million. Since
the annualized capital cost for 1974 is $1,194 mil-
lion, depreciation expense is $460 million (i.e., the
difference between annualized cost and the interest
component of annualized cost). This procedure is re-
peated to determine interest and depreciation for each
year through 1990 (see the last three rows of Table A-
11).
The three tables above all present costs (and in-
termediate calculations) assuming a five percent in-
terest rate. As noted above, the Project Team also
employed rates of three percent and seven percent to
calculate costs. Those calculations and intermediate
results are not replicated here. The method employed,
however, is identical to that employed to derive the
five percent results (with the only difference being
the interest rate employed in the annuity factor calcu-
lation). Table A-12 presents a summary of expendi-
tures and annualized costs at the three interest rates.
Tabic A-12. Compliance Expenditures and
Annualized Costs. 1973-1990 ($1990
millions).
A n mi ii liyc (1 f 'osts
Vpar
Kvpiwul.
¦At V>/„
¦At ?»/,,
at 7%
1973
19,635
10,957
11,042
11,134
1974
21,405
13,231
13,435
13,655
1975
24,425
13,314
13,638
13,988
1976
24,139
14,123
14,611
15,139
1977
24,062
15,253
15,904
16,608
1978
22,593
14,963
15,776
16,653
1979
24,837
17,309
18,282
19,331
1980
25,741
19,666
20,812
22,046
1981
24,367
19,590
20,905
22,321
1982
21,555
18,643
20,125
21,720
1983
20,148
19,095
20,734
22,498
1984
21,560
20,133
21,909
23,819
1985
22,903
22,516
24,447
26,523
1986
20,831
21,109
23,161
25,364
1987
20,615
ii,mi
24,237
26,562
1988
19,805
22,012
24,281
26,719
1989
19,817
22,916
25,288
27,836
1990
19,019
23,598
26,066
28,717
Discounting Costs and Expenditures
The stream of costs from 1973 to 1990 can be
expressed as a single cost number by discounting all
costs to a common year. In this analysis, all costs and
benefits are discounted to 1990 (in addition, all costs
and benefits are converted to 1990 dollars, removing
the effects of price inflation)24 There is a broad range
22 Similar calculations were performed for mobile source control capital costs, where the assumed amortization period is ten years.
23 One might, for example, wish to examine the relative importance of the "time value" component of the computed capital costs.
24 Unlike most cost-benefit analyses, where future expected costs and benefits are discounted back to the present, this exercise
brings past costs closer to the present. That is, the discounting procedure used here is actually compounding past costs and benefits.
A-19
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
of opinion in the economics profession regarding the
appropriate discount rate to use in analyses such as
this. Some economists believe that the appropriate rate
is one that approximates the social rate of time pref-
erence — three percent, for example (all rates used
here are "real", i.e., net of price inflation impacts).
Others believe that a rate that approximates the op-
portunity cost of capital (e.g., seven percent or greater)
should be used. A third school of thought holds that
some combination of the social rate of time prefer-
ence and the opportunity cost of capital is appropri-
ate, with the combination effected either by use of an
intermediate rate or by use of a multiple-step proce-
dure which uses the social rate of time preference as
the "discount rate," but still accounts for the cost of
capital. The section 812 Project Team chose to use a
range of discount rates (three, five, and seven per-
cent) for the analysis.
Expenditures and annualized costs discounted to
1990 are found on Table A-13. Expenditures are dis-
counted at all three rates; annualized costs are dis-
counted at the rate corresponding to that used in the
annualization procedure (i.e., the "annualized at 3%"
cost stream is discounted to 1990 at three percent).
The final row presents the result of an explicit combi-
nation of two rates: Capital costs are annualized at
seven percent, then the entire cost stream is discounted
to 1990 at three percent.
Tabic A-13. Costs Discounted to 1990 ($1990
millions).
Wn
7»/n
Expenditures
520,475
627,621
760.75 1
Annualized Costs
416,804
522,906
657.003
Annualizedat 7%
476,329
Indirect Economic Effects of the CAA
In addition to imposing direct compliance costs
on the economy, the CAA induced indirect economic
effects, primarily by changing the size and composi-
tion of consumption and investment flows. Although
this analysis does not add these indirect effects to the
direct costs and include them in the comparison to
benefits, they are important to note. This section sum-
marizes the most important indirect economic effects
of the CAA, as estimated by the J/W macroeconomic
simulation.
GNP and Personal Consumption
Under the no-control scenario, the level of GNP
increases by one percent in 1990 relative to the con-
trol case (see Table A-14). During the period 1973-
1990, the percent change in real GNP rises monotoni-
cally from 0.26 percent to 1.0 percent. The increase
Tabic A-14. Differences in Gross
National Product Between the Control and
No-control Scenarios.
Nominal %.
Real %
Year
Chan ge
Change
1973
-0.09
0.26
1974
-0.18
0.27
1975
-0.10
0.44
1976
-0.00
0.49
1977
-0.10
0.54
1978
-0.16
0.56
1979
-0.16
0.63
1980
-0.14
0.69
1981
-0.14
0.73
1982
-0.19
0.74
1983
-0.19
0.78
1984
-0.17
0.84
1985
-0.12
0.95
1986
-0.14
0.98
1987
-0.15
1.01
1988
-0.20
1.00
1989
-0.21
0.99
1990
-0.18
1.00
in the level of GNP is attributable to a rapid accumu-
lation of capital, which is driven by changes in the
price of investment goods. The capital accumulation
effect is augmented by a decline in energy prices rela-
tive to the base case. Lower energy prices that corre-
spond to a world with no CAA regulations decreases
costs and increases real household income, thus in-
creasing consumption.
Removing the pollution control component of new
capital is equivalent to lowering the marginal price of
investment goods. Combining this with the windfall
gain of not having to bring existing capital into com-
pliance leads to an initial surge in the economy's rate
of return, raising the level of real investment. The in-
A-20
-------
Appendix A: Cost andMacroeconomic Modeling
vestment effects are summarized in Figure A-l. More
rapid (ordinary) capital accumulation leads to a de-
cline in the rental price of capital services which, in
turn, stimulates the demand for capital services by pro-
ducers and consumers. The capital rental price reduc-
tions also serve to lower the prices of goods and ser-
vices and, so, the overall price level. Obviously, the
more capital intensive sectors exhibit larger price re-
ductions.25 The price effects from investment changes
are compounded by the cost reductions associated with
releasing resources from the operation and mainte-
nance of pollution control equipment and by the elimi-
nation of higher prices due to regulations on mobile
sources.
To households, no-control scenario conditions are
manifest as an increase in permanent future real earn-
ings which supports an increase in real consumption
in all periods and, generally, an increase in the de-
mand for leisure (see Table A-15). Households mar-
ginally reduce their offer of labor services as the in-
come effects of
^u higher real earn-
ings dominate the
substitution ef-
fects of lower
goods prices.
The increase in
consumption is
dampened by an
increase in the
rate of return that
produces greater
investment (and
personal sav-
ings).
Finally, tech-
nical change is a
very important
aspect of the sup-
ply-side adjust-
ments under the
no-control sce-
nario. Lower fac-
tor prices in-
crease the endog-
enous rates of
Table A-15. Difference in Personal
Consumption Between the Control
and No-Control Scenarios.
Nominal %
Real %
Year
Chan ge
C'haniie
1973
-0.02
0.33
1974
-0.01
0.43
1975
-0.10
0.24
1976
-0.10
0.39
1977
-0.10
0.54
1978
-0.09
0.63
1979
-0.11
0.68
1980
-0.12
0.71
1981
-0.13
0.74
1982
-0.12
0.81
1983
-0.13
0.85
1984
-0.15
0.86
1985
-0.19
0.88
1986
-0.19
0.94
1987
-0.19
0.98
1988
-0.17
1.03
1989
-0.17
1.04
1990
-0.18
1.01
technical change in those industries that are factor-
using. Lower rental prices for capital benefit the capi-
tal-using sectors, lower materials prices benefit the
materials-using sectors, and lower energy prices ben-
efit the energy-using sectors. On balance, a signifi-
cant portion of the increase in economic growth is
attributable to accelerated productivity growth. Un-
der the no-control scenario, economic growth aver-
ages 0.05 percentage points higher over the interval
1973-1990. The increased availability of capital ac-
counts for 60 percent of this increase while faster pro-
ductivity growth accounts for the remaining 40 per-
cent. Thus, the principal effect arising from the costs
associated with CAA initiatives is to slow the
economy's rates of capital accumulation and produc-
tivity growth. This finding is consistent with recent
analyses suggesting a potential association between
higher reported air, water, and solid waste pollution
abatement costs and lower plant-level productivity in
some manufacturing industries (Gray and Shadbegian,
1993 and 1995).
As with the cost and expenditure data presented
above, it is possible to present the stream of GNP and
consumption changes as single values by discounting
the streams to a single year. Table A-16 summarizes
the results of the discounting procedure, and also in-
cludes discounted expenditure and annualized cost
data for reference. Accumulated (and discounted to
1990) losses to GNP over the 1973-1990 period were
half again as large as expenditures during the same
period, and approximately twice as large as annual-
ized costs. Losses in household consumption were
approximately as great as annualized costs.
Tabic A-1.6. GNP and Consumption Impacts
Discounted to 1990 ($1990 billions).
3%
5%
7%
Expenditures
520
628
761
Annualized Costs
417
523
657
GNP
880
1005
1151
Household Consumption
500
569
653
HH and Gov't Consumption
676
769
881
Source: Expenditures and annualized costs from above;
macroeconomic impacts from Jorgenson et al. (1993),
Table 4.1
25 Not surprisingly, at the industry level, the principal beneficiaries in the long run of eliminating the costs associated with air
pollution abatement are the most heavily regulated industries. The largest changes in industry prices and outputs occur in the motor
vehicles industry. Other industries that benefit significantly from the elimination of environmental controls are refined petroleum
products, electric utilities, and other transportation equipment. Turning to manufacturing industries, metal mining and the primary
metals have the largest gains in output from elimination of air pollution controls.
A-21
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Figure A-l. Percent Difference in Real Investment Between Control and No-control Scenarios.
2.5%
2.0%
1.5%
1.0%
0.5%
0.0%
1973 1974 1975 1976 1977 1978 1979
1980
1981
1982 1983
1984
1985
1986 1987
1988
1989
1990
Year
Figure A-2. Percent Difference in Price of Output by Sector Between Control and No-control
Scenario for 1990.
1%
0%
-1%
c
o
CD
CL
-2%
-3%
III
1 2 3 4 5 6 7
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Sector
A-22
-------
Appendix A: Cost andMacroeconomic Modeling
Although they have value as descriptors of the
magnitude of changes in economic activity, neither
GNP nor consumption changes are perfect measures
of changes in social welfare. A better measure is
Equivalent Variations (EVs), which measure the
change in income that is equivalent to the change in
(lifetime) welfare due to removal of the CAA. As part
of its macroeconomic exercise, EPA measured the EV s
associated with removal of the CAA. Elimination of
CAA compliance costs (disregarding benefits) repre-
sents a welfare gain of $493 billion to $621 billion,
depending on assumptions used in the analysis.26 This
result does not differ greatly from the range of results
represented by expenditures, anualized costs, and con-
sumption changes.
Prices
One principal consequence of the Clean Air Act
is that it changes prices. The largest price reductions
accrue to the most heavily regulated industries which
are the large energy producers and consumers (see
Table A-17). But these are also the most capital in-
tensive sectors and it is the investment effects that are
the dominant influences in altering the course of the
economy. Focusing on energy prices, under the
no-control scenario the price of coal in 1990 declines
by 1.3 percent, refined petroleum declines by 3.03
Tabic A-17. Percentage Difference in Energy Prices
Between the Control and No-control Scenarios.
Refined
Electric
Cias
r
final
Petrnlenm
Utilities
1 il ilil ii-s
1973
-0.44
-5.99
-2.11
-0.32
1974
-0.47
-4.X4
-2.53
-0.44
1975
-0.42
-4.2X
-2.19
-0.31
1976
-0.57
-3.X3
-2.12
-0.44
1977
-0.74
-3.43
-2.22
-0.59
.1978
-0.X6
-3.2X
-2.39
-0.6X
1979
-0.91
-2.92
-2.XI
-0.71
19X0
-0.94
-2.76
-2.97
-0.69
19X1
-0.97
-2.50
-2.76
-0.71
19X2
-0.9X
-2.42
-2.63
-0.77
19X3
-1.09
-2.35
-2.5X
-0.X5
19X4
-1.12
-2.26
-2.49
-0.91
19X5
-1.21
-2.X9
-2.62
-0.97
19X6
-1.27
-3.35
-2.69
-1.12
19X7
-1.31
-3.50
-2.7X
-1.1X
19XX
-1.30
-3.61
-2.75
-1.19
19X9
-1.31
-3.45
-2.74
-1.19
1990
-1.30
-3.03
-2.75
-1.20
percent, electricity from electric utilities declines by
2.75 percent, and the price of natural gas from gas
utilities declines by 1.2 percent. The declining price
of fossil fuels induces substitution toward fossil fuel
energy sources and toward energy in general. Total
Btu consumption also increases.
Sectoral Effects: Changes in Prices and
Output by Industry
At the commodity level, the effect of the CAA
varies considerably. Figure A-2 shows the changes in
the supply price of the 35 commodities measured as
changes between the no-control case and the control-
case for 1990.
In 1990, the largest change occurs in the price of
motor vehicles (commodity 24), which declines by
3.8 percent in the no-control case. Other prices show-
ing significant effects are those for refined petroleum
products (commodity 16) which declines by 3.0 per-
cent, and electricity (commodity 30) which declines
2.7 percent. Eight of the remaining industries have
decreases in prices of 1.0 to 1.4 percent under the
no-control scenario. The rest are largely unaffected
by environmental regulations, exhibiting price de-
creases between 0.3 and 0.8 percent.
To assess the intertemporal consequences of the
CAA, consider the model's dynamic results and the
adjustment of prices between 1975 and 1990. Initially,
in 1975, the biggest effect is on the price of output
from petroleum refining (sector 16), which declines
by 4.3 percent. But by 1990, the price of petroleum
refining is about 3.0 percent below control scenario
levels. In contrast, the price of motor vehicles (sector
24) is about 2.4 percent below baseline levels in 1975,
but falls to about 3.8 percent below baseline levels in
1990.
The price changes affect commodity demands,
which in turn determine how industry outputs are af-
fected. Figure A-3 shows percentage changes in quan-
tities produced by the 35 industries for 1990. As noted
earlier, the principal beneficiaries under the no-control
scenario are the most heavily regulated industries:
motor vehicles, petroleum refining, and electric utili-
ties.
In 1990, the motor vehicle sector (sector 24) shows
the largest change in output, partly due to the fact that
the demand for motor vehicles is price elastic. Recall
26 Jorgenson et al, 1993.
A-23
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Figure A-3. Percent Difference in Quantity of Output by Sector Between Control and No-
control Scenario for 1990.
1
ll- ¦_ 1
-
i ¦¦ 11
~n
U u 1
1 11—=1
¦
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Sector
Figure A-4. Percent Difference in Employment by Sector Between Control and No-control
Scenario for 1990.
¦-I ¦ II
¦
pi
¦ill
III
II 1
|i
1 'I1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Sector
A-24
-------
Appendix A: Cost andMacroeconomic Modeling
that the largest increase in prices also occurred in the
motor vehicles sector. The 3.8 percent reduction in
prices produces an increase in output of 5.3 percent
relative to the base case.
Significant output effects are also seen in the pe-
troleum refining sector (sector 16) with a 3.2 percent
increase, in electricity (sector 30) with a 3.0 percent
increase, and in other transportation equipment (sec-
tor 25) with a 1.6 percent increase. The large gains in
output for these industries are mostly due to the de-
cline in their prices. In manufacturing, the sectors
exhibiting the most significant output effects are metal
mining (sector 2) with a 2.0 percent increase, and pri-
mary metals (sector 20) with a 1.8 percent increase.
Twenty of the remaining industries exhibit increase
in output of less than 0.9 percent after pollution con-
trols are removed.
While most sectors increase output under the
no-control scenario, a few sectors decline in size in
the absence of air pollution controls. The most no-
table of these are food and kindred products (sector
7) which decline by 0.5 percent, furniture and fixtures
(sector 12) which decline by 0.6 percent, and rubber
and plastic products (sector 17) which decline by 0.3
percent. These sectors are among the least capital in-
tensive, so the fall in the rental price of capital ser-
vices has little effect on the prices of outputs. Buyers
of the commodities produced by these industries face
higher relative prices and substitute other commodi-
ties in both intermediate and final demand. The rest
of the sectors are largely unaffected by environmen-
tal regulations.
Changes in Employment Across
Industries
The effect of the CAA on employment presents a
much more complicated picture. Although Jorgenson-
Wilcoxen is a full-employment model and cannot be
used to simulate unemployment effects, it is useful
for gaining insights about changes in the patterns of
employment across industries. Percentage changes in
employment by sector for 1990 are presented in Fig-
ure A-4.
For 1990, the most significant changes in the level
of employment relative to the control scenario occur
in motor vehicles (sector 24) which increases 1.2 per-
cent, other transportation equipment (sector 25) which
increases 0.8 percent, electric utilities (sector 30)
which increases 0.7 percent, and primary metals (sec-
tor 20) which increases 0.6 percent. The level of em-
ployment is higher relative to the control case in 10
other industries.
For a few sectors, the no-control scenario results
in changes in real wages which cause reductions in
employment. The most notable reductions in employ-
ment under the no-control scenario occur in tobacco
manufacturing (sector 8) which declines 1.2 percent,
furniture and fixtures (sector 12) which declines 0.8
percent, rubber and plastic products (sector 17) which
declines 0.8 percent, food and kindred products (sec-
tor 7) which declines 0.7 percent, stone, clay and glass
products (sector 19) which declines 0.6 percent, and
instruments (sector 26) which declines 0.6 percent.
These sectors are generally those in which the level
of output was lower in 1990 relative to the control
scenario, since they are among the least capital inten-
sive and the fall in the rental price of capital services
has little effect on the prices of outputs. Buyers of the
commodities produced by these industries face higher
relative prices and substitute other commodities in
both intermediate and final demand. It is interesting
to note that several of the least capital intensive sec-
tors experience insignificant employment effects in
the short run (1975) under the no-control scenario,
but increasingly adverse effects over the 20-year pe-
riod of analysis. These include food and kindred prod-
ucts, furniture and fixtures, rubber and plastic prod-
ucts, stone, clay and glass products, and instruments.
Examination of the transition of employment in
the economy from the initial equilibrium to 1990 re-
veals that the employment effects of the CAA on motor
vehicles, transportation equipment, electric utilities,
and primary metals persist over the entire period of
analysis. Employment varies from: an increase of 1.7
percent in 1975 to 1.2 percent in 1990 in motor ve-
hicles; an increase of 0.7 in 1975 to 0.8 percent in
1990 in transportation equipment; an increase of 1.2
percent in 1975 to 0.7 percent in 1990 in electric utili-
ties; and an increase of 0.8 percent in 1975 to 0.6 per-
cent in 1990.
A-25
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Uncertainties in the Cost
Analysis
Potential Sources of Error in the Cost
Data
Because of the importance of the Cost of Clean
data for this assessment, the project team investigated
potential sources of error due to the use of industry's
self-reported costs of compliance with air pollution
abatement requirements. Concerns about the accuracy
of responses include (1) misreporting by firms in re-
sponse to federal agency surveys, and (2) omission of
important categories of compliance cost from the data
collected or reported by these federal agencies.27 Table
A-18 contains a summary of the results of the analy-
sis. This analysis is consistent with the findings of
two recent studies comparing combined air, water, and
solid waste pollution abatement costs, as reported in
federal abatement cost surveys, to their observed ef-
fects on productivity levels. These studies suggest that,
since observed productivity decreases exceed those
expected to result from the reported abatement costs,
there may be additional pollution abatement costs not
captured or reported in the survey data, and that total
abatement costs for the three manufacturing indus-
tries studied may be under-reported by as much as a
factor of two in the most extreme case (Gray and
Shadbegian, 1993 and 1995; Gray, 1996).
The major finding from this analysis indicates that
total O&M costs are likely to be under-reported due
to exclusion of private research and development
Tabic A-18. Potential Sources of Error and Their Effect on Total Costs of Compliance.
Source of Error
Effect on Capital Costs
Effect on O&M Costs
Lack of Data at Firm Level
Under-reported
Percent Unknown
Under-reported
Percent Unknown
Misallocation of Costs:
Inclusion of OSHA and Other
Regulatory Costs
Over-reported
Percent Unknown
Over-reported
Percent Unknown
Exclusion of Solid Waste Disposal Costs
Related to Air Pollution Abatement
Under-reported
Percent Unknown
Exclusion of Costs:
Exclusion of Private R&D Expenses
—
Under-reported by 14 to 17%
(varies by year)
Exclusion of Energy Use by Pollution
Abatement Devices®
—
Under-reported by 1 to 3%
(varies by year)
Exclusion of Depreciation Expenses®
—
Under-reported by 1 to 2%
(varies by year)
Exclusion of Recovered Co sts
—
Over-reported by 1% Plus
Omission of Small Firms
Under-reported by 1 to 2%
Under-reported by 1 to 2%
NET EFFECT
Under-reported
Under-reported
'¦> Energy outlays are part of the data on O&M costs and depreciation expenses are not. Accordingly, in the J/W model, energy outlays are
considered along with other operating expenditures in terms of their impacts on unit costs. Depreciation is represented fully in the capital
accumulation process, as the undepreciated capital stock at the beginning of any period gives rise to the flow of capital services available to
producers and consumers.
Source: IndustrialEconomics, Incorporated, memorandum to Jim DeMocker, EPA/OAR," Sources of Error in
Reported Costs of Compliance with Air Pollution Abatement Requirements," October 16, 1 991.
27 Memorandum from Industrial Economics, Incorporated to Jim DeMocker (EPA/OAR) dated 10/16/91 and entitled "Sources
of Error in Reported Costs of Compliance with Air Pollution Abatement Requirements."
A-26
-------
Appendix A: Cost andMacroeconomic Modeling
(R&D) expenditures. Note, however, that although
these costs were excluded from those used for the
macroeconomic modeling, they were included in the
overall direct cost estimate of the CAA; see "Other
Direct Costs," above. These costs are excluded from
the macromodeling because they cannot be disaggre-
gated by industry and, more importantly, because there
is no information on what was purchased or obtained
as a result of these expenditures.
Based on the need indicated by the IEc review,
modifications to the BEA data were made to remedy
some of the biases noted above. In particular, recov-
ered costs for stationary source air pollution, e.g. sul-
fur removed using scrubbers that is then sold in the
chemical market, have been accounted for in the data
set used in the model runs.
Tabic A-19. Stationary Source O&M
Expenditures as a Percentage of Capital Stock
(millions of 1990 dollars).
O&M rlivirlerl hy
K stnr* Npt K OftM
K stnnk Npt K
1973
6,521 6,521 3,936
0.60 0.60
1974
14,880 14,684 4,778
0.32 0.33
1975
23,533 22,876 5,154
0.22 0.23
1976
32,773 31,372 5,768
0.18 0.18
1977
41,331 38,869 6,527
0.16 0.17
1978
49,448 45,612 6,991
0.14 0.15
1979
57,299 51,776 7,959
0.14 0.15
1980
65,763 58,232 8,791
0.13 0.15
1981
74,366 64,469 8,785
0.12 0.14
1982
82,381 69,740 7,855
0.10 0.11
1983
89,937 74,173 8,168
0.09 0.11
1984
95,879 76,606 8,505
0.09 0.11
1985
101,723 78,587 8,617
0.08 0.11
1986
107,082 79,71 3 8,477
0.08 0.11
1987
112,225 80,249 8,602
0.08 0.11
1988
117,269 80,300 8,143
0.07 0.10
1989
122,182 79,819 8,259
0.07 0.10
1990
127,394 79,2 1 7 8,842
0.07 0.11
"K stock" is the accumulated undepreciated stationary
source control capital stock available at the beginning of
eachyear, fromTable A-10.
"Net K" is the stationary source control capital stock less
depreciation implied by amortization at 5%; from Table
A-10.
"O&M" is the stationary source control O&M
expenditures; fromTableA-9.
The final two columns are ratios: O&M divided by capital
stock; and O&M divided by net capital.
An additional set of concerns relates directly to
reporting of costs by firms. Some have noted an un-
expected temporal pattern of stationary source con-
trol expenditures in the BEA data that might lead one
to question the accuracy of the Census survey re-
sponses. One would expect that stationary source
O&M expenditures over time would be roughly pro-
portional to the accumulated stationary source con-
trol capital stock. Yet, as illustrated in Table A-19,
O&M expenditures as a fraction of accumulated capi-
tal stock decline over time (even if one discounts the
first few years because of the dramatic percentage
increases in capital stock during those years). It is true
that the ratio of O&M expenditures to the depreci-
ated capital stock (in the far right column, labeled "net
K") is reasonably stable after 1981. The depreciation
shown here, however, is a financial depreciation only,
depicting the declining value of a piece of equipment
over time, rather than a measure of physical asset
shrinkage. Assuming a twenty-year useful lifetime,
all of the stationary source control capital stock put in
place since 1972 could conceivably still be in place in
1990. If anything, one would expect the O&M/K ra-
tio to increase as the capital depreciates (i.e., ages),
until the equipment is scrapped, because aging equip-
ment requires increasing maintenance. Consequently,
one might infer from this information that firms have
systematically under-reported O&M expenditures, or
have over-reported capital expenditures.
The apparent anomaly might be explained by an
examination of the types of O&M expenditures re-
ported. If more than a token percentage of O&M ex-
penditures are unrelated to "operation and mainte-
nance" of pollution control devices, then the observed
O&M/K ratio would not appear unusual.
The Census PACE survey28 required respondents
to report air pollution abatement O&M expenses in
the following categories: salaries and wages; fuel and
electricity; contract work; and materials, leasing, and
"miscellaneous."29 In later versions of the survey,
additional information relating to the types of expenses
to report was provided as a guide to respondents. The
types of expenses listed that are relevant to air pollu-
tion abatement include:
28 Pollution Abatement Costs and Expenditures, various years.
29 Census also requested a reporting of "depreciation" expenses as a component of O&M. BEA, however, removed depreciation
expense from the reported O&M costs because retaining depreciation would have amounted to double-counting, since BEA also
reported capital expenditures.
A-27
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
(1) operating and maintaining pollution abate-
ment equipment;
(2) fuel and power costs for operating pollution
abatement equipment;
(3) parts for pollution abatement equipment re-
placement and repair;
(4) testing and monitoring of emissions;
(5) incremental costs for consumption of envi-
ronmentally preferable materials and fuels;
(6) conducting environmental studies for devel-
opment or expansion;
(7) leasing of pollution abatement equipment;
(8) compliance and environmental auditing;
(9) salaries and wages for time spent completing
environmental reporting requirements; and
(10) developing pollution abatement operating
procedures.30
The magnitude of the expenditures associated with
the first three items should be correlated with the size
of the existing stock of air pollution abatement capi-
tal. Expenditures associated with items four through
ten, however, should be independent of the size of the
existing capital stock (expenditures associated with
item seven, leasing of pollution abatement equipment,
could be negatively correlated with the size of the
capital stock), //'items four through ten account for a
non-negligible proportion of total O&M expenditures,
and if respondents included these cost categories even
though they were not explicitly listed in the survey
instructions before 1991, then one would expect to
see the O&M/K ratio declining during the study pe-
riod. Thus, even though it is possible that O&M ex-
penditures are underreported (or that capital expendi-
tures are overreported), one cannot be certain.
Mobile Source Costs
For the section 812 analysis, EPA used the best
available information on the estimated cost of mobile
source air pollution control. Several other sources of
cost estimates exist, however, including a cost series
produced by the Department of Commerce Bureau of
Economic Analysis (BEA). The BEA cost series is
summarized in Table A-20. The BEA estimates dif-
fer significantly from EPA estimates, particularly with
respect to estimates of capital costs and the "fuel price
penalty" associated with the use of unleaded gaso-
line.
EPA's capital cost estimates are based on esti-
mates of the cost of equipment required by mobile
30 Pollution Abatement Costs and Expenditures, 1992, pg. A-9.
Tabic A-20. Comparison of EPA and BEA Stationary
Source Expenditure Estimates (millions of current
dollars).
IV. till
Vpar
<•11 pi tn 1
OA,M
i-ii pi tn I
O&M
Kvpi'ii H.
EPA Estimates
1986
4,090
7,116
312
140
1 1.658
1987
4,179
7,469
277
130
12.055
1988
4,267
7,313
243
161
1 1.984
1989
4,760
7,743
235
173
12.91 1
1990
4,169
8,688
226
154
13.237
BEA Estimates
1986
4,090
7,072
312
182
1 1.656
1987
3,482
5,843
246
141
9.712
1988
3,120
6,230
121
161
9.632
1989
3,266
6,292
229
152
9.939
1990
4.102
6.799
200
154
1 1.255
"Recovered Costs" arenotincluded in this table.
Sources for "BEA Estimates": for 1986, "Pollution Abatement and Control
Eypen Hitnres " Survey of Current Business (RF.A) June 1989, Table 7; for
1987-90, BEA May 1995, Table 8.
source regulations. BEA's estimates are based on sur-
vey data from the Bureau of Labor Statistics (BLS)
that measures the increase in the per-automobile cost
(relative to the previous model year) due to pollution
control and fuel economy changes for that model year.
The difference in approach is significant: BEA's an-
nual capital cost estimates exceed EPA's by a factor
of (roughly) two. EPA may underestimate costs to the
extent that engineering cost estimates of components
exclude design and development costs for those com-
ponents. The BLS estimates add the incremental an-
nual costs to all past costs to derive total current-year
costs. Such an approach overestimates costs to the
extent that it fails to account for cost savings due to
changes in component mixes over time.
Some mobile source pollution control devices re-
quired the use of unleaded fuel. Unleaded gasoline is
more costly to produce than is leaded gasoline, and
generally has a greater retail price, thus imposing a
cost on consumers. EPA estimated the "fuel price pen-
alty" by using a petroleum refinery cost model to deter-
mine the expected difference in production cost be-
tween leaded and unleaded gasoline. BEA's "fuel price
penalty" was the difference between the retail price
of unleaded gasoline and that of leaded gasoline.
A detailed description of the data sources, ana-
lytic methods, and assumptions that underlie the EPA
and BEA mobile source cost estimates can be found
in McConnell et al. (1995).
A-28
-------
Appendix A: Cost andMacroeconomic Modeling
Stationary Source Cost Estimate
Revisions
As noted above, the costs used for stationary
sources in the macro-modeling (and retained in this
cost analysis) were projected for several years in the
late 1980s. Since that time, BEA has released histori-
cal expenditure estimates for those years based on
survey data. A comparison of the expenditure series
can be found in Table A-21. Apparently, EPA's pro-
jections overestimated stationary source compliance
expenditures by approximately $2 billion per year for
the period 1987-1990. Since expenditures from all
sources are estimated to be $18 billion -$19 billion
(current dollars) per year during 1987-1990, this im-
plies that EPA has overestimated compliance expen-
ditures by more than ten percent during this period.
Although a substantial overstatement for those years,
the $2 billion per year overestimate would have little
impact (probably less than two percent) on the dis-
counted present value, in 1990 dollars, of the 1973-
1990 expenditure stream.
Tabic A-21. BEA Estimates of Mobile Source Costs.
Endogenous Productivity Growth in the
Macro Model
For each industry in the simulation, the JW model
separates price-induced changes in factor use from
changes resulting strictly from technical change. Thus,
simulated productivity growth for each industry has
two components: (a) an exogenous component that
varies over time, and (b) an endogenous component
that varies with policy changes. Some reviewers have
noted that, although not incorrect, use of endogenous
productivity growth is uncommon in the economic
growth literature. EPA conducted a sensitivity run of
the J/W model, setting endogenous growth parameters
to zero (i.e., removing endogenous productivity
growth from the model)31
Endogenous productivity growth is an important
factor in the J/W model. For example, for the period
1973-1990, removal of the endogenous productivity
growth assumptions reduces household income by 2.9
to 3.0 percent (depending on whether one uses a world
with CAA or one without CAA as the baseline). In
comparison, removal of CAA compliance costs re-
sults in a 0.6 to 0.7 percent change in household in-
come (depending on whether one uses, as a baseline,
a world with or one without endgenous productivity
growth). That is, use of the endogenous productivity
growth assumption has four to five times the impact
of that of CAA compliance costs.
Although very important to the simulated growth
of the economy within any policy setting, the endog-
enous productivity growth assumption is less impor-
tant across policy settings. Under the base (i.e., "with
endogenous productivity growth") scenario, the ag-
gregate welfare effect (measured as EVs, see above)
of CAA compliance costs and indirect effects is esti-
mated to be 493 billion to 621 billion in 1990 dollars.
If one removes the endogenous productivity growth
assumption, the aggregate welfare effect declines to
the range 391 billion to 494 billion in 1990 dollars
(Jorgenson et al., 1993, pg. 6-15), a reduction of about
twenty percent.
( :i pitii 1
.Net
Fuel Price
Fuel Fcouomv
Vi'ii r
l v p.
WtM*
IVn.iltv
IVn.iltv
1973
1,013
1,104
697
1974
1,118
1,380
5
1,180
1975
2,131
1,520
97
1,344
1976
2,802
1,420
309
1,363
1977
3,371
1,289
701
1,408
1978
3,935
1,136
1,209
1,397
1979
4,634
931
1,636
1,792
1980
5,563
726
2,217
2,320
1981
7,529
552
2,996
2,252
1982
7,663
409
3,518
1,876
1983
9,526
274
4,235
1,582
1984
11,900
118
4,427
1,370
1985
13,210
165
4,995
1,133
1986
14,368
(331)
4,522
895
1987
13,725
(453)
3,672
658
1988
16,157
(631)
3,736
420
1989
15,340
(271)
1,972
183
1990
14,521
(719)
1,370
(55)
* Inspection and maintenance costs less fuel density savings and
maintenance savings.
31 For greater detail, see Jorgenson et al., 1993.
A-29
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Amortization Period for Stationary
Source Plant and Equipment
In developing annualized costs, stationary source
capital expenditues were amortized over a twenty-year
period. That is, it was assumed that plant and equip-
ment would depreciate over twenty years. It is pos-
sible that stationary source plant and equipment has,
on average, a useful lifetime significantly greater than
twenty years. The Project Team tested the sensitivity
of the cost analysis results to changes in stationary
source capital amortization periods.
Table A-22 presents total annualized compliance
costs assuming a 40-year amortization period for sta-
tionary source capital expenditures (all other cost com-
ponents are unchanged from the base analysis). All
costs are in 1990-value dollars, ad three alternative
discount rates are used in the annualization period.
Table A-23 presents the results discounted to 1990,
and compared to the base case results (i.e., using a
twenty-year amortization period). Doubling the am-
ortization period to 40 years decreases the 1990 present
value of the 1973-1990 cost stream by approximately
40 billion dollars. This represents a change of six per-
cent to nine percent, depending on the discount rate
employed.
Tabic A-22. Annualized Costs
Assuming 40-Year Stationary Source
Capital Amortization Period. 1973-
1990 (millions of 1990 dollars).
Wii »•
lit V>/„
¦At *<>/„
¦At 7<>/„
.1 973
10,801
10,899
11,008
1974
12,875
13,108
13,366
1 975
12,751
13,121
13,532
1 976
13,338
13,891
14,504
1977
14,263
14,996
15,807
1 978
13,778
14,690
15,695
1 979
15,936
17,024
18,220
19X0
18,091
19,368
20,771
19X1
17,809
19,272
20,880
19X2
16,670
18,316
20,123
1 9X3
16,941
18,759
20,754
19X4
17,836
19,803
21,960
19X5
20,079
22,213
24,551
19X6
18,544
20,809
23,288
19X7
19,384
21,772
24,387
19XX
19,203
21,706
24,446
19X9
19,989
22,604
25,467
1 990
20,546
23,268
26,247
Tabic A-23. Effect of Amortization
Periods on Annualized Costs Discounted
to 1990 (billions of 1990 dollars).
Discount rate
1%.
OLji ZU
20-yr amortization
period
417
523 657
40-yr amortization
period
379
483 617
A-30
-------
Cost and Macroeconomic
Modeling References
Chase Econometrics Associates, Inc. 1976. "The
Macroeconomic Impacts of Federal Pollution
Control Programs: 1976 Assessment." Report
prepared for the Council on Environmental
Quality and the Environmental Protection
Agency.
Congressional Budget Office. 1990. Carbon Charges
as a Response to Global Warming: The Ef-
fects of Taxing Fossil Fuels. Washington, DC,
U.S. Government Printing Office.
Data Resources, Inc. 1979. "The Macroeconomic
Impacts of Federal Pollution Control Pro-
grams: 1978 Assessment," Report prepared
for the Environmental Protection Agency and
the Council on Environmental Quality.
Data Resources, Inc. 1981. "The Macroeconomic
Impact of Federal Pollution Control Pro-
grams: 1981 Assessment," Report prepared
for the Environmental Protection Agency.
July 17.
Economic Report of the President. 1995. U.S. Gov-
ernment Printing Office, Washington, DC.
February.
Farber, Kit D. and G. Rutledge. 1989. "Pollution
Abatement and Control Expenditures: Meth-
ods and Sources for Current-Dollar Esti-
mates." Unpublished Paper for U.S. Depart-
ment of Commerce, Bureau of Economic
Analysis. October.
Freeman, A.M. 1978. "Air and Water Pollution
Policy," in P.R. Portney (ed.), Current Issues
in U.S. Environmental Policy. Johns Hopkins
University Press, Baltimore.
Gray, Wayne B. 1996. Personal communication with
Michael Hester of Industrial Economics, Inc.
December 4.
Gray, Wayne B. and Ronald J. Shadbegian. 1993.
"Environmental Regulation and Manufactur-
ing Productivity at the Plant Level," Center
for Economic Studies Discussion Paper, CES
93-6. March.
Appendix A: Cost and Macroeconomic Modeling
Gray, Wayne B. and Ronald J. Shadbegian. 1995.
"Pollution Abatement Costs, Regulation, and
Plant-Level Productivity," National Bureau of
Economic Research, Inc., Working Paper
Series, Working Paper No. 4994. January.
Hazilla, M. and R.J. Kopp. 1990. "Social Cost of En-
vironmental Quality Regulations: A General
Equilibrium Analysis," Journal of Political
Economy, Vol. 98, No. 4. August.
Industrial Economics, Incorporated. 1991. "Sources
of Error in Reported Costs of Compliance
with Air Pollution Abatement Requirements,"
memorandum to Jim DeMocker, EPA/OAR.
October 16.
Jorgenson, Dale W. and Barbara M. Fraumeni. 1989.
"The Accumulation of Human and Nonhu-
man Capital, 1948-1984," in R.E. Lipsey and
H.S. Tice, eds., The Measurement of Saving,
Investment, and Wealth. University of Chi-
cago Press, Chicago, II.
Jorgenson, Dale W. and Barbara M. Fraumeni. 1981.
"Relative Prices and Technical Change," in
E. Berndt and B. Field, eds., Modeling and
Measuring Natural Resource Substitution.
MIT Press, Cambridge, MA.
Jorgenson, Dale W., Richard J. Goettle, Daniel
Gaynor, Peter J. Wilcoxen, and Daniel T.
Slesnick. 1993. "The Clean Air Act and the
U.S. Economy," Final report of Results and
Findings to the U.S. EPA. August.
Jorgenson, Dale W. and Peter J. Wilcoxen. 1990a.
"Environmental Regulation and U.S. Eco-
nomic Growth," in RAND Journal of Econom-
ics, Vol. 21, No. 2, pp. 314-340.
Jorgenson, Dale W. and Peter J. Wilcoxen. 1990c.
"Intertemporal General Equilibrium Model-
ing of U.S. Environmental Regulation," in
Journal of Policy Modeling, Vol. 12, No. 4,
pp.715-744.
Jorgenson, Dale W. and Peter J. Wilcoxen. 1993. "En-
ergy, the Environment and Economic
Growth," in Handbook of Natural Resource
and Energy Economics, Allen V. Kneese and
James L. Sweeney, eds., Volume 3, Chapter
27. North-Holland, Amsterdam, forthcoming.
A-31
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Kokoski, Mary F. and V. Kerry Smith. 1987. "A Gen-
eral Equilibrium Analysis of Partial-Equilib-
rium Welfare Measures: The Case of Climate
Change," American Economic Review, Vol.
77, No. 3, pp. 331-341.
McConnell, Virginia, Margaret A. Walls, and Win-
ston Harrington. 1995. "Evaluating the Costs
of Compliance with Mobile Source Emission
Control Requirements: Retrospective Analy-
sis," Resources for the Future Discussion Pa-
per.
Schwartz, Joel. 1991. "Fuel Economy Benefits."
Memorandum to Joe Somers and Jim
DeMocker. December 12.
Somers, J.H. 1991. "Fuel Economy Penalties for Sec-
tion 812 Report." Memorandum to Anne
Grambsch and Joel Schwartz. December 23.
U.S. Department of Commerce. Government Fi-
nances, various issues. Bureau of the Census.
U.S. Department of Commerce. "Pollution Abatement
and Control Expenditures," Survey of Current
Business, various issues. Bureau of Economic
Analysis.
U.S. Department of Commerce. "Pollution Abatement
Costs and Expenditures," Current Industrial
Reports, various issues. Bureau of the Cen-
sus.
U.S. Environmental Protection Agency (EPA). 1985.
Costs and Benefits of Reducing Lead in
Gasolines: Final Regulatory Impact Analy-
sis. Office of Policy Analysis, EPA-230-05-
85-006. February.
U.S. Environmental Protection Agency (EPA). 1990.
Environmental Investments: The Cost of a
Clean Environment, Report to the Congress.
Office of Policy, Planning and Evaluation.
EPA-230-12-90-084. December.
Verleger, Philip K., Jr. 1992. "Clean Air Regulation
and the L.A. Riots," The Wall Street Journal,
Tuesday, May 19. p. A14.
Walsh, M.P. 1991. "Motor Vehicles and Fuels: The
Problem." in EPA Journal, Vol. 17, No. l,p.
12.
Wilcoxen, Peter J. 1988. The Effects of Environmen-
tal Regulation and Energy Prices on U.S.
Economic Performance, Doctoral thesis pre-
sented to the Department of Economics at
Harvard University, Cambridge, MA. Decem-
ber.
A-32
-------
Appendix B: Emissions
Modeling
Introduction
This appendix provides additional details of the
methodologies used to estimate control and no-control
scenario emissions and the results obtained by these
methods. Methodological information and results are
provided for each of the six principal emission sec-
tors: industrial combustion, industrial processes, elec-
tric utilities, on-highway vehicles, off-highway ve-
hicles, and commercial/residential sources.
The initial section of this appendix assesses the
emissions projections presented in this analysis by (1)
comparing the 1970 to 1990 control scenario projec-
tions with recent EPA Trends report estimates for the
same years and (2) comparing the 1970 to 1990 trend
in no-control scenario projections with 1950 to 1970
emissions as reported in Trends. The first compari-
son indicates that control scenario emissions projec-
tions approximate, but do not precisely match, the EPA
Trends data. The reason forthis mismatch is discussed
below. The second comparison is useful for demon-
strating that pre-1970 emissions trends would not pro-
vide a satisfactory basis for extrapolating emissions
trends into the 1970 to 1990 period. The inability to
simply extrapolate pre-1970 trends provides further
justification for applying the present modeling meth-
odologies to generate no-control scenario emissions
projections.
The remainder of the appendix provides further
details of the emissions modeling conducted in sup-
port of the present analysis, and is largely adapted
from the draft report "The Impact of the Clean Air
Act on 1970 to 1990 Emissions; section 812 retro-
spective analysis," March 1, 1995 by Pechan Associ-
ates. The draft Pechan report surveys the methodolo-
gies and results associated with the sector-specific
emission modeling efforts by Argonne National Labo-
ratory (ANL), ICF Resources Incorporated (ICF), Abt
Associates (Abt), and the Environmental Law Insti-
tute (ELI).
Comparison of Emissions
Projections with Other EPA Data
Control Scenario Projections Versus
EPA Trends Projections
The control scenario emission results are similar,
but not identical, to official EPA historical emission
estimates provided by the EPA National Air Pollut-
ant Emission Trends Reports.1 Comparisons between
the current estimates and the Trends data for S02, NOx,
VOC, CO, and TSP are presented in Figures B-l, B-
2, B-3, B-4, and B-5 respectively. More detailed tables
providing emission estimates by sector and by target
year for TSP, S02, NOx, VOC, CO, and Lead are pre-
sented in Tables B-l6, B-l7, B-l8, B-l9, B-20, and
B-21, respectively, at the end of this appendix.
Though the EPA Trends and the present study
emission profiles are similar to each other, they should
not be expected to match precisely. This is because
the emission estimates developed forthe present study
are based on modeled macroeconomic and emission
sector conditions. Even though the macroeconomic
and sector models themselves are constructed and
calibrated using historical data, modeled replications
of historical trends would not be expected to precisely
capture actual historical events and conditions which
affect emissions. Relying on modeled historical sce-
narios is considered reasonable for the present analy-
sis since its purpose is to estimate the differences be-
tween conditions with and without the CAA. Com-
paring actual historical emissions with modeled no-
control emissions would lead to an inconsistent basis
for comparisons between scenarios. Using models for
both scenarios allows potential model biases to es-
sentially cancel out.
In general, however, these comparisons show
close correspondence between control scenario and
Trends estimates with the largest differences occur-
1 EPA/OAQPS, "National Air Pollutant Emission Trends 1900 - 1994," EPA-454/R-95-011, October 1995.
B-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Figure B-l. Comparison of Control, No-control, and
Trends SO, Emission Estimates.
Figure B-2. Comparison of Control, No-control, and
Trends NCL Emission Estimates.
40
Control
No-Contro!
TRENDS
^ Control
No-Contro]
+ TRENDS
x
GO
J3
(*>
10 -
-A
1960
1970
Year
1980
1970
Year
1990
Figure B-3. Comparison of Control, No-control, and
Trends VOC Emission Estimates.
40
30
G
0
H
V.
-§ c
.3 I 20
1 i
o
I
10
J L
J L
J L
1950
1960
1970
Year
1980
1990
^ Control
. No-Control
.TRENDS
Figure B-4. Comparison of Control, No-control, and
Trends CO Emission Estimates.
200
150
e
£
I a
.s I100
E/l S
c 2
o
50
w
J L
J L
. Control
. No-Control
.TRENDS
1950
1960
1970
Year
1980
1990
Figure B-5. Comparison of Control, No-control, and
Trends TSP Emission Estimates.
C/l
e
o
40
30
B
w
t:
2 C/5
w g
c | 20
10
1950
I L
1960
f Control
. No-Contro]
.TRENDS
1970
Year
1980
1990
B-2
-------
Appendix B: Emissions Modeling
ring for VOC and CO emissions. The Trends report
VOC estimates are generally higher than the control
scenario estimates due to the inclusion of Waste Dis-
posal and Recycling as a VOC source in the Trends
report. This inconsistency is of no consequence since
Waste Disposal and Recycling sources were essen-
tially uncontrolled by the historical CAA and there-
fore do not appear as a difference between the control
and no-control scenarios. The higher CO emission
estimates in the Trends Report are primarily associ-
ated with higher off-highway vehicle emissions esti-
mates. Again, since off-highway emissions do not
change between the control and no-control scenario
in the present analysis, this inconsistency is of no con-
sequence.
No-Control Scenario Projections Versus
Historical EPA Trends Data
Comparisons between the control scenario emis-
sions estimates generated for the present study and
1970 to 1990 emissions estimates obtained from the
Trends Report are useful for assessing the reasonable-
ness of the control scenario estimates. As indicated
above, there is close correspondence between the con-
trol scenario and the Trends Report. It may also be
useful to compare the pre-1970 historical emissions
data from the Trends Report2 with the no-control sce-
nario estimates presented herein to assess whether
these pre-1970 trends can be reasonably extrapolated
to the 1970 to 1990 period. In addition, examination
of any significant changes in emissions trends between
the pre-1970 Trends data and post-1970 no-control
projections might indicate flaws in the emissions
modeling conducted for the present study.
For S02, the 1950 to 1970 Trends data in Figure
B-l demonstrate the effects of the huge increase in
fossil fuel combustion between 1960 and 1970. This
net increase occurred, despite the obsolescence of coal-
fired locomotives and reductions in coal refuse burn-
ing, largely because utility emissions nearly doubled
between 1950 and 1960, and nearly doubled again
between 1960 and 1970.3 Although no-control sce-
nario projections for the post-1970 period show sig-
nificant additional increases in S02 emissions, the rate
of growth is markedly slower than during the 1950 to
1970 period.
The Trends data for 1950 to 1970 NO shown in
X
Figure B-2 indicate the steady increase in emissions
resulting from increased combustion of natural gas
and gasoline.4 The post-1970 emissions estimates
derived for the present study reflect a continuation of
this trend.
Emissions ofVOCs increased steadily over the
1950 to 1970 period, as shown in Figure B-3, prima-
rily due to increases in industrial production and ve-
hicular travel.5 The no-control scenario emission es-
timates continue this trend throughout the 1970to 1990
period, with some acceleration of the rate of change
due to the rapid increase in VMT projected under this
scenario.
The Trends data shown in Figure B-4 for CO in-
dicate an overall increase between 1950 and 1970. This
increase occurred despite significant reductions in
emissions from stationary source fuel combustion and
industrial processes because mobile source emissions
nearly doubled during this period.6 Under the no-con-
trol scenario of the present study, additional reduc-
tions from stationary sources are not available to off-
set the transportation-related increases; therefore, the
rate of increase in CO emissions after 1970 under the
no-control scenario reflects the rapid increase in mo-
bile source emissions caused by increases in vehicle
miles traveled.
Finally, Figure B-5 demonstrates a directional
shift in emissions of primary particulates between the
1950 to 1970 Trends data and the post-1970 no-con-
trol scenario. The declining trend from 1950 to 1970
indicated by the Trends data, however, is largely due
to reductions in use of coal-fired locomotives, reduc-
tions in residential coal-burning, coarse (i.e., visible)
particle emissions controls installed on fossil fuel com-
bustors and industrial processes, and reductions in
forest fires and other open burning.7 Since the reduc-
tions achievable from these sources were largely
2 While 1970 to 1990 Trends data were obtained from more recent Trends reports, the 1950 to 1970 data were obtained from the
November 1991 report since this was the last year the Trends report series included data for this period.
3 U.S. EPA, "National Air Pollutant Emission Estimates, 1940 - 1990", EPA-450/4-91-026, November 1991, Table 4, p. 16.
4 U.S. EPA, "National Air Pollutant Emission Estimates, 1940 - 1990", EPA-450/4-91-026, November 1991, p. 42.
5 U.S. EPA, "National Air Pollutant Emission Estimates, 1940 - 1990", EPA-450/4-91-026, November 1991, p. 42.
6 U.S. EPA, "National Air Pollutant Emission Estimates, 1940 - 1990", EPA-450/4-91-026, November 1991, Table 7, p. 19.
7 U.S. EPA, "National Air Pollutant Emission Estimates, 1940 - 1990", EPA-450/4-91-026, November 1991, Table 3, p. 15.
B-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
achieved by 1970, they are no longer available to off-
set the increases observed from other source catego-
ries (e.g., highway vehicles). The no-control scenario
therefore shows a steady increase in overall emissions
of primary particulates after 1975.
The following sections of this appendix summa-
rize the methodologies used to model control and no-
control scenario emissions for each of the six major
emission sectors. Additional details can be found in
the supporting documents listed in the References sec-
tion of this appendix.
Industrial Boilers and Processes
For the purposes of the retrospective analysis, the
industrial sector was divided into two components:
(1) boilers; and (2) industrial processes and process
heaters. The factors affecting emissions from these
two source types are different, and, as a result, sepa-
rate methods were used to calculate control and
no-control scenario emissions in each of the target
years. To analyze the change in emissions from in-
dustrial boilers, ANL used the ICE model (Hogan,
1988). This model was developed under the auspices
of NAPAP to forecast State-level fuel choice and
emissions from conventional, steam raising, industrial
boilers. For the retrospective analysis of industrial
processes and fuel use emissions from process heat-
ers, ELI used the EPA Trends methods and the ANL
MSCET database (EPA, 1991; Kohoutetal., 1990).
The Trends report contains estimates of national emis-
sions for a variety of industrial sources for the time
period of interest. The MSCET data base provided
the spatial distribution used to calculate State-level
emissions.
The distinction between industrial boilers and non-
boiler industrial processes was necessitated by the
structure of the CAA regulations and by the factors
affecting emission levels from these two source types.
Boilers are regulated differently from processes and
process heaters. Emissions from industrial processes
are primarily a function of levels of industrial activ-
ity. The emissions from fuel combustion, however,
are a function of energy use and fuel choice as well as
industrial activity. Fossil fuel emissions in the absence
of the CAA are not proportional to industrial output,
since the level of energy use is a decision variable for
the firm in its production process. Therefore, in the
ICE model simulations used to estimate no-control
scenario boiler emissions, the level (and type) of en-
ergy use were determined first, and then the effects of
emission regulation were taken into account.
Overview of Approach
Industrial Boilers
ICE model inputs include fuel prices, total boiler
fossil fuel demand by industry type, and environmen-
tal control costs. The outputs of the ICE model were
S02, NOx, and TSP emissions by State, industry, and
boiler size class. The model runs in 5-year increments
and has a current base year of 1985.
The model required boiler demand input data at
the State level. Seven industry types were included in
the ICE model: Standard Industrial Classification (SIC
) codes 20, 22, 26, 28, 29, 33, and "other manufactur-
ing." ANL's approach assumed that industrial boiler
fuel use occurs only in the manufacturing sector. The
model also required fuel price data in each of the tar-
get years at the Federal Region level. Prices by grade
of coal and petroleum product, such as sulfur content
and heating value, were used by the model to deter-
mine the cost of compliance, and to determine emis-
sions when the regulations are not binding.
Control costs were computed by engineering sub-
routines in the model. These costs were used by the
ICE model's fuel choice component to determine the
effect of CAA-related costs on the market share of a
particular fuel. This fuel choice decision only applies
to new industrial boilers, since the cost of existing
emission controls are not in the ICE data base and
fuel choice is not re-evaluated for existing boilers.
Industrial Processes and In-Process Fuel
Combustion
The calculation of historical emissions from in-
dustrial processes uses EPA Trends methods to esti-
mate national emissions for the analysis years, then
allocates these emissions to States using the State
shares from the MSCET data base.
MSCET uses a variety of methods to estimate his-
torical emissions forthe various industrial sectors. For
industrial process emissions, MSCET is based on his-
torical data on industrial activity to allocate emissions
based on the State level distribution of the polluting
activities. The State level distribution and benchmark
B-4
-------
Appendix B: Emissions Modeling
is based on the 1985 NAPAP Inventory (EPA, 1989).
This approach implies that the MSCET data corre-
sponds directly to the 1985 NAPAP Inventory, and
that, for any State, the sum of the emissions from
Source Classification Codes (SCCs ) that comprise
the MSCET industry sector are equal to the MSCET
data for that State and sector. Data from Trends are
used by MSCET to provide information on changes
in the aggregate level of control for years other than
the 1985 benchmark. Since no direct correspondence
existed between the Trends data and MSCET, a rela-
tionship was developed to link MSCET sectors to
Trends industry categories and to industry categories
in the J/W model, which was used to change activity
levels for the no-control scenario.
Table B-l shows the relationship between the sec-
tor definition used by MSCET, Trends, and the J/W
model. The mapping from MSCET to J/W and Trends
is used to provide the changes in aggregate activity
and emission control for the calculation of no-control
scenario emissions.
Establishment of Control Scenario Emissions
Energy use and corresponding emissions were
broken down between boilers and non-boiler indus-
trial processes. The latter category includes furnaces,
kilns, internal combustion engines (e.g., compressors),
and other non-steam types of process heat. The focus
of this analysis is on boiler emissions, which were
subject to increasingly stringent regulations over the
1970 to 1990 period. (Emissions from some types of
industrial processes were also regulated, but regula-
tion of non-boiler sources was targeted on the emis-
sions from the industrial process itself, not on its fuel
combustion) For this study, ANL assumed that only
boiler fuel use is affected by emission regulations. The
non-steam boiler portion of industrial fuel use is not
directly affected by the CAA. This portion of the
emissions may be affected indirectly by changes in
industry activity level and fuel consumption. The
emissions from non-boiler industrial processes were
calculated separately by ELI.
Control Scenario Boiler Emissions
Control scenario boiler S02, NOx, and TSP emis-
sions were calculated by the ICE model. The MSCET
database provided an estimate of historical emissions
for total fossil fuel combustion by industry. Since
MSCET does not identify the two required compo-
nents of boiler and non-boiler emissions, ANL de-
fined the residual of the ICE model control scenario
and MSCET as the non-boiler or in-process fuel use
emissions. For the relevant study period, MSCET pro-
vided a control scenario estimate of total boiler and
non-boiler emissions, which was used to calculate the
control scenario State-level boiler emissions based on
a special run of the ICE model.8
In order to use ICE to model the historical emis-
sions path, it was necessary to construct a new ICE
model base year file and new user input file so that
the model could begin its calculations from 1975 con-
ditions. Construction of the base year file was com-
pleted in two stages, using two different data sources,
as discussed below. The user input file has several
elements, including energy prices and historical boiler
fuel use; its construction is discussed in the next sec-
tion. The model base year file provided the energy
use in boilers and corresponding emission control
regulations (State Implementation Plans -SIPs- for
example) by several categories. These categories in-
clude:
State;
Industry group (one of seven);
Fuel type (natural gas, distillate or residual
fuel oil, and coal);
• Boiler size class (MMBTU/hr, one of eight
categories);
Utilization rate (one of five categories); and
Air quality control region (AQCR).
For the purposes of ANL's analysis, only the first
three categories were assumed to vary. In other words,
for each State, industry, and fuel type combination,
the distribution of boiler size, utilization rate, and
AQCR was assumed to be constant. Over time, how-
ever, changes in the aggregate composition of State,
industry, and fuel type would cause corresponding
changes in the aggregate composition of the other three
characteristics. As mentioned previously, the current
base year file was 1985. The retrospective analysis
required a 1975 base year. Because of data limita-
tions, the approach to construct a new base year was
achieved in the following two steps: the construction
of a 1980 interim base year file from the 1985 file,
and then the construction of the 1975 file from the
interim 1980 file.
8 MSCET does not provide State-level estimates of TSP, while ICE does. To estimate total regional TSP from fuel combustion,
the Trends model was employed. These national emissions estimates were allocated to the States based on the State-level shares of
TSP from the NAPAP inventory.
B-5
-------
Tabic B-1. Correspondence Between Process Emissions Categories Used by MSCET. Trends, and J/W Industrial Sectors and Identifier Codes.
MSCET Category
MSCET Code
Trends Industry Category
J/W Code
J/W Industry Category
FoodProc. andAgric. Operations
FOODAG
Cattle Feed Lots (0211)
1
Agriculture/forestry/fisheries
Cotton Ginning (0724)
Feed and Grain Milling (204)
Grain Elevators (4421,5153)
Mining Operations
MINING
Metallic Ore Mining (10)
2
Metal Mining
Coal Mining (1211)
Oil and Gas Extraction
OILGAS
Crude Oil Production, Storage, and Transfer (1211,4463)
3
Oil & Gas Extraction
Natural Gas Production (1311)
Mining Operations
MINING
Crushed Stone (142)
5
Nonfuel mining
Sand and Gravel (144)
Clays (145)
Potash/Phosphate Rock (1474,1475)
Degreasing
DEGRS
Degreasing
NA
Manufacturing
Misc. Industrial Processes
MISIND
Adhesives
Indus. Organic Solvent Use, Misc.
SOLV
Other Organic Solvent Use
Solvent Extraction Processes
Surface Coating
SRFCT
Surface Coating
NA
Durable Goods
Misc. Industrial Processes
MISIND
Lumber and Plywood (24)
11
Lumber & Wood Products
Cement Production
CEMNT
Cement (3241)
19
Stone, Clay, & Glass Products
Glass Manufacturing
GLASS
Glass (321,322)
19
Stone, Clay, & Glass Products
Lime Manufacturing
LIME
Concrete, Lime, Gypsum (327)
19
Stone, Clay, & Glass Products
Lime (3274)
Mineral Products Processing
MINRL
Clay Sintering (3295)
19
Stone, Clay, & Glass Products
Brick and Tile (3251)
Iron and Steel Production
IRNST
Iron and Steel (3312)
20
Primary Metal Industries
Ferroalloys (3313)
Iron and Steel Foundries (332)
Other Primary Metals Smelting
OTHMET
20
Primary Metal Industries
Primary Aluminum Smelting
PALUM
Primary Aluminum (3334)
20
Primary Metal Industries
Primary Copper Smelting
PCOPR
Primary Copper (3331)
20
Primary Metal Industries
Primary Lead and Zinc Smelting
PLDZC
Primary Lead and Zinc (3332,3333)
20
Primary Metal Industries
Other Sec. Metal Smelting and Refining
SECMET
Primary Nonferrous Smelters (333)
20
Primary Metal Industries
Other Sec. Metal Smelting and Refining
SECMET
Secondary Nonferrous Smelters (334,336)
20
Primary Metal Industries
Secondary Lead Refining
SLEAD
Secondary Lead (3341)
20
Primary Metal Industries
FoodProc. andAgric. Operations
FOODAG
Food and Beverages (20)
7
Food & Kindred Products
Misc. Industrial Processes
MISIND
Textiles (22)
9
Textile Mill Products
Paper and Pulp Mills Operations
PAPER
Pulp Mills (261,262)
13
Paper & Allied Products
Misc. Industrial Processes
MISIND
Graphic Arts (27)
14
Printing & Publishing
Printing Operations
PRINT
14
Printing & Publishing
Organic Chemicals Manufacture
ORGCM
Organic Chemicals (286)
15
Chemicals & Allied Products
Carbon Black (2895)
Other Chemicals Manufacture
OTHCM
Ammonia (2873)
15
Chemicals & Allied Products
Nitric Acid (2873)
Chemicals (28)
Sulfuric Acid (2819)
Petroleum Refining
PTREF
Petroleum Refining (2911)
16
Petroleum & Coal Products
Asphalt Paving and Roofing (295)
16
Petroleum & Coal Products
Plastics Production
PLAST
Plastics (2821,3079)
17
Rubber and plastic products
Rubber and Misc. Plastics Manufacture
RUBR
Rubber Tires (3011)
17
Rubber and plastic products
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Appendix B: Emissions Modeling
Estimates of boiler fossil fuel consumption in
1980 for each State and major fuel type were pro-
vided by Hogan (Hogan, 1988). These estimates are
based on the assumption that the industry mix, size,
utilization, and AQCR distribution within a State are
constant. Through assuming this relationship, the 1985
ICE base year was scaled to match the data for 1980,
thus forming the 1980 interim base year data.
To construct the 1975 base year file, the assump-
tion of a constant industry mix for a State and fuel
type was no longer necessary, since detailed data on
each industry for 1980 and 1975 were available from
PURchased Heat And Power (PURHAPS) model data
files (Werbos, 1983). These PURHAPS data files were
derived from the Annual Survey of Manufactures:
Fuels and Electric Energy Purchased for Heat and
Power (DOC, 1991). The available data in these files
were for total fuel use not boiler fuel use. To make
use of these data, it was necessary to assume that the
fraction of fuel used in boilers, for any given State
and industry, remained constant from 1975 to 1980.
To the extent that the fraction of boilers' heat versus
process heat applications is a function of the specific
industrial production process, this assumption is rea-
sonable.
Based on the assumption of constant boiler fuel
fraction of total fuel use, the ratio of 1975 to 1980
energy use for each State, industry, and fuel type was
applied to the corresponding record of the 1980 in-
terim base year file to produce 1975 base year files.
Control Scenario Industrial Process Emissions
To estimate boiler emissions of sulfur oxides
(SOx), NOx, and VOC from industrial processes, data
from Trends were used. The percentage change in
national emissions by Trends category was applied to
the appropriate sector from MSCET to obtain State-
level emissions. In some cases there are several cat-
egories in Trends that match directly with MSCET
categories (see Table B-l). In these cases, the Trends
sectors were aggregated and the percentage change
was computed. It was assumed that the level of con-
trol in each industry sector implied by Trends was
uniform across States. The changes in emissions in
each State are not equal to those at the national level,
since the industry composition in each State varies.
Development of Economic Driver
Data for the Control Scenario -
Industrial Boilers and Processes
The results of the J/W model were the primary
source of activity in the ICE model driver data. These
results were also used by ELI to produce the national
results for industrial processes from Trends. Both ICE
and Trends use the forecasted change in industrial
activity that results under the no-control scenario.
These data were in the form of industry specific
changes in energy consumption and industrial output,
for boilers and industrial processes.
Economic Driver Data for Industrial
Boiler Approach
Using the 1975 base year file as a starting point,
the ICE model estimated fuel choice and emissions
based on a user input file containing total boiler en-
ergy demand and regional energy prices. The 1975,
interim 1980, and original 1985 base year files con-
tained the required information on energy demand for
each industry group and State, so the data in these
three files were aggregated across fuel type, and other
boiler characteristics (for example, size). These ag-
gregated data provided the energy demand for three
of the target years. Since 1990 State-level data on
energy use by industry group were not available at
the time of the study, the NAPAP base case forecast
for the ICE model for 1990 was used to provide the
demand data for this year.
The user input file for ICE also requires a price
input for each target year. These prices were input by
Federal Region for distillate oil, 4 grades of residual
oil (by sulfur content), natural gas, and 11 grades of
coal (by sulfur content and coal rank, i.e., bituminous
and sub-bituminous). Prices for 1985 and 1990 were
obtained from the NAPAP base case user input file.
The prices for 1975 and 1980 are from U.S. Depart-
ment of Energy (DOE) data on State-level industrial
energy prices (DOE, 1990). Regional prices of natu-
ral gas, distillate oil, steam coal, and residual oil were
constructed by aggregating expenditures across States
within each region and dividing by total British ther-
mal unit (BTU) consumption forthe years 1975,1980,
and 1985. Since prices by sulfur content grade are not
reported by this DOE source, ANL assumed that the
sulfur premium implied by the 1985 ICE model input
file was proportional to the average price. Based on
this assumption, the ratio of the regional coal and re-
B-7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
sidual oil price in 1975 and 1980 to the 1985 price
was applied to the 1985 price in the ICE model base
case file for each grade of fuel. To provide additional
consistency between the NAPAP analysis and ANL's
study, the distillate oil and natural gas prices were
benchmarked to the 1985 ICE model prices as well.
One possible inconsistency arises using this pro-
cedure. The residual oil and natural gas markets are
closely linked, particularly for industrial customers.
These markets, specifically the gas market, underwent
tremendous changes over the study period. To model
the effect of these structural changes on the sulfur pre-
miums in residual oil would require a detailed oil and
gas supply model that was beyond the scope of this
project. Moreover, the CAA regulations themselves
create the potential for sulfur premiums. This poten-
tial effect of the CAA was not captured, though, be-
cause of the assumption of proportional fuel sulfur
premiums on residual fuel oil. The relationship be-
tween market driven sulfur premiums in the coal mar-
ket and the CAA was given additional consideration
in this analysis through the use of an explicit coal sup-
ply model.
The J/W data for industrial energy consumptions
was supplied in the form of percentage change in cost
shares. In order to compute the percentage change in
the quantity of energy used, ANL used the following
identity:
1,1 (7wT}= 1,1 ^ + 1,1 (E)' 1,1 {1>'J x ~}-or(1)
1,1 (lf~o~}'1,1 + 111 (,'>j x (~)} = 1,1 {h:)-or(2)
Q
The percentage change in E is the percentage
change in cost share, minus the change in price, plus
the change in value of shipments. These calculations
were performed for each energy type and industry
sector in the J/W model. The ICE model requires to-
tal fuel use, so the fuel specific percentages were
weighted by historical fuel consumption to produce
an aggregate change in fuel consumption to apply to
the ICE model input data files.9
ICE also uses energy prices to simulate boiler fuel
choices. The control scenario forecasts of energy
prices in ICE were adjusted based on the percentage
changes in energy prices, by coal, oil and natural gas.
This implicitly assumes that the oil and coal fuel sul-
fur premiums, by region, are proportional to the aver-
age national price. To test this assumption forthe coal
market, additional modeling of the coal prices was
performed using the coal market component of the
ARGUS model.
It is possible that in some regions low sulfur coal
prices to the industrial sector may be lower than the
national average. This was not found to be the case.
For example, in 1990, delivered regional industrial
coal prices change by less than two-thirds of one per-
cent. In most cases, the percentage change was near
zero. This result appears to occur because of the highly
regional nature of the coal market. While the artifi-
cial demand for low sulfur coal may fall, power plants
near low sulfur coal reserves now find it advantageous
to buy this local coal, which raises the price back to
an equilibrium level near to that of the control sce-
nario. This is even more likely to be true of industrial
delivered prices, since industrial prices are more af-
fected by transportation costs than are the utility prices.
No additional ICE modeling was performed.
Economic Driver Data for the Industrial
Process Approach
The J/W model was also used to account for ac-
tivity level changes in the calculation of industrial
process emissions under the no-control scenario. The
correspondence between Trends, MSCET, and the J/
W model was used to apply changes in industrial ac-
tivity in each target year to each industrial process.
No-control Scenario Emissions
Industrial Boiler Emissions of S02, NOx, and TSP
The CAA imposed different regulations, SIPs, and
New Source Performance Standards (NSPS) that ap-
ply to industrial boilers of varying size. The primary
effect of CAA regulations on industrial boilers was
simulated by defining the Air Quality Control Region
(AQCR), the resulting SIPs, and subsequent NSPS for
boilers. The industrial boiler SIP regulations were in-
cluded in the ICE base year file discussed in the pre-
vious section. Since the ICE model estimates new
boiler emissions for each target year, the boiler NSPS
are input through the ICE user files. Industrial NSPS
were implemented in two phases. The 1971 regula-
tions are imposed forthe study years 1975 and 1980.
9 ICE uses six of the manufacturing industries from the J/W model directly. The remaining industries' percentage changes were
weighted to produce the "other" category.
B-8
-------
Appendix B: Emissions Modeling
The 1984 NSPS revisions are imposed in the study
years 1985 and 1990. Forthe no-control scenario, ANL
set the SIPs and NSPS to a flag that indicated "no
regulation."
Industrial Boiler Emissions of CO and VOC
Two of the criteria pollutants emitted by indus-
trial fuel combustors, CO and VOC, were not included
as outputs of the ICE model. Therefore, CO and VOC
emissions were analyzed separately using Trends
methods. Control scenario CO and VOC emissions
were taken directly from Trends.
To estimate CO and VOC emissions from indus-
trial combustion for the no-control scenario, fuel use
for industrial manufacturing was adjusted, reflecting
fuel consumption changes estimated by the J/W model.
These changes in the level of fuel consumption by
industrial combustion were also used in ANL's ICE
boiler model. Changes in industrial combustion fuel
use by manufacturing between the control and
no-control scenarios are reported in Table B-2. These
estimates represent an average of several sectors,
which were developed by ANL as part of the model-
ing process for ICE.
No-control scenario emissions were computed
using 1970 emission factors. Since there were no add-
Tablc B-2. Fuel Use Changes Between
Control and No-control Scenarios.
Year
Fuel Type
Fuel Use Changes
Coal
-.0042
1975
Oil
+.031 1
Gas
-.0064
Coal
-.0061
1980
Oil
+.0 107
Gas
-.0095
Coal
-.0061
1985
Oil
+.0089
Gas
-.0097
Coal
-.0079
1990
Oil
+.0091
Gas
-.0099
on controls for industrial combustion VOC and CO
emissions, it was not necessary to adjust the no-con-
trol scenario for changes in control efficiency.
Emission estimates were regionalized using State-
level emissions data from industrial boilers recorded
in MSCET. Forthe control scenario estimates, VOCs
were regionalized using the MSCET State-level shares
for industrial fuel combustion. In the no-control sce-
nario, the State-level shares were held constant. The
control scenario emissions of CO were regionalized
using the control scenario NOx emissions from the ICE
model. This approach assumes that CO emissions are
consistent with NO emissions. The no-control see-
X
nario CO emission estimates from industrial combus-
tion sources were regionalized using no-control NOx
emission estimates from industrial combustion
sources.
Industrial Process Emissions
A wide range of controls were imposed on indus-
trial processes. These emission limits are embodied
in the assumptions of control efficiencies in the Trends
model. Data on national no-control scenario emissions
from industrial processes were provided by EPA.
These data were combined with MSCET to produce
regional-level results.
Lead Emissions
Estimates of lead emissions from industrial boil-
ers and industrial processes were completed by Abt
Associates. The methods used for calculating lead
emissions from industrial processes and industrial
boilers were similar. The starting point was the TRI,
which provides air toxics emissions data for manu-
facturing facilities with more than 10 employees. To
estimate lead emissions from industrial boilers and
processes, 1990 facility-level lead emissions data were
extracted from the TRI. These data were then adjusted
to create estimates of lead emissions from industrial
sources under the control and no-control scenarios for
each of the target years. Forthe control scenario, lead
emissions for 1975, 1980, and 1985 were obtained by
extracting an emission factor and a control efficiency
for each lead-emitting industrial process in the Trends
database. These emission factors and control efficien-
cies were multiplied by the economic activity data
for each year for each process as reported in Trends
to yield estimated control scenario emissions by in-
dustrial process. Each industrial process was assigned
B-9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
a code to correspond with energy consumption data
by industrial process compiled in the National Energy
Accounts (NEA) by the Bureau of Economic Analy-
sis, and emissions were summed over all processes to
obtain a total for each target year.
For consistency with the other emission estimates
in this analysis, industrial process no-control scenario
lead emissions were adjusted for changes in indus-
trial output, and for changes in emissions per unit of
output due to control technology applications. Changes
in industrial output were accounted for using results
from the J/W model. Lead-emitting industrial pro-
cesses in the Trends data base were assigned to a J/W
sector. For each sector, the percentage change in eco-
nomic output was used to adjust the economic activ-
ity data for that process from the Trends data base.
These adjusted economic output figures were used
with the 1970 emission factors and control efficien-
cies to derive the estimated no-control scenario lead
emissions for each industrial process in each target
year. The process-level emissions were then aggre-
gated to the NEA-code level as in the control sce-
nario.
The lead emission estimates from industrial pro-
cesses, by NEA code, were used to derive percentage
changes in emissions under the control and no-control
scenarios by NEA code for application to the TRI
emissions data. Since TRI data are reported by SIC
code, NEA codes were "mapped" to the appropriate
SIC codes, and then the percentage change for each
NEA code was used to represent the percentage change
for all SIC codes covered by that NEA code.
To calculate lead emissions from industrial boil-
ers, Abt Associates developed estimates of lead emis-
sions from industrial combustion under the CAA for
each of the target years. The Trends data base con-
tains national aggregate industrial fuel consumption
data by fuel type. For each fuel type, the fuel con-
sumption estimate was disaggregated by the share of
that fuel used by each NEA industrial category. The
Trends data base also contains emission factors for
industrial fuel use, by fuel type, as well as control
efficiencies. The lead emissions from industrial com-
bustion for each NEA category were derived by mul-
tiplying the fuel-specific combustion estimate for each
NEA category by the emission factor and control ef-
ficiency for that fuel type. The result was emissions
of lead by NEA code and by fuel type. Emissions from
all fuel types were then summed by NEA code. The
NEA data were used to disaggregate the industrial fuel
consumption figures, based on the assumption that the
ICE are the same among all industries covered by a
given NEA code.
To estimate no-control scenario lead emissions,
the macroeconomic effect of the CAA and the change
in emissions per unit of output that resulted from spe-
cific pollution control mandates of the CAA were both
taken into account. As in the control scenario, the na-
tional aggregate industrial fuel consumption estimate
by fuel type was disaggregated by the share of that
fuel used by each NEA industrial category. The fuel
use was then adjusted in two ways: some NEA codes
were specifically modeled by the ICE model, and for
the remaining NEA codes, J/W percentage changes
in fuel use were applied. These fuel use estimates were
then combined with the 1970 emission factors and
control efficiencies for industrial combustion by fuel
type from the Trends data base to obtain no-control
scenario combustion-related lead emissions from in-
dustrial boilers by NEA code. These estimates of to-
tal lead emissions by NEA codes were matched to
SIC codes, and then to the data in the TRI data base.
This approach assumed that an average emission value
was assigned to all reporting TRI facilities in a given
SIC code.
Off-Highway Vehicles
The off-highway vehicle sector includes all trans-
portation sources that are not counted as highway ve-
hicles. Therefore, this sector includes marine vessels,
railroads, aircraft, and off-road internal combustion
engines and vehicles. As a whole, off-highway ve-
hicle emissions are a relatively small fraction of total
national anthropogenic emissions.
Overview of Approach
The process used by ELI to determine the national
level of emissions from the off- highway transporta-
tion sector is similar to the procedure outlined above
for industrial processes. To estimate the emissions of
criteria air pollutants from these sources under the
no-control scenario, the historical activity levels were
held constant, rather than attempting to calculate a
new no-control scenario level of off-highway vehicle
activity. This assumption was necessary since the off-
highway activity indicators (amount of fuel consumed,
and landing and take-off cycles for aircraft) do not
B-10
-------
Appendix B: Emissions Modeling
have direct correspondence with a given J/W category.
The national no-control scenario emissions of criteria
air pollutants from these sources were simply derived
by recalculating emissions using 1970 emission fac-
tors.
Development of Control Scenario
To estimate control scenario emissions, the analy-
sis relied on Trends methods, using historical activity
indicators, emission factors, and control efficiencies.
Essentially, the estimates of off-highway emissions
under the control scenario represent the historical es-
timates from the Trends data base.
No-control Scenario Emissions Estimates
The calculation of off-highway emissions for the
no-control scenario required the Trends data to be
adjusted to reflect changes in controls and economic
activity in each of the target years. Linking source
activity changes with economic activity for this sec-
tion is not straightforward. The economic activity data
for off-highway engines and vehicles are expressed
either in terms of amount of fuel consumed, or in terms
of landing and take-off cycles for aircraft. Neither of
these off-highway activity indicators has a direct cor-
respondence with a given J/W sector, making the sort
of direct linkage between Trends categories and J/W
sectoral outputs that was used for industrial processes
inappropriate.
In the absence of a link between the economic
factors that are determinants of emissions from this
sector and the available economic activity forecasts,
the no-control scenario emissions of criteria air pol-
lutants from off-highway mobile sources were esti-
mated based on the same historical activity levels used
for the control scenario. Although there were changes
in sectoral output and personal income that might have
had an effect on off-highway vehicle usage, these
changes were deemed to be small and not likely to
have a major effect on the emissions from this sector.
Emission factors for each of the off-highway
sources were also held constant at 1970 levels to cal-
culate no-control scenario emissions for each target
year. The national emissions of criteria air pollutants
from these sources were then recalculated using 1970
emission factors.
National and State-Level Off-Highway
Emission Estimates
Table B-3 summarizes national-level emission
estimates for off-highway sources. The emission es-
timates derived from using the methodology discussed
above yielded results that seem counter-intuitive. The
emissions from off-highway sources, in particular the
emissions from aircraft, are lower in the no-control
scenario than those projected for the control scenario
for most pollutants. This is a result of calculating
emissions using 1970 emission factors, since the 1970
emission factors for aircraft are lower than the air-
craft emission factors in later years.
ELI identified several potential sources of uncer-
tainty in the emission estimates for this sector. First,
the assumption that the total level of off-highway ve-
hicle fuel consumption is constant between the two
scenarios may be flawed. Second, the use of 1970
emission factors in the no-control scenario may fail
to capture significant changes in technology. These
technological changes are implicitly captured in the
control scenario and it is possible that these techno-
logical changes may also have occurred under a
no-control scenario.
One possible response to the biases created by the
use of 1970 emission factors for all years in the
no-control scenario is to test how results might differ
if the emission factors used for the control scenario,
which would include technological change, were also
used for the no-control scenario. However, using this
treatment of emission factors, the emissions projec-
tions from the adopted methodology from non-high-
way sources in the no-control scenario would be iden-
tical to the emissions projections under the control
scenario. The reason for this is that the economic ac-
tivity levels were not adjusted for the calculation of
emissions under the no-control scenario.
In order to disaggregate the national data to a State
level, the methodology used the MSCET data base,
which is described earlier. Emissions of VOC, SO ,
' x'
and NOx were regionalized using the State-level shares
from the MSCET methodology. The emissions ofTSP
were regionalized by using the State-level shares for
SOx reported by MSCET, and the emissions of CO
were regionalized using the State-level shares for NOx,
also reported by MSCET. The potential bias that this
introduces is likely to be small, due to the relative
homogeneity of off-highway vehicle emission sources.
B-ll
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic B-3. Difference in Control and No-control Scenario Off-Highway Mobile Source
Emissions.
1975
1980
1985
1990
TSP
Control Scenario:
268.6
281.1
268.7
280.9
No-Control Scenario:
260.8
268.8
261.2
266.9
Percentage Increase:
-3%
¦4%
-3%
-4%
NO,
Control Scenario:
1,987.6
2,176.7
2,077.5
2,085.9
No-Control Scenario:
1,974.6
2,150.5
2,042.7
2,058.9
Percentage Increase:
-1%
-1%
-2%
-1%
SO,
Control Scenario:
364.6
531.1
406.4
392.5
No-Control Scenario:
363.2
528.6
403.0
386.9
Percentage Increase:
0%
0%
-1%
-1%
CO
Control Scenario:
8,512.8
8,101.4
7,881.9
8,079.0
No-Control Scenario:
8,511.0
8,071.2
7,880.2
8,077.7
Percentage Increase:
0%
0%
0%
0%
VOCs
Control Scenario:
1,374.9
1,370.8
1,334.8
1,405.0
No-Control Scenario:
1,385.9
1,416.1
1,388.6
1,485.8
Percentage Increase:
1%
3%
4%
6%
Note: Emission estim ates are expressed in thousands of short tons. Percentage increase is the differential between
scenarios divided by the Control Scenario projection.
As with regionalization of industrial process emis-
sions, the State-level shares are held constant between
the two scenarios. To the extent that the distribution
of economic activity between States was not constant
over the period of the analysis, holding State-level
emission shares constant may bias the results, although
the direction and magnitude of the potential bias is
unknown.
On-Highway
This section addresses the highway vehicle por-
tion of the transportation sector. Highway vehicle
emissions depend on fuel type, vehicle type, technol-
ogy, and extent of travel. Emissions from these ve-
hicles have been regulated through Federal emission
standards and enforced through in-use compliance
programs, such as State-run emission inspection pro-
grams. Vehicle activity levels are related to changes
in economic conditions, fuel prices, cost of regula-
tions, and population characteristics. Emissions are a
function of vehicle activity levels and emission rates
per unit activity.
TEEMS was employed by ANL to analyze the
transportation sector. The modeling system links sev-
eral models, disaggregate and aggregate, to produce
State-level estimates of criteria pollutants. The sys-
tem is subdivided into two modules: an activity/en-
ergy module and an emissions module. Each module
contains multiple models. TEEMS has been docu-
mented in several reports and papers (Mintz and Vyas,
1991; Vyas and Saricks, 1986; Saricks, 1985). It has
been used for several policy analyses and assessment
studies for DOE and NAPAP. This section presents
an overview of the approach used to conduct the analy-
sis of the transportation sector. Also included in this
section is a summary of the methodology used by Abt
Associates to estimate changes in lead emissions from
highway vehicles in each target year.
B-12
-------
Appendix B: Emissions Modeling
Overview of Approach
TEEMS has two modules: an activity/energy
module and an emissions module. The activity/energy
module calculates emissions based on: (1) personal
travel; (2) goods movement; and (3) other transporta-
tion activity inputs.
Personal Travel
Personal travel activity and resulting fuel con-
sumption were calculated for each target year using
procedures that disaggregate households by demo-
graphic and economic attributes. Economic driver
data, developed from U.S. Government data and mac-
roeconomic model(s) of the domestic economy,
formed the basis for household disaggregation. Mod-
eling procedures were employed by ANL to project
movement of households between various attribute
classes, and vehicle holdings were projected in terms
of the number and type of vehicles held by each house-
hold type. National totals were then developed by
aggregating the vehicle holding estimates for each
household type, accounting for the number of house-
holds of that type. Travel estimates, in terms ofVMT,
were calculated using the same approach, and based
on the VMT of each household type. The basis for
household transportation activity projection has been
empirically established through analysis of the 1983-
84 Nationwide Personal Transportation Survey
(NPTS) (FHWA, 1986; Mintz and Vyas, 1991). VMT
are projected using this empirical relationship, and es-
timates of the elasticity ofVMT to vehicle operating
cost are then made. Energy consumption was esti-
mated in each target year using VMT, shares ofVMT
by vehicle type, and exogenously developed vehicle
characteristics.
The following three models and an accounting
procedure were employed to develop target year per-
sonal travel activity projections:
1. The first model projected the target year dis-
tribution of households by their attributes.
This model employed an iterative proportional
fitting (IPF) technique and projected the num-
ber of households in each cell of the house-
hold matrix - each of which is defined by vari-
ous categories within six household attributes.
2. The second model projected changes in ve-
hicle ownership resulting from changes in
income and cost of vehicle operation. The
model applied estimated ownership changes
to each target year household matrix such that
the control values within each of the house-
hold attributes, excepting vehicle ownership,
remained unchanged.
3. The third model estimated the composition
of household vehicle fleet by type (cars and
trucks), size, technology, and fuel.
4. An accounting procedure applied VMT per
vehicle to vehicle ownership in each combi-
nation of household attributes. VMT and en-
ergy consumption were accumulated by ve-
hicle type, size, and fuel.
Each of these models is described separately in
the following subsections.
Iterative Proportional Fitting (IPF)
This IPF model modified a control scenario ma-
trix of household counts. A household matrix was
developed from the 1983 NPTS data and upgraded to
the year 1985 using published aggregate data. The
procedure used in constructing the 1985 household
matrix has been documented elsewhere (Appendix B
of Mintz and Vyas, 1991). The matrix is defined by
six attributes: (1) residential location (central city,
suburb, rural); (2) household income; (3) age of house-
holder; (4) household size; (5) number of drivers; and
(6) number of vehicles. The household matrix has
3,072 cells, some of which are illogical (such as 1
person, 2 drivers). Illogical cells were replaced with
zeros.
Household shares within each attribute in each
target year were developed exogenously using data
from the Bureau of the Census and selected macro-
economic model runs. The projected total of house-
holds and shares of households in each category of an
attribute were supplied to the IPF model. The model
modified the control scenario household matrix to
match the specified shares and total number of house-
holds.
The IPF model treated household distribution
within each attribute as a set of vectors. These vectors
were scaled to match the specified shares and house-
hold total. Following the initial scaling, a gradual scal-
ing technique was used to move in the direction of the
target shares. The scaling process was repeated until
closure was achieved for all attribute classes. Since
B-13
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
vehicle ownership levels were estimated by the ve-
hicle ownership model (described in the next section),
shares within the sixth household attribute (number
of vehicles held) were not specified, leaving it uncon-
trolled. This flexibility of an uncontrolled attribute
helped to facilitate the model operation. The number
of households in each class of vehicle ownership
within the output matrix represents distribution of
households using the control scenario (1985) relation-
ship of vehicle ownership to other household at-
tributes.
Vehicle Ownership Projection (VOP)
The VOP model projected the changes in vehicle
ownership resulting from changes in the number of
licensed drivers, disposable personal income, and an-
nual fuel cost of vehicle operation. The model is based
on historical household ownership rates. A target per-
driver ownership rate was computed using disposable
income and fuel cost. This target rate represented de-
sired ownership if income and fuel cost were the only
determinants. A parameter representing ownership
responsibilities such as acquisition effort, disposal
effort, parking requirements, and other indirect aspects
was applied to adjust this target. The new ownership
rate was used to estimate the number of household
vehicles.
The household matrix created by the IPF model
was revised to match the projected household vehicle
ownership. Household shares within the first five at-
tributes remain constant while those within the sixth
attribute (i.e., number of vehicles) were variable. A
deviation measure was defined and its value for each
class within the first five attributes was minimized. A
set of simultaneous equations was solved using
Lagrangian multipliers.
Projection of Vehicle Fleet Composition
The composition of household vehicles was pro-
jected for each household matrix cell using a vehicle
choice model called the Disaggregate Vehicle Stock
Allocation Model (DVSAM ). Vehicles are defined
by type (auto, light truck), size (small, mid-size, full-
size auto; small pickup, small utility/minivan, stan-
dard pickup, large utility/standard van; or any other
size classification), fuel (gasoline, diesel, methanol,
ethanol, or compressed natural gas), and technology
(stratified charge, direct injection, electric, fuel cell,
or Brayton).
The model computed vehicle composition based
on an individual vehicle's utility to households and
household needs. A menu of vehicles classified by
the previously mentioned vehicle attributes was sup-
plied to the model. The menu specified characteris-
tics of each vehicle available to households. Vehicles
were characterized by price, operating cost, seating
capacity, curb weight, and horsepower. These vari-
ables formed the basis for computing "utility" (analo-
gous to consumer satisfaction). The household ma-
trix provided demographic and economic attributes
which, when combined with vehicle usage in miles,
define household needs. Vehicle usage (VMT) was
computed as a function of income, number of drivers,
and number of vehicles. A logit model was applied to
compute vehicle ownership shares. Several model en-
hancements facilitated modeling of limited range ve-
hicles, and representation of supply constraints and/
or regulated market penetration.
Activity/Energy Computation
An accounting procedure was applied to compute
personal travel activity in terms of VMT by vehicle
type. Control scenario VMT per vehicle estimates for
each cell in the household matrix were developed from
the 1983 NPTS. These rates were adjusted within the
procedure on the basis of changes in average vehicle
operating cost per mile for each cell. The vehicle com-
position projection model computes ownership shares
and share-weighted change in vehicle operating cost.
Elasticity values were applied to this change.
ANL assumed that VMT per vehicle remained
nearly unchanged for a household matrix cell over time
(with the exception of the effect of changes in vehicle
operating cost). In other words, variation of VMT
across household types is far greater than within house-
hold types. VMT per household vehicle remained
stable during the period from 1977 to 1984 (Klinger
and Kuzmyak, 1986). Some increases were observed
in recent years, which were attributed to lower fuel
prices and increased household income (DOC, 1991;
FHWA, 1992). (A portion of the increase could be
attributed to the method of computing average VMT
per vehicle.) The assumption that VMT per vehicle
for each cell remained nearly constant and was elas-
tic relative to vehicle operating cost is reasonable. As
households move from one cell of the matrix to an-
other, they "acquire" the VMT per vehicle rate of that
cell. Thus, this approach accounted for changes in
VMT per vehicle due to increased household afflu-
ence, increased rate of driver licensing, changes in
fuel price, and changes in vehicle technology.
B-14
-------
Appendix B: Emissions Modeling
Goods Movement
Energy and activity demand resulting from move-
ment of 24 aggregate categories of commodities is
estimated by this subcomponent of the TEEMS activ-
ity module. Changes in commodity demand/produc-
tion were provided by growth indexes by two-digit
SIC generated by a macro model. A model that
projects shifts in mode shares among truck, rail, ma-
rine, air, and pipeline modes was used, followed by a
procedure to compute ton miles of travel for each
mode, VMT by fuel type for trucks, and energy con-
sumption by operation type for non-highway modes.
The model used 1985 control scenario data, which
were compiled from railroad waybill sample and pub-
lications, waterborne commerce publications, trans-
portation statistics, and other sources. The procedure
used in developing the 1985 control scenario freight
data has been documented in an ANL report
(Appendix A of Mintz and Vyas, 1991).
This goods movement model was not used for this
retrospective analysis because of funding and time
constraints. A procedure to estimate truck VMT by
fuel type was employed in its place. Published his-
torical VMT values (FHWA, 1988; 1992) were used
along with VMT shares by fuel and truck type from
Truck Inventory and Use Surveys (TIUS) (DOC, 1981;
1984; 1990).
Other Transportation Activities
The activity/energy module also has other mod-
els for developing activity and energy use projections
for air, fleet automobiles, and bus modes. Fleet auto-
mobile activity estimates from an earlier study (Mintz
and Vyas, 1991) were used while other modes were
not analyzed.
Lead Emissions
Estimates of lead emissions in the transportation
sector were developed by Abt Associates based on
changes in reductions of lead in gasoline. This esti-
mation required the estimates of lead in gasoline con-
sumed over the period from 1970 to 1990 and the
amount of lead content in gasoline that would have
been consumed in the absence of the CAA. These
values were calculated using the quantity of both
leaded and unleaded gasoline sold each year and the
lead concentration in leaded gasoline in each target
year. Data on annual gasoline sales were taken from a
report by ANL that presented gasoline sales for each
State in each target year. For the control scenario, data
on the fraction of gasoline sales represented by leaded
gasoline were used. For the no-control scenario, all of
the gasoline sold was assumed to be leaded. Data on
the lead content of gasoline was obtained from ANL
for 1975 through 1990. For 1970 through 1975, the
analysis assumed that the 1974 lead content was used.
Estimation of No-control Scenario
Emissions
TEEMS emissions projections were carried out
by ANL in the following three steps:
1. Development of emission factors;
2. Allocation of highway activity to States; and
3. Development of highway pollutant estimates.
The following subsections describe the procedures
used for computing highway vehicle emissions.
Development of Emission Factors
EPA's MOBILE5a Mobile Source Emission Fac-
tor model was used to provide all of the highway ve-
hicle emission factors used to estimate 1975 to 1990
emission rates (EPA, 1994b). Documentation of the
MOBILE5a model is found in the User's Guide for
the MOBILE5 model.10
Although the actual emission factors used by ANL
are not documented in either the original ANL TEEMS
model report or in the Pechan summary report, the
Project Team provided direction that defined the emis-
sion factors to be used. For the control scenario, ANL
was directed to use the official EPA emission factors
prevailing at the time for each target year. For ex-
ample, the official EPA emission factor being used in
1980 for on-highway vehicle NOx was to be used to
estimate 1980 control scenario on-highway vehicle
NO emissions. For the no-control scenario, the offi-
X '
cial EPA emission factors used to estimate emissions
in 1970 were to be used throughout the 1970 to 1990
period.
It is important to note that using the 1970 on-high-
way vehicle emission factors to estimate no-control
scenario emissions for the entire 1970 to 1990 period
may bias scenario emission differentials upward. This
is because it is possible that technological changes to
on-highway vehicles unrelated to CAA compliance
10 EPA/OAR/OMS, "User's Guide to MOBILE5," EPA-AA-AQAB-94-01, May 1994; see also 58 FR 29409, May 20, 1993.
B-15
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
strategies may have yielded incidental reductions in
emissions. However, EPA Office of Mobile Sources
(EPA/OMS) experts indicate that the two major tech-
nological changes in vehicles occurring during the
period of the analysis -electronic ignition and elec-
tronic fuel injection- would have yielded negligible
emission reductions in the absence of catalytic con-
verters.11
Another potential bias is introduced by assuming
the CAA had no substantial effect on vehicle turn-
over. However, two factors render this potential bias
negligible. First and foremost, under the no-control
scenario retired vehicles would be replaced by new
but equally uncontrolled vehicles. Second, no-control
scenario vehicle use is greater in terms of VMT per
year. This means no-control scenario vehicles would
reach the end of their service lives earlier, offsetting
to some extent the alleged incentive to retire vehicles
later due to costs imposed by CAA control require-
ments.
Allocation of Highway Activity to States
TEEMS' activity module generated national ac-
tivity and energy estimates. These activity totals were
allocated to States through a regionalization algorithm
that used time series data on historical highway activ-
ity shares by State. A trend extrapolation methodol-
ogy was used that stabilizes shifts after 5 years in the
future. For the retrospective analysis, historical high-
way activity shares for each target year were devel-
oped using data published by the Federal Highway
Administration (FHWA) (FHWA, 1988; 1992).
Development of Highway Pollutant Estimates
Highway emission estimates were calculated in
both scenarios for each target year using VMT esti-
mates generated by TEEMS and emission factors from
MOBILE5a. Control scenario activity levels were
adjusted for the no-control scenario using economic
forecasts and historical data.
Control Scenario Emissions Calculation
Control scenario data for the transportation sec-
tor were compiled from several sources. Household
counts and shares of households by six attributes were
obtained from various editions of the Statistical Ab-
stracts of the United States. Household income infor-
mation was obtained from the control scenario run of
the J/W model. Fuel prices were obtained from the
Annual Energy Review (DOE, 1992) while vehicle fuel
economy and aggregate VMT per vehicle were ob-
tained from Highway Statistics (FHWA, 1988; 1992).
B-4 lists data sources for the control scenario run.
Table B-5 shows household shares prepared for
the IPF model. The total number of households in-
creased from 63.4 million in 1970 to 93.3 million in
1990. A gradual shift from rural to urban was observed
with movement to suburbs within urban areas. The
effect of economic downturns in 1975 and 1980 was
an increase in share for the lowest income category;
more households moved to the highest income group
from 1970 to 1990, while the lower middle income
group share expanded and the upper middle income
share declined. The rate of household formation was
high during the 1970's, which resulted in increases in
smaller and younger households. The trend in younger
households reversed after 1980 as household forma-
tion slowed. Average household size dropped from
3.2 in 1970 to 2.67 in 1990. The number of licensed
drivers increased throughout the analysis period as
more and more young people were licensed to drive.
Data for the VOP model included disposable in-
come per capita, fuel price, overall personal vehicle
fuel economy, and annual usage in terms of VMT.
Table B-6 shows these data for each year in the analy-
sis period.
Data preparation for the model that projected
household vehicle composition was limited to char-
acterization of existing technology vehicles. Seven
vehicle size and type combinations were character-
ized for 1975 and 1980 while one vehicle, minivan/
small utility, was added for 1985 and 1990. Control
scenario vehicle characteristics are tabulated in Table
B-7. TEEMS' activity and energy computation pro-
cedure was executed to produce personal vehicle travel
and energy consumption estimates.
Commercial truck travel was not modeled but,
historical data published by the FHWA (FHWA, 1987;
1991) were used. FHWA publishes truck travel by
three categories: 1) 2-axle, 4-tire trucks; 2) single unit
11 Telephone conversation between Jim DeMocker, EPA/OAR and EPA/OMS/Ann Arbor Laboratory staff (date unknown).
Nevertheless, the Project Team did consider reviewing emission factors for European automobiles to attempt to estimate no-control
scenario emission factors for 1975 through 1990 reflecting the use of electronic fuel injection and electronic ignition but no catalytic
converter. However, the Project Team concluded that differences in fuel/air mix ratios used in Europe would probably obscure any
differences in emission rates attributable to the use of electronic fuel injection and electronic ignition.
B-16
-------
Appendix B: Emissions Modeling
trucks; and 3) combination trucks. All 2-axle, 4-tire
trucks were treated as light-duty trucks. VMT by per-
sonal light trucks were subtracted from the published
totals to arrive at commercial light truck VMT. Die-
sel truck VMT shares of total VMT were obtained
from TIUS (DOC, 1981; 1984; 1990). TIUS data were
also used to split VMT by single unit and combina-
tion trucks. All combination trucks were assumed to
be the heaviest, class 7 and class 8, while single unit
trucks could be of any size class 3 through 8. Gaso-
line and diesel VMT totals were developed for these
heavy-duty trucks and were kept constant for the con-
trol and no-control scenarios.
Tabic B-4. Sources of Data for Transportation Sector Control Scenario Activity Projection.
Data Item
iVIodd
Household total, population, household
shares by four attributes (location, income,
age of head, and household size).
IPF
Statistical Abstract of the United States, editions 96th,
98th, 103 rd, 104th, 108th, and 113th.
Household shares by number of drivers.
IPF
Statistical Abstracts and FHWA Highway Statistics
provided total drivers. The with CAA distribution of
households trended.
Personal and Disposable income.
VOP
J/W model output and Statistical Abstracts.
Vehicle fleet on-road fuel economy.
VOP
DVSAM
FHWA Highway Statistics.
Fuel Prices
VOP
DVSAM
Energy Information Administration's (EIA) Annual
Energy Review.
Vehicle Price
DVSAM
Ward's Automotive Yearbooks 1975-1983, Automotive
News Market Data Book 1985.
IPF - Iterative Proportional Fitting
VOP - Vehicle Ownership Projection
DVSAM - Disaggregate Vehicle Stock Allocation Model
FHWA - Federal Highway Administration
EI A - Energy Information Administration
B-17
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic Bo. Distribution of Households by Demographic Attributes for Control Scenario.
Household (Million)
63.4
71.1
80.8
86.8
93.3
Population (Million)
204.0
215.5
227.2
237.9
249.5
Attribute
Household Percentage, by Year
1970
1975
1980
1985
1990
Location
Central City
33.2
32.0
31.9
31.6
31.4
Suburbs
33.6
36.0
37.0
38.1
38.3
Rural
33.2
32.0
31.1
30.3
30.3
Income (1990 $)*
<$13,000
25.9
26.5
26.6
25.9
25.5
$13,000 - $33,000
34.0
37.2
37.4
37.7
38.0
$33,000 -$52,500
27.6
22.7
22.4
22.2
22.2
>$52,500
12.5
13.6
13.6
14.2
14.3
Age of Householder (YR)
<35
25.4
29.1
31.1
29.3
27.4
35 -44
18.6
16.7
17.3
20.1
22.1
45 - 64
36.3
34.0
31.2
29.6
29.0
> = 65
19.7
20.2
20.4
21.0
21.5
Household Size
1
17.2
19.5
22.7
23.7
24.6
2
29.0
30.7
31.3
31.6
32.2
3 - 4
33.0
33.0
33.2
33.5
32.8
> = 5
20.8
16.8
12.8
11.2
10.4
Licensed Drivers
0
9.1
8.5
8.1
7.2
6.6
1
27.8
27.3
27.0
26.2
26.0
2
48.1
49.2
50.5
52.5
53.5
> = 3
15.0
15.0
14.4
14.1
13.9
Note: * Approximated to 1 990 dollars.
B-18
-------
Appendix B: Emissions Modeling
Tabic B-6. Economic and Vehicle Usage Data lor Vehicle Ownership Projection ~
Control Scenario.
Year
Disposable Income
per Capita (84 S)
Fuel Price
(84 S)/Gallon
Miles/Gallon
VMT/Vehicle
1970
7,597
0.92
13.5
10.143
1971
7,769
0.88
13.5
10.246
1972
7,990
0.84
13.4
10.350
1973
8,436
0.84
13.3
10,184
1974
8,270
1.06
13.4
9.563
1975
8,340
1.03
13.5
9.729
1976
8,553
1.02
13.5
9.833
1977
8,742
1.01
13.8
9,936
1978
9,070
0.97
14.0
10.143
1979
9,154
1.21
14.4
9.522
1980
9,052
1.53
15.5
9.212
1981
9,093
1.55
15.9
9.212
1982
9,050
1.38
16.7
9,41 9
1983
9,239
1.27
17.1
9.419
1984
9,691
1.20
17.8
9,550
1985
9,881
1.09
18.2
9.568
1986
10,139
0.88
18.3
9,672
1987
10,174
0.88
19.2
10.090
1988
10,564
0.86
19.9
10.100
1989
10,713
0.90
20.3
9.819
1990
10,903
1.00
20.8
9.780
B-19
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic B-7. Control Scenario Personal Characteristics.*
1975
19$®)
VeMdfeTyje
and Sia:
((Seafa))
Can nib
Wdi#t
m
Limine
P®*©ir
(hp)
Fuel
lk-
-------
Appendix B: Emissions Modeling
Tabic B-8. Distribution of Households by Income Class
for No-control Scenario.
Attribute
Household Shares (%), by Year
1975
1980
1985 1990
Income (1990 S)*
<$13,000
26.3
26.2
25.3 24.7
$13,000-33,000
37.3
37.6
38.4 38.4
$33,000-52,000
22.8
22.6
22.0 22.6
>$52,000
13.6
13.6
14.3 14.3
Mile: * Approximated to 1990 dollars.
No-control Scenario Emissions
The control scenario data were modified to re-
flect no-control scenario emissions using economic
changes predicted by the J/W model, EPA, and ANL.
The J/W model predicted a slight loss of employment
and drop in GNP in terms of nominal dollars. How-
ever, the lower rate of inflation coincided with a real
GNP rise. ANL's information from the model did not
include any indexes for converting nominal income
to real income. ANL assumed real income changes to
be similar to those of real GNP and modified house-
hold shares by income classes accordingly. The model
also predicted a slight drop in refined petroleum price
beginning in 1973. The predicted drop was the larg-
est (5.35 percent) in 1973, reached the lowest level
(2.16 percent) in 1984, then increased to a second peak
(3.44 percent) in 1988, and dropped again from 1989
to 1990. Since these changes were inconsistent with
historical patterns of leaded and unleaded gasoline
price change, ANL developed an estimate of changes
in fuel price resulting from the cost of removal of lead
from gasoline and other infrastructure costs involved
with distributing a new grade of fuel. Subsequently,
EPA provided a set of fuel costs for use in the analy-
sis. Both ANL and EPA fuel prices followed a similar
pattern, although their magnitudes differed. The
no-control scenario was analyzed with EPA fuel
prices. ANL also established a relationship with cost
of regulation/emission control technology, and the
effect of costs on vehicle price and fuel economy di-
rectly from the EPA publication Cost of A Clean En-
vironment (EPA, 1990). These changes were used in
the analysis.
The IPF model was executed for target years 1975,
1980, 1985, and 1990 using a set of revised house-
hold shares by income class. Table B-8 shows the re-
vised shares. Comparing Table B-8 no-control sce-
nario shares with those in Table B-5 for the control
scenario, there seems to be a slight shift away from
travel by the lowest income group and toward the
middle income groups.
The vehicle ownership projection model was ex-
ecuted for the above four target years using the data
listed in Table B-9. Changes in fleet characteristics
are summarized in Table B-10.
B-21
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic B-9. Economic and Vehicle Usage Data for Vehicle Ownership
Projection ~ No-control Scenario.
Year
Disposable
Income per
Capita (84 S)
Fuel Price
(84 $)/Gallon
Miles/
Gallon
VMT/Vehicle
1970
7,597
0.91
13.5
10.143
1971
7,769
0.88
13.5
10.247
1972
7,990
0.83
13.4
10,353
1973
8,463
0.84
13.3
10,189
1974
8,297
1.06
13.4
9,569
1975
8,406
1.02
13.5
9.736
1976
8,600
1.01
13.5
9,854
1977
8,795
1.01
13.8
9.963
1978
9,126
0.96
14.0
10.174
1979
9,216
1.19
14.4
9.557
1980
9,114
1.51
15.5
9.234
1981
9,158
1.53
16.0
9,234
1982
9,116
1.36
16.8
9.447
1983
9,312
1.25
17.2
9.450
1984
9,775
1.18
17.9
9.582
1985
9,976
1.06
18.3
9.607
1986
10,244
0.84
18.4
9.73 8
1987
10,282
0.86
19.4
10,201
1988
10,676
0.83
20.1
10.214
1989
10,827
0.88
20.5
9,902
1990
11,019
0.97
21.0
9.849
Note- The effect of reductions in vehicle price and vehicle operating cost, and increases in fuel economy
and horsepower were reflected in the menu of the vehicle choice model (DVSAM). Vehicle weight and
seating capacity were kept unchanged from the with CAA run. Table IV-7 shows the changes in various
vehicle attributes.
B-22
-------
Appendix B: Emissions Modeling
Tabic B-.1.0. Percent Changes in Key Vehicle Characteristics Between
the Control and No-control Scenarios.
1975
1980
Vehicle
Price
mps
HP
Price
mpj>
HP
Small Auto
-2.35
0.01
0.59
-2.76
0.22
1.81
Compact Auto
-2.35
0.01
0.59
-2.76
0.22
1.81
Midsize Auto
-2.35
0.01
0.59
-2.76
0.22
1.81
Large Auto
-2.35
0.01
0.59
-2.76
0.22
1.81
Small Truck
-1.30
0.01
0.59
-2.71
0.22
1.81
Ski Truck
-1.30
0.01
0.59
-2.71
0.22
1.81
Std Van/Util
-1.30
0.01
0.59
-2.71
0.22
1.81
M Vn/Sm
IJtilitv
1985
1990
Vehicle
Price
mpj>
HP
Price
mpj>
HP
Small Auto
-3.25
0.62
2.20
-2.94
0.95
2.77
Compact Auto
-3.25
0.62
2.20
-2.94
0.95
2.77
Midsize Auto
-3.25
0.62
2.20
-2.94
0.95
2.77
Large Auto
-3.25
0.62
2.20
-2.94
0.95
2.77
Small Truck
-2.53
0.62
2.20
-2.58
0.95
2.77
Std Truck
-2.53
0.62
2.20
-2.58
0.95
2.77
Std Van/Util
-2.53
0.62
2.20
-2.58
0.95
2.77
M Vn/Sm
Utility
-2.53
0.62
2.20
-2.58
0.95
2.77
Note- *Average change for each vehicle size and type combination.
B-23
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Utilities
The electric utility industry retrospective analy-
sis was prepared using two different utility simula-
tion models. ICF utilized its CEUM to estimate con-
trol and no-control scenario emissions for S02, TSP,
and NOx in each of the target years. ANL's ARGUS
model was used to estimate electric utility CO and
VOC emissions for the same period. This mix of mod-
eling approaches was used because, while CEUM was
determined to be a better tool for examining fuel shifts
that were affected by the CAA than ARGUS, the
CEUM model was not initially set-up to evaluate CO
or VOC emissions. Although CEUM can be (and even-
tually was) configured to provide emission estimates
for pollutants other than S02, NOx, and PM, ARGUS
was already configured to provide VOC and CO emis-
sions. However, it should also be noted that VOC and
CO emissions from utilities are quite low, as efficient
fuel combustion reduces both pollutants. Thus, for this
sector, the presence or absence of the CAA would not
produce any different VOC or CO control techniques.
VOC and CO emission rates for this sector differ pri-
marily based on the fuel and boiler type. Therefore, a
simpler modeling approach was judged to be accept-
able and appropriate for these two pollutants. This
chapter presents the methodology used to estimate
utility emissions underthe control and no-control sce-
nario using the CEUM and ARGUS models. The
method used by Abt Associates to estimate lead emis-
sions from utilities is also presented.
Overview of Approach
The CEUM model uses industry capacity data and
specific unit-by-unit characteristics, operating costs
data, electricity demand estimates under the control
and no-control scenario, and historical fuel prices to
estimate SO,, TSP, and NO emissions for 1980,1985,
2' ' x ' '
and 1990. Changes in electric utility emissions, costs,
and regional coal production were developed using
ICF's CEUM with a calibration to historical electric-
ity generation, fuel use, and emissions. The ARGUS
model, which was used by ANL to estimate utility
VOC and CO emissions, is driven by operating costs,
industry capacity and generation data, demand for
coal, and unit-level operating characteristics. The J/
W model is used to incorporate predicted changes in
electricity demand underthe no-control scenario. Fi-
nally, Abt Associates relied upon energy use data, the
Trends data base, and the Interim 1990 Inventory to
calculate utility lead emissions based on coal con-
sumption. The approaches used by each of these three
contractors are discussed individually in the follow-
ing sections.
Establishment of Control Scenario Emissions
A common feature of the approaches taken by ICF
and ANL was to identify conditions that are inputs to
the CEUM and ARGUS models, respectively, in the
control scenario. Later in the analysis, these variables
were revised to reflect no-control scenario conditions.
The next section discusses the specific assumptions
used in the CEUM analysis.
Key Assumptions in the Development of the
ICF Analysis
At EPA's direction, ICF made several assump-
tions in conducting this analysis for purposes of con-
sistency with other ongoing EPA efforts assessing the
effects of the CAA. These include the macroeconomic
assumptions regarding the effects of the CAA on eco-
nomic growth, or more specifically, electricity de-
mand, developed from other EPA commissioned ef-
forts. Each is described briefly below.
Pollution Control Equipment Costs
Only limited actual data were available for this
analysis on the historical capital and operating costs
of pollution control equipment. Accordingly, for this
analysis, the actual capital and operating costs of
scrubbers were estimated using EPA scrubber cost
assumptions adjusted to reflect actual data from a sur-
vey of scrubbed power plants with scrubbers installed
during the 1970s and early 1980s. For those power
plants with actual survey data, actual capital costs were
used. For other pre-1985 scrubbers, ICF relied on the
average costs from the survey data. For particulate
control equipment (primarily electrostatic precipita-
tors, or ESPs), costs were estimated based on limited
actual data, and a 1980 Electric Power Research In-
stitute (EPRI) study of ESP and baghouse costs. Based
on this information, ESPs were estimated to cost an
average of $50 per kilowatt (in 1991 dollars). The
development of more detailed data on actual power
plant pollution control costs was beyond the scope of
ICF's analysis. ICF concluded that such an effort
would not significantly change the national or regional
cost estimates developed by its approach.
B-24
-------
Appendix B: Emissions Modeling
Electricity Demand and Fuel Prices
Consistent with other EPA ongoing analyses, ICF
assumed that the CAA resulted in a reduction in elec-
tricity demand of 3.27 percent in 1980, 2.77 percent
in 1985, and 2.97 percent in 1990. Also consistent
with these studies, ICF assumed that natural gas prices
and oil prices would not be affected by the CAA. Coal
prices were estimated to change in line with increases
and decreases in demand for specific coal supplies
(and consistent with ICF's detailed modeling of coal
supply and demand). The average prices of all residual
oils consumed were also estimated to change due to a
greater use of more expensive lower sulfur residual
oils under the CAA.
Coal, Nuclear, Hydro, and Oil/Gas Capacity
At EPA's direction, ICF's approach was based
on the assumption that no changes in the amount of
nuclear, coal, hydro, or oil/gas stream or combined
cycle capacity would be built or in place in 1980,1985,
or 1990. Given that the driving factors associated with
the actual decisions to build new baseload capacity
were not based solely on economics but entailed fi-
nancial, regulatory, and political factors as well, the
actual effect of the CAA on these build decisions is
very uncertain. To the extent that more coal-fired
power plants would be built and fewer oil/gas-fired
power plants constructed, the actual emissions reduc-
tions associated with the CAA would be greater than
those estimated by ICF, while the estimated costs of
the CAA would be greater (because fewer, lower-cost,
coal-fired power plants would be on line under the
CAA). However, the CAA had virtually no effect on
the costs of constructing new coal-fired power plants
that came on line prior to about 1975 and a relatively
moderate cost effect on coal-fired power plants that
came on line through the early 1980s (since these
power plants were not required to install scrubbers).
Since a large majority of coal-fired power plant ca-
pacity came on line prior to 1975, ICF concluded that
the effect of the CAA on the amount of total coal-
fired capacity was not expected to be very large.
Natural Gas Consumption
The analysis assumed that the amount of natural
gas consumed under the no-control scenario could not
exceed the actual amount of consumption in 1980,
1985, and 1990. In part, because of natural gas price
regulation and the oil price shocks of the 1970s, natu-
ral gas was often unavailable to electric utilities in the
early 1980s. Since the CAA is relatively unrelated to
the questions of supply availability and price regula-
tion of natural gas, ICF assumed that no additional
gas supplies would be available if the CAA had never
been adopted. It is possible, however, that in the ab-
sence of the CAA, industrial and commercial users of
natural gas would have used more oil or coal. To the
extent that this would have occurred, there would have
been more natural gas supplies available to the elec-
tric utility sector. This increase in supply would have
resulted in an increase in the estimated costs of the
CAA, and a corresponding decrease in the estimated
emission reductions. ICF concluded, however, that this
effect would not be very significant.
State and Local Environmental Regulations
At EPA's direction, ICF assumed that there would
be no State and local emission limits or other emis-
sion control requirements under the no-control sce-
nario. Accordingly, ICF assumed that there would be
no SO„, NO , or TSP emission limits under the
27 x7
no-control scenario and that all scrubbers, NO con-
7 x
trols, and ESPs/baghouses (at coal-fired power plants)
were installed as a result of the CAA. (The more lim-
ited amount of particulate control equipment installed
at oil-fired plants was assumed to have been installed
prior to the passage of the CAA.) In the case of par-
ticulate control equipment, some ESPs and other
equipment were installed at coal plants prior to the
1970 CAA. To the extent that this is the case, the es-
timates of the costs of meeting the CAA have been
overstated. ICF concluded, however, that the amount
of such capacity was not substantial.
Retirement Age
The analysis assumed that unit retirement age was
constant between the control and no-controls sce-
narios. Adoption of this assumption might bias the
emission reduction estimates upward to the extent
turnover rates of older (and presumably higher-emit-
ting) units may be slower under the control scenarios,
because more significant CAA control requirements
focused on new units. However the vast majority of
existing coal and oil capacity was built after 1950 and
it is generally acknowledged that a relatively short
technical plant lifetime would be about 40 years. As
such, even if the no-control scenarios resulted in no
life-extension activity, there would be virtually no
effect over the 1970 to 1990 timeframe of the analy-
sis.
B-25
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
ICF1975 Control Scenario Emissions
The 1975 emissions under both scenarios were
calculated differently than emissions in 1980, 1985,
and 1990. In calculating or estimating 1975 S02 emis-
sions forthe control scenario (i.e., "actual" 1975), the
weighted average emission rates at the State level, in
the year 1975 were estimated, based on plant level
average sulfur content of fuel deliveries from Federal
Energy Regulatory Commission (FERC ) Form 423
and assumed AP-42 sulfur retention in ash. These
weighted average emission rates were then applied to
actual State-level electric utility fuel consumption in
the year 1975 (DOE, 1991). In the case of NOx emis-
sions, first, an estimate of Statewide NO emissions
' ' x
in the year 1975 was derived based on the use of the
same NOx emission rates, by fuel type, as developed
forthe 1980 no-control scenario modeling runs. These
emission rates were specific to the fuel type (coal, oil,
or natural gas). These Statewide NOx emission rates
or factors were then applied to actual fuel consumed
by electric utilities in the year 1975, in order to obtain
estimated "actual" 1975 emissions. As before, the fuel
consumption at a State level was derived from the State
Energy Data Report (DOE, 1991). ICF calculated the
weighted average heat content (BTU/lb) by State from
the 1975 FERC Form 423 data and used these figures
with the TSP emission factors (lbs/ton) to derive emis-
sion rates by State (lbs/MMBTU). These emission
rates were then applied to 1975 fuel consumption es-
timates obtained from the State Energy Data Report.
Forthe control scenario 1975 estimates, ICF used the
1975 factors.
For the remaining target years, ICF used the re-
sults of CEUM runs that provided fuel consumption
figures in 1980, 1985, and 1990, respectively. Emis-
sions were then calculated using the appropriate emis-
sion factors for each year.
ARGUS Modeling Assumptions
The portion of the electric utility sector analysis
conducted by ANL with the ARGUS model is de-
scribed in this subsection. ARGUS contains four ma-
jor components: BUILD, DISPATCH, the Emissions
and Cost Model, and the Coal Supply and Transpor-
tation Model (CSTM). An overview of ARGUS can
be found in Vcsclkat/a/ (1990). Only the DISPATCH
and CSTM modules were used for the present analy-
sis. A brief description of the ARGUS components
used in this analysis is found in the following subsec-
tions.
DISPATCH Module
The DISPATCH module contains a probabilistic
production-cost model called the Investigation of
Costs and Reliability in Utility Systems (ICARUS ).
This module calculates reliability and cost informa-
tion for a utility system. ICARUS represents detailed,
unit-by-unit operating characteristics such as fuel cost,
forced outage rate, scheduled maintenance, heat rate,
and fixed and variable operating and maintenance
(O&M ) costs. These components are used to effi-
ciently compute system reliability (such as loss-of-
load probability and unserved energy) and production
costs.
The input data required by ICARUS include
monthly load duration curves, annual peak demands,
and, for both new and existing units, unit sizes, capi-
tal costs, fixed and variable O&M costs, fuel types
and costs, heat rates, scheduled maintenance, and
equivalent forced outage rates. The output from
ICARUS includes annual summaries of capacity, gen-
eration, cost, and reliability for the entire generating
system.
CSTM Module
The CSTM module determines the least-cost com-
bination, on a per BTU basis, of coal supply sources
and transportation routes for each demand source.
First, it estimates coal market prices based on regional
demands for coal from all economic sectors. To gen-
erate market prices, CSTM estimates regional coal
production patterns and coal transportation routes. The
CSTM input data are grouped into three major cat-
egories: demand, supply, and transportation. CSTM
uses supply curves from the Resource Allocation and
Mine Costing (RAMC ) Model (DOE, 1982). Every
region has a separate curve for one or more of the 60
different coal types that may be produced in that re-
gion. CSTM modifies the original RAMC supply
curve by dividing the single RAMC curve into two
curves, one representing deep mines and the other rep-
resenting surface mines, but still uses the same ranges
for heating values and mine prices that define the sup-
ply curves in RAMC. Prices fluctuate as a result of
different mining methods, size of mining operations,
reserve characteristics, and depletion effects.
The transportation data defines the network that
connects 32 coal supply origins with 48 demand cen-
ters. Transportation cost is affected by distance, ter-
rain, congestion, variable fuel costs, cost escalators
B-26
-------
Appendix B: Emissions Modeling
for fuels and facility upgrades, and competition.
CSTM first computes the production cost for each coal
supply region and coal type. It then matches supply
sources with transportation routes to find the lowest
delivered costs.
Coal demand for a particular region is based on
the amount, geographic region, economic sector, and
range of coal types. There are 44 domestic demand
regions. CSTM allows demand to be met by one, or a
combination of, different supply regions.
The ARGUS input data for existing units are based
on the Argonne Power Plant Inventory (APPI). APPI
is a data base of operating and planned generating units
in the United States that was current through 1988 at
the time of ANL's analysis. This data base is updated
annually based on information in the regional North
American Electric Reliability Council (NERC ) re-
ports, reports from the Energy Information Adminis-
tration (EIA), and other sources. Unit operating char-
acteristics (fixed O&M, variable O&M, heat rate,
forced outage rate, and scheduled maintenance) are
based on regional data as defined in the EPRI report
on regional systems and other historic data (EPRI,
1981).
ANL used the 1988 inventory to generate a 1990
inventory. The 1990 inventory was then used to gen-
erate a separate unit inventory for the target years
1975,1980 and 1985. The target year inventories were
generated by removing units whose on-line year was
greater than the target year, from their respective in-
ventory. The regional capacity totals in these prelimi-
nary inventories were tabulated by major fuel category
(nuclear, coal, oil and gas steam) and compared to the
regional historic NERC totals. This review identified
capacity differences, especially in 1975 and 1980 in-
ventories. The original plan was to add phantom units
to match the regional historic totals. However, based
on the need for State-level emissions, it was decided
that a more thorough review of the unit inventories
was required.
ANL's detailed review included an examination
of the nuclear and coal units greater than 100 mega-
watt equivalent (MWe) in each target year. Missing
units, with the appropriate unit size and State code,
were added so that the regional totals were compa-
rable. The availability of coal units was based on the
on-line year of the unit as reported in the EIA report
Inventory of Power Plants in the United States (DOE,
1986). The coal units were also checked against the
EIA Cost and Quality Report (EIA, 1985) to verify
the existence of flue gas desulfurization (FGD ) sys-
tems in each of the target years. The nuclear unit in-
ventories were verified with the EIA report An Analy-
sis of Nuclear Power Plant Operating Costs (DOE,
1988). The review also included oil and gas steam
units greater than 100 MWe. The total capacity of the
oil and gas steam units were compared because many
units switched primary fuel from oil to gas during the
relevant time period. The oil and gas units were com-
pared to historic inventories based on information pro-
vided by Applied Economic Research. In addition to
thermal generation, the hydro and exchange energy
was reviewed. For each target year, the hydro genera-
tion and firm purchase and sale capacity data was ad-
justed to reflect the historic levels. These two compo-
nents, hydro and firm purchase and sales, are ac-
counted for first in the loading order. If these vari-
ables are overestimated, there will be less generation
from coal units. Likewise, if they are underestimated,
there will be too much coal generation. The hydro and
firm purchases and sales can vary significantly from
year to year because of weather conditions and other
variables. Therefore, it was important that they be
accurately represented.
No-control Scenario Emissions
In order to calculate utility emissions under the
no-control scenario, inputs to both the CEUM and
ARGUS models were adjusted to reflect no-control
scenario conditions. The changes made to each
model's base year input files are discussed separately
in the following sections.
ICF Estimates of SO,, TSP, and NO Emissions
2' ' x
in the No-control Scenario
As described earlier, ICF utilized a different meth-
odology to calculate 1975 emission estimates. Rather
than relying on the use of detailed modeling runs, ICF
based the 1975 emission estimation on historic fuel
consumption and sulfur content data in 1975. This
subsection first outlines the process used to calculate
no-control scenario emissions in 1975 and then pre-
sents the methods used for the remaining target years.
1975 Utility SOj NO^, and TSP Emissions
To develop State-level no-control scenario utility
S02 emissions, ICF developed no-control scenario S02
emission rates. A reasonable surrogate for these emis-
sion rates is S02 rates just prior to the implementa-
B-27
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
tion of the SIPs under the CAA. ICF developed 1972
rates (based on the earliest year available for FERC
Form 423) and compared these with 1975 rates. In
each State, the greater of 1972 or 1975 rates was used
in the calculation of S02 emissions in the absence of
the CAA. To develop State-level no-control scenario
S02 emissions, no-control scenario fuel consumption
data were needed. ICF assumed that the demand for
electricity in 1975 would be 2.73 percent higher than
the actual energy sales in 1975. This assumption is
identical to the no-control scenario electricity demand
projections derived from the J/W projections. For the
purpose of this analysis, it was further assumed that
this increment in demand would have been met in 1975
from the oil and coal-fired plants in each State. The
increase in consumption of these fuels was assumed
to be in the same proportion as their share in the 1975
total energy mix for electricity generation in that State.
It was assumed that the generation of nuclear, gas-
fired, and other electricity generation would not
change. A sensitivity case without an assumed elec-
tricity demand change was also calculated. (The sen-
sitivity analysis results are presented later in this ap-
pendix.)
For NO emissions under the no-control scenario,
X '
it was also assumed that the 1975 electricity sales
would have been 2.73 percent higher than was the case
in 1975. No-control scenario TSP emissions in 1975
were based on national emission rate numbers from
EPA that were converted to pounds per million BTU
using the average energy content of fuels in each State.
No-control scenario TSP emissions were calculated
based on 1970 emission factors (Braine, Kohli, and
Kim, 1993).
1980, 1985, and 1990 Utility Emissions
For 1980, 1985, and 1990, ICF calculated
no-control scenario emissions based on fuel consump-
tion figures from the CEUM runs, and 1970 emission
factors from EPA.
Electric utility S02 emission estimates are ap-
proximately 10 million tons (or about 38 percent)
lower by 1990 under the control scenario than under
the no-control scenario. Most of this estimated differ-
ence results from the imposition of emission limits at
existing power plants through the SIPs under the 1970
CAA. Most of these SIPs were effective by 1980 (with
some not fully effective until 1985). Most of the ad-
ditional reductions that occurred during the 1980s were
the result of the electric utility NSPS, which required
the installation of 70 to 90 percent S02 removal con-
trol equipment.
By contrast, electric utility NOx emission esti-
mates under the control scenario are only about 1.2
million tons, or 14 percent, lower than under the
no-control scenario by 1990. This occurs because,
under the implementation of the 1970 CAA, only a
few existing power plants were subject to NOx emis-
sion limits. Virtually all of the estimated reductions
are the result of NOx NSPS, which generally required
moderate reductions at power plants relative to un-
controlled levels. In addition, electricity demand is
estimated to be about 3 percent lower under the con-
trol scenario. This decrease reduces the utilization of
existing power plants and also contributes to lower
NOx emissions (and other pollutants as well).
Electric utility annualized costs (levelized capi-
tal, fuel, and O&M) are estimated to be $0.2 billion
lower in 1980, $1.5 billion higher in 1985, and $1.9
billion higher in 1990 under the control scenario. Note,
however, that this reflects the effects of two offset-
ting factors: (1) the higher utility compliance costs
associated with using lower sulfur fuels, and the in-
creased O&M and capital costs associated with scrub-
bers and particulate control equipment; and (2) lower
utility generating costs (fuel, operating and capital
costs) associated with lower electricity demand re-
quirements. In 1980, the increase in fuel costs due to
higher generation requirements (under the no-control
scenario), was larger than the decrease in capital and
O&M costs and thus yielded a cost increase over the
control case.
However, lower electricity demand for the utility
sector would translate into higher costs in other sec-
tors (as electricity substitutes are used). This effect
was captured to some extent by the original J/W mac-
roeconomic modeling conducted for the present analy-
sis.
Average levelized U.S. electricity rate estimates
are approximately 3 percent higher under the control
scenario during the 1980s. Note that year by year, elec-
tric utility revenue requirements and capital expendi-
tures (not estimated by ICF) would be estimated to
have increased by a greater percentage particularly in
the 1970s and early 1980s as incremental capital ex-
penditures for scrubbers and ESPs were brought into
the rate base.
B-28
-------
Appendix B: Emissions Modeling
Significant shifts in regional coal production are
estimated to have occurred between the control and
no-control scenarios. High sulfur coal producing re-
gions such as Northern Appalachia and the Midwest/
Central West are estimated to have lower production
under the control scenario, while lower sulfur coal
producing regions such as Central and Southern Ap-
palachia are estimated to have higher coal produc-
tion.12
ARGUS No-control Scenario
Regional fuel prices, for the thermal units, were
based on historic information from the EIA Form 423
data for the year 1977, 1980 and 1985. The 1977 data
was used for 1975. Fixed and variable O&M costs
were adjusted from the 1988 level, and all cost data
were converted to 1985 dollars.
The load data were based on regional historic
NERC data for each of the target years. The shapes of
the monthly load duration curves are the result of
modifications based on the data in the EPRI report on
regional systems (EPRI, 1981). The shapes were modi-
fied to match the projected 1988 monthly load factors
for the NERC regions. These load shapes were held
constant for all years.
The actual peak-loads were selected from historic
information and used with the existing load duration
curves. The system was dispatched so that the calcu-
lated generation could be compared with historic data.
Discrepancies were resolved by adjusting the peak
load so that the annual generation was on target. This
procedure was repeated for each of the target years.
The electric utilities were expected to have an in-
crease in generation as identified by the J/W data.
Table B-ll identifies the increase in national level
generation by year. The national level increase in gen-
eration was applied to each power pool.
In addition to load changes, coal units with FGD
equipment were modified. These units had their FGD
equipment removed along with a 3 percent decrease
in heat rate, a 2 percentage point decrease in forced
outage rate, and a 50 percent decrease in their fixed
and variable O&M costs. These changes were incor-
Tablc B-11. J/W Estimates of
Percentage Increases in National
Electricity Generation Under
No-control Scenario.
Year
Percentage
Increase
1975
2.7%
1980
3.3%
1985
2.8%
1990
3.0%,
porated into the ARGUS model for each of the target
years. Model runs were then conducted to arrive at
estimates ofVOC and CO emissions in the no-control
scenario.
Estimation of Lead Emissions from
Utilities
In order to estimate lead emissions from electric
utilities in each of the target years, data from three
different sources were used. Energy use data for the
control and no-control scenarios were obtained from
the national coal use estimates prepared for the sec-
tion 812 analysis by ICF (Braine and Kim, 1993). The
Trends data base provided emission factors and con-
trol efficiencies, and the Interim 1990 Inventory iden-
tified utility characteristics. The ICF data bases pro-
vided the amount of coal consumed for both the con-
trol and no-control scenarios in each of the target years.
A correspondence between the Interim Inventory and
the ICF data base was achieved through the plant name
variable. Using emission factors for lead and control
efficiencies for electric utilities, estimates of lead
emissions per plant per year were calculated. These
factors were obtained from the Trends data base. It
was assumed that pollution control on coal-burning
power plants under the no-control scenario would be
the same as the pollution control level in 1970. There-
fore, the control efficiency from 1970 is used as the
basis for the no-control case.
12 At EPA's direction, ICF's analysis did not estimate the effect of shifts in non-utility coal consumption on regional coal
production, nor did it consider the possibility that fewer new coal powerplants might have been built due to the CAA as discussed
earlier. Both of these factors could result in a greater estimated change in total U.S. coal production than estimated herein although the
difference is not likely to be very significant.
B-29
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
CEUM Sensitivity Case
In addition to comparing actual (control scenario)
historical costs and emissions with the higher elec-
tricity demand under the no-control scenario, ICF also
evaluated emissions in a sensitivity case without the
CAA (i.e., under the no-control scenario) with the
same electricity demand (versus the no-control sce-
nario with higher demand). The purpose of this sensi-
tivity analysis was to isolate the incremental electric
utility compliance costs and reductions in emissions
associated with the CAA from the lower resulting
generation costs and emissions due to lower estimated
electricity demand under the CAA. The incremental
effects of the CAA when compared with this case in-
dicate:
Estimated reductions in emissions due to the
CAA are somewhat lower if measured against
the sensitivity case without the CAA with the
same electricity demand than the emissions
without the CAA with lower demand. This
occurs because lower electricity demand un-
der the no-control scenario sensitivity results
in lower utilization of existing coal and oil
plants which, in turn, results in lower emis-
sions. As noted above, in some sense, the
changes in emissions represent the effects of
electric utility compliance actions under the
CAA, absent the effect of lower resultant de-
mand for electricity.
When measured against the sensitivity case
without the CAA (with the same electricity
demand), electric utility annualized costs are
estimated to have increased by about $5 to $6
billion during the 1980 to 1990 period. This
reflects the following cost factors: (1) higher
annualized capital costs associated primarily
with scrubbers and ESPs installed by electric
utilities to comply with the CAA; (2) higher
O&M costs associated with the additional air
pollution control equipment; and (3) higher
fuel costs associated with using lower sulfur
coal and oil in order to meet the emission limit
requirements of the CAA.
Commercial/Residential
The Commercial and Residential Simulation Sys-
tem (CRESS) model was developed by ANL as part
of the Emissions and Control Costs Integrated Model
Set and used in the NAPAP assessment (Methods for
Modeling Future Emissions and Control Costs, State
of Science and Technology, Report 26) (McDonald
and South, 1984). CRESS is designed to project emis-
sions for five pollutants: SOx, NOx, VOC, TSP, and
CO. The CRESS output is aggregated into residential
and commercial subsectors related to both economic
activity and fuel use. The introductory material pro-
vided in this appendix about CRESS describes the base
year as being 1985. It appears in this way because
CRESS was originally developed to operate using the
1985 NAPAP Emission Inventory as its base year data
set. For the five pollutants reported by CRESS, emis-
sion estimates are provided for the following sectors:
~ Commercial/institutional
• coal, including point and area categories of
anthracite and bituminous boilers;
• liquid fuel, including boiler and space heat-
ing uses of residual, distillate, LPG, and
other fuels;
• natural gas boilers, space heaters, and in-
ternal combustion engines;
• wood used in boilers and space heaters; and
• other mixed or unclassified fuel use.
~ Residential
• coal, including area sources of anthracite
and bituminous;
• liquid fuel, composed of distillate and re-
sidual oil;
• natural gas; and
• wood.
~ Miscellaneous
• waste disposal, incineration, and open burn-
ing; and
• other, including forest fires, managed and
agricultural burning, structural fires, cut-
back asphalt paving, and internal combus-
tion engine testing.
In addition, VOC emissions are projected forthese
source categories:
~ Service stations and gasoline marketing;
~ Dry-cleaning point and area sources; and
B-30
-------
Appendix B: Emissions Modeling
~ Other solvents, including architectural surface
coating, auto-body refinishing, and consumer/
commercial solvent use.
This section describes the use of CRESS to esti-
mate control and no-control scenario emissions from
the commercial/residential sector.
Control Scenario Emissions
For the NAPAP assessment, 1985 CRESS output
corresponded to the 1985 NAPAP Inventory (EPA,
1989), which served as the benchmark for any pro-
jections. The design of CRESS is such that emissions
by NAPAP SCC are input for each State, then pro-
jected to future years by scaling them to economic
data such as energy demand. In estimating emissions,
differences in emission controls associated with new,
replacement, and existing equipment are taken into
account where such differences are considered sig-
nificant. The basic modeling approach is shown in
the following equation:
Q= (^) - h x (-^) x (/;(x /, ) (3)
where:
Q = emissions in year t or the base year, year 0
E = emission factor for the source category b
in the base year, or for a subcategory j sub-
ject to controls in year t (this takes into
account changes in emission rates that may
occur as a result of emission regulations or
technology changes)
D = driver data indicating activity levels in the
base and future years
f = fraction of total activity in year t differen-
tially affected by emission controls
The calculations are carried out in two subroutines,
one for SO,, NO , TSP and CO, and one for VOC.
2' x' '
Typically S02, NOx, TSP, and CO emissions are
projected by multiplying the 1985 NAPAP SCC data
or base year data by the ratio of the driver data (activ-
ity level) value in the projection year to its value in
the base year. Because there are few controls on SOx
or NOx emissions from the sources covered by CRESS,
projected emissions for most sectors are proportional
to the expected activity levels. Thus,
a = VnX(jr) (4)
There are a few source types, such as commer-
cial/institutional boilers, for which emission controls
are mandated. These are modeled by multiplying the
1985 emission data by the ratio of the controlled emis-
sion factor to the base-year emission factor. Emission
factors for each source type are weighted by the pro-
portion of base year activity in each subsector to which
controls are expected to apply.
Q,.h = a, + (-£--)x + Kj I )
where:
g = the fraction of base-year activity accounted
for by existing source b, replacement
source r, or new source n in year t
The effective emission factor (Et,n) for the sector
is calculated by weighing the portions of sectoral
emissions subject to NSPS controls and those likely
to continue at existing levels. An appropriate Internal
Revenue Service-based rate at which new equipment
replaces existing sources is applied to each sector in
the model. This is done to estimate how emissions
might change as older sources are retired and replaced
by new sources that emit at lower rates.
The SO /NO /TSP/CO subroutine varies in new
x x
and replacement emission-source fractions subject to
NSPS controls. These fractions are applied to the
emission-source replacement rates. In addition, ratios
for new source emission factors are varied by State.
However, emission ratios for any pollutant/source type
combination do not vary over the projection period.
The VOC estimation methodology is similar, but
allows variation in emission factors overtime. Emis-
sion ratios are calculated from files of replacement
and existing source emission factors weighted by the
replacement rate for each sector and new source fac-
tors by State. These are input for each 5-year projec-
tion interval. For most source categories, VOC con-
B-31
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
trols are not envisioned, and the 1985 NAPAP emis-
sions for the category are simply scaled proportion-
ally to changes in the driver (activity level) data.
For sources to which controls apply, a variation
on the following equation is employed:
C^ = (-|--A)x(y^ + ^XAJ| (6)
In equation 6, the emission factors for new and
existing sources are effectively weighted by the pro-
portion of total activity in year t to which controls
apply.
In using CRESS for the CAA retrospective analy-
sis, the base year was 1975. CRESS requires emis-
sions information by State and NAPAP source cat-
egory as input. Since detailed information on emis-
sion levels for 1975 by NAPAP source category were
not available, the data were developed from a combi-
nation of sources. The procedure for calculating 1975
emissions based on the 1985 NAPAP inventory is
described below. The emissions module uses these
initial values in conjunction with activity estimates to
project control and no-control scenario emissions.
Emissions Data
Since the starting point for the analysis was 1975,
emissions data by State and SCC for S02, NOx, VOC,
TSP, and CO were required. Available emissions in-
formation for this year was not at the level of detail
needed by CRESS. The 1985 NAPAP Inventory,
which contains the necessary level of detail, in con-
junction with information from EPA's National Air
Pollutant Emission Estimates, 1940-1990 (Trends) and
ANL's MSCET, was used to construct an emissions
inventory for 1975. The model then uses these emis-
sions as a benchmark for the analysis.
The method for constructing the 1975 emissions
data base was consistent for all pollutants; however,
two different sources of emissions data were neces-
sary in order to obtain time series information on all
pollutants. MSCET contains monthly State-level emis-
sion estimates from 1975 to 1985 by emission source
group for S02, NOx, and VOC. Therefore, MSCET
information was used for S0o, NO , and VOC, while
2' x' '
Trends data were used for TSP and CO. Emission
source groups from MSCET were matched with 1985
NAPAP Inventory SCCs. The MSCET methodology
is benchmarked to the 1985 NAPAP Inventory and
uses time series information from Trends in conjunc-
tion with activity information to estimate State-level
emissions for S02, NOx, and VOC. Although the level
of detail contained in the NAPAP Inventory could not
be preserved because of the aggregation needed to
match with MSCET emissions sources, MSCET pro-
vided the State-level spatial detail required by CRESS.
Once the 1985 emissions by SCC and State from
the 1985 NAPAP Inventory were matched with emis-
sion source groups and States from the MSCET data
base, an estimate of 1975 emissions was computed
by multiplying the 1985 NAPAP Inventory emissions
value by the ratio of 1975 MSCET emissions to 1985
MSCET emissions. Ratios were computed and applied
for each combination of State, pollutant, and MSCET
emission source group.
This method of constructing an emissions inven-
tory for 1975 utilizes the State estimates from MSCET,
thus capturing the spatial shifts that occurred over the
analysis period. It is assumed that NAPAP provides
the most reliable point and area source information in
terms of the level of 1985 emissions (which is also
the assumption of the MSCET methodology). Note
that if there were a 1-to-l correspondence between
MSCET and NAPAP, this method would be equiva-
lent to using the MSCET methodology directly for
constructing 1975 emission levels.
A similar method was used for TSP and CO, but
since these pollutants are not included in MSCET, the
Trends ratio of 1975 to 1985 emissions for these two
pollutants was used. Thus, for TSP and CO, all States
were assumed to have experienced the same change
in emissions as indicated by the national figures.
It should be noted that in addition to the loss in
spatial detail, the Trends source groups generally
spanned several NAPAP source categories. The
strength in the Trends information is the consistency
of emissions estimates over time. It is considered to
be the most reliable data for tracking changes in emis-
sions over the time period of the analysis, and was
therefore chosen for developing 1975 estimates for
TSP and CO.
The 15 source categories reported in Trends were
matched with those in the 1985 NAPAP Inventory.
The ratios of 1975 emissions to 1985 emissions by
source category that were applied to the 1985 NAPAP
emissions data are shown in B-12. The 1975 emis-
B-32
-------
Appendix B: Emissions Modeling
Tabic B-12. Trends Source Categories and (1975 to
1985) Scaling Factors forTSPand CO.
Trends Source Category
ISP*
CO*
Commercial/Institutional Fuel
Combustion:
Coal
2.11
0.59
Natural Gas
1.00
0.91
Fuel Oil
2.35
1.43
Other
1.83
0.67
Residential Fuel Combustion:
Coal
1.33
1.47
Natural Gas
1.17
1.00
Fuel Oil
1.11
1.76
Wood
0.49
0.49
Miscellaneous: Forest Fires
0.67
0.62
Solid Waste Disposal:
Incineration
3.00
0.64
Open Burning
1.50
1.44
Miscellaneous Other Burning
1.00
1.33
Industrial Processes: Paving
2.71
0.56
Asphalt Paving and Roofing
2.71
0.56
Miscellaneous Other
1.83
0.67
Note: *These values are the ratios of 1985 Trends emissionsto
1975 Trends emissions for each source category. For example,
the commercial/ institutional fuel combustion: cod emission
ratio of 2.11 is computed as the ratio of the 1975 TSP emissions
of 40 gigagrams per year to the corresponding 1985 emissions of
19 gigagrams per year.
sions data estimated from the above procedure served
as the benchmark and initial value for the CRESS
emissions module for both scenarios.
CAA regulation of commercial/ residential emis-
sions was limited and largely confined to fuel com-
bustion sources (S02, NOx, TSP), gasoline marketing
(VOC), dry cleaning (VOC), and surface coating
(VOC). NSPS regulations of small (over 29 MW ca-
pacity) fuel combustors were promulgated in 1984 and
1986. For purposes of emissions calculations, the
stipulated NSPS for S02, NOx, and TSP were incor-
porated into the control scenario for 1985 and 1990.
Emission rates for source categories subject to VOC
regulation were similarly adjusted.
Energy Data
Nearly 75 percent of the source categories in
CRESS use energy consumption by State and sector
as the driver for the emissions calculation. State-level
energy consumption statistics are published by EIA
in State Energy Data Report, Consumption Estimates,
1960-1989, and are electronically available as part of
the State Energy Data System (SEDS ) (DOE, 1991).
The SEDS database contains annual energy consump-
tion estimates by sector for the various end-use sec-
tors: residential, commercial, industrial and transpor-
tation, and electric utilities.
Seven fuel-type categories are used in CRESS:
coal, distillate oil, residual oil, natural gas, liquid pe-
troleum gas, wood, and electricity. The model assumes
zero consumption of residual fuel oil in the residen-
tial sector and zero consumption of wood in the com-
mercial sector. Energy consumption for each fuel-type
was expressed in BTUs for purposes of model calcu-
lations. With the exception of wood consumption, all
of the energy consumption statistics used in CRESS
were obtained from SEDS.
Residential wood consumption estimates were
derived from two data sources. State-level residential
sector wood consumption estimates for 1975 and 1980
were obtained from Estimates of U.S. Wood Energy
Consumption from 1949 to 1981 (EIA, 1982). State-
level wood consumption, however, was not available
for 1985 and 1990, therefore, regional information
from an alternative publication, Estimates of U.S.
Biofuels Consumption 1990 (EIA, 1990), was used to
derive State-level residential wood use figures. Re-
gional 1985 and 1990 wood consumption was distrib-
uted among States using 1981 State shares. All wood
consumption figures were converted to BTU's using
an average value of 17.2 million BTU per short ton.
Economic/Demographic Data
Emissions from slightly more than 25 percent of
the CRESS source categories follow State-level eco-
nomic and demographic activity variables. The de-
mographic variables used by CRESS include State-
level population, rural population, and forest acreage.
State population is the activity indicator for six emis-
sions source categories for S02, NOx, TSP, and CO,
and 13 VOC source categories. State population data
were assembled from the SEDS database. Rural popu-
lation, which is the indicator of residential open burn-
ing activity, is computed as a fraction of total State
B-33
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
population. Forest wildfires and managed open burn-
ing activity are related to 1977 State-level forest acre-
age. The demographic information is assumed to be
invariant to CAA regulations and thus is the same in
the control and no-control scenarios.
Car stock (or vehicle population), the driver vari-
able for the auto body refinishing, is approximated by
State motor vehicle registrations. Highway Statistics,
an annual publication by the FHWA, was the source
for data on State motor vehicle registrations. The three
source categories connected with gasoline marketing
are driven by State-level gasoline sales in gallons. State
gasoline consumption was obtained from the SEDS
data base. Housing starts and 10 percent of the exist-
ing housing stock were combined to form the activity
indicator for architectural surface coating emissions.
Housing data compiled by the U.S. Bureau of the
Census were available in the Statistical Abstract of
the United States (DOC, 1975; 1977; 1982; 1983;
1987; 1993). Regional-level data for 1975 was allo-
cated to the States based on the 1980 State distribu-
tion.
No-control Scenario Emissions
Adjustments to control scenario emissions in each
of the target years to reflect conditions un-
der the no-control scenario were achieved
through emission factors, energy input data,
and economic/demographic data. The adjust-
ments made to each of these variables to gen-
erate no-control scenario emissions are dis-
cussed individually in the following subsec-
tions.
Energy Data
State-level energy demand for the residential and
commercial sectors for the no-control scenario was
estimated from the J/W model forecast. Final energy
demand estimates for the household sector were cal-
culated by an EPA contractor for the purposes of the
no-control scenario analysis. State allocation of the
national-level estimates was based on historic State
shares, i.e., this assumes that there is no change in the
distribution of energy demand across States as a re-
sult of removing regulations. In addition, the J/W
model estimates an aggregate refined petroleum cat-
egory and does not distinguish among liquid petro-
leum gas, distillate oil, and residual oil. The relative
shares among these three categories of petroleum prod-
ucts remained constant between the control and
no-control scenarios. The information on percentage
change in energy demand by fuel type as provided by
the J/W model is listed in Table B-13.
The differential for commercial sector final en-
ergy demand was calculated from the combination of
four intermediate product flow categories from the J/
W forecast. The National Income and Product Ac-
counts (NIPA ) for the commercial sector correspond
to J/W SIC categories 32 through 35:
Table B-13. Percentage Change in Real Energy Demand bv
Households from Control to No-control Scenario.
Emissions Data
CAA regulation of the commercial/resi-
dential sector was minimal. For regulated
source categories, emission factors were re-
vised to reflect pre-regulation emission rates.
Six commercial/residential source categories were
regulated for VOC emissions: Service Stations Stage
I Emissions, Service Stations Stage II Emissions, Dry
Cleaning (perchloroethylene), Gasoline Marketed, Dry
Cleaning (solvent), and Cutback Asphalt Paving.
Commercial-Institutional boilers were regulated for
S02 and TSP and internal combustion sources were
regulated for NOx emissions. All NSPS were removed
for these sources to estimate no-control scenario emis-
sions levels.
Year
(,'Oill
Roll noil Petroleum
Klcctric
Natural
Cias
1975
1.48
4.76
3.62
2.42
1980
1.50
3.84
4.26
2.12
1985
1.98
3.90
3.88
2.41
1990
2.23
4.33
4.18
2.77
(32) Wholesale and Retail Trade;
(33) Finance, Insurance, and Real Estate;
(34) Other Services; and
(35) Government Services.
Percentage change information from the J/W fore-
cast for energy cost shares, value of output, and en-
ergy prices was used to calculate the differential in
commercial sector energy demand for the no-control
scenario. The energy cost share is defined as the cost
B-34
-------
Appendix B: Emissions Modeling
of energy input divided by the value of the output. In
order to calculate the percentage change in commer-
cial sector energy demand, the change in energy price
was subtracted from the percentage change in energy
cost, and added to the change in the value of output.
Each of these variables was available from the J/W
model results. This calculation was performed for each
of the four energy types, and each of the four NIPA
categories. The change in commercial sector energy
demand was obtained by taking the weighted average
of the four NIPA categories. Since data on relative
energy demand for NIPA categories were not readily
available, square footage was used as a proxy for cal-
culating the weights. These data were taken from the
Nonresidential Buildings Energy Consumption Sur-
vey, Commercial Buildings Consumption and Expen-
diture 1986 (EIA, 1989). The resulting estimate for
commercial sector changes in energy demand is pro-
vided in Table B-14.
Tabic B-14. Percentage Change in Commercial Energy Demand
from Control to No-control Scenario.
Year
(,'Oill
Refined
Petroleum
Klectric
Niitu nil
Gas
1975
-0.13
3.36
1.30
-0.80
1980
0.31
1.90
2.06
-0.82
1985
0.48
1.98
1.72
-0.40
1990
0.39
2.26
1.74
-0.22
The national-level change in commercial sector
energy demand was allocated to the States using his-
toric shares. Implicit is the assumption that removal
of CAA regulations does not alter the State distribu-
tion of energy use.
Economic/Demographic Data
State population was assumed not to vary as a re-
sult of CAA regulations, thus only the economic vari-
ables were revised for the no-control scenario.
No-control scenario housing starts and car stock were
derived from J/W forecast information on construc-
tion and motor vehicles. The differential for catego-
ries 6 (construction) and 24 (motor vehicles and equip-
ment) was applied to control scenario values to ob-
tain no-control scenario levels. The percentage change
from the J/W forecast is given in Table B-15.
State-level gasoline sales is one of the activities
forecasted by the transportation sector model. The
percentage change in gasoline sales calculated by the
TEEMS model was used in the no-control scenario as
a CRESS model input.
Table B-15. J/W Percent Differential in
Economic Variables Used in CRESS.
Year
Construction
Motor
Vehicles
1975
0.70
5.04
1980
0.14
4.79
1985
0.41
6.07
1990
0.29
6.25
B-35
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic B-16. TSP Emissions Under the Control and No-control Scenarios by Target Year (in
thousands of short tons).
Whh lbcCAA
Without the CAA
lDiHRfeiremKCg:
in 1990
Sector
1OT5
19B0
«5
1990
1975
Wffl
1985
1990
Kill issions
Transportation:
Highway Vehicles
700
760
770
820
770
910
1,030
1,180
(30%)
Off-Highway Vehicles
270
280
270
280
260
270
260
270
5%
Stationary Sources:
Electric Utilities
1,720
880
450
430
3,460
4,480
5,180
5,860
(93%)
Industrial Processes
5,620
3,650
3,040
3,080
11,120
12,000
11,710
12,960
(76%)
Industrial Boilers
740
480
250
240
780
550
360
400
(41%)
Commercial/Residential
2,020
2,510
2,680
2,550
2,020
2,520
2,700
2,560
(1%)
TOTAL'"
11,070
8,550
7,460
7,390
18,410
20,730
21,250
23,23 0
(68%)
Notes: The estim ates of em ission levels with and without the CAA were developed specifically for this section 812 analysis using
models designed to simulate conditions in the absence of the CAA. These numbers should not be interpreted as actual historical
emission estimates.
*Totals may differ slightly from sums due to rounding.
Table B-17. SO2 Emissions Under the Control and No-control Scenarios by Tan>ct Year (in thousands
of short tons).
SiOdkM'
WiM«tiaieCAA
Without the CAA
TDlrifflfeirtainffire
ik 11®$®
llmisswns
1975
1980
1985
1990
W5
BSffl
1985
IWffl
Transportation:
Highway Vehicles
380
450
500
570
380
450
500
560
1%
Off-Highway Vehicles
370
530
410
390
360
530
400
390
1%
Stationary Sources:
Electric Utilities
18,670
17,480
16,050
16,510
20,690
25,620
25,140
26,730
00
Industrial Processes
4,530
3,420
2,730
2,460
5,560
5,940
5,630
6,130
(60",,)
Industrial Boilers
3,440
3,180
2,660
2,820
3,910
4,110
4,020
4,610
(39%)
Comm ercial/Re sidentia 1
1,000
800
590
690
1,000
810
610
710
(3%)
TOTAL*
28,380
25,860
22,950
23,440
31,900
37,460
36,310
39,140
(40%)
Not ps- The estimates of emission levels with and without the CAA were developed specifically for this section 812 analysis using
models designed to simulate conditions in the absence of the CAA. These numbers should not be interpreted as actual historical
emission estimates.
*Totals may differ slightly from sums due to rounding.
B-36
-------
Appendix B: Emissions Modeling
Tabic B-18. NOx Emissions Under the Control and No-control Scenarios by Target Year (in
thousands of short tons).
SedtaM-
Wiitlh tike CAA
Without Ihc CAA
n2nfl5ar£mion>
1975
1980
1985
1990
WIS
119M
19X5
IWffl
Transportation:
Highway Vehicles
12,220
10,770
9,470
7,740
14,620
16,460
19,800
23,010
Off-Highway Vehicles
1,380
1,370
1,340
1,410
1,390
1,420
1,390
1,490
<5"„)
S tat io n ar y S o u re es:
Electric Utilities
20
30
30
40
20
30
30
40
<7"„)
Industrial Processes
5,910
6,780
6,230
5,630
6,130
7,930
7,290
6,810
(1 7"o)
Industrial Boilers
150
150
150
150
150
150
140
150
0"o
Comm ercial/Re sidentia 1
4,980
5,480
5,820
5,870
4,980
5,700
6,080
6,130
<4"„)
TOTAL*
24,660
24,580
23,030
20,8 4 0
27,290
31,680
34,730
37,630
<45"„)
Nntfis' The estimates of emission levels with and without the CAA were developed specifically for this section 812 analysis using
models designed to simulate conditions in the absence of the CAA. These numbers should not be interpreted as actual historical
emission estimates.
*Totals may differ slightly from sums due to rounding.
B-37
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic B-20. CO Emissions Under the Control and No-control Scenarios by Target Year (in thousands
of short tons).
Sector
Wilh llicCAA
Wiitlkoxiilt ttllnc CAA
inn
Emissions
1975
1980
1*5
1990
1975
1980
19X5
1990
Transportation:
Highway Vehicles
83,580
79,970
72,490
65,430
90,460
105,530
131,420
149,280
(5 6»„)
Off-Highway Vehicles
8,510
8,100
7,880
8,080
8,510
8,070
7,880
8,080
0"o
Stationary Sources:
Electric Utilities
240
280
290
370
250
290
300
380
(3%)
Industrial Processes
7,580
6,990
4,840
5,140
9,240
9,120
8,860
10,180
(4 9",,)
Industrial Boilers
720
710
670
740
720
710
620
740
0"o
Comm ercial/Re siden tial
10,250
13,130
14,140
13,150
10,250
13,170
14,200
13,210
0%
TOTAL*
110,880
109,170
100,300
92,900
119,430
136,880
163,280
181,860
(4 9",,)
VnUv The estimates of emission levels with and without the CAA were developed specifically for this section 812 analysis using
models designed to simulate conditions in the absence of the CAA These numbers should not be interpreted as actual historical
emission estimates.
*T otals may differ slightly from sums due to rounding.
Table B-21. Lead (Pb) Emissions Under the Control and No-control Scenarios by Tan>ct Year (in
thousands of short tons).
S!©dt®r
Wltk tike CAA
Without the CAA
DiiTarcinicc
in 1990
EnmmsBMtnns
1975
1980
1985
199®
1975
1980
19X5
1990
Transportation:
Highway Vehicles
180
86
22
2
203
207
214
223
(99%)
Stationary Source:
Industrial Processes
3
1
1
1
7
7
6
5
(87%)
Industrial Combustion
4
2
0
0
5
5
5
5
(96%)
I itililies
1
1
0
0
2
3
4
4
(95%)
TOTAL'"
190
90
23
3
217
221
228
237
(99%)
Notes; The estimates of emission levels with and without the CAA were developed specifically for this section 812 analysis using
models designed to simulate conditions in the absence of the CAA These numbers should not be interpreted as actual historical
emission estimates.
*Totals may differ slightly from sums due to rounding.
B-38
-------
Appendix B: Emissions Modeling
Emissions Modeling References
Abt Associates Inc. (Abt). 1995. The Impact of the
Clean Air Act on Lead Pollution: Emissions
Reductions, Health Effects, and Economic
Benefits from 1970 to 1990, Final Report,
Bethesda, MD, October.
Argonne National Laboratory (ANL). 1990. Current
Emission Trends for Nitrogen Oxides, Sulfur
Dioxide, and Volatile Organic Chemicals by
Month and State: Methodology and Results,
Argonne, IL, August.
Argonne National Laboratory (ANL). 1992. Retro-
spective Clean Air Act Analysis: Sectoral
Impact on Emissions from 1975 to 1990,
(Draft), Argonne, IL, July.
Braine, Bruce and P. Kim. 1993. Fuel Consumption
and Emission Estimates by State, ICF Re-
sources, Inc., Fairfax, VA, memorandum to
Jim DeMocker, EPA. April 21.
Braine, Bruce, S. Kohli, and P. Kim. 1993.1975 Emis-
sion Estimates with and without the Clean Air
Act, ICF Resources, Inc., Fairfax, VA, memo-
randum to Jim DeMocker, EPA, April 15.
DeMocker, J. Personal Communication with Office
of Mobile Sources Staff, Ann Arbor, Michi-
gan. Date unknown.
Energy Information Administration (EIA). 1982. Es-
timates of U.S. Wood Energy Consumption
from 1949 to 1981. DOE/EIA-0341, U.S.
Department of Energy. August.
Energy Information Administration (EIA). 1985. Cost
and Quality of Fuels for Electric Utility
Plants. DOE/EIA-0091(85), U.S. Department
of Energy.
Energy Information Administration (EIA). \9%9.Non-
residential Buildings Energy Consumption
Survey: Commercial Buildings Consumption
and Expenditures 1986. DOE/EIA-0318(86),
U.S. Department of Energy. May.
Energy Information Administration (EIA). 1990. Es-
timates of U.S. Biofuels Consumption 1990.
DOE/EIA-0548(90), U.S. Department of En-
ergy. October.
Electric Power Research Institute (EPRI). 1981. The
EPRI Regional Systems, EPRI-P-1950-SR,
Palo Alto, CA.
Federal Highway Administration (FHWA). 1986.
1983-1984 Nationwide Personal Transporta-
tion Survey, U.S. Department of Transporta-
tion, Washington, DC.
Federal Highway Administration (FHWA). 1988.
Highway Statistics 1987, PB89-127369, U.S.
Department of Transportation, Washington,
DC.
Federal Highway Administration (FHWA). 1992.
Highway Statistics 1991, FHWA-PL-92-025,
U.S. Department of Transportation, Washing-
ton, DC.
Gschwandtner, Gerhard. 1989.Procedures Document
for the Development of National Air Pollut-
ant Emissions Trends Report, E.H. Pechan &
Associates, Inc., Durham, NC. December.
Hogan, Tim. 1988. Industrial Combustion Emissions
Model (Version 6.0) Users Manual, U.S. En-
vironmental Protection Agency, EPA-600/8-
88-007a.
ICF Resources, Inc. 1992. Results of Retrospective
Electric Utility Clean Air Act Analysis -1980,
1985 and 1990, September 30.
Jorgenson, D.W. and P. Wilcoxen. 1989.Environmen-
tal Regulation and U.S. Economic Growth,
Harvard University Press, Cambridge, MA.
Klinger, D. and J.R. Kuzmyak. 1986. Personal Travel
in the United States, Vol. I: 1983-84 Nation-
wide Personal Transportation Study, U.S.
Department of Transportation, Federal High-
way Administration, Washington, DC. Au-
gust.
Kohout, Ed. 1990. Current Emission Trends for Ni-
trogen Oxides, Sulfur Dioxide, and Volatile
Organic Compounds by Month and State:
Methodology and Results," Argonne National
Laboratory, ANL/EAIS/TM-25, Argonne, IL.
B-39
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Lockhart, Jim. 1992. Projecting with and without
Clean Air Act Emissions for the Section 812
Retrospective Analysis: A Methodology Based
Upon the Projection System used in the Of-
fice of Air Quality Planning and Standards
National Air Pollutant Emission Estimate
Reports," (Draft Report), Environmental Law
Institute, November 16.
McDonald, J.F. and D.W. South. 1984. The Commer-
cial and Residential Energy Use and Emis-
sions Simulation System (CRESS): Selection
Process, Structure, and Capabilities, Argonne
National Laboratory, ANL/EAIS/TM-12,
Argonne, IL. October.
Mintz, M.M. and A.D. Vyas. 1991 .Forecastof Trans-
portation Energy Demand through the Year
2010, Argonne National Laboratory, ANL/
ESD-9, Argonne, IL. April.
Pechan Associates. 1995 .The Impact ofthe Clean Air
Act on 1970 to 1990 Emissions; Section 812
Retrospective Analysis. Draft Report. March.
Saricks, C.L. 1985. The Transportation Energy and
Emissions Modeling System (TEEMS): Selec-
tion Process, Structure, and Capabilities,
Argonne National Laboratory, ANL/EES-
TM-295, Argonne, IL. November.
Veselka, T.D., etal. 1990.Introduction to the Argonne
Utility Simulation (ARGUS) Model," Argonne
National Laboratory, ANL/EAIS/TM-10,
Argonne, IL. March.
Vyas, A.D. and C.L. Saricks. 1986. Implementation
of the Transportation Energy and Emissions
Modeling System (TEEMS) in Forecasting
Transportation Source Emissions for the 1986
Assessment, Argonne National Laboratory,
ANL/EES-TM-321, Argonne, IL. October.
U.S. Department of Commerce (DOC). 1975. Statis-
tical Abstract of the United States: 1975 (96th
Edition), Bureau of the Census, Washington,
DC, September.
U.S. Department of Commerce (DOC). 1977. Statis-
tical Abstract of the United States: 1977 (98th
Edition), Bureau of the Census, Washington,
DC, September.
U.S. Department of Commerce (DOC). 1981. 1977
Truck Inventory and Use Survey, Bureau of
the Census, TC-77-T-52, Washington, DC,
August.
U.S. Department of Commerce (DOC). 1982. Statis-
tical Abstract of the UnitedStates: 1982-1983
(103rdEdition), Bureau ofthe Census, Wash-
ington, DC, December.
U.S. Department of Commerce (DOC). 1983. Statis-
tical Abstract of the United States: 1984
(104th Edition), Bureau of the Census, Wash-
ington, DC, December.
U.S. Department of Commerce (DOC). 1984. 1982
Truck Inventory and Use Survey, Bureau of
the Census, TC-82-T-52, Washington, DC,
August.
U.S. Department of Commerce (DOC). 1987. Statis-
tical Abstract of the United States: 1988
(108th Edition), Burea of the Census, Wash-
ington, DC, December.
U.S. Department of Commerce (DOC). 1990. 1987
Truck Inventory and Use Survey," Bureau of
the Census, TC87-T-52, Washington, DC,
August.
U.S. Department of Commerce (DOC). \99\.Annual
Survey of Manufactures: Purchased Fuels and
Electric Energy Used for Heat and Power by
Industry Group, Bureau of the Census,
M87(AS)-1, Washington, DC.
U.S. Department of Commerce (DOC). 1993. Statis-
tical Abstract of the United States: 1993
(113th Edition)," Bureau of the Census,
Washington, DC.
U.S. Department of Energy (DOE). 1982. Documen-
tation of the Resource Allocation and Mine
Costing (RAMC) Model. DOE/NBB-0200.
Energy Information Administration.
U.S. Department of Energy (DOE). 1986. Inventory
of Power Plants in the United States 1985.
DOE/EIA-0095(85), Energy Information
Administration, Washington, DC, August.
B-40
-------
Appendix B: Emissions Modeling
U.S. Department of Energy (DOE). 1988. An Analy-
sis of Nuclear Power Plant Operating Costs.
DOE/EIA-0511(88), Energy Information
Administration.
U.S. Department of Energy (DOE). 1990. State En-
ergy Price and Expenditure Report 1988.
DOE/EIA-0376(88), Energy Information
Administration, Washington, DC, September.
U.S. Department of Energy (DOE). 1991. State En-
ergy Data Report: Consumption Estimates -
1960-1989. DOE/EIA-0214(89), Energy In-
formation Administration, Washington, DC,
May.
U.S. Department of Energy (DOE). 1992. Annual
Energy Review 1991. DOE/EIA-03 84(91),
Energy Information Administration, Wash-
ington, DC.
U.S. Environmental Protection Agency (EPA). 1985.
Compilation of Air Pollutant Emission Fac-
tors, Volume I: Stationary Point and Area
Sources, AP-42, Fourth Edition, GPO No.
055-000-00251-7, Research Triangle Park,
NC. September.
U.S. Environmental Protection Agency (EPA). 1989.
The 1985 NAPAP Emissions Inventory, EPA-
600/7-89-012a, Research Triangle Park, NC.
November.
U.S. Environmental Protection Agency (EPA). 1990.
The Cost of a Clean Environment, EPA-230-
11-90-083. November.
U.S. Environmental Protection Agency (EPA). 1991.
Office of Air Quality Planning and Standards,
National Air Pollutant Emissions Estimates,
1940-1990, EPA-450/4-91-026, Research
Triangle Park, NC. November.
U.S. Environmental Protection Agency (EPA). 1992.
1990 Toxics Release Inventory, EPA-700-S-
92-002, Washington, DC.
U.S. Environmental Protection Agency (EPA). 1994a.
National Air Pollutant Emission Trends,
1900-1993, EPA-454/R-94-027, Office of Air
Quality Planning and Standards, Research
Triangle Park, NC. October.
U.S. Environmental Protection Agency (EPA). 1994b.
Office of Mobile Sources, User's Guide to
MOBILE5 (Mobile Source Emission Factor
Model), EPA-AA-AQAB-94-01, Ann Arbor,
MI. May.
U.S. Environmental Protection Agency (EPA). 1995.
National Air Pollutant Emission Trends 1900-
1994, EPA-454/R-95-011. Office of Air Qual-
ity Planning and Standards, Research Triangle
Park, NC. October.
Werbos, Paul J. 1983. A Statistical Analysis of What
Drives Energy Demand: Volume III of the
PURHAPSModel Documentation, U.S. De-
partment of Energy, Energy Information Ad-
ministration, DOE/EIA-0420/3, Washington,
DC.
B-41
-------
Appendix C: Air Quality Modeling
Introduction
This appendix describes in greater detail the vari-
ous methodologies used to translate differences in
control and no-control scenario emission estimates
into changes in air quality conditions. Summary char-
acterizations of the results of the air quality modeling
efforts for 1990 are provided here and in the main
text. Further details and discussion of key analytical
and modeling issues can be found in a number of sup-
porting documents. These documents, which provide
the analytical basis for the results presented herein,
are:
~ ICF Kaiser/Systems Applications Interna-
tional, "Retrospective Analysis of Ozone Air
Quality in the United States Final Report,
May 1995. (Hereafter referred to as "SAI
Ozone Report (1995).")
~ ICF Kaiser/Systems Applications Interna-
tional, "Retrospective Analysis of Particulate
Matter Air Quality in the United States Draft
Report, September 1992. (Hereafter referred
to as "SAI PM Report (1992).")
~ ICF Kaiser/Systems Applications Interna-
tional, "Retrospective Analysis of Particulate
Matter Air Quality in the United States", Fi-
nal Report, April 1995. (Hereafter referred to
as "SAI PM Report (1995).")
~ ICF Kaiser/Systems Applications Interna-
tional, "PM Interpolation Methodology for
the section 812 retrospective analysis",
Memorandum from J. Langstaff to J.
DeMocker, March 1996. (Hereafter referred
to as "SAI PM Interpolation Memo (1996).")
~ ICF Kaiser/Systems Applications Interna-
tional, "Retrospective Analysis of S02, NOx
and CO Air Quality in the United States
Final Report, November 1994. (Hereafter re-
ferred to as "SAI SO,, NO and CO Report
(1994).")
~ ICF Kaiser/Systems Applications Interna-
tional, "Retrospective Analysis of the Impact
of the Clean Air Act on Urban Visibility in
the Southwestern United States Final Re-
port, October 1994. (Hereafter referred to as
"SAI SW Visibility Report (1994).")
~ Dennis, Robin L., US EPA, ORD/NERL,
"Estimation of Regional Air Quality and
Deposition Changes Under Alternative 812
Emissions Scenarios Predicted by the Re-
gional Acid Deposition Model, RADM", Draft
Report, October 1995. (Hereafter referred to
as "RADM Report (1995).")
The remainder of this appendix describes, for each
pollutant or air quality effect of concern, (a) the basis
for development of the control scenario air quality
profiles; (b) the air quality modeling approach used
to estimate differences in air quality outcomes for the
control and no-control scenario and the application of
those results to the derivation of the no-control sce-
nario air quality profiles; (c) the key assumptions,
caveats, analytical issues, and limitations associated
with the modeling approach used; and (d) a summary
characterization of the differences in estimated air
quality outcomes for the control and no-control sce-
narios.
Carbon Monoxide
Control scenario carbon monoxide
profiles
As described in the preceding general methodol-
ogy section, the starting point for development of con-
trol scenario air quality profiles was EPA's AIRS da-
C-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic C-l. Summary of CO Monitoring Data.
Year
Number of
Monito rs
Number of
Counties
Percent
Population
Covered
Number of
Samples
Mean
Number of
Samples per
Monito r
1970
82
54
n/a
408,524
4.982
1975
503
246
n/a
2,667,525
5.303
1980
522
250
50 %
3,051,599
5.846
1985
472
232
n/a
3,533,286
7.486
1990
506
244
55 %
3,788,053
7,486
Data Source: SAI S02, NOx and CO Report (1994).
tabase. Hourly CO air quality monitoring data were
compiled for all monitors in the 48 contiguous states
for the study target years of 1970, 1975, 1980, 1985,
and 1990. Although the CO monitoring network was
sparse in 1970, by 1990 506 monitors in 244 counties
provided monitoring coverage for 55 percent of the
population in the conterminous U.S. Table C-l sum-
marizes the CO monitoring data derived from AIRS.
Additional data regarding the EPA Region location,
land use category, location-setting category, and ob-
jective category of the monitors providing these data
are described in the SAI S02, NOx, and CO Report
(1994).
The next step in constructing the control scenario
air quality profiles was to calculate moving averages,
for a variety of time periods, of the hourly CO data
for each monitor. For CO, moving averages of 1, 3, 5,
7, 8, 12, and 24 hours were calculated. Daily maxi-
mum concentrations observed at each monitor for each
of these averaging periods were then calculated. Fi-
nally, profiles were developed to reflect the average
and maximum concentrations for each of the seven
averaging periods. However, profiles were only de-
veloped for a given monitor when at least 10 percent
of its theoretically available samples were actually
available. The purpose of applying this cutoff was to
avoid inclusion of monitors for which available sample
sizes were too small to provide a reliable indication
of historical air quality.
As discussed in the air quality modeling chapter
of the main text, development of representative dis-
tributions for these profiles was then necessary to pro-
vide a manageable char-
acterization of air qual-
ity conditions. Initially,
two-parameter lognor-
mal distributions were
fitted to the profiles
based on substantial evi-
dence that such distribu-
tions are appropriate for
modeling air quality
data. However, given
the relative importance
of accurately modeling
higher percentile obser-
vations (i.e., 90th per-
centile and higher), a
three-parameter model-
ing approach was used
to isolate the effect of
observations equal, or very close, to zero. In this ap-
proach one parameter defines the proportion of data
below a cutoff close to zero and the remaining two
parameters describe the distribution of data above the
cutoff value. Several other studies have already dem-
onstrated good fit to air quality modeling data with a
three-parameter gamma distribution, and both lognor-
mal and gamma distributions using a three-parameter
approach were developed for the present study. As
documented in the SAI S02, NOx, and CO Report
(1994), a cutoff of 0.05 ppm was applied and both the
three-parameter lognormal and three-parameter
gamma distributions provided a good fit to the em-
pirical data. For CO, the gamma distribution provided
the best fit.
The control scenario air quality profiles are avail-
able on diskette. The filename for the CO Control
Scenario profile database is COCAA.DAT, and adopts
the format presented in Table C-2.
No-control scenario carbon monoxide
profiles
To derive comparably configured profiles repre-
senting CO air quality in the no-control scenario, con-
trol scenario profile means and variances were ad-
justed in proportion to the difference in emissions es-
timated under the two scenarios. Specifically, for all
control scenario air quality observations predicted by
the three-parameter distributions falling above the
"near-zero" cutoff level, comparable no-control esti-
mates were derived by the following equation:
C-2
-------
Appendix C: Air Quality Modeling
Tabic C-2, Format of Air Quality Profile Databases.
Columns
Konii at
Doscri ption
1 - 2
Integer
Year (70, 75, 80, 85, 90)
4-6
Integer
Averaging time (1, 3, 5, 7, 8, 12, 24 hours)
8-9
Integer
State FIPS code
11-13
Integer
County FIPS code
15 -19
Integer
Monitor number (digits 6-10 of monitor id)
21 -30
Real
Latitude
32-41
Real
Longitude
43-44
Integer
Latitude/longitude flag'
46-55
Real (F10.3)
Hourly intermittency parameter p
56 - 65
Real (F10.3)
Hourly lognormal parameter (ib
66-75
Real (F10.3)
Hourly lognormal parameter a
76-85
Real (F10.3)
Hourly gamma parameter a
86-95
Real (F10.3)
Hourly gamma parameter pb
96-105
Real (F10.3)
Daily max intermittency parameter p°
106- 115
Real (F10.3)
Daily max lognormal parameter (ib
116- 125
Real (F10.3)
Daily max lognormal parameter a
126- 135
Real (F10.3)
Daily max gamma parameter ab
136- 145
Real (F10.3)
Daily max gamma parameter pb
aValues for flag: 1= actual latitude/longitude values
2= latitude/longitude values from collocated monitor or previous monitor
location (monitor parameter occurrence code 1)
-9 = latitude/longitude missing (county center substituted)
b Units of concentration are ppm for CO and ppb for SO2, NO2 and NO.
Srmrr.p- SAI SO2, NOx and CO Report (1994).
^c =
V
E,
V J
{Xr-b) + b
(i)
where
XNC = air quality measurement for the no-control scenario,
Xc = air quality measurement for the control scenario,
Enc = emissions estimated for the no-control scenario,
Ec = emissions estimated for the control scenario, and
b = background concentration.
C-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
The adjustment for background concentration is
made to hold ambient background concentrations of
the pollutant constant between the control and
no-control scenarios. To the extent background con-
centrations are affected by transport of anthropogenic
pollutants from upwind sites, and to the extent up-
wind emissions may have been controlled under the
control scenario, assuming a fixed background con-
centration represents a conservative assumption in this
analysis. As discussed in the SAI S02, NOx, and CO
Report (1994), the CO background concentration used
for this analysis was 0.2 ppm, which equals the low-
est typical concentration observed in the lower 48
states.
In the SAI S02,NOx, and CO Report (1994) docu-
menting the CO air quality modeling effort, reference
is made to using county-level emission estimates as
the basis for deriving the no-control profiles. Deriva-
tion of these county-level results is described in more
detail in the appendix on emissions estimation. It is
important to emphasize here, however, that the county-
level CO emissions data were derived for both the
control and no-control scenarios by simple popula-
tion-weighted disaggregation of state-level emission
totals. Although CO emission estimates were needed
at the county level to support the ozone air quality
modeling effort, differences in state-level emissions
estimates are what drive the difference in the control
and no-control air quality profiles for CO. In other
words, the to i? , ratios used to derive the
' NCAA CAA
no-control profiles according to Equation (1) above
are essentially based on state-level emissions estimates
for CO.
As for the control scenario air quality profiles,
the no-control scenario air quality profiles are avail-
able on diskette. The filename for the CO No-control
Scenario profile database is CONCAA.DAT. The
same data format described in Table C-2 is adopted.
Summary differences in carbon
monoxide air quality
Figure C-l. Frequency Distribution of Estimated Ratios
for 1990 Control to No-control Scenario 95th Percentile
1-Hour Average CO Concentrations, by Monitor.
300
s 100
j i i i [_
0.05 0.25 0.45 0.65 0.85 1.05 1.25
Ratio of CAA:No-CAA 95th Percentile 1-Hour Average
which the ratio of 1990 control to no-control scenario
95th percentile 1-hour average concentrations falls
within a particular range. The x-axis values in the
graph represent the midpoint of each bin. The results
indicate that, by 1990, CO concentrations under a no-
control scenario would have been dramatically higher
than control scenario concentrations.
Key caveats and uncertainties for
carbon monoxide
A number of important uncertainties should be
noted regarding the CO air quality estimates used in
this analysis. First and foremost, CO is a highly local-
ized, "hot spot" pollutant. As such, CO monitors are
often located near heavily-used highways and inter-
sections to capture the peak concentrations associated
with mobile sources. Since this analysis relies on state-
level aggregate changes in CO emissions from all
sources, the representativeness and accuracy of the
predicted CO air quality changes are uncertain. There
is no basis, however, for assuming any systematic bias
which would lead to over- or under-estimation of air
quality conditions due to reliance on state-wide emis-
sion estimates.
While the control and no-control scenario air qual-
ity profiles are too extensive to present in their en-
tirety in this report, a summary indication of the dif-
ference in control and no-control scenario CO con-
centrations is useful. Figure C-l provides this sum-
mary characterization. Specifically, the air quality
indicator provided is the 95th percentile observation
of 1990 CO concentrations averaged over a 1-hour
period. The graph shows the number of monitors for
A second source of uncertainty is the extent to
which the three-parameter distributions adequately
characterize air quality indicators of concern. Appen-
dix C of the SAI S02, NOx, and CO Report (1994)
presents a number of graphs comparing the fitted ver-
sus empirical data for one-hour and 12-hour averag-
ing periods. In the case of CO, the gamma distribu-
tion appears to provide a very reasonable fit, though
clearly some uncertainty remains.
C-4
-------
Appendix C: Air Quality Modeling
Tabic C-3. Summary of SO2 Monitoring Data.
Yen r
iNiim bcr of
Monitors
INu in bcr of
Counties
l'e rcen t
l'opu hi t i o n
Covered
Nuill bcr of
S;i m pies
Me ii n iN u ill bcr
ofSa ill pics per
Monitor
1970
86
56
n/a
399,717
4.648
1975
847
340
n/a
4,280,303
5.053
1980
1,113
440
60 %
6,565,589
5.899
1985
926
401
n/a
6,602,615
7.130
199(1
769
374
50 %
5,810,230
7.556
Data Source: SAI S02, NOx and CO Report (1994).
Finally, a central
premise of this analy-
sis is that changes in
CO emissions should
be well-correlated
with changes in CO
air quality. Strong
correlation between
the state-level emis-
sions estimates used
in this analysis and
empirical air quality
measurements would
not be expected due to
inconsistencies be-
tween the state-level
scale of modeled
emissions versus the
monitor-level scale of the air quality data, and between
the modeled control scenario emissions inventories
and actual historical air quality measurements. Under
these circumstances, it is particularly important to
focus on the primary objective of the current analy-
sis, which is to estimate the difference in air quality
outcomes between scenarios which assume the ab-
sence or presence of historical air pollution controls.
In the process of taking differences, some of the un-
certainties are expected to cancel out. No attempt is
made in the overall analysis to predict historical air
quality, or hypothetical air quality in the absence of
the Clean Air Act, in absolute terms.
Sulfur Dioxide
Sulfur dioxide (S02) emissions lead to several air
quality effects, including secondary formation of fine
particle sulfates, long range transport and deposition
of sulfuric acid, and localized concentrations of gas-
eous sulfur dioxide. The first two effects are addressed
later in this appendix, under the particulate matter and
acid deposition sections. The focus of this section is
estimation of changes in local concentrations of sul-
fur dioxide.
The methodology applied to estimation of local
sulfur dioxide air quality is essentially identical to the
one applied for carbon monoxide. As such, this sec-
tion does not repeat the "roll-up" modeling method-
ological description presented in the CO section, but
instead simply highlights those elements of the sulfur
dioxide modeling which differ from carbon monox-
ide.
Control scenario sulfur dioxide profiles
Unlike the CO monitoring network, the number
of monitors as well as the population coverage of the
S02 monitoring network shrank during the 1980's.
Table C-3 summarizes the S02 monitoring data used
as the basis for development of the control scenario
air quality profiles.
As for CO, air quality profiles reflecting average
values and daily maxima for 1, 3, 5, 7, 8, 12, and 24
hour averages were compiled from AIRS for moni-
tors in the lower 48 states which had at least 10 per-
cent of their potential samples available. Applying a
cutoff of 0.1 ppb to isolate the zero and near-zero ob-
servations, three-parameter lognormal and gamma
distributions were fitted to these empirical profiles.
In the case of S02, the three-parameter lognormal dis-
tribution was found to provide the best fit.
The control scenario S02 air quality profiles are
available on diskette, contained in a file named
S02CAA.DAT. The same data format described in
Table C-2 is adopted.
No-control scenario sulfur dioxide
profiles
The no-control air quality profiles for S02 are
derived using Equation 1, the same equation used for
CO. For S02, the background concentration was as-
sumed to be zero. Although anthropogenic emissions
contribute only small amounts to total global atmo-
spheric sulfur, measured background concentrations
C-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
for the continental U.S. range from only 0.1 to 1.3
ppb. Background S02 is discussed in more detail in
the supporting document SAI S02, NOx, and CO Re-
port (1994).1
The no-control scenario S02 air quality profiles
are available on diskette, contained in a file named
S02NCAA.DAT. The data format is described in
Table C-2.
Summary differences in sulfur dioxide
air quality
As for CO, reporting differences in control and
no-control scenario air quality projections for each
monitor covered in the analysis is impractical due to
the large amount of data involved. However, Figure
C-2 provides an illustration of scenario differences
similar to the one provided for CO. Specifically, the
graph shows the distribution of 1990 control to no-
control scenario 95th percentile 1-hour average con-
centrations ratios at S02 monitors. By 1990, S02 con-
centrations under the no-control scenario were sub-
stantially higher than those associated with the con-
trol scenario.
Figure C-2. Frequency Distribution of Estimated Ratios
for 1990 Control to No-control Scenario 95th Percentile
1-Hour Average S02 Concentrations, by Monitor.
300
¦a 200
S 100
J L-
0.05 0.25 0.45 0.65 0.85 1.05 1.25
Ratio of CAA:No-CAA 95th Percentile 1-Hour Average
Key caveats and uncertainties for sulfur
dioxide
The height of stacks used to vent flue gases from
utility and industrial fossil fuel-fired boilers has a sig-
nificant effect on the dispersion of sulfur dioxide and
on the formation and long-range transport of second-
ary products such as particulate sulfates. Under a no-
control scenario, it is conceivable that some sources
might have built taller stacks to allow higher emis-
sion rates without creating extremely high ground-
level concentrations of flue gases. On the other hand,
it is also conceivable that, in the absence of post-1970
air pollution control programs, sources might have
built shorter stacks to avoid incurring the higher costs
associated with building and maintaining taller stacks.
To the extent facilities would have adopted different
stack height configurations under a no-control sce-
nario, both local exposures to sulfur dioxides (and
other emissions from fossil fuel combustion) and long-
range transport, deposition, and exposure associated
with secondary formation products may have been
different. However, this analysis assumes that both
the location of individual facilities and the height and
configuration of emission stacks are constant between
the two scenarios. If, in fact, stack heights were raised
under the historical case due to CAA-related concerns,
increases in local S02 concentrations under the
no-control scenario may be overestimated. However,
this same assumption may at the same time lead to
underestimation under the no-control scenario of long-
range transport and formation of secondary particu-
lates associated with taller stacks. For stacks built
lower under a no-control scenario, local S02 expo-
sures would have been higher and long-range effects
lower. Finally, the comments on uncertainties for car-
bon monoxide apply as well to S02.
Nitrogen Oxides
Similarly to sulfur dioxide, emissions of nitro-
gen oxides (NOx) -including nitrogen dioxide (N02)
and nitrous oxide (NO)- lead to several air quality
effects. These effects include secondary formation of
fine particle nitrates, formation of ground-level ozone,
long range transport and deposition of nitric acid, and
localized concentrations of both N02 and NO. The
first three effects are addressed later in this appen-
dix, under the particulate matter, ozone, and acid
deposition sections. The focus of this section is esti-
mation of changes in local concentrations of N02 and
NO.
The methodology applied to estimation of local
nitrogen oxides air quality is essentially identical to
the one applied for carbon monoxide and sulfur diox-
ide. As such, this section does not repeat the "roll-up"
modeling methodological description presented in the
CO section, but instead simply highlights those ele-
1 SAI S02, NOx, and CO Report (1994), page 4-9.
C-6
-------
Appendix C: Air Quality Modeling
24 hour N02 and NO averages were compiled from
AIRS for monitors in the lower 48 states which had at
least 10 percent of their potential samples available.
Applying a cutoff of 0.5 ppb to both N02 and NO to
isolate the zero and near-zero observations, three-pa-
rameter lognormal and gamma distributions were fit-
ted to these empirical profiles. For N02 and NO, the
three-parameter gamma distribution was found to pro-
vide the best fit.
The control scenario N02 and NO air quality pro-
files are available on diskette, contained in files named
N02CAA.DAT andNOCAA.DAT, respectively. The
same data format described in Table C-2 is adopted.
Table C-4. Summary of NO: Monitoring Data
Year
Number of
Monitors
Number of
Counties
Percent
Population
Covered
Number of
Samples
Mean Number
of Samples per
Monitor
1970
45
32
n/a
275,534
6.123
1975
308
155
n/a
1,574,444
5.112
.1980
379
205
45%
1,984,128
5.235
1985
305
182
n/a
2,142,606
7,025
1990
346
187
40%
2,456,922
7.101
Efata Source: SAI SO>, NCI and CO Report (1994).
Table C-5. Summary of NO Monitoring Data.
V Oil r
iNum her of
Monitors
iNum her of
Counties
lVrt'cn t
Population
Covered
iNum her of
Sn in pi cs
Mean iNum ber
of Samples per
Monitor
1970
39
28
n/a
246262
6.314
1975
206
94
n/a
1,101,051
5.345
1980
224
124
30 %
1,023,834
4.571
1985
139
86
n/a
956,425
6.881
1990
145
81
15 %
999,808
6.895
Data Source: SAI S02, NOx and CO Report (1994).
ments of the nitrogen oxides modeling which differ
from carbon monoxide.
Control scenario nitrogen oxides
profiles
After peaking around 1980, the number of N02
and NO monitors, their county coverage, and their
population coverage shrank between 1980 and 1990.
Tables C-4 and C-5 summarize, respectively, the N02
and NO monitoring data used as the basis for devel-
opment of the control scenario air quality profiles.
As for CO and S02, air quality profiles reflecting
average values and maxima for 1, 3, 5, 7, 8, 12, and
C-7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
No-control scenario nitrogen oxides
profiles
Key caveats and uncertainties for
nitrogen oxides
The no-control air quality profiles for N02 and
NO are derived using Equation 1, the same equation
used for CO and S02. As discussed in detail in the
SAI S02, NOx, and CO Report (1994),2 nitrogen ox-
ides are emitted almost entirely from anthropogenic
sources and they do not have long atmospheric resi-
dence times. Therefore, global background concen-
trations are very low, on the order of 0.1 or 0.2 ppb.
For the present analysis, background concentrations
of N02 and NO were assumed to be zero.
The no-control scenario N02 and NO air quality
profiles are available on diskette, contained in files
namedN02NCAA.DAT andNONCAA.DAT, respec-
tively. The data format is described in Table C-2.
Summary differences in nitrogen oxides
air quality
Figure C-3 provides a summary indication of the
differences in control and no-control scenario air qual-
ity for N02. As for CO and S02, the graph shows the
distribution of 1990 control to no-control scenario 95th
percentile 1-hour average concentration ratios atN02
monitors. These ratios indicate that, by 1990, no-con-
trol scenario N02 concentrations were significantly
higher than they were under the control scenario. The
changes for NO are similar to those for N02.
Figure C-3. Frequency Distribution of Estimated Ratios
for 1990 Control to No-control Scenario 95th Percentile
1-Hour Average N02 Concentrations, by Monitor.
300
§ 200
100
J I I I l_
_l I I L-
0.05 0.25 0.45 0.65 0.85 1.05 1.25
Ratio of CAA:No-CAA 95th Percentile 1-Hour Average
A number of caveats and uncertainties specific to
modeling NOx should be noted. First, stack height and
stack height control strategies likely to have influenced
local concentrations of S02 may also have influenced
local concentrations of N02 and NO. (For a fuller dis-
cussion of the stack heights issue, refer to the section
"Key caveats and uncertainties for S02") In addition,
the earlier discussion of uncertainties resulting from
the use of state-level emissions and the cancellation
of uncertainties resulting from analyzing only differ-
ences or relative changes also applies to NOx.
Acid Deposition
The focus of air quality modeling efforts described
above for carbon monoxide, sulfur dioxide, and ni-
trogen oxides was to estimate the change in ambient
concentrations of those pollutants as a result of
changes in emissions. Particularly since the emissions
modeling was driven by modeled macroeconomic
conditions, rather than actual historical economic ac-
tivity patterns, neither the emissions inventories nor
the resultant air quality conditions developed for this
analysis would be expected to match historical out-
comes. The need to focus on relative changes, rather
than absolute predictions, becomes even more acute
for estimating air quality outcomes for pollutants sub-
ject to long-range transport, chemical transformation,
and atmospheric deposition. The complexity of the
relationships between emissions, air concentrations,
and deposition is well-described in the following para-
graph from the RADM report document developed
by Robin Dennis of US EPA's National Exposure
Research Laboratory in support of the present analy-
sis:
"Sulfur, nitrogen, and oxidant species in the
atmosphere can be transported hundreds to
thousands of kilometers by meteorological
forces. During transport the primary
emissions, SO NO and volatile organic
emissions (VOC) are oxidized in the air or in
cloud-water to form new, secondary
compounds, which are acidic, particularly
sulfate and nitric acid, or which add to or
subtract from the ambient levels of oxidants,
such as ozone. The oxidizers, such as the
hydroxyl radical, hydrogen peroxide and
' SAI S02, NOx, and CO Report (1994), page 4-9.
C-8
-------
Appendix C: Air Quality Modeling
ozone are produced by reactions ofVOC and
NOx. The sulfur and nitrogen pollutants are
deposited to the earth through either wet or
dry deposition creating a load of pollutants
to the earth's surface... However, the
atmosphere is partly cleansed of oxidants
through a number of physical processes
including deposition (e.g., ozone is removed
by wet and dry deposition). Dry deposition
occurs when particles settle out of the air onto
the earth or when gaseous or fine particle
species directly impact land, plants, or water
or when plant stomata take up gaseous
species, such as S0T In wet deposition,
pollutants are removed from the atmosphere
by either rain or snow. In addition, fine
particles or secondary aerosols formed by the
gas- and aqueous-phase transformation
processes scatter or absorb visible light and
thus contribute to impairment of visibility. "3
Control scenario acid deposition
profiles
The derivation of control scenario emission in-
ventory inputs to the RADM model is succinctly de-
scribed in this excerpt from the RADM Report (1995):
The RADM model requires a very detailed
emissions inventory in both time and space.
The emissions fields are also day-specific to
account for the temperature effects on the
volatile organics and the wind and
temperature effects on the plume rise of the
major point sources. At the time of the 812
retrospective study RADM runs, these
inventories had been developed for 1985,
using the 1985 NAPAP (National Acid
Precipitation Assessment Program)
inventory, and adjusted for point source
The complexity and nonlinearity
of the relationships between localized
emissions of precursors, such as S02
and VOC, and subsequent regional
scale air quality and deposition effects
are so substantial that the simple "roll-
up" modeling methodology used for
estimating local ambient concentra-
tions of SO,, NO . and CO is inad-
T x7
equate, even for a broad-scale, aggre-
gate assessment such as the present
study. For sulfur deposition, and for
a number of other effects addressed
in subsequent sections of this appen-
dix, a regional air quality model was
required. After careful review of the
capabilities, geographic coverage,
computing intensity, and resource re-
quirements associated with available
regional air quality models, EPA de-
cided to use various forms of the Re-
gional Acid Deposition Model
(RADM) to estimate these effects.4
Figure C-4 shows the geographic do-
main of the RADM.
Figure C-4. Location of the High Resolution RADM 20-km Grid Nested
Inside the 80-km RADM Domain.
3 Dennis, R. RADM Report (1995), p. 1.
4 For a detailed description of the various forms of the RADM and its evaluation history, see the Dennis, R. RADM Report (1995).
C-9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
emissions to 1988 for the Eulerian Model
Evaluation Field Study funded by NAPAP.
These RADM emissions inventories had
county-level and detailed SCC and species-
level information incorporated into them to
provide the 80- and 20-km detail. The 812
Study emissions are principally computed at
the state level. While the 1985 812 Study
emissions are close to the NAPAP inventory,
they do not exactly match, nor do they have
the spatial, nor economic sector, nor species
detail within a state needed to run RADM. To
connect the 812 Study emissions to the RADM
emissions, the following approach was
followed: An industry/commercial-level
disaggregation (including mobile sources)
was developedfor the 812 emissions to allow
different sectors in a state to change their
emissions across time without being in lock
step and the detailed NAPAP emissions for
every 80- and 20-km RADM grid-cell were
grouped by state to the same level of industry/
commercial aggregation for an exact
correspondence. Then it was assumed that the
812 Study 1985 control emissions were
effectively the same as the 1985 NAPAP
emissions. Relative changes in emissions
between the 812 1985 control and any other
scenario (e.g., 1985 no-control, or 1990
control, or 1980 no-control, etc.) were then
applied to the 1985 NAPAP state-level
industry/commercial groups in the
appropriate 80- and 20-km grid cells. Thus,
state-level emissions for each group would
retain the same state-level geographic pattern
in the different scenarios years, but the mix
across groups could change with time. In this
way, the more detailed emissions required by
RADM were modeled for each scenario year
using the 812 Study emissions data sets.5
Although the focus of the present analysis is to
estimate the differences between the control and no-
control scenarios, it is useful to illustrate the abso-
lute levels of acid deposition associated with the two
scenarios. It is particularly important to demonstrate
the initial deposition conditions to preclude possible
misinterpretations of the maps showing percent
change in deposition. A relatively high percentage
change in a particular region, for example, may oc-
cur when initial deposition is low, even when the
change in deposition is also modest. The RADM-
5 Dennis, R. RADM Report (1995).
Figure C-5. RADM-Predicted 1990 Total Sulfur Deposition
(Wet + Dry; in kg/ha) Under the Control Scenario.
iBSls
f
gipiiS
fc/
Wr
BllliM
:
t. *
LEGEND:
0-5
10
0 15
¦ i
115 - 20
w
¦ > 20
Figure C-6. RADM-Predicted 1990 Total Nitrogen Deposi-
tion (Wet + Dry; in kg/ha) Under the Control Scenario.
C-10
-------
Appendix C: Air Quality Modeling
Figure C-7. RADM-Predicted 1990 Total Sulfur Deposition
(Wet + Dry; in kg/ha) Under the No-control Scenario.
iHIHiS
| (
¦gyHH
LEGEND!
D - 5
5-10
ID - 15
i
115 - 20
w
1 > 20
Figure C-8. RADM-Predicted 1990 Total Nitrogen Deposi-
tion (Wet + Dry; in kg/ha) Under the No-control Scenario.
modeled 1990 control scenario wet and dry sulfur
deposition pattern is shown in Figure C-5. A com-
parable map for nitrogen deposition is presented in
Figure C-6. Maps of the RADM-predicted 1990 no-
control scenario sulfur and nitrogen deposition are
presented in Figures C-7 and C-8, respectively.
No-control scenario acid deposition
profiles
Configuration of the RADM model for the
present analysis —including allocation of emission
inventories to model grid cells, design of meteoro-
logical cases, treatment of biogenic versus anthro-
pogenic emissions, and temporal, spatial, and spe-
cies allocation of emissions— are described in de-
tail in the RADM Report (1995). The remainder of
this section provides a summary description of the
acid deposition modeling effort.
For sulfur deposition, the RADM Engineering
Model (RADM/EM), which focuses on sulfur com-
pounds, was used to derive annual average total (wet
plus dry) deposition of sulfur in kilograms sulfur
per hectare (kg-S/ha) under both the control and
no-control scenarios. The relative changes in an-
nual average total sulfur deposition for each of the
80-km RADM/EM grid cells for 1975, 1980, 1985,
and 1990 were then compiled.
Nitrogen deposition was calculated in a differ-
ent manner. Since nitrogen effects are not included
in the computationally fast RADM/EM, nitrogen
deposition had to be derived from the full-scale,
15-layer RADMruns. Because of the cost and com-
putational intensity of the 15-layer RADM, nitro-
gen deposition estimates were only developed for
1980 and 1990. As for sulfur deposition, the rela-
tive changes in annual average total (wet plus dry)
nitrogen deposition, expressed as kg-N/ha, were cal-
culated for each 80-km grid cell and for each of the
two scenarios. It is important to note that ammonia
depositin contributes significantly to total nitrogen
deposition. However, the activities of sources as-
sociated with formation and deposition of ammo-
nia, such as livestock farming and wildlife, were
essentially unaffected by Clean Air Act-related con-
trol programs during the 1970 to 1990 period of
this analysis. Therefore, ammonia deposition is held
constant between the two scenarios.
C-ll
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Summary differences in acid
deposition
Figure C-9 is a contour map showing the esti-
mated percent increase in sulfur deposition under
the no-control scenario relative to the control sce-
nario for 1990. Figure C-10 provides comparable
information for nitrogen deposition. These maps
indicate that by 1990 acid deposition would have
been significantly higher across the RADM domain
under the no-control scenario.
Examination of the percent change sulfur depo-
sition map indicates relatively large percentage
changes in the upper Great Lakes and the Florida-
Southeast Atlantic Coast areas. This result may ap-
pear somewhat surprising to readers familiar with
the historical patterns of acid deposition. However,
a review of the emission data and the control sce-
nario sulfur deposition map reveal the reasons for
this result.
First, Figure C-5 shows that control scenario
deposition rates are relatively low. As described
above, even a small absolute increase in deposition
leads to a large percentage increase in areas with
low initial rates of deposition. Second, the scenario
differences in SO emission rates for these areas
x
were substantial. For example, 1990 no-control sce-
nario total SOx emissions for Michigan were ap-
proximately 1.8 million tons but control scenario
emissions for the same year were less than 600,000
tons; a reduction of over two-thirds. Similarly, 1990
no-control scenario emissions for Florida were over
2.3 million tons, compared to approximately
800,000 tons under the control scenario; also a re-
duction of about two-thirds. Almost 1 million tons
of the Michigan reduction and approximately 1.3
million tons of the Florida reduction were associ-
ated with utilities. Emission reductions of these
magnitudes would be expected to yield significant
reductions in rates of acid deposition.
Key caveats and uncertainties for acid
deposition
Regional-scale oxidant and deposition model-
ing involves substantial uncertainty. This uncer-
tainty arises from uncertainties in modeling atmo-
spheric chemistry, incomplete meteorological data,
normal seasonal and temporal fluctuations in atmo-
spheric conditions, temporal and spatial variability
Figure C-9. RADM-Predicted Percent Increase in Total
Sulfur Deposition (Wet + Dry; in kg/ha) Under the No-
control Scenario.
—
LE.GEND-
0-25
25 - 30
30 - 35
135 - 40
I > 40
Figure C-10. RADM-Predicted Percent Increase in Total
Nitrogen Deposition (Wet + Dry; in kg/ha) Under the No-
control Scenario.
LEGEND
0-10
10 - 15
15 - 20
120-25
> 25
C-12
-------
Appendix C: Air Quality Modeling
in emissions, and many other factors. Uncertainties
specific to the RADM model, and this particular ex-
ercise, are discussed in detail in the RADM Report
(1995). It is important, however, to highlight some of
the potential sources of modeling uncertainty unique
to this analysis.
The first source of uncertainty specific to this
analysis is associated with the spatial and geographic
disaggregation of emissions data. As discussed in the
RADM Report, the RADM model requires emission
inventory inputs which are highly disaggregated over
both time and space. The ideal emissions inventory
fed into the RADM model includes day-specific emis-
sions to account for temperature effects on VOCs and
the significance of localized meteorological conditions
around major point sources. Given the broad-scale,
comprehensive nature of the present study, such de-
tailed emissions inventories were not available. How-
ever, the industry/commercial-level disaggregation ap-
proach developed for the present analysis would not
be expected to introduce any systematic bias, and the
contribution of this disaggregation of emissions would
not be expected to contribute significantly to the over-
all uncertainty of the larger analysis.
The acid deposition estimates included in the
present analysis are limited in that only the eastern 31
of the 48 coterminous states are covered. Although
acid deposition is a problem primarily for the eastern
U.S., acid deposition does occur in states west of the
RADM domain. The magnitude of the benefits of re-
ducing acid deposition in these western states is likely
to be small, however, relative to the overall benefits
of the historical Clean Air Act.
Particulate Matter
Developing air quality profiles for particulate
matter is significantly complicated by the fact that
"particulate matter" is actually an aggregation of dif-
ferent pollutants with varying chemical and aerody-
namic properties. Particulate species include chemi-
cally inert substances, such as wind-blown sand, as
well as toxic substances such as acid aerosols; and
include coarse particles implicated in household soil-
ing as well as fine particles which contribute to hu-
man respiratory effects. In addition, emissions of both
primary particulate matter and precursors of second-
arily-formed particulates are generated by a wide va-
riety of mobile and stationary sources, further com-
plicating specification of particulate air quality mod-
els. Finally, particulate air quality models must take
account of potentially significant background concen-
trations of atmospheric particles.
Modeling multiple species and emission sources,
however, is not the only major challenge related to
particulate matter which is faced in the present study.
Over the 1970 to 1990 period being analyzed, under-
standing of the relative significance of fine versus
coarse particles evolved significantly. Up until the
mid-1980s, particulate air quality data were collected
as Total Suspended Particulates (TSP). However, dur-
ing the 1980s, health scientists concluded that small,
respirable particles, particularly those with an aero-
dynamic diameter of less than or equal to 10 microns
(PM10), were the component of particulate matter pri-
marily responsible for adverse human health effects.
As of 1987, federal health-based ambient air quality
standards for particulate matter were revised to be ex-
pressed in terms of PMJ0 rather than TSP. Starting in
the mid-1980s, therefore, the U.S. began shifting away
from TSP monitors toward PM|( monitors. As a re-
sult, neither TSP nor PM|u are fully represented by
historical air quality data over the 1970 to 1990 pe-
riod of this analysis. Furthermore, a large number of
U.S. counties have no historical PM monitoring data
at all, making it difficult to estimate changes in ambi-
ent concentrations of this significant pollutant for ar-
eas containing roughly 30 percent of the U.S. popula-
tion.
Given the relative significance of particulate mat-
ter to the bottom-line estimate of net benefits of the
historical Clean Air Act, it was important to develop
methodologies to meet each of these challenges. The
methodologies developed and data used are described
primarily in the two supporting documents SAI PM
Report (1992) and SAI PM Report (1995).6 To sum-
marize the overall approach, historical TSP data were
broken down into principal component species, in-
cluding primary particulates, sulfates, nitrates, organic
particulates, and background particulates. Historical
data were used for the control scenario. To derive the
no-control profiles, the four non-background compo-
nents were scaled up based on corresponding
no-control to control scenario ratios of emissions and /
or modeled atmospheric concentrations. Specifically,
the primary particulate component was scaled up by
the ratio of no-control to control emissions of PM.
6 In addition, SAI memoranda and reports which supplement the results and methodologies used in this analysis are included in
the references.
C-13
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Organic constituents were scaled up by the ratio of
no-control to control VOC emissions. In the eastern
31 states where RADM sulfate and nitrate data were
available, values for S04 and N03 from an appropri-
ate RADM grid cell were assigned to the relevant
county and used to scale these components of PM.
For the western states not covered by RADM, sul-
fates were scaled up by the change in S02 emissions
and nitrates were scaled up the change in NOx emis-
sions. No-control scenario profiles were then con-
structed by adding these scaled components to back-
ground concentrations.
To resolve the problem of variable records of TSP
and PMJ0 data, both TSP and PM10 profiles were gen-
erated for the entire 20 year period. Missing early year
data for PM10 were derived by applying region-spe-
cific, land use category-specific PM10to TSP ratios to
the historical TSP data. Missing recent year TSP data
were derived for those areas where PM10 monitors
replaced TSP monitors by applying the reciprocal of
the relevant PM10 to TSP ratio. The methodology is
described in detail in the SAI PM Report (1995).
In addition, to increase the geographic coverage
of estimates of air quality, an interpolation methodol-
ogy7 was developed to predict air quality for the con-
trol scenario in counties without measured data. PM
concentrations were estimated by first estimating the
components of PM (i.e., sulfate, nitrate, and organic
particulate, and primary particulate). The methodol-
ogy for developing the concentrations of components
within a county differed depending upon whether the
county was within or outside the RADM domain.
ent TSP and PM10 to describe these constituents in
counties without data. Control scenario PM profiles
were developed by adding the RADM-estimated sul-
fate particulate levels to the statewide average nitrate,
VOC, and primary particulate levels, and background.
For counties outside the RADM domain, an al-
ternate procedure was used. Using the primary and
secondary particulate estimates for counties with data,
statewide average sulfate, nitrate, VOC, and primary
particulate concentrations were determined. Control
scenario PM10 was predicted by adding the statewide
averages of all primary and secondary particulate, and
background. Using this method, all counties that did
not have monitors and are in the same state are as-
signed the same PM concentration profiles. These in-
terpolated results are clearly less certain than results
based on actual historical monitoring data and are
therefore presented separately.
Control scenario particulate matter
profiles
The number of TSP monitors peaked in 1977 and
declined throughout the 1980s. Table C-6 summarizes
the daily (i.e., 24-hour average) TSP monitoring data
used as the basis for development of the control sce-
nario air quality profiles. Most of the TSP and PM10
monitors collected samples every six days (i.e., 61
samples per year).
Daily PM10 data were also collected for each year
between 1983 and 1990. Table C-7 summarizes the
daily PM10 monitoring data used for the control sce-
nario air quality profiles.
For those counties within the
RADM domain, the RADM modeled
concentrations for 1980 and 1990 were
used to predict sulfate air quality. Re-
lationships based on linear regressions
that related 1980 and 1990 RADM sul-
fate concentrations to estimated sulfate
particulate concentrations were calcu-
lated for counties with AIRS data. Sul-
fate particulate concentrations were
then calculated for all counties in the
domain by applying the regression re-
sults to the RADM grid cell concen-
tration located over the county center.
Statewide average nitrate, VOC, and
primary particulate concentrations
were calculated from measured ambi-
Table C-6. Summary of TSP Monitoring Data
Year
Number of
Monitors
Number of
Counties
Number of
Samples
Mean Number
of Samples per
Monitor
1970
751
245
56,804
76
1975
3,467
1,146
221,873
64
1980
3,595
1,178
234,503
65
1985
2,932
1,018
189,344
65
1990
923
410
59,184
64
Data Source: SAI PM Report (1995).
7 The interpolation methodology is described in detail in SAI, 1996. Memo from J. Langstaff to J. DeMocker. PM Interpolation
Methodology for the section 812 retrospective analysis. March 1996.
CM4
-------
Appendix C: Air Quality Modeling
Tabic C-7. Summary of PMm Monitoring Data.
V oil i'
iNum her of
Monitors
iNum lier of
Counties
iNum her of
Sil 111 pi OS
Me an iNum her
of Samples per
Monitor
1985
303
194
22,031
73
1990
1,249
556
98,904
79
allow differentiation between urban and
rural locations for coarser particles.
The TSP and PMJ0 control scenario
profiles developed based on this meth-
odology are available on diskette, un-
der the filenames listed in Table C-10.
No-control scenario
particulate matter profiles
Data Source: SAI PM Report (1995).
Further speciation of TSP and PM10 air quality
data serves two purposes in the present analysis. First,
speciation of TSP into PM10 and other fractions al-
lows derivation of PM10:TSP ratios. Such ratios can
then be used to estimate historical PM10 for those years
To derive the no-control TSP and
PM10 air quality profiles, individual
component species were adjusted to
reflect the relative change in emissions or, in the case
of sulfates and nitrates in the eastern U.S., the rela-
tive change in modeled ambient concentration. The
following excerpt from the SAI PM Report (1995)
and monitors which had TSP data but no PM,n data. describes the specific algorithm used:1
The reciprocal ratio is also applied in this analysis to
expand 1985 and 1990 TSP data to cover those areas
which monitored PM10 but not TSP. The second pur-
pose served by speciation of particulate data is, as
described earlier, to provide a basis for scaling up
concentrations of each species to derive no-control
scenario TSP and PM10 profiles.
To break the TSP and PM10 data down into com-
ponent species, speciation factors were applied to the
PM fractions with aerodynamic diameters below 2.5
microns (PM25) and from 2.5 to 10 microns (PM10).
The PM2 5 speciation factors were drawn from a Na-
tional Acid Precipitation Assessment Program
(NAPAP) report on visibility which reviewed and
consolidated speciation data from a number of stud-
ies.8 These factors are presented in Table C-8. In the
table, fine particle concentrations are based on par-
ticle mass measured after equilibrating to a relative
humidity of 40 to 50 percent; and organics include
fine organic carbon.
To develop speciation factors for coarser particles
(i.e., in the PM25 to PM10 range), SAI performed a
review of the available literature, including Conner et
al. (1991), Wolff and Korsog (1989), Lewis and
Macias (1980), Wolff etal. (1983), Wolff etal. (1991),
and Chow et al. (1994).9 These speciation factors are
summarized in Table C-9. Data were too limited to
"For the retrospective analysis, the no-CAA
scenario TSP and PMW air quality was
estimated by means of the following
algorithm:
• Apportion CAA scenario TSP and PMI0
to size categories and species;
• Adjust for background concentrations;
• Use a linear scaling to adjust the non-
background portions of primary
particulates, sulfate, nitrate, and organic
components based on emissions ratios of
PM, S02, NOx and VOC, and Regional
Acid Deposition Model (RADM) annual
aggregation results for S04 and NO};
• Add up the scaled components to estimate
the no-CAA scenario TSP and PMJ0
concentrations. "
The specific procedures and values used for the
linear rollback, speciation, fine to coarse particle ra-
tio, scaling, and background adjustment steps are de-
scribed in detail in the SAI PM report (1995).11 Table
C-l 1 lists the names of the electronic data files con-
taining the TSP and PM10 profiles for the no-control
scenario.
8 J. Trijonis, "Visibility: Existing and Historical Conditions—Causes and Effects," NAPAP Report 24, 1990.
9 This literature review, and complete citations of the underlying studies, are presented in the SAI PM Report (1995), pp. 4-2 to
4-6 and pp. R-l to R-2, respectively.
10 SAI PM Report (1995), p. 5-1.
11 SAI PM Report (1995), pp. 5-2 to 5-15.
C-15
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic C-8. Fine Particle (PM25) Chemical Composition by U.S. Region.
Com pone lit
Units
iNuni her of
Data Sets
AH tli me tic
Mean
Range of
Values
RURAL EAST
Fine particle concentration
Hg/ttl3
19
18
6-46
Ammonium sulfate
% Fine particles
19
52
41 -66
Ammonium nitrate
% Fine particles
3
1
1
Organics
% Fine particles
5
24
9-34
URBANEAST
Fine particle concentration
Hg/m3
3
36
29-43
Ammonium sulfate
% Fine particles
3
55
53-57
Ammonium nitrate
% Fine particles
2
1
1
Organics
% Fine particles
2
24
15-32
RURAL WES'T
Fine particle concentration
Hg/m3
25
5
1 -11
Ammonium sulfate
% Fine particles
25
35
15-56
Ammonium nitrate
% Fine particles
17
4
1 -17
Organics
% Fine particles
25
27
14-41
URBAN WEST
Fine particle concentration
|ig/1113
16
35
13-74
Ammonium sulfate
% Fine particles
16
16
3-35
Ammonium nitrate
% Fine particles
14
15
2-37
Organics
% Fine particles
16
42
25-79
Data Sources: SAI PM Report(1995); and J. Trijonis, "Visibility: Existing and Historical Conditions—Causes and
Effects, "NAPAP Report24, 1990.
Summary differences in particulate
matter air quality
Figure C-ll provides one indication of the esti-
mated change in particulate matter air quality between
the control and no-control scenarios. Specifically, the
graph provides data on the estimated ratios of 1990
control to no-control scenario annual mean TSP con-
centrations in monitored counties. The X-axis values
represent the mid-point of the ratio interval bin, and
the Y-axis provides the number of counties falling into
each bin. Figure C-ll indicates that annual average
TSP concentrations would have been substantially
higher in monitored counties under the no-control sce-
nario.
Key caveats and uncertainties for
particulate matter
There are several important caveats and uncer-
tainties associated with the TSP and PM10 air quality
profiles developed for this study. Although further
C-16
-------
Appendix C: Air Quality Modeling
Tabic C-9. Coarse Particle (PIVfc..s to PMio) Chemical Composition by U.S. Region.
Colli poiicnt
I in its
iNuni her
of Data
Sets
Arithmetic
Me si it
Kiin^e of
Values
EAST
Coarse particle concentration
(Xg/rnj
1
5.5
5.5
Ammonium sulfate
% Coarse particles
3
3
1 -4
Ammonium nitrate
% Coarse particles
1
4
4
Organics
% Coarse particles
2
10
7 -13.8
WEST
Coarse particle concentration
Hg/ni3
18
24
7.7-56.7
Ammonium sulfate
% Coarse particles
18
6
2.1 -10.39
Ammonium nitrate
% Coarse particles
18
18
2.33-28.52
Organics
% Coarse particles
18
14
8.41-25.81
Data Source: SAIPM Report (1995)
Tabic C-10. PM Control Scenario Air Quality Profile Filenames.
(,'oni pone lit
Indicator
Mleiiiinie
TSP
Annual Mean
TSPCMEAN.DAT
TSP
2nd Highest Daily
TSPCHI2.DAT
TSP
(X)th Percentile
TSPC(X).DAT
PMnffl
Annual Mean
PM10CMEA.DAT
I'M'
2nd Highest Daily
PM10CHI2.DAT
PMi®
(X)th Percentile
PM10C(X).DAT
Note: "(X)" refers to percentiles from 5 to 95, indicating 19 percentile data files available
forTSPand 19 files available for PM10; for example, the filename for the 50th percentile
TSP air quality data profile for the control scenario is named TSPC50.DAT.
C-17
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic C-11. PM No-Control Scenario Air Quality Profile
Filenames.
Colli polio lit
Indicsi tor
M1 oil n in e
TSP
Annual Mean
TSPCNMEA.DAT
TSP
2nd Highest Daily
TSPNCHI.DAT
TSP
(X)th Percentile
TSPNC(X).DAT
PM i
Annual Mean
PM10NCME.DAT
PMi»
2nd Highest Daily
PM10NCHI.DAT
I'M..
(X)th Percentile
PM10NC(X).DAT
Note: "(X)" refers to percentiles from 5 to 95, indicating 19 peicentile-based data files
available for TSP and 19 similar files available for PMio; for example, the filenam efor the
50th percentile TSP air quality data profile for the no-control scenario is named
TSPNC50.DAT.
reductions in these uncertainties were not possible for
this study given time and resource limitations, the rela-
tive importance of particulate matter reduction con-
tributions towards total benefits of the Clean Air Act
highlights the importance of these uncertainties.
A number of uncertainties were introduced in the
process of speciating and rolling up individual com-
ponents of particulate matter. First, temporal and spa-
tial variability in the size and chemical properties of
particulate emissions are substantial. These charac-
teristics change from day to day at any given loca-
tion. Second, using changes in proxy pollutant emis-
Figure C-ll. Distribution of Estimated Ratios for 1990
Control to No-control Annual Mean TSP Concentra-
tions, by Monitored County.
50
40
C
g 30
o
10
0.00 0.20 0.40 0.60 0.80 1.00
Ratio of CAA:No-CAA Annual Mean TSP (interval midpoint)
sions, such as using S02 as a surrogate for S04
in the western states, to roll up individual PM
components may introduce significant uncer-
tainty. Third, even assuming a satisfactorily
high degree of correlation between target and
surrogate pollutants, relying on predicted
changes in emissions at the state level further
compounds the uncertainty. Finally, and per-
haps most important, using PM10 to TSP ratios
derived from late 1980s monitoring data may
lead to significant underestimation of reduc-
tions in fine particulates achieved in earlier
years. This is because historical Clean Air Act
programs focused extensively on controlling
combustion sources of fine particulates. As a
result, the share of TSP represented by PM10
observed in the late 1980s would be lower due
to implementation of controls on combustion
sources. This would lead, in turn, to underesti-
mation of baseline PM10 concentrations, as a
share of TSP, in the 1970s and early 1980s. Ifbaseline
PM10 concentrations in these early years are underes-
timated, the reductions in PM10 estimated by linear
scaling would also be underestimated.12
Ozone
Nonlinear formation processes, long-range atmo-
spheric transport, multiple precursors, complex atmo-
spheric chemistry, and acute sensitivity to meteoro-
logical conditions combine to pose substantial diffi-
culties in estimating air quality profiles for ozone.
Even in the context of an aggregated, national study
such as this, the location-specific factors controlling
ozone formation preclude the use of roll-up modeling
based on proxy pollutants or application of state-wide
or nation-wide average conditions. Such simplifica-
tions would yield virtually meaningless results for
ozone.
Ideally, large-scale photochemical grid models —
such as the Urban Airshed Model (UAM)— would
be used to develop control and no-control scenario
estimates for ozone concentrations in rural and urban
areas. Such models provide better representations of
the effects of several important factors influencing air
quality projections such as long-range atmospheric
transport of ozone. However, the substantial comput-
ing time and data input requirements for such models
precluded their use for this study.13 Instead, three sepa-
12 See SAI PM Report (1995), p. 5-9.
13 For a description of the extensive data inputs required to operate UAM, see SAI Ozone Report (1995), p. 1-1.
C-18
-------
Appendix C: Air Quality Modeling
Tabic C-12. Urban Areas Modeled with OZIPM4.
Albany, NY
Fort Wayne, IN
Owensboro, KY
Albuquerque, NM
Grand Rapids, MI
Paikersburg, WV
A1 lentown, PA-NJ
Greeley, CO
Pascagoula, MS
Altoona, PA
Green Bay, WI
Pensacola, FL
Anderson, IN
Greensboro, NC
Peoria, IL
Appleton, WI
Greenville, SC
Philadelphia, PA
Asheville, NC
Harrisburg, PA
Phoenix, AZ
Atlanta, GA
Hartford, CT
Portland, OR
Atlantic City, NJ
Houston, TX
Portsmouth, NH
Auburn, M E
Huntington, WV-KY
Raleigh, NC
Augusta, GA-SC
Huntsville, AL
Reading, PA
Austin, TX
Indianapolis, IN
Reno, NV
Baltimore, MD
Iowa City, IA
Richmond, VA
Baton Rouge, LA
Jackson, MS
Roanoke, VA
Beaumont, TX
Jacksonville, FL
Rochester, NY
Bellingham, WA
Janesville Rock Co, WI
Rockford, IL
Billings, MT
Johnson City, TN-VA
Sacramento, CA
Birmingham, AL
Johnstown, PA
Salt Lake City, UT
Boston, MA
Kansas City, MO
San Antonio, TX
Boulder, CO
Knoxville, TN
San Diego, CA
Canton, OH
Lafayette, IN
San Francisco, CA
Cedar Rapids, IA
Lafayette, LA
San Joaquin Valley, CA
Champaign, IL
Lake Charles, LA
Santa Barbara, CA
Charleston, SC
Lancaster, PA
Sarasota, FL
Charleston, WV
Lansing MI
Scranton, PA
Charlotte, NC
Las Cruces, NM
Seattle, WA
Chattanooga, TN-GA
Las Vegas, NV
Sheboygan, WI
Chicago, IL
Lexington, KY
Shreveport, LA
Cincinnati, OH
Lima, OH
South Bend, IN
Cleveland, OH
Little Rock AR
Springfield, IL
Colorado Springs, CO
Longview, TX
Springfield, MO
Columbia, SC
Los Angeles, CA
Springfield, OH
Columbus, GA-AL
Louisville, KY
St Louis, MO
Columbus, OH
Lynchburg, VA
Steubenville, OH-WV
Corpus Christi, TX
Medford, OR
Syracuse, NY
Cumberland, MD-WV
Memphis, TN
Tallahassee, FL
Dallas, TX
Miami, FL
Tampa, FL
Davenport, IA-IL
Minneapolis, MN-WI
Terre Haute, IN
Decatur, IL
Mobile, AL
Toledo, OH
Denver, CO
Monroe, LA
T ucson, AZ
Detroit, MI
Montgomery, AL
Tulsa, OK
El Paso, TX
Nashville, TN
Utica-Rome, NY
Erie, PA
New Orleans, LA
Ventura County, CA
Eugene, OR
New York NY
Victoria, TX
Evansville, IN
Norfolk, VA
Washington, DC
Fayetteville, NC
Oklahoma City, OK
Wheeling, WV-OH
Flint, MI
Omaha, NE-IA
Wichita, KS
Fort Collins, CO
Orange Co, CA
York, PA
Fort Smith, AR-OK
Orlando, FL
Youngstown, OH-PA
rate modeling efforts were conducted to provide ur-
ban and rural ozone profiles for those areas of the lower
48 states in which historical ozone changes attribut-
able to the Clean Air Act may be most significant.
First, for urban areas the Ozone Isopleth Plotting
with Optional Mechanisms-IV (OZIPM4) model was
14 See SAI Ozone Report (1995), p. 1-1.
run for 147 urban areas. Table C-12 lists the urban
areas modeled with OZIPM4. Although it requires
substantially less input data than UAM, the OZIPM4
model provides reasonable evaluations of the relative
reactivity of ozone precursors and ozone formation
mechanisms associated with urban air masses.14 Three
to five meteorological episodes were modeled for each
C-19
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
of the 147 urban areas; and for each of these, four
model runs were performed to simulate the 1980 and
1990 control and no-control scenarios. The outputs of
these model runs were peak ozone concentrations for
each of the target year-scenario combinations. The
differentials between the control and no-control sce-
nario outputs were averaged over meteorological epi-
sodes and then applied to scale up historical air qual-
ity at individual monitors to obtain no-control case
profiles. As for the other pollutants, the control sce-
nario profiles were derived by fitting statistical distri-
butions to actual historical data for individual moni-
tors.
Second, the 15-layer RADM runs for 1980 and
1990 were used to estimate the relative change in ru-
ral ozone distributions for the eastern 31 states. In ad-
dition, a limited number of 20-km grid cell high-reso-
lution RADM runs were conducted to benchmark the
15-layer, 80-km RADM median ozone response and
to estimate high ozone response. The relative changes
in modeled median and 90th percentile rural ozone
were then assumed to be proportional to the changes
in, respectively, the median and 90th percentile ozone
concentrations. The domain of the high-resolution
RADM is shown in Figure C-4 and the general RADM
domain is shown in Figure C-12.
Finally, the SARMAP Air Quality Model
(SAQM) was run for EPA by the California Air Re-
sources Board (CARB) to gauge the differences in
peak ozone concentrations in key California agricul-
tural areas for 1980 and 1990. No-control profiles were
developed for ozone monitors in these areas by as-
suming the relative change in peak ozone concentra-
tion also applies to the median of the ozone distribu-
tion. The domain of the SAQM is shown in Figure C-
12.
Figure C-12. RADM and SAQM Modeling Domains, with Rural Ozone Monitor Locations.
C-20
-------
Appendix C: Air Quality Modeling
Control scenario ozone profiles
For ozone, air quality profiles were developed
from historical AIRS data and calculated for individual
monitors based on 1,2, 6, 12, and 24 hour averaging
times. Profiles based on the daily maximum concen-
trations for these averaging times were also calculated.
Given the significance of seasonal and diurnal ozone
formation, twelve separate profiles of hourly ozone
distributions were also developed for six 2-month
periods and for daytime and nighttime hours. The
2-month periods are January-February, March-April,
and so forth. The diurnal/nocturnal profiles are divided
at 7 A.M. and 7 P.M. Local Standard Time. All of
these profiles are based on constructing 1, 2, 6, 12,
and 24-hour moving average profiles from the hourly
ozone data from each monitor.15 A two-parameter
gamma distribution is then fitted to characterize each
of these air quality profiles.16 The functional form of
the gamma distribution, the basis for deriving the
monitor-specific values for mean and variance, and
an analysis of the goodness of fit to the data are pre-
sented in the SAI Ozone Report (1995).
Table C-13 summarizes the ozone monitoring data
used as the basis for the control scenario profiles. The
distribution of these monitors among urban, subur-
Table C-13. Summary of Ozone Monitoring
Data.
V oil i'
iNum hor of
Monitors
iNum bcr of
Counties
1970
1
1
1975
467
240
1980
791
415
1985
719
415
1990
834
477
Data Source: SAI Ozone Report (1995).
ban, and rural locations is presented in Table C-2 of
the SAI Ozone Report (1995).
Given the substantial number of alternative air
quality profiles for ozone, approximately 20 high-den-
sity disks are required to hold the profiles, even in
compressed data format. Resource limitations there-
fore preclude general distribution of the actual pro-
files. As discussed in the caveats and uncertainties
subsection below, however, the substantial uncertain-
ties associated with model results for any given area
preclude application of these profiles in contexts other
than broad-scale, aggregated assessments such as the
present study. The historical ozone monitoring data
used as the basis for this study are, nevertheless, avail-
able through EPA's Aerometric Information Retrieval
System (AIRS).
No-control scenario ozone profiles
The specific modeling methodologies for the
OZIPM4 runs —including emissions processing, de-
velopment of initial and boundary conditions, meteo-
rological conditions, simulation start and end times,
organic reactivity, and carbon fractions— are de-
scribed in detail in the SAI Ozone Report (1995).
Assumptions and modeling procedures not otherwise
described in the SAI report were conducted in accor-
dance with standard EPA guidance.17
Similarly, the RADM modeling methodology
used to estimate changes in day-time rural ozone dis-
tributions in the eastern 31 states are described in de-
tail in the RADM Report (1995). The referenced re-
port also provides complete citations of the literature
associated with development, standard application
procedures, and evaluation of RADM by the National
Acid Precipitation Assessment Program (NAPAP).
To derive the no-control scenario results for key
California agricultural areas, the California Air Re-
sources Board and US EPA's Region 9 office agreed
to conduct three runs of the SAQM. For the 1990 con-
trol scenario, the 1990 SARMAP base case scenario
adopted for California State Implementation Plan
modeling was adopted.18 Derivation of 1990
15 For the nighttime profiles, only 1, 2, 6, and 12-hour averaged concentrations are derived.
16 Normal and lognormal distributions were also developed and tested for goodness of fit; however, the gamma distribution provided
a better representation of the concentration distribution. See SAI Ozone Report (1995), page 4-2.
17 US EPA, Office of Air Quality Planning and Standards, "Procedures for Applying City-Specific EKMA," EPA-450/4-89-012, 1989.
18 Documentation of the SARMAP Air Quality Model and the SARMAP 1990 base case can be found in the SAQM references listed
at the end of this appendix.
C-21
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
no-control and 1980 control and no-control scenarios
was based on adjusting the aggregate mobile, point,
and area source VOC and NO emissions associated
x
with each of these cases. For example, the 1980
no-control results were derived by, first, multiplying
the 1990 SARMAP base case mobile source VOC
emissions by the ratio of 1980 no-control scenario to
1990 control scenario mobile source VOC emissions
derived for the present study. Similar adjustments were
made for point and area sources, and for NOx. The
SAQM was then re-run holding fixed all other condi-
tions associated with the 1990 SARMAP base case,
including meteorology, activity patterns, and other
conditions. The specific emission ratios used to modify
the 1990 SARMAP base case are presented in Table
C-14. The ratios themselves were derived by adding
on-highway and off-highway emissions to represent
the mobile source category; adding utility, industrial
process, and industrial combustion emissions to rep-
resent point sources; and using commercial/residen-
tial emissions to represent area sources. The no-control
scenarios were then derived by adjusting the peak and
median of the control scenario ozone distribution
based on the ratio of SARMAP-predicted peak ozone
concentrations under the control and no-control sce-
narios.
The relative results of the control and no-control
scenario runs of the OZIPM4, RADM, and SAQM
models were then used to derive the no-control case
air quality profiles. For the urban monitors relying on
OZIPM4 results, only ozone-season daytime concen-
trations could be calculated directly from OZIPM4
results. This is because OZIPM4 provides only the
maximum hourly ozone concentration. However, to
estimate all the various physical consequences of
changes in ambient ozone concentrations, the current
study requires estimation of the shift in the entire dis-
tribution of ozone concentrations. Since it is daytime
ozone season concentrations which are most sensi-
tive to changes in VOC and NOx emissions, the pre-
dicted shifts in the most important component of the
ozone concentration distribution are reasonably well-
founded. The method adopted for this analysis in-
volved applying the no-control to control peak con-
centration ratio to all concentrations in the distribu-
tion down to a level of 0.04 ppm. The 0.04 ppm level
is considered at the high end of hypothetical ambient
ozone concentrations in the absence of all anthropo-
genic ozone precursor emissions. A ratio of 1.0 is used
for ozone concentrations at or near zero. The method-
ology is described in more detail in the SAI Ozone
Report (1995) on page 4-6.
Estimating changes in rural ozone concentrations
is required primarily for estimating effects on agri-
cultural crops, trees, and other vegetation. For this
reason, only the differences in daytime, growing sea-
son ozone concentrations are derived for the present
study. As described in detail in the SAI Ozone Report
(1995) on page 4-7, the no-control rural ozone pro-
files are calculated by, first, taking the ratio of the
average daytime growing season ozone concentrations
simulated by RADM or SAQM (whichever is relevant
for that monitor). The ratio of no-control to control
scenario average ozone concentration is then applied
to all the hourly concentrations from that monitor.
Table C-14. Apportionment of Emissions Inventories for SAQM Runs.
Source
Category
1980 Control
to 1990 Control Ratio
1980 iNo-Control to
1990 Control Ratio
1990 No-Control to
1990 Control Ratio
VOC
Mobile
1.344
1.955
3.178
Area
0.820
0.901
1.106
Point
1.284
1.439
1.232
MX
Mobile
1.042
1.148
1.677
Area
0.731
0.738
1.058
Point
0.987
1.339
1.159
C-22
-------
Appendix C: Air Quality Modeling
Profiles based on 1, 2, 6, 12, and 24-hour averages
are then calculated for the control case; and averages
for daytime hours are calculated for the no-control
case.19 Even though the control and no-control sce-
nario off-season profiles are held constant, profiles
for the no-control scenario are developed for all
months of the year since the ozone season varies
throughout the country.
Summary differences in ozone air
quality
Figure C-13 presents a summary of the results of
the 1990 OZIPM4 results for all 147 of the modeled
urban areas. Specifically, the graph depicts a fre-
quency distribution of the ratio of control to no-control
scenario peak ozone. While the vast majority of simu-
lated peak ozone concentration ratios fall below 1.00,
eight urban areas show lower simulated peak ozone
for the no-control scenario than for the control sce-
nario. For these eight urban areas, emissions of pre-
cursors were higher under the no-control scenario;
however, the high proportion of ambient NOx com-
pared to ambient non-methane organic compounds
(NMOCs) in these areas results in a decrease in net
ozone production when NOx emissions increase. Fig-
ures C-14 and C-15 present frequency distributions
for control to no-control ratios of average ozone-sea-
son daytime ozone concentrations at rural monitors
as simulated by RADM and SAQM, respectively.
These figures indicate that, by 1990, no-control
scenario ozone concentrations in the modeled areas
would have been generally higher in both urban and
rural areas. Rural area concentrations differences are
not as great as urban area differences due to (a) the
differentially greater effect of CAA emission controls
in high population density areas, and (b) potential dif-
ferences in the models used for urban and rural areas.
Ozone reductions in both rural and urban areas
projected in this analysis are not as proportionally large
as the estimated reductions in emissions of ozone pre-
cursors for at least four reasons. First, current knowl-
edge of atmospheric photochemistry suggests that
ozone reductions resulting from emissions changes
will be proportionally smaller than the emissions re-
ductions. Second, biogenic emissions of VOCs, an
important ozone precursor, are significant and are held
constant for the control and no-control scenarios of
this analysis. Biogenic emissions are important be-
cause they contribute roughly half of the total
Figure C-13. Distribution of Estimated Ratios for 1990
Control to No-control OZIPM4-Simulated 1-Hour Peak
Ozone Concentrations, by Urban Area.
30
< 20
I10
a a H
w IliL,
0.00 0.20 0.40 0.60 0.80 1.00
Ratio ofCA A :No-CA A Peak Ozone (interval midpoint)
1.20
Figure C-14. Distribution of Estimated Ratios for 1990
Control to No-control RADM-Simulated Daytime Aver-
age Rural Ozone Concentrations, by RADM Grid Cell.
200
u
*©
C
o
"ra
t-i
(§
s
Q
2
H
150
100
50
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Ratio ofCA A :No-CA A Ozone-Season Daytime Average Ozone (intervalmidpoint)
Figure C-15. Distribution of Estimated Ratios for 1990
Control to No-control SAQM-Simluated Daytime Aver-
age Ozone Concentrations, by SAQM Monitor.
10
C
o
2
s
<
<73
0
0.00 0.20 0.40 0.60 0.80 1.00 1.20
Ratio ofCA A :No-CA A Ozone-Season Daytime Average Ozone (intervalmidpoint)
' The no-control scenario nighttime profiles are assumed to be the same as the control scenario profiles.
C-23
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
(manmade plus natural) VOC emissions nationwide.
Due to this abundance of VOC loading and the inher-
ent nonlinearity of the ozone-precursor response sys-
tem,20 historical reductions in anthropogenic VOC
emissions can yield minimal reductions in ozone, es-
pecially in rural environments. Third, this rural effect
also influences urban areas receiving substantial ozone
transported in from surrounding areas. Consequently,
the effect of emission controls placed in urban areas
often is reduced since much of the urban area ozone
is imported. Thus, the problem is truly regionalized
given the importance of transport, biogenic emissions
and associated urban-rural interactions, all contribut-
ing toward a relatively non-responsive atmospheric
system.21 Finally, physical process characterizations
within OZIPM4 are severely limited and incapable of
handling transport, complex flow phenomena, and
multi-day pollution events in a physically realistic
manner. Consequently, it is possible that the OZIPM4
method used herein produces negative bias tenden-
cies in control estimations. Additional discussion of
uncertainties in the ozone air quality modeling is pre-
sented in the following section.
Key caveats and uncertainties for ozone
There are a number of uncertainties in the overall
analytical results of the present study contributed by
the ozone air quality modeling in addition to the po-
tential systematic downward bias discussed above.
First, there are substantial uncertainties inherent in any
effort to model ozone formation and dispersion. These
uncertainties are compounded in the present study by
the need to perform city-specific air quality modeling
using OZIPM4, which is less sophisticated than an
Eulerian model such as the Urban Airshed Model.
However, while the absolute ozone predictions for any
given urban area provided by OZIPM4 may be quite
uncertain, the process of aggregating results for a num-
ber of cities and meteorological episodes should sig-
nificantly reduce this uncertainty.22 Urban areas for
which ozone changes may be overpredicted are offset
to some degree by urban areas for which the change
in ozone concentrations may be underpredicted. In
weighing the significance of this source of uncertainty,
it is important to consider the central purpose of the
present study, which is to develop a reasonable esti-
mate of the overall costs and benefits of all historical
Clean Air Act programs. All analyses are based on
relative modeled results, and ratios of the model pre-
dictions for the control and no-control scenarios, rather
than the absolute predictions. As a result of this, the
effect of any bias in the model predictions is greatly
reduced due to partial cancellation.
Additional uncertainty is contributed by other
limitations of the models, the supporting data, and the
scope of the present analysis. Relying on linear inter-
polation between 1970 and modeled 1980 results to
derive results for 1975, and between modeled results
for 1980 and 1990 to derive results for 1985, clearly
adds to the uncertainty associated with the RADM-
based rural ozone estimates. Assuming that changes
in peak concentration predicted by OZIPM4 and
SAQM can be applied to scale hourly ozone values
throughout the concentration distribution also contrib-
utes to uncertainty. Resource and model limitations
also required that night-time ozone concentrations be
held constant between the scenarios. This leads to an
underestimation of the night-time component of ozone
transport. Finally, changes in rural ozone in areas not
covered by RADM or SAQM could not be estimated.
As a result, potentially significant changes in ambi-
ent ozone in other major agricultural areas, such as in
the mid-west, could not be developed for this analy-
sis. The Project Team considered using an emissions
scaling (i.e., a roll-back) modeling strategy to develop
crude estimates of the potential change in rural ozone
concentrations in monitored areas outside the RADM
and SAQM domains. However, the Project Team con-
cluded that such estimates would be unreliable due to
the nonlinear effect on ozone of precursor emission
changes. Furthermore, the team concluded that
baseline levels of ozone and changes in precursor
emissions in these areas are relatively low. The deci-
sion not to spend scarce project resources on estimat-
ing ozone changes in these rural areas is further sup-
ported by the relatively modest change in rural ozone
concentrations estimated within the RADM and
SAQM domains.
20 Nonlinear systems are those where a reduction in precursors can result in a wide range of responses in secondary pollutants
such as ozone. Ozone response often is "flat" or nonresponsive to reductions of VOCs in many rural areas with significant natural
VOC emissions. Also, ozone can increase in response to increases in NOx emissions in certain localized urban areas.
21 Both the 1990 CAA and EPA's and the National Academy of Science's Section 185B Report to Congress recognized the
consequences of biogenics, transport and the need to conduct regionalized assessments, as reflected in organizational structures such
as the Ozone Transport Commission and the North American Research Strategy for Tropospheric Ozone (NARSTO).
22 Note that aggregating individual urban area results may reduce the effect of uncertainty in individual city projections (i.e.,
overestimated cities would offset underestimated cities). However, aggregation of individual urban area results would not reduce
potential errors caused by systematic biases which arise due to, for example, misestimated emissions inventories.
C-24
-------
Appendix C: Air Quality Modeling
Visibility
Two separate modeling approaches were used to
estimate changes in visibility degradation in the east-
ern and southwestern U.S. These are the two regions
of the coterminous U.S. for which Clean Air Act pro-
grams were expected to have yielded the most sig-
nificant reductions in visibility degradation. Visibil-
ity changes in the eastern 31 states were estimated
based on the RADM/EM results for sulfates; and
changes in visibility in 30 southwestern U.S. urban
areas were calculated using a linear emissions scaling
approach. Despite the potential significance of Clean
Air Act-related visibility changes in southwestern U.S.
Class I areas, such as National Parks, resource limita-
tions precluded implementation of the analysis
planned for these areas.
The RADM/EM system includes a post-proces-
sor which computes various measures of visibility
degradation associated with changes in sulfate aero-
sols.23 The basic approach is to allocate the light ex-
tinction budget for the eastern U.S. among various
aerosols, including particulate sulfates, nitrates, and
organics. The change in light extinction from sulfates
is provided directly by RADM, thereby reflecting the
complex formation and transport mechanisms asso-
ciated with this most significant contributor to light
extinction in the eastern U.S. Nitrates are not estimated
directly by RADM. Instead, RADM-estimated con-
centrations of nitric acid are used as a surrogate to
provide the basis for estimating changes in the par-
ticulate nitrate contribution to light extinction. The
organic fractions were held constant between the two
scenarios. Standard outputs include daylight distribu-
tion of light extinction, visual range, and DeciViews24
for each of RADM's 80-km grid cells. For the present
study, the RADM visibility post-processor was con-
figured to provide the 90th percentile for light extinc-
tion and the 10th percentile for visual range to repre-
sent worst cases; and the 50th percentile for both of
these to represent average cases. More detailed docu-
mentation of the RADM/EM system and the assump-
tions used to configure the visibility calculations are
presented in the RADM Report (1995).
To estimate differences in control and no-control
scenario visibility in southwestern U.S. urban areas,
a modified linear rollback approach was developed
and applied to 30 major urban areas with population
greater than 100,000.25 For each of the 30 urban cen-
ters, seasonal average 1990 air quality data was com-
piled for key pollutants, including N02 and PM|(. con-
tributing to visibility degradation in southwestern U.S.
coastal and inland cities. PM10 was then speciated into
its key components using city-specific annual aver-
age PM10 profile data. After adjusting for regional —
and for some species, city-specific— background lev-
els, concentrations of individual light-attenuating spe-
cies were scaled linearly based on changes in emis-
sions of that pollutant or a proxy pollutant.26 Using
the same approach used for the 1993 EPA Report to
Congress on effects of the 1990 Clean Air Act Amend-
ments on visibility in Class I areas, light extinction
coefficients for each of these species were then mul-
tiplied by their respective concentrations to derive a
city-specific light extinction budget.27 This process
was repeated for pre-1990 control and all no-control
scenarios by scaling 1990 results by the relative change
in annual county-level emissions of SOx, NOx, and
PM. Based on the city-specific light extinction bud-
get calculations, measures for total extinction, visual
range, and DeciView were calculated for each sce-
nario and target year.
Control scenario visibility
Unlike the other air quality conditions addressed
in the present study, modeled visibility conditions are
used as the basis for the control scenario rather than
actual historical conditions. However, like the other
air quality benefits of the historical Clean Air Act, it
is the differences between modeled visibility outcomes
for the control and no-control scenarios which are used
23 A complete discussion, including appropriate references to other documents, of the RADM and RADM/EM modeling
conducted for the present study is presented in the subsection on acid deposition earlier in this appendix.
24 The DeciView Haze Index (dV) is a relatively new visibility indicator aimed at measuring visibility changes in terms of human
perception. It is described in detail in the SAI SW Visibility Report (1994), pp. 4-2 to 4-3. See also Pitchford and Malm (1994) for
the complete derivation of the DeciView index.
25 Complete documentation of the linear scaling modeling, speciation methodologies, spatial allocation of emissions, and other
data and assumptions are provided by the SAI SW Visibility Report (1994).
26 For example, sulfate (S04) concentrations were scaled based on changes in sulfur oxide (SO ) emissions.
27 The term "light extinction budget" refers to the apportionment of total light attenuation in an area to the relevant pollutant
species.
C-25
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
to estimate visibility benefits. Nevertheless, 1990 ab-
solute levels of eastern U.S. visibility predicted by
RADM under the control scenario are presented in
Figure C-16 to provide a sense of initial visibility con-
ditions.
For the southwestern urban areas, 1990 control
scenario annual average light extinction budget, vi-
sual range, and DeciView conditions are listed in Table
C-15. These 1990 results are presented to give the
reader a sense of the initial visibility conditions in
absolute, albeit approximate, terms.
No-control scenario visibility
The no-control scenario visibility results for the
eastern U.S. area covered by RADM are presented in
Figure C-17. No-control scenario 1990 outcomes for
the 30 southwestern U.S. urban areas are presented in
Table C-16.
Summary differences in visibility
DeciView Haze Index
The DeciView Haze Index (dV) has recently been
proposed as an indicator of the clarity of the atmo-
sphere that is more closely related to human percep-
tion than visual range (VR) or total extinction (bext)
(Pitchford and Malm, 1994). It is defined by the equa-
tion:
dV= H)lne(^) (2)
where:
bext= total extinction in inverse megameters
(Mm1)
This index has the value of approximately 0 when
the extinction coefficient is equal to the scattering
coefficient for particle-free air (Rayleigh scattering)
and increases in value by approximately one unit for
each 10 percent increase in bext. Since the apparent
change in visibility is related to the percent change in
bext (Pitchford et al., 1990), equal changes in dV cor-
respond to approximately equally perceptible changes
in visibility. Recent research indicates that, for most
observers, a "just noticeable change" in visibility cor-
responds to an increase or decrease of about one to
two dV units.
Figure C-16. RADM-Predicted Visibility Degradation,
Expressed in Annual Average DeciView, for Poor Visibility
Conditions (90th Percentile Under the Control Scenario.
Figure C-17. RADM-Predicted Visibility Degradation,
Expressed in Annual Average DeciView, for Poor Visibility
Conditions (90th Percentile Under the No-control Scenario.
36 - 38
> 38
C-26
-------
Appendix C: Air Quality Modeling
Tabic C-15. 1990 Control Scenario Visibility
Conditions lor 30 Southwestern U.S. Cities.
City
Light
Extinction
Budget (b,
Mm')
Visual
Range
(fan)
Deci
View
(clV)
Los Angeles, CA
197.6
15.2
29.8
San Bernardino, CA
201.7
14.9
30.0
Riverside, CA
208.3
14.4
30.4
Anaheim, CA
170.1
17.6
28.3
Ventura, CA
113.3
26.5
24.3
San Diego, CA
126.9
23.6
25.4
Santa Barbara, CA
112.8
26.6
24.2
Bakersfield, CA
215.1
13.9
30.7
Fresno, CA
211.7
14.2
30.5
Modesto, CA
148.8
20.2
27.0
Stockton, CA
153.1
19.6
27.3
San Francisco, CA
120.8
24.8
24.9
Oakland, CA
117.5
25.5
24.6
San Jose, CA
154.6
19.4
27.4
Monterey, CA
84.7
35.4
21.4
Sacramento, CA
119.1
25.2
24.8
Redding, CA
83.2
36.1
21.2
Reno, NV
147.4
20.3
26.9
Las Vegas, NV
157.9
19.0
27.6
Salt Lake City, UT
117.5
25.5
24.6
Provo, UT
107.8
27.8
2v8
Fort Collins, CO
80.7
37.2
20.9
Greeley, CO
84.2
35.6
21.3
Denver, CO
153.4
19.6
27.3
Co lo r ad o S p r ings,
CO
83.3
36.0
21.2
Pueblo, CO
88.1
34.1
21.8
Albuquerque, NM
91.1
32.9
22.1
El Paso, TX
109.3
27.5
2v9
Tucson, AZ
85.6
35.0
21.5
Phoenix, AZ
125.3
23.9
25.3
Data Source: SAI SW Visibility Report (1994).
Table C-16. 1990 No-control Scenario Visibility Conditions for
30 Southwestern U.S. Cities.
City
Light Extinction
Budget (b, Mm ')
Visual
Range (km)
DeciView
(dV)
Los Angeles, CA
333.4
9.0
35.1
San Bernardino, CA
337.3
8.9
35.2
Riverside, CA
343.2
8.7
35.4
Anaheim, CA
286.3
10.5
33.5
Ventura, CA
194.8
15.4
29.7
San Diego, CA
210.1
14.3
30.4
Santa Barbara, CA
183.2
16.4
29.1
Bakersfield, CA
356.4
8.4
35.7
Fresno, CA
349.0
8.6
35.5
Modesto, CA
240.1
12.5
31.8
Stockton, CA
248.1
12.1
32.1
San Francisco, CA
197.3
15.2
29.8
Oakland, CA
188.6
15.9
29.4
San Jose, CA
253.0
11.9
32.3
Monterey, CA
141.4
21.2
26.5
Sacramento, CA
189.2
15.9
29.4
Redding, CA
128.6
23.3
25.5
Reno, NV
416.6
7.2
37.3
Las Vegas, NV
CO
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Both VR and dV are measures of the value of b
ext
at one location in the atmosphere. Both are unaffected
by the actual variability of the compositions and illu-
mination of the atmosphere, so neither is closely linked
to the human perception of a particular scene. The
isolation of these parameters from site-specific varia-
tions and temporal fluctuations of the atmospheric il-
lumination increases their usefulness for comparing
the effects of air quality on visibility across a range of
geographic locations for a range of time periods. Each
parameter attempts to scale the bext data so that changes
in air quality can be used to provide an indication of
changes in the human perception of a scene.
Modeling Results
The differences in modeled 1990 control and
no-control scenario visibility conditions projected by
the RADM/EM for the eastern U.S. are presented in
Figure C-18. The map shows the percent increase in
modeled annual average visibility degradation under
poor conditions for 1990 when moving from the con-
trol to the no-control scenario. The results indicate
perceptible differences in visibility between the con-
trol and no-control scenario throughout the RADM
domain. The relatively large increase in visibility im-
pairment in the Gulf Coast area is a reflection of the
Figure C-18. RADM-Predicted Increase in Visibility
Degradation, Expressed in Annual Average DeciView,
for Poor Visibility Conditions (90th Percentile) Under the
No-control Scenario.
I
significant increases in 1990 sulfate concentrations
associated with the no-control scenario. (See the ear-
lier discussion of effects in this region in the sections
dealing with acid deposition.)
The differences in modeled 1990 control and
no-control scenario visibility conditions in the 30
southwestern U.S. urban areas projected by linear roll-
back modeling are presented in Table C-17. When
reviewing these visibility degradation differentials for
the 30 southwestern U.S. urban areas, it is important
to consider that while estimated differences in visual
range were in many cases very large, changes in the
DeciView Haze Index (dV) may be relatively small.
This is because the perception of visibility degrada-
tion measured by dV may be small when baseline vis-
ibility is high.28 Even so, the results indicate that, by
1990, visibility in southwestern U.S. urban areas
would be noticeably worse under the no-control sce-
nario.
Key caveats and uncertainties for
visibility
There are several sources of uncertainty in the
RADM and southwestern U.S. linear scaling model
analyses. For RADM, the use of nitric acid as a surro-
gate for estimating changes in light-attenuating ni-
trate particles ignores the interaction effects of ni-
trates, sulfates, and ammonia. As a result, increases
in nitrates may be overestimated by the model when
both sulfates and nitric acid increase. However, the
significance of this potential overestimation is miti-
gated to some extent by the relative insignificance
of nitrate-related visibility degradation relative to
sulfates which prevails in the eastern U.S.
Several important uncertainties in the south-
western U.S. urban area visibility analysis are de-
scribed in detail in the SAI SW Visibility Report
(1994). First, the need to use seasonal average con-
ditions leads to underestimation of extreme visibil-
ity impairment episodes associated with high hu-
midity, since particle growth due to water absorp-
tion is highly nonlinear. Second, although the use
of city-specific light extinction and PM speciation
data is significantly better than reliance on regional
averages, uncertainties in city-specific data may
contribute to overall uncertainty in the estimates.
However, overall uncertainty associated with these
factors will be reduced to some extent since over-
estimation of visibility degradation in some cities
: See SAI SW Visibility Report (1994), page 5-3.
C-28
-------
Appendix C: Air Quality Modeling
will be offset by underestimations in other cities. Fi-
nally, the linear scaling used to estimate the pre-1990
control scenarios and the no-control scenarios was
based on changes in county-wide or air basin emis-
sions. Uncertainties associated with apportionment of
state-wide emission changes to individual counties or
air basins may contribute significantly to overall un-
certainty in the visibility change estimates. Such ap-
portionment is particularly difficult for SOx emission
changes, since emission reductions achieved by the
Clean Air Act tended to be at relatively remote utility
and smelter plants. However, sulfates are a relatively
minor source of light attenuation in western urban
areas.
An important overall limitation of the visibility
analysis conducted for the present study is that only
southwestern urban areas and the eastern 31 states
were included. The Clean Air Act may have contrib-
uted toward significant reductions in visibility degra-
dation in other areas. For example, Clean Air Act pro-
grams to reduce ambient particulate matter may have
motivated reductions in silvicultural burning in some
northwestern states. Perhaps the greatest deficiency
in geographic coverage by the present study is the
omission of visibility changes in Class I areas in the
west.
Tabic C-17. Summary of Relative Change in
Visual Range and Dcci View Between 1990 Control
and No-control Scenario Visibility Conditions lor
30 Southwestern U.S. Cities.
City
Visual Range
(%)
DeciView
(dV)
Los Angeles, CA
69
-5
San Bernardino, CA
67
-5
Riverside, CA
65
-5
Anaheim, CA
68
-5
Ventura, CA
72
-5
San Diego, CA
65
-5
Santa Barbara, CA
62
-5
Bakersfieid, CA
66
-5
Fresno, CA
65
-5
Modesto, CA
61
-5
Stockton, CA
62
-5
San Francisco, CA
63
-5
Oakland, CA
61
-5
San Jose, CA
64
-5
Monterey, CA
67
-5
Sacramento, CA
59
-5
Redding, CA
55
-4
Reno, NV
183
-10
Las Vegas, NV
308
-14
Salt Lake City, UT
58
-5
Provo, UT
48
-4
Fort Collins, CO
137
-9
Greeley, CO
39
-3
Denver, CO
85
-6
Colorado Springs, CO
111
-7
Pueblo, CO
240
-12
Albuquerque, NM
93
-7
El Paso, TX
153
-9
Tucson, AZ
218
-12
Phoenix, AZ
243
-12
Data Source: SAI SW Visibility Report (1994).
C-29
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Air Quality Modeling References
Chang. 1995. SARMAP Air Quality Model (SAQM).
Final report to San Joaquin Valley wide Air
Pollution Study Agency.
DaMassa, Tanrikulu, and Ranzier. 1996. Photochemi-
cal Modeling of August 3-6, 1990, Ozone
Episode in Central California Using the
SARMAP Air Quality Model. Part II: Sensi-
tivity and Diagnostic Testing. Preprints, Ninth
Joint Conference on the Applications of Air
Pollution Meteorology with Air Waste Man-
agement Association. January 28 - February
2, 1996, Atlanta, Georgia.
Dennis, R. 1995. Estimation of Regional Air Quality
and Deposition Changes Under Alternative
812 Emissions Scenarios Predicted by the
Regional Acid Deposition Model, RADM.
Draft Report for U.S. Environmental Protec-
tion Agency, ORD/NERL. October 1995.
ICF Kaiser/Science Applications International. 1996.
PM Interpolation Methodology for the Sec-
tion 812 Retrospective Analysis. Memoran-
dum from J. Langstaff to Jim DeMocker.
ICF Kaiser/Systems Applications International. 1994.
Retrospective Analysis of the Impact of the
Clean Air Act on Urban Visibility in the South-
western United States. Final Report.
ICF Kaiser/Systems Applications International. 1995.
Retrospective Analysis of Ozone Air Quality
in the United States. Final Report.
ICF Kaiser/Systems Applications International. 1992.
Retrospective Analysis of Particulate Matter
Air Quality in the United States. Draft Re-
port.
ICF Resources Incorporated. 1992. Results of Retro-
spective Electric Utility Clean Air Act Analy-
sis -1980, 1985, and 1990. September 30.
Pitchford, Marc L. and William C. Malm. 1994. "De-
velopment and Applications of a Standard
Visual Indent" Atmospheric Environment, vol.
28, no. 5.pp.1049-1054.
Seaman and Stauffer. 1995 .Development and Design
Testing of the SARMAP Meteorological
Model. Final report to San Joaquin Valley
wide Air Pollution Study Agency.
Seaman, Stauffer, and Lario-Gibbs. 1995. "A Multi-
Scale Four Dimensional Data Assimilation
System Applied in the San Joaquin Valley
During SARMAP. Part I: Modeling Design
and Basic Performance Characteristics."Jour-
nal of Applied Meteorology. Volume 34. In
press.
Tanrikulu, DaMassa, and Ranzieri. 1996 .Photochemi-
cal Modeling of August 3-6, 1990 Ozone Epi-
sode in Central California Using the
SARMAP Air Quality Model. Part I: Model
Formulation, Description and Basic Perfor-
mance. Preprints. Ninth Joint Conference on
the Application of Air Pollution Meteorology
with Air Waste Management Association.
January 28 - February 2, 1996. Atlanta, Geor-
gia.
Trijonis. 1990. Visibility: Existing and Historical
Conditions—Causes and Effects. NAPAP Re-
port 24. 1990.
U.S. Environmental Protection Agency (EPA). 1989.
Procedures for Applying City-Specific EKMA.
EPA-450/4-89-012. Office of Air Quality
Planning and Standards.
ICF Kaiser/Systems Applications International. 1995.
Retrospective Analysis of Particulate Matter
Air Quality in the United States. Final Report.
ICF Kaiser/Systems Applications International. 1994.
Retrospective Analysis of SO , NO and CO
Air Quality in the United States. Final Report.
C-30
-------
Appendix D: Human Health and Welfare Effects
of Criteria Pollutants
Introduction
In responding to the mandate of section 812, EPA
conducted a comprehensive benefits analysis to iden-
tify and estimate the quantifiable health and welfare
benefits enjoyed by Americans due to improved air
quality resulting from the CAA. Health benefits re-
sulted from avoidance of air pollution-related health
effects, such as mortality, respiratory illness, and heart
disease. Welfare benefits accrued where improved air
quality averted damage to ecological health and mea-
surable resources, such as agricultural production,
building materials, and visibility.
This appendix presents an overview of EPA's
approach for modeling human health and welfare ef-
fects. It provides an outline of the principles used to
guide the benefits analysis, details methods used to
quantify criteria air pollutant exposure nationwide
across the study period (1970 to 1990), and discusses
several critical conceptual and implementation issues
for using health and welfare effect information. Mod-
eling results, estimates of avoided incidences of ad-
verse health and welfare effects, are then presented.
Ecological and agricultural benefits are examined in
more detail in Appendices E and F, respectively. Ap-
pendix I details the approach used to translate health
and welfare effects into monetary benefits.
Principles for the Section 812
Benefits Analysis
Estimating the effects of even modest shifts in
environmental releases involves complex chemical,
environmental, biological, psychological and eco-
nomic processes. The task of estimating the broad
changes associated with adoption and implementation
of the Clean Air Act challenges the limits of scien-
tific knowledge and modeling capability to synthe-
size available information and techniques into a prac-
tical framework. A pragmatic plan for a comprehen-
sive assessment must fairly reflect the complexities
and uncertainties, but still produce a policy-relevant
analysis in a timely fashion. In order to achieve this
ambitious goal, the following principles have been
used to guide the section 812 benefits assessment.
Comprehensiveness: The assessment should in-
clude as many benefit categories as are reasonably
believed to be affected by implementation of the Clean
Air Act. Comprehensiveness requires assessing effects
with which greater levels of scientific confidence are
associated, as well as less well-understood effects. The
degree of relative certainty among effects must be
carefully described in order to fairly present a broad
portrayal of the physical and social benefits accruing
to the nation from implementing the Act. In addition,
section 812 of the 1990 CAA Amendments explicitly
directs a comprehensive benefits coverage that pro-
hibits a default assumption of zero value for identi-
fied benefits unless a zero value is supported by spe-
cific data.
Quantification Where Feasible: The central goal
of the present study is to evaluate and compare the
benefits and costs of historical CAA-related programs.
Effective comparison of the variety of human health,
welfare, and ecological benefits with the associated
compliance costs requires that these consequences be
measured in terms of a common metric. Expressing
the value of these various effects in economic terms
is the most efficient way to accomplish this objec-
tive, and is consistent with standard practices associ-
ated with economic benefit-cost analysis. Expressing
these effects in economic terms requires quantifying
and presenting estimated effects in both physical and
monetized economic terms. Pursuant to this paradigm,
the emphasis in the present study is largely on cat-
egories having direct and perceptible effects on hu-
man health. That is, the emphasis of the analysis is on
categories such as symptoms and diseases rather than
on physical changes (such as cell level changes) that
do not directly result in a decreased health status no-
ticeable to the individual.
D-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Efficient Use of Previous Research Results: Sig-
nificant research effort has been spent to understand
and quantify the complex relationships between air
pollution and human health. The present study has
relied as much as possible on available research re-
sults, making adjustments as necessary to apply the
existing results to the current analysis.
Incorporate Uncertainty: To properly convey the
results of any benefits assessment, it is important to
include an evaluation and characterization of how
much confidence the analysts have in the estimates.
Ideally this would include a formal quantitative as-
sessment of the potential for error, and the sources,
directions, and potential significance of any resultant
biases. A method for considering and reporting un-
certainty must be built into the fundamental design of
the assessment. Such a framework was developed and
applied in the present study, and was supplemented
where necessary by expert judgment regarding the
sources and potential significance of errors in each
analytical step.
General Modeling Approach
Consistent with these principles, the EPA devel-
oped an approach for quantifying the effects of re-
duced pollutant exposure, with particular focus on
those effect categories for which monetary benefits
could be estimated. As described previously, the study
design adopted for the section 812 assessment links a
sequence of analytical models. The macroeconomic
modeling (Appendix A) estimated economy-wide ef-
fects of CAA expenditures. These effects provided a
basis for the modeling of criteria pollutant emissions
under the two scenarios considered (the factual con-
trol scenario and the hypothetical no-control scenario),
as documented in Appendix B. The emissions esti-
mates were used as input to the air quality models
(Appendix C). Ambient pollutant concentrations es-
timated by the air quality models were used as inputs
to the health and welfare benefits model, the focus of
this appendix.
The approach developed to model health and wel-
fare benefits is known as a "reduced form" or "em-
bedded model" approach. The concept of a reduced
form model is to use simplified versions of previously
constructed complex models to characterize the im-
pact of a series of linked physical and socioeconomic
processes. The health and welfare benefits model is
characterized as a reduced form model because it re-
lies on summaries of the data output from the air qual-
ity models, which rely on emissions summaries and
summaries of macroeconomic conditions, succes-
sively. Although results of the independent models
are used in series, the models themselves have not
been integrated into the health and welfare benefits
model.
In general, the reduced form health and welfare
benefits model relies on two fundamental inputs: (1)
nationwide changes in pollutant exposures across the
study period, and (2) the association between changes
in exposure and expected changes in specific health
and welfare effects. These inputs are discussed be-
low.
Quantifying Changes in Pollutant
Exposures
Estimating changes in pollutant exposures re-
quires characterization of nationwide air quality im-
provements across the study period, as well as the
populations exposed to the different levels of improve-
ment.
Air Quality
As discussed in Appendix C, the section 812
analysis estimated ambient concentrations for both the
control and no-control scenarios for the following
pollutants and air quality parameters:
Particulate matter, less than 10 microns in
diameter (PM10)
Ozone (03)
Nitrogen dioxide (N02)
Sulfur dioxide (S02)
Carbon monoxide (CO)
Visibility measures (light extinction and
DeciView)1
Lead (Pb)
Generally, this analysis adopted actual historical
air pollution monitoring data to represent control sce-
nario air quality. No-control scenario profiles were
1 While the visibility measures listed are not criteria air pollutants, they provide important measures of a significant welfare
effect resulting from air pollution, visibility degradation. Light extinction (which is related to DeciView, a haziness index) results
from light scattered by fine particles in the atmosphere, especially sulfates and ammonium nitrates. As atmospheric concentrations of
such particles increase, light is attenuated and visibility diminishes.
D-2
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
derived by running the control and no-control scenario
emissions inventories through a suite of air quality
models and then using the differences in these mod-
eled outcomes to adjust the historical profiles. Since
lead was treated differently than the other pollutants,
the analysis of the CAA impacts on atmospheric lead
concentrations is documented in Appendix G.
With respect to the distribution of air quality data
across the two decades considered, it should be noted
that both the number and location of monitors track-
ing air quality changed over time. Table D-l depicts
the number of monitors for each pollutant across the
period of this analysis. The number of monitors gen-
erally increased throughout the 1970s and leveled off
or declined at varying points during the 1980s, de-
pending on the pollutant.
Tabic D-l. Criteria Air Pollutant Monitors
in the U.S.. 1970 - 1990.
Pollutant
Year
£Ml0
£h
m
SO.
CO
1970
245
i
43
86
82
1975
1,120
321
303
827
494
1980
1,131
546
375
1,088
511
1985
970
527
305
916
458
1990
720
627
345
753
493
For the section 812 modeling, the non-lead pol-
lutants have been characterized as either county-level
or monitor-level pollutants. The distinction was im-
portant for quantifying the population exposed to dif-
ferent levels of air quality improvements, as discussed
below. PM10 is considered a county-level pollutant,
since historical concentrations in monitored counties
have been synthesized into a single concentration for
each county.2 In contrast, 03, N02, NO, S02, and CO
were reported at specific monitor locations, given by
latitude/longitude coordinates. Finally, visibility was
treated as a county-level pollutant in the western U.S.
and a monitor-level pollutant in the eastern U.S.3 Air
quality data for PM: 0 and ozone were reported for each
year of the study period; data for the remaining pol-
lutants were reported only for 1975, 1980, 1985, and
1990.
In order to reduce the volume of air quality data
necessary to describe pollutant concentrations for two
scenarios nationwide over twenty years, annual con-
centration profiles were reduced to frequency distri-
butions. That is, annual pollutant concentrations for a
variety of averaging times (e.g., 1-hour, 6-hour, daily)
were summarized as a distribution of values across
the year. This approach reduced data management
requirements significantly, while adequately captur-
ing air quality improvements between the control and
no-control scenarios.
Population Distribution
Health and some welfare benefits resulting from
air quality improvements are distributed to popula-
tions in proportion to the reduction in exposure each
enjoys. Predicting population exposures, then, is a
necessary step in estimating health effects. Doing so
for the section 812 analysis required not only an un-
derstanding of where air quality improved as a result
of the CAA, but also how many individuals were af-
fected by varying levels of air quality improvements.
Thus, a critical component of the benefits analysis
required that the distribution of the U.S. population
nationwide be described in a manner compatible with
the air quality data. Described below is the method
used to allocate U.S. Census data to a symmetrical
grid overlying the country.
Census Data
Three years of U.S. Census data were used to rep-
resent the geographical distribution of U.S. residents:
1970, 1980, and 1990. Population data were supplied
at the census block group level, with approximately
2 Two different measures of ambient concentrations of particulate matter were used in the United States during the period 1970
to 1990. Prior to 1987, the indicator for the National Ambient Air Quality Standard for PM was total suspended particulates (TSP). In
1987, the indicator was changed to PM (particles less than 10 (iM in diameter). Widespread PM10 monitoring did not begin until
1985; prior to that only TSP data is available. Because the recent scientific literature reports primarily the relationship between PM10
and adverse health and welfare effects, PM10 data is preferred, if available. Where only TSP is available, PM10 concentrations were
estimated using PM10:TSP ratios that vary by area of the country and the urban/rural characterization of the area.
3 In the western U.S., visibility was modeled using a linear-rollback model and extinction budget approach for 30 major urban
centers (SAI, 1994). The modeling results, reported in DeciView, were applied to the counties in the vicinity of the urban centers and
considered to share a common air basin. In the eastern U.S., Regional Acid Deposition Model (RADM) runs provided visibility
estimates in terms of light extinction coefficients. These were modeled across a 60 km. X 60 km. grid, approximately covering the
eastern half of the country. Since the extinction coefficients were reported at the grid cell centroids, for which the coordinates were
known, visibility in the east was treated as a monitor-level pollutant.
D-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
290,000 block groups nationwide. Allocating air qual-
ity improvements to the population during intermedi-
ate years necessitated interpolation of the three years
of population data. Linear interpolation was performed
at the block group level in order to preserve the vari-
ability in growth rates throughout the country.
Gridding U.S. Population
To ease computational burden, block group popu-
lation estimates were aggregated to a rectangular grid
structure. The grid, comprised of ten kilometer by ten
kilometer gridcells, spanned the entire area of the con-
tentional United States. This grid size generated
46,885 populated gridcells throughout the U.S.
The entire population of each block group was
assumed to reside at the geographical centroid of the
block group area, the coordinates of which were avail-
able from the U.S. Bureau ofthe Census. Block group
populations were aggregated to gridcells according
to the block group centroids encompassed by each cell.
In addition to the population of each gridcell, the state
and county names for each gridcell were retained,
permitting aggregation of data at the state and county
level, as well as nationwide.
Allocating Exposure Estimates to the Population
Two alternative modeling strategies were used to
allocate air quality improvements to the U.S. popula-
tion. They differed in terms of both the certainty of
the estimates and the geographic coverage:
Tabic D-2. Population Coverage in the "Within
50 km" Model Runs (percent of continental U.S.
population).
1975
1980
1985
1990
CO
67.4%
67.9%
68.4%
70.4%
EXT
73.2%
72.3%
72.3%
72.2%
NOi
53.3%
58.8%
60.8%
61.5%
Cfe
55.5%
70.5%
71.5%
74.4%
PMio
78.5%
79.5%
75.8%
67.8%
SO2
64.7%
73.3%
73.0%
70.6%
Pb
100%
100%
100%
100%
Method One
Air quality improvements (difference between
control and no-control scenarios) were applied to in-
dividuals living in the vicinity of air quality monitors.
For pollutants with monitor-level data, it was assumed
that the individuals in a gridcell were exposed to air
quality changes estimated at the nearest monitor, as
long as the monitor was within 50 kilometers. Like-
wise, for PMJ0 (for which data was available at the
county level) the population of each monitored county
was assumed to be exposed to the air quality changes
reported for that county.4 The remainder of the popu-
lation was excluded from the analysis.
Unfortunately, by limiting the quantitative analy-
sis to populations within 5 0 km of a monitor (or within
a monitored county, for PM), a significant portion of
the U.S. population was left out of the analysis (see
Table D-2). For most pollutants in most years (ex-
cepting lead), less than three-quarters of the popula-
tion lived within 50 km of a monitor (or within a PM-
monitored county). Clearly, an analysis that excluded
25 percent of the population from the benefits calcu-
lations (thus implicitly assuming that the CAA had
no impact on that population) would understate the
physical effects of the CAA. Conversely, ascribing
air pollution reduction benefits to persons living great
distances from air quality monitors is a speculative
exercise, and could overstate benefits.
Method Two
As an alternative modeling strategy, air quality
improvements were applied to almost all individuals
nationwide. Where monitor data were not available
within 50 kilometers, data from the closest monitor,
regardless of distance, were used. Similarly, PM10
concentrations were extrapolated using regional air
quality models to all counties (even those for which
monitoring data was unavailable) and applied to the
populations of those counties.
Although subject to less certain air quality data,
the second alternative extrapolates pollutant exposure
estimates to almost the entire population using the
closest monitoring data available (see Table D-3).5
This second alternative was chosen as the preferred
approach in the benefits analysis. The sensitivity of
4 Since the lead (Pb) analysis, which was handled separately from that of the other criteria pollutants, did not require air quality
modeling data, the issue of proximity to monitors is irrelevant. The Pb analysis extended to 100 percent of the population.
5 While this alternative captures the vast majority of the U.S. population, it does not model exposure for everyone. To improve
computational efficiency, those gridcells with populations less than 1,000 were not modeled; these cells account for less than five
percent of the U.S. population.
D-4
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
the benefits estimate to the extrapolation of air qual-
ity data beyond monitored areas is explored in Ap-
pendix I.
Tabic D-3. Population Coverage lor
"Extrapolated to All U.S." Model Runs (percent
of continental U.S. population).
1975
1980
1985
1990
CO
97.2%
97.2%
98.7%
100.0%
EXT
75.6%
74.8%
74.7%
74.7%
NO2
97.2%
97.2%
98.7%
100.0%
O3
96.6%
97.2%
98.7%
100.0%
PM10
95.9%
95.8%
97.2%
98.5%
SO2
95.4%
95.6%
97.0%
98.4%
Pb
100%
100%
100%
100%
Estimating Human Health Effects
of Exposure
It is impossible to estimate all of the physical ef-
fects that would have occurred without the Clean Air
Act. While scientific information is available that
makes it possible to estimate certain effects, many
other, potentially very important, health and welfare
effects cannot be estimated at this time. Other physi-
cal effects can be quantified, but it is impossible to
assess the economic value of those endpoints based
on the current economics literature. Table D-4 shows
the health and welfare effects for which quantitative
analysis has been prepared, as well as some of the
health effects that have not been quantified in the
analysis.
In order to translate the reductions in pollutant
exposure estimated to result from the CAA into health
benefits, it is necessary to quantify the relationship
between such exposures and adverse health effects.
As indicated below, this analysis relies on concentra-
tion-response relationships published in the scientific
literature which provide estimates of the number of
fewer individuals that incur an adverse health effect
per unit change in air quality. Such relationships are
combined with the air quality improvement and popu-
lation distribution data to estimate changes in the in-
cidence of each health endpoint. By evaluating each
concentration-response function for every gridcell
throughout the country, and aggregating the resulting
incidence estimates, it was possible to generate na-
tional estimates of avoided incidence.
It should be noted that a slightly different approach
was used to compute health effects associated with
exposure to gasoline lead. Instead of relating health
outcomes to ambient pollutant concentrations, the
concentration-response functions for lead-induced
effects link changes in health effects directly to
changes in the population's mean blood lead level.
This value is directly related to the concentration of
lead in gasoline in a particular year. Appendix G docu-
ments both the methods used to characterize mean
blood lead levels and the approach for estimating hu-
man health effects from lead exposure.
The discussion below outlines the types of health
studies considered for this analysis, and issues criti-
cal to selecting specific studies appropriate for use in
the section 812 context. Next, details regarding use of
the results of the studies are explored. Finally, the
concentration-response functions used to model health
benefits from reductions in non-lead criteria pollut-
ants are outlined.
Types of Health Studies
Scientific research about air pollution's adverse
health impacts uses a broad array of methods and pro-
cedures. The research methods used to investigate the
health effects of air pollution have become consider-
ably more sophisticated over time, and will continue
to evolve in the future. This progress is the result of
better available research techniques and data, and the
ability to focus further research more sharply on key
remaining issues based on the contributions of earlier
work.
The available health effects studies that could
potentially be used as the basis of the section 812 as-
sessment are categorized into epidemiology studies
and human clinical studies. Epidemiological research
in air pollution investigates the association between
exposure to air pollution and observed health effects
in the study population. Human clinical studies in-
volve examination of human responses to controlled
conditions in a laboratory setting. Research has been
conducted on health effects from exposure to pollu-
tion using each approach, and studies using these tech-
niques have been considered in various formal regu-
latory proceedings. Each type of study (as it is used
D-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Table D-4. Human Health Effects of Criteria Pollutants.
Pollute lit
Quantified Ilenlth Kffects
In quantified Mesilth Kffects
Other Possible Kffects
O/onc
Mortality*
Respiratory symptoms
Minor restricted activity days
Respiratory restricted activity
days
Hospital admissions
Asthma attacks
Changes in pulmonary function
Chronic Sinusitis & Hay Fever
Increased airway responsiveness
to stimuli
Ce ntro ac inar fib rc> s i s
Inflammation in the lung
Immunologic changes
Chronic respiratory diseases
Extrapulmonary effects (e.g.,
changes in structure,
function of other organs)
Pn t'ticu In to Mutter/
ISP/ Sulfates
Mortality*
Bronchitis - Chronic and Acute
Hospital admissions
Lower respiratory illness
Upper respiratory illness
Chest illness
Respiratory symptoms
Minor restricted activity days
All restricted activity days
Days of work loss
Moderate or worse asthma status
(asthmatics)
Changes inpulmonary function
Chronic respiratory diseases
other than chronic bronchitis
Inflammation in the lung
( jirbon Monoxide
Hospital Admissions -
congestive heart failure
Decreased time to onset ofangina
Behavioral effects
Other hospital admissions
Other cardiovascular effects
Developmental effects
Nitrogen Oxides
Respiratory illness
Increased airway responsiveness
Decreased pulmonary function
Inflammation in the lung
Immunological changes
Sulfur Dioxide
In exercising asthmatics:
Changes in pulmonary function
Respiratory symptoms
Combined responses of
respiratory symptoms and
pulmonary function changes
Respiratory symptoms in non-
asthmatics
Hospital admissions
I A' s i d
Mortality
Hypertension
Non-fatal coronary heart disease
Non-fatal strokes
IQ loss effect on lifetime earnings
IQ loss effects onspecial
educ ation n ee ds
Health effects for individuals in
age ranges other than those
studied
N euro behavioral function
Other cardiovascular diseases
Reproductive effects
Fetal effects from maternal
expo sure
Delinquent and anti-social
behavior in children
* This analysis estimates excess mortality using PMi0 as an indicator of the pollutant mix to which
individuals were exposed.
for air pollution research) is described below, and the
relative strengths and weaknesses for the purposes of
the section 812 assessment are examined.
Epidemiological Studies
Epidemiological studies evaluate the relationship
between exposures to ambient air pollution and health
effects in the human population, typically in a "natu-
ral" setting. Statistical techniques (typically variants
of multivariate regression analysis) are used to esti-
mate quantitative concentration-response (or expo-
sure-response) relationships between pollution levels
and health effects.
Epidemiology studies can examine many of the
types of health effects that are difficult to study using
a clinical approach. Epidemiological results are well-
suited for quantitative benefit analyses because they
provide a means to estimate the incidence of health
effects related to varying levels of ambient air pollu-
tion without extensive further modeling effort. These
estimated relationships implicitly take into account
at least some of the complex real-world human activ-
ity patterns, spatial and temporal distributions of air
pollution, synergistic effects of multiple pollutants and
other risk factors, and compensating or mitigating
behavior by the subject population. Suspected rela-
tionships between air pollution and the effects of both
D-6
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
long-term and short-term exposure can be investigated
using an epidemiological approach. In addition, ob-
servable health endpoints are measured, unlike clini-
cal studies which often monitor endpoints that do not
result in observable health effects (e.g. forced expira-
tory volume). Thus, from the point of view of con-
ducting a benefits analysis, the results of epidemio-
logical studies, combined with measures of ambient
pollution levels and the size of the relevant popula-
tion, provide all the essential components for associ-
ating measures of ambient air pollution and health sta-
tus for a population in the airshed being monitored.
Two types of epidemiological studies are consid-
ered for dose-response modeling: individual level
cohort studies and population level ecological stud-
ies. Cohort-based studies track individuals that are
initially disease-free over a certain period of time, with
periodic evaluation of the individuals' health status.
Studies about relatively rare events such as cancer
incidence or mortality can require tracking the indi-
viduals over a long period of time, while more com-
mon events (e.g., respiratory symptoms) occur with
sufficient frequency to evaluate the relationship over
a much shorter time period. An important feature of
cohort studies is that information is known about each
individual, including other potential variables corre-
lated to disease state. These variables, called con-
founders, are important to identify because if they are
not accounted for in the study they may produce a
spurious association between air pollution and health
effect.
A second type of study used in this analysis is a
population-level ecological study. The relationship
between population-wide health information (such as
counts for daily mortality, hospital admissions, or
emergency room visits) and ambient levels of air pol-
lution are evaluated. One particular type of ecologi-
cal study, time-series, has been used frequently in air-
pollution research. An advantage of the time-series
design is that it allows "the population to serve as its
own control" with regard to certain factors such as
race and gender. Other factors that change over time
(tobacco, alcohol and illicit drug use, access to health
care, employment, and nutrition) can also affect health.
However, since such potential confounding factors are
unlikely to vary over time in the same manner as air
pollution levels, or to vary over periods of months to
several years in a given community, these factors are
unlikely to affect the magnitude of the association
between air pollution and variations in short-term
human health responses.
Drawbacks to epidemiological methods include
difficulties associated with adequately characterizing
exposure, measurement errors in the explanatory vari-
ables, the influence of unmeasured variables, and cor-
relations between the pollution variables of concern
and both the included and omitted variables. These
can potentially lead to spurious conclusions. However,
epidemiological studies involve a large number of
people and do not suffer extrapolation problems com-
mon to clinical studies of limited numbers of people
from selected population subgroups.
Human Clinical Studies
Clinical studies of air pollution involve exposing
human subjects to various levels of air pollution in a
carefully controlled and monitored laboratory situa-
tion. The physical condition of the subjects is mea-
sured before, during and after the pollution exposure.
Physical condition measurements can include general
biomedical information (e.g., pulse rate and blood
pressure), physiological effects specifically affected
by the pollutant (e.g., lung function), the onset of
symptoms (e.g., wheezing or chest pain), or the abil-
ity of the individual to perform specific physical or
cognitive tasks (e.g., maximum sustainable speed on
a treadmill). These studies often involve exposing the
individuals to pollutants while exercising, increasing
the amount of pollutants that are actually introduced
into the lungs.
Clinical studies can isolate cause-effect relation-
ships between pollutants and certain human health
effects. Repeated experiments altering the pollutant
level, exercise regime duration and types of partici-
pants can potentially identify effect thresholds, the
impact of recovery (rest) periods, and the differences
in response among population groups. While cost con-
siderations tend to limit the number of participants
and experimental variants examined in a single study,
clinical studies can follow rigorous laboratory scien-
tific protocols, such as the use of placebos (clean air)
to establish a baseline level of effects and precise
measurement of certain health effects of concern.
There are drawbacks to using clinical studies as
the basis for a comprehensive benefits analysis. Clini-
cal studies are appropriate for examining acute symp-
toms caused by short-term exposure to a pollutant.
While this permits examination of some important
health effects from air pollution, such as
bronchoconstriction in asthmatic individuals caused
by sulfur dioxide, it excludes studying more severe
D-7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
effects or effects caused by long term exposure. An-
other drawback is that health effects measured in some
well-designed clinical studies are selected on the ba-
sis of the ability to measure precisely the effect, for
example forced expiratory volume, rather than a larger
symptom. The impact of some clinically measurable
but reversible health effects such as lung function on
future medical condition or lifestyle changes are not
well understood.
Ethical limits on experiments involving humans
also impose important limits to the potential scope of
clinical research. Chronic effects cannot be investi-
gated because people cannot be kept in controlled
conditions for an extended period of time, and be-
cause these effects are generally irreversible. Partici-
pation is generally restricted to healthy subjects, or at
least to exclude people with substantial health condi-
tions that compromise their safe inclusion in the study.
This can cause clinical studies to avoid providing di-
rect evidence about populations of most concern, such
as people who already have serious respiratory dis-
eases. Ethical considerations also limit the exposures
to relatively modest exposure levels, and to examin-
ing only mild health effects that do no permanent dam-
age. Obviously for ethical reasons human clinical evi-
dence cannot be obtained on the possible relationship
between pollution and mortality, heart attack or stroke,
or cancer.
One potential obstacle to using dose-response in-
formation from clinical research methods in a ben-
efits assessment is the need for an exposure model.
The dose-response functions developed from clinical
research are specific to the population participating
in the study and the exposure conditions used in the
laboratory setting. It is therefore difficult to extrapo-
late results from clinical settings to daily exposures
faced by the whole population. For example, many
clinical studies evaluate effects on exercising individu-
als. Only a small portion of the population engages in
strenuous activity (manual labor or exercise) at any
time. Reflecting these fundamental differences be-
tween the laboratory setting and the "real world" im-
poses a formidable burden on researchers to provide
information about human activity patterns, exercise
levels, and pollution levels. This requirement adds an
additional step in the analytical process, introducing
another source of uncertainty and possible error.
To apply the clinical results to model the general
population, two decisions must be made. First, how
far can the conditions in the clinical setting be ex-
panded? For example, if the subjects in the clinical
study were healthy male college students, should the
results be applied to the entire population, including
children? Second, how many people in the general
population are exposed to conditions similar to those
used in the clinical setting? Frequently, clinical stud-
ies are conducted at relatively high exercise levels (in-
creasing the dose, or the quantity of pollutants actu-
ally delivered to the lungs). In the general population
few people experience these conditions very often,
and people do not reach these exercise levels with
equal frequencies during the day and night.
In addition, the analyst must determine the num-
ber of people that are exposed to the levels of ambient
conditions seen in the laboratory. Air quality varies
throughout a city and is typically reported by data from
monitors located at various places throughout the city.
However, people are not exposed to the conditions at
any one monitor all day. As people move around in
the city, they are exposed to ambient air quality con-
ditions represented by different monitors at different
times during the day. To further compound the prob-
lem, air quality also varies between indoors and out-
doors, within a car or garage, and by such factors as
proximity to a roadway or major pollution source (or
sink). The exposure model must account for the am-
bient conditions in the "microenvironments" that the
population actually experiences.
The issues of study subjects, exercise and mi-
croenvironments can influence the choice of clinical
studies selected forthe section 812 assessment. Clini-
cal studies that use exposure regimes and exercise lev-
els more similar to what larger groups of the popula-
tion see are easier to apply in a benefits model than
are more narrow studies. Similarly, studies that use a
diverse group of subjects are easier to apply to the
general population than are more narrow studies.
Given the major advantages of epidemiological
studies—exposures do not need to be modeled and
health effects are observed in a large, more heteroge-
neous population—epidemiological studies are used
as the basis for determining the majority of health ef-
fects and dose-response curves. The diverse activity
patterns, microenvironments, and pollution levels are
already considered in the aggregate through the con-
centration-response functions derived from epidemio-
logical studies. Clinical studies are used if there are
health effects observed in clinical studies not observed
in epidemiological studies.
D-8
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
Issues in Selecting Studies To Estimate
Health Effects
A number of issues arise when selecting and link-
ing the individual components of a comprehensive
benefits analysis. The appropriate procedure for han-
dling each issue must be decided within the context
of the current analytical needs, considering the broader
analytical framework. While more sophisticated or
robust studies may be available in some circumstances,
the potential impact on the overall analysis may make
using a simpler, more tractable approach the pragmatic
choice. In considering the overall impact of selecting
a study for use in the section 812 assessment, impor-
tant factors to consider include the likely magnitude
the decision will have on the overall analysis, the bal-
ance between the overall level of analytical rigor and
comprehensiveness in separate pieces of the analysis,
and the effect on the scientific defensibility of the
overall project.
This section discusses ten critical issues in select-
ing health information for use in the section 812 as-
sessment: use of peer-reviewed research, confound-
ing factors, uncertainty, the magnitude of exposure,
duration of exposure, threshold concentrations, the
target population, statistical significance of relation-
ships, relative risks, and the need for baseline inci-
dence data. The previous discussion about the types
of research methods available for the health informa-
tion alluded to some of these issues, as they are po-
tentially important factors in selecting between stud-
ies using different methods. Other issues address how
scientific research is used in the overall analytical
framework.
Peer-Review of Research
Whenever possible, peer reviewed research rather
than unpublished information has been relied upon.
Research that has been reviewed by the EPA's own
peer review processes, such as review by the Clean
Air Science Advisory Committee (CASAC) of the
Science Advisory Board (SAB), has been used when-
ever possible. Research reviewed by other public sci-
entific peer review processes such as the National
Academy of Science, the National Acidic Precipita-
tion Assessment Program, and the Health Effects In-
stitute is also included in this category.
Research published in peer reviewed journals but
not reviewed by CASAC has also been considered for
use in the section 812 assessment, and has been used
if it is determined to be the most appropriate avail-
able study. Research accepted for publication by peer
reviewed journals ("in press") has been considered to
have been published. Indications that EPA intends to
submit research to the CASAC (such as inclusion in a
draft Criteria Document or Staff Paper) provide fur-
ther evidence that the journal-published research
should be used.
Air pollution health research is a very active field
of scientific inquiry, and new results are being pro-
duced constantly. Many research findings are first
released in University Working Papers, dissertations,
government reports, non-reviewed journals and con-
ference proceedings. Some research is published in
abstract form in journals, which does not require peer
review. In order to use the most recent research find-
ings and be as comprehensive as possible, unpublished
research was examined for possible use in the section
812 assessment. Any unpublished research used is
carefully identified in the report, and treated as hav-
ing a higher degree of uncertainty than published re-
sults. The peer review of the section 812 assessment
by the Advisory Council on Clean Air Compliance
Analysis provides one review process for all compo-
nents of the assessment, as well as for the way in which
the components have been used.
Confounding Factors
Confounding can occur when the real cause of
disease is associated with a number of factors. If only
one contributing factor is evaluated in an epidemio-
logical study, a false association may occur. For ex-
ample, in epidemiology studies of air pollution, it is
important to take into account weather conditions,
because weather is associated with both air pollution
and health outcomes. If only air pollution is evalu-
ated, a false association between air pollution and
health could result; one may incorrectly assume that
a reduction in air pollution is exclusively responsible
for a reduction in a health outcome. Potential con-
founders include weather-related variables, age and
gender mix of the subject population, and pollution
emissions other than those being studied. Studies that
control for a broad range of likely confounders can
offer a more robust conclusion about an individual
pollutant, even if the statistical confidence interval is
larger due to the inclusion of more variables in the
analysis.
D-9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
In many cases, several pollutants in a "pollutant
mix" are correlated with each other—that is, they tend
to occur simultaneously. Therefore, although there
may be an association between a health effect and each
of several pollutants in the mix, it may not be clear
which pollutant is causally related to the health effect
(or whether more than one pollutant is causally re-
lated). This analysis includes epidemiological mod-
eling of the health effects that have been associated
with exposure to a number of pollutants. In most cases
where the health effect is being modeled for the sev-
eral correlated pollutants of interest, regression coef-
ficients based on PM as a surrogate for the mixture
were chosen in preference to multiple pollutant mod-
els and single pollutant models. The most important
example of this occurs in estimating mortality effects.
There is substantial evidence that exposure to criteria
pollutants, either individually or collectively, is sig-
nificantly associated with excess mortality. Generally,
this association is related to particulate matter. There-
fore, even though particulate matter cannot be shown
to be the sole pollutant causing pollution-related ex-
cess mortality, it can be used as an indicator of the
pollutant mixture which appears to result in excess
mortality. This analysis estimates excess mortality (for
all criteria pollutants other than lead) using PM as an
indicator of the pollutant mix to which individuals
were exposed. This issue is discussed further below,
where details on estimating mortality effects are ex-
plored.
The one exception to the use of single pollutant
regression models is estimating hospital admissions.
Both PM and ozone are generally found to have a sta-
tistically significant and separate association with
hospital admissions. Using separate regressions (from
single pollutant models) for each pollutant may over-
state the number of effects caused by each pollutant
alone. On the other hand, using PM as a single indica-
tor of the pollutant mix could underestimate the total
hospital admissions caused by different mechanisms.
Separate PM and ozone coefficients for hospital ad-
missions are selected from regression models that
consider the effects of both pollutants simultaneously.
Uncertainty
The stated goal of the section 812 assessment is
to provide a comprehensive estimate of benefits of
the Clean Air Act. To achieve this goal, information
with very different levels of confidence must be used.
Benefit categories are not to be omitted simply be-
cause they are highly uncertain or controversial, but
those benefit categories that are reasonably well un-
derstood must be distinguished from those which are
more tentative.
The ideal approach to characterizing uncertainty
is to conduct a formal quantitative uncertainty analy-
sis. A common approach develops an estimated prob-
ability distribution for each component of the analy-
sis. A Monte Carlo procedure draws randomly from
each of these distributions to generate an estimate of
the result. Evaluating the result for many such ran-
dom combinations, creates a distribution of results that
reflects the joint uncertainties in the analysis.
The most serious obstacle to preparing a formal
quantitative uncertainty analysis is identifying all the
necessary distributions for each component of the
analysis. The Monte Carlo procedure requires that all
components of the model be rerun many times. How-
ever, the section 812 project links the outputs from
independent modeling activities. It would be imprac-
tical to simultaneously rerun the macroeconomic,
emissions, air quality, and exposure models because
of the diverse origins of the models. Therefore, in-
stead of a complete formal uncertainty analysis, the
section 812 assessment includes a less rigorous analy-
sis of the inherent uncertainties in the modeling ef-
fort. The uncertainty analysis combines quantitative
and qualitative elements designed to sufficiently de-
scribe the implications of the uncertainties. A primary
goal of the sensitivity/uncertainty analysis is to iden-
tify the health effects that make a sizable contribution
to the overall assessment of the monetary benefits.
There may be situations where there are significant
differences in the available information used to pre-
dict the incidence of a particular health effect (i.e.,
the uncertainty bounds are large). It is important to
alert the reader to situations where using the lower
incidence estimates may portray the health effect as
only modestly contributing to the overall total ben-
efits, but using reasonable alternative higher estimated
incidence figures (or higher monetized values) would
substantially impact not only the monetized value of
the individual health effect, but actually make a no-
ticeable difference in the total benefits assessment.
Consideration of the overall uncertainties inher-
ent in the section 812 assessment has several impor-
tant implications for health study selection. It was im-
portant to carefully examine the balance between the
level of uncertainties in the analysis and the need for
D-10
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
comprehensive coverage of all benefit categories.
There were frequently situations in which a direct
tradeoff existed between more comprehensive cover-
age and the restriction of the analysis to more certain
information. Also, the relationship between the un-
certainty in other parts of the analysis and the uncer-
tainty for each particular health effect was carefully
considered.
Magnitude of Exposure
One component of the section 812 analysis esti-
mates the air pollution levels that would have occurred
in the absence of the Clean Air Act. These estimates
are larger than currently observed levels of U.S. air
pollution, and perhaps even levels currently observed
elsewhere in the world. This aspect of the analysis
poses difficulties for the application of concentration-
response functions that have been based on exposures
at much lower pollution levels. The shape of the con-
centration-response function much above observed
exposures levels is unknown. It is possible that bio-
logical mechanisms affecting response that are unim-
portant at low levels of exposure may dominate the
form of response at higher levels, introducing
nonlinearity to the mathematical relationship. In gen-
eral, studies that include exposure levels spanning the
range of interest in the section 812 assessment are
preferable to studies at levels outside of the range, or
that only include a narrow part of the range. A pos-
sible drawback to this approach is that studies which
fit this criterion have often been conducted outside
the U.S. The application of foreign studies to U.S.
populations introduces additional uncertainties regard-
ing the representativeness of the exposed population
and the relative composition of the air pollution mix
for which the single pollutant is an indicator. These
difficult issues were considered in selecting studies
for the benefits analysis.
Duration of Exposure
Selection of health studies for the section 812 as-
sessment must consider the need to match the health
information to the air quality modeling conducted for
the assessment. For example, information on the health
effects from short term (five minute) exposure to sul-
fur dioxide cannot be readily combined with infor-
mation on average daily sulfur dioxide levels. In se-
lecting studies for the benefits analysis, preference was
shown for studies whose duration of exposure matched
one of the averaging times of the air quality data.
Thresholds
Exposure-response relationships are conceptual-
ized as either exhibiting a threshold of exposure be-
low which adverse effects are not expected to occur,
or as having no response threshold, where any expo-
sure level theoretically poses a non-zero risk of re-
sponse to at least one segment of the population. The
methods employed by health researchers to charac-
terize exposure-response relationships may or may not
explicitly analyze the data for the existence of a thresh-
old. Studies may analyze relationships between health
and air pollution without considering a threshold. If a
threshold for population risk exists but is not identi-
fied by researchers, then Clean Air Act benefits could
be overestimated if CAA levels are below the thresh-
old, because the risk reduction from the no-control
scenario could be overstated. On the other hand, if a
threshold is artificially imposed where one does not
exist, the relative benefits of the Clean Air Act may
be underestimated. In general, those studies that ex-
plicitly consider the question of a threshold (whether
a threshold is identified or not) provide stronger evi-
dence; consideration of this question is a positive fea-
ture when selecting studies for this analysis.
Target Population
Many of the studies relevant to quantifying the
benefits of air pollution reductions have focused on
specific sensitive subpopulations suspected to be most
susceptible to the effects of the pollutant. Some of
these effects may be relevant only for the studied sub-
population; effects on other individuals are either un-
known, or not expected to occur. For such studies, the
challenge of the analysis is to identify the size and
characteristics of the subpopulation and match its oc-
currence to exposure. Other studies have examined
specific cohorts who may be less susceptible than the
general population to health effects from air pollu-
tion (e.g., healthy workers), or who differ in age, gen-
der, race, ethnicity or other relevant characteristics
from the target population of the benefits analysis.
Extrapolating results from studies on nonrepresenta-
tive subpopulations to the general population intro-
duces uncertainties to the analysis, but the magnitude
of the uncertainty and its direction are often unknown.
Because of these uncertainties, benefit analyses often
limit the application of the dose-response functions
only to those subpopulations with the characteristics
of the study population. While this approach has merit
in minimizing uncertainty in the analysis, it can also
D-ll
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
severely underestimate benefits if, in fact, similar ef-
fects are likely to occur in other populations. For these
reasons, studies that examine broad, representative
populations are preferable to studies with narrower
scope because they allow application of the functions
to larger numbers of persons without introducing ad-
ditional uncertainty.
Many studies included in the section 812 analy-
sis focus on a particular age cohort of the population
for the identification of health effects. The choice of
age group is often a matter of convenience (e.g., ex-
tensive Medicare data may be available for the eld-
erly population) and not because the effects are, in
reality, restricted to the specific age group (even
though their incidence may vary considerably over
the life span). However, since no information is avail-
able about effects beyond the studied population, this
analysis applies the given concentration-response re-
lationships only to those age groups corresponding to
the cohorts studied. Likewise, some studies were per-
formed on individuals with specific occupations, ac-
tivity patterns, or medical conditions because these
traits relate to the likelihood of effect. In these cases,
application of dose-response functions has been re-
stricted to populations of individuals with these same
characteristics.
Statistical Significance of Exposure-Response
Relationships
The analysis includes as many studies related to a
given health effect as possible, except for studies in-
applicable to the current analysis. For some endpoints,
the group of adequate studies yielded mixed results,
with some showing statistically significant responses
to pollutant concentrations and others with insignifi-
cant associations. Unless study methods have been
judged inadequate, dose-response functions with both
statistically significant and insignificant coefficients
have been included to characterize the possible range
of risk estimates. Excluding studies exclusively on the
basis of significance could create an upward bias in
the estimates by not reflecting research that indicates
there is a small, or even zero, relationship between
pollution and specific health effects. It should be noted,
however, that some studies that found insignificant
effects for a pollutant could not be used because they
did not report the insignificant coefficient values.
In some cases, a single study reported results for
multiple analyses, yielding both significant and non-
significant results, depending on the nature of the in-
put parameters (e.g., for different lag periods or con-
current exposures). In these cases, only significant
results were included.
Relative Risks
Many studies reported only a relative risk value
(defined as the ratio of the incidence of disease in two
groups exposed to two different exposure levels). The
analysis required conversion of these values to their
corresponding regression coefficients when the coef-
ficients were not reported. When converting the rela-
tive risk to a coefficient value, the analysis used the
functional form of the regression equation reported
by the authors of the study.
The coefficients from a number of studies mea-
sured the change in the number of health effects for
the study population rather than a change per indi-
vidual. These coefficients were divided by the size of
the study population to obtain an estimate of change
per individual. The coefficient could then be multi-
plied by the size of the population modeled in the cur-
rent analysis to determine total incidence of health
effects.
Baseline Incidence Data
Certain dose-response functions (those expressed
as a change relative to baseline conditions) require
baseline incidence data associated with ambient lev-
els of pollutants. Incidence data necessary for the cal-
culation of risk and benefits were obtained from na-
tional sources whenever possible, because these data
are most applicable to a national assessment of ben-
efits. The National Center for Health Statistics pro-
vided much of the information on national incidence
rates. However, for some studies, the only available
incidence information come from the studies them-
selves; in these cases, incidence in the study popula-
tion is assumed to represent typical incidence nation-
ally.
Studies were excluded if health endpoints could
not be defined in the U.S. population. For example, in
Pope and Dockery (1992) the authors developed a
unique definition of symptomatic children in Utah
which has no correlation in the incidence data bases
which were available; consequently, the results could
not be applied to the general population.
D-12
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
Estimating Mortality Effects
Using PM as an Indicator
There is substantial evidence that exposure to cri-
teria pollutants, either individually or collectively, is
significantly associated with excess mortality. This
association is most closely and consistently related to
the ambient air concentrations of PM.
Several studies have found small but statistically
significant relationships between ozone and mortal-
ity, while other studies have not found a significant
relationship. There is inconclusive evidence whether
ozone has an effect independent of the effect of other
pollutants (e.g., PM or CO), has a synergistic effect
in combination with other effects, or is a confounder
in the relationship between mortality and other pol-
lutants. For example, in a recent study HEI (1996)
found a significant and relatively stable ozone coeffi-
cient for most of the model specifications presented
in the study. However, the measured ozone effect was
largest and most significant in the winter and autumn,
when ozone levels are low.
This analysis estimates excess mortality (for all
criteria pollutants other than lead) using PM as an in-
dicator of the pollutant mix to which individuals
were exposed. Even if particulate matter exposure
cannot be shown to be an independent causal factor
of excess mortality, it is, at a minimum, a good indi-
cator measure of the exposure to the pollutant mix-
ture that has been shown to be related to excess mor-
tality. Because PM is used as an indicator, the con-
centration-response functions from single pollutant
models (i.e., statistical models including PM as the
only pollutant) are preferred. To the extent that ozone
is correlated with PM, the effect of ozone, either as an
independent association or acting in combination with
other pollutants, will be captured by this approach.
Estimating the Relationship Between PM and
Premature Mortality
Long-term exposure versus short-term exposure
studies and the degree of prematurity of mortality.
Both long-term exposure (cohort) studies and short-
term exposure (longitudinal or time-series) studies
have estimated the relationship between exposure to
PM and premature mortality. While there are advan-
tages and disadvantages to each type of study (as dis-
cussed above), the long-term studies may capture more
of the PM-related premature mortality, as well as pre-
mature mortality that is more premature, than the
short-term studies.
The degree of prematurity of pollution-related
death may be an important uncertainty in the effort to
estimate the benefits of reducing pollution concentra-
tions, as discussed in Appendix I. The willingness to
pay to save a few days of life may be significantly
less than the willingness to pay to save a few, or many,
years of life. Evidence concerning the degree of pre-
maturity of pollution-related death would, in this case,
be crucial. Such evidence is, however, still scarce.
There is some limited evidence that the relative risk
of mortality from exposure to PM is higher for older
individuals than for younger individuals. This, com-
bined with the fact that the baseline incidence of mor-
tality consists disproportionately of people 65 and
over, suggests that PM-related mortality is dispropor-
tionately among older individuals. The extent to which
prematurity of death among older individuals is on
the order of days or weeks versus years, however, is
more uncertain. The short-term exposure studies can
provide little information on this. It is possible that
premature deaths on high pollution days would have
occurred only days later, if the individuals were sick
and therefore particularly susceptible. The fact that
the long-term exposure mortality studies found sub-
stantially larger relative risks, however, suggests that
not all of the premature mortality is on the order of
days or even weeks. Shortening of life of such a small
duration would not be detectable in a long-term epi-
demiology study, ensuring that the effects detected in
such studies must represent longer periods of life short-
ening. This suggests that at least some of the prema-
ture mortality associated with exposure to PM may
reduce lifespans by substantially longer amounts of
time.
Even if an individual's PM-related premature
mortality is of very short duration, on the order of
days, however, it may be misleading to characterize
such a PM-related loss as only those few days if the
individual's underlying susceptibility was itself ex-
acerbated by chronic exposure to elevated levels of
pollution. Suppose, for example, that long-term ex-
posure to elevated PM levels compromises the car-
diopulmonary system, making the individual more
susceptible to mortality on peak PM days than he oth-
erwise would have been. If this is the case, then the
underlying susceptibility would itself be either caused
by chronic exposure to elevated PM levels or exacer-
D-13
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
bated by it. Characterizing the individual's loss as a
few days could, in this case, be a substantial underes-
timate.
In addition, the long-term studies estimate sig-
nificantly more PM-related mortality than the annual
sum of the daily estimates from the short-term stud-
ies, suggesting that the short-term studies may be
missing a component of PM-related mortality that is
being observed in the long-term studies. For example,
if chronic exposure to elevated PM levels causes pre-
mature mortality that is not necessarily correlated with
daily PM peak levels, this type of mortality would be
detected in the long-term studies but not necessarily
in the short-term studies. Two of the long-term expo-
sure studies suggest, moreover, that the association
between ambient air pollution and mortality cannot
be explained by the confounding influences of smok-
ing and other personal risk factors.
Uncertainties surround analyses based on epide-
miological studies of PM and mortality. In addition
to the uncertainty about the degree of prematurity of
mortality, there are other uncertainties surrounding
estimates based on epidemiological studies of PM and
mortality. Although epidemiological studies are gen-
erally preferred to human clinical studies, there is
nevertheless uncertainty associated with estimates of
the risk of premature mortality (and morbidity) based
on studies in the epidemiological literature. Consid-
ering all the epidemiological studies of PM and mor-
tality, both short-term and long-term, there is signifi-
cant interstudy variability as well as intrastudy un-
certainty. Some of the difference among estimates
reported by different studies may reflect only sam-
pling error; some of the difference, however, may re-
flect actual differences in the concentration-response
relationship from one location to another. The trans-
ferability of a concentration-response function esti-
mated in one location to other locations is a notable
source of uncertainty.
Although there may be more uncertainty about
the degree of prematurity of mortality captured by
short-term exposure studies than by long-term expo-
sure studies, certain sources of uncertainty associated
with long-term exposure studies require mention. Al-
though studies that are well-executed attempt to con-
trol for those factors that may confound the results of
the study, there is always the possibility of insuffi-
cient or inappropriate adjustment for those factors that
affect long-term mortality rates and may be con-
founded with the factor of interest (e.g., PM concen-
trations). Prospective cohort studies have an advan-
tage over ecologic, or population-based, studies in that
they gather individual-specific information on such
important risk factors as smoking. It is always pos-
sible, however, that a relevant, individual-specific risk
factor may not have been controlled for or that some
factor that is not individual-specific (e.g., climate) was
not adequately controlled for. It is therefore possible
that differences in mortality rates that have been as-
cribed to differences in average PM levels may be
due, in part, to some other factor or factors (e.g., dif-
ferences among communities in diet, exercise,
ethnicity, climate, industrial effluents, etc.) that have
not been adequately controlled for.
Another source of uncertainty surrounding the
prospective cohort studies concerns possible histori-
cal trends in PM concentrations and the relevant pe-
riod of exposure, which is as yet unknown. TSP con-
centrations were substantially higher in many loca-
tions for several years prior to the cohort studies and
had declined substantially by the time these studies
were conducted. If this is also true for PM10 and or
PM^, it is possible that the larger PM10 and or PM25
coefficients reported by the long-term exposure stud-
ies (as opposed to the short-term exposure studies)
reflect an upward bias. If the relevant exposure pe-
riod extends over a decade or more, then a coefficient
based on PM concentrations at the beginning of the
study or in those years immediately prior to the study
could be biased upward if pollution levels had been
decreasing markedly for a decade or longer prior to
the study.
On the other hand, if a downward trend in PM
concentrations continued throughout the period of the
study, and if a much shorter exposure period is rel-
evant (e.g., contained within the study period itself),
then characterizing PM levels throughout the study
by those levels just prior to the study would tend to
bias the PM coefficient downward.
The relevant exposure period is one of a cluster
of characteristics of the mortality-PM relationship that
are as yet unknown and potentially important. It is
also unknown whether there is a time lag in the PM
effect. Finally, it is unknown whether there may be
cumulative effects of chronic exposure — that is,
whether the relative risk of mortality actually increases
as the period of exposure increases.
Estimating the relationship between PMandpre-
mature mortality. The incidence of PM-related mor-
tality used for estimating the benefits of the CAA is
D-14
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
based on the concentration-response relationship re-
ported by one of the two recent long-term exposure
(prospective cohort) studies (Pope et al., 1995, and
Dockery et al., 1993). Because it is based on a much
larger population and many more locations than
Dockery et al. (1993), the concentration-response
function from Pope et al. (1995) was used in this analy-
sis. The results of Pope et al. are consistent with those
of Dockery et al., which reported an even larger re-
sponse, but in only six cities. Moreover, Pope et al. is
also supported by several ecological cross-sectional
studies of annual mortality based on 1960 and 1970
census data (using either TSP or sulfate as indicators
of PM), including the work of Lave and Seskin (1977)
and Lipfert (1984).
Numerous short-term exposure (time series) stud-
ies have also reported a positive and statistically sig-
nificant relationship between PM and mortality. Of
the fourteen studies that estimated the relationship
between daily PM10 concentrations and daily mortal-
ity listed in Table 12-2 of the PM Criteria Document,
twelve reported positive and statistically significant
findings (Pope et al., 1992; Pope and Kalkstein, 1996;
Dockery et al., 1992; Schwartz, 1993a; Ozkaynak et
al., 1994; Kinney etal., 1995: Itoetal., 1995; Ostroet
al., 1996; Saldiva et al., 1995; Styer et al., 1995; Ito
and Thurston, 1996; Schwartz et al., 1996). While
these studies lend substantial support to the hypoth-
esis that there is a relationship between PM10 and
mortality, they may be capturing only the portion of
that relationship involving short-term effects. For this
reason, they are considered in this analysis only as
supporting evidence to the results of the study by Pope
et al.
The Pope et al. study has several further advan-
tages. The population followed in this study was
largely white and middle class, decreasing the likeli-
hood that interlocational differences in premature mor-
tality were due in part to differences in socioeconomic
status or related factors. In addition, the generally
lower mortality rates and possibly lower exposures to
pollution among this group, in comparison to poorer
minority populations, would tend to bias the PM co-
efficient from this study downward, counteracting a
possible upward bias associated with historical air
quality trends discussed above.
Another source of downward bias in the PM co-
efficient in Pope et al. is that intercity movement of
cohort members was not considered in this study.
Migration across study cities would result in expo-
sures of cohort members being more similar than
would be indicated by assigning city-specific annual
average pollution levels to each member of the co-
hort. The more intercity migration there is, the more
exposure will tend toward an intercity mean. If this is
ignored, differences in exposure levels, proxied by
differences in city-specific annual median PM levels,
will be exaggerated, resulting in a downward bias of
the PM coefficient (because a given difference in mor-
tality rates is being associated with a larger differ-
ence in PM levels than is actually the case).
In summary, because long-term exposure studies
appear to have captured more of the PM-related pre-
mature mortality, as well as premature mortality that
is more premature, they are preferable to the short-
term exposure studies. Among the long-term expo-
sure studies, the Pope et al. study has several advan-
tages, as discussed above, which are likely to reduce
the possibility of a key source of confounding and
increase the reliability of the concentration-response
function from that study. For these reasons, the con-
centration-response function estimated in this study
is considered the most reasonable choice for this analy-
sis.
Matching PM Indices in the Air Quality Profiles
and Concentration-Response Function. The Pope et
al. study examined the health effects associated with
two indices of PM exposure: sulfate particles and fine
particles (PM25). The reported mortality risk ratios are
slightly larger for PM2 5 than for sulfates (1.17 versus
1.15 for a comparison between the most polluted and
least polluted cities). The PM2 5 relationship is used in
this analysis because it is more consistent with the
PM10 air quality data selected for the analysis. Esti-
mated changes in PM25 air quality must be matched
with the PM2 _ mortality relationship. However, only
PM10 profiles were used for the entire 20 year period.
Therefore, the same regional information about the
PM10 components (sulfate, nitrate, organic particulate
and primary particulate) used to develop the PM10 pro-
files were used to develop regional PM2 5/PM10 ratios.
Although both urban and rural ratios are available,
for computational simplicity, only the regional urban
ratios were used to estimate the PM25 profiles from
the PM10 profiles used in the analysis. This reflects
the exposure of the majority of the modeled popula-
tion (i.e., the urban population), while introducing
some error in the exposure changes for the rural popu-
lation. In the east and west, where the rural ratio is
larger than the urban ratio, the change in PM2 5 expo-
sure will be underestimated for the rural population.
D-15
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
In the central region the PM25 change will be overes-
timated. These ratios were used in each year during
1970-1990, introducing another source of uncertainty
in the analysis. Table D-5 summarizes the PM25/PM1Q
ratios used in this analysis.
Tabic D-5. PM2VPM10 Ratios Used to Estimate
PM2.S Data Used With Pope ct al. (1995)
Mortality Relationship.
East
Central
West National
Urban
0.59
0.58
0.48 0.55
Rural
0.68
0.53
0.49 0.57
Prematurity of Mortality: Life-Years Lost as a Unit
of Measure
Perhaps the most important health effect that is
examined in this analysis is mortality. Although this
analysis does not take into account the degree of pre-
maturity of death (that is, the ages of those individu-
als who die prematurely from exposure to PM are not
considered), considerable attention has been paid to
this issue and, in particular, to life-years lost as an
alternative to lives lost as a measure of the mortality-
related effects of pollution.
Because life-years lost is of potential interest and
because there is a substantial potential for confusion
in understanding apparently disparate estimates of life-
years lost from pollution exposure, this section at-
tempts to present a clear discussion of the various
possible measures of life-years lost, what they depend
on, and how they are related to each other.
Because the actual number of years any particu-
lar individual is going to live cannot be known, "life-
years lost" by an individual actually refers to an ex-
pected loss of years of life by that individual. The
expected loss of years of life by an individual depends
crucially on whether the expectation is contingent on
the individual only having been exposed to PM or on
the individual actually having died from that expo-
sure.
An ex ante estimate of life-years lost per indi-
vidual is contingent not on the individual having died
prematurely but only on the individual having been
exposed. Suppose, for example, that a 25 year old has
a life expectancy of 50 more years in the absence of
exposure and only 49 more years in the presence of
exposure. Given (chronic) exposure from the age of
25 on, the 25 year old exposed to (some elevated level
of) PM might expect a shortening of life expectancy
of one year, for example. That is one expected life-
year lost due to chronic exposure. This is the life-years
lost that can be expected by every exposed individual.
An ex post estimate of life-years lost per individual
is contingent on the individual actually having died
from exposure to PM. When an individual dies of
exposure to PM, he is said to have lost the number of
years he would have been expected to live, calculated,
for example, from age- and gender-specific life ex-
pectancy tables. Suppose that the life expectancy of
25 year olds is 75 — that is, a 25 year old can expect
to live 50 more years. A 25 year old who dies from
exposure to PM has therefore lost 50 expected years
of life. This is the life-years lost that can be expected
by every 25 year old affected individual (i.e., every
25 year old who actually dies from exposure to PM).
Estimates of the total life-years lost by a popula-
tion exposed to PM depend on several factors, includ-
ing the age distribution and the size of the exposed
population, the magnitude of the change (or changes)
in PM being considered, the relative risk assumed to
be associated with each change in PM, and the length
of time exposure (i.e., the change in PM) is presumed
to occur. A population chronically exposed to a given
increase in PM will lose more life-years than a popu-
lation exposed to the same increase in PM for only a
year or two.6 A population that is generally older will
lose fewer life-years, all else equal, than one that is
generally younger, because older individuals have
fewer (expected) years of life leftto lose. And apopu-
lation exposed to a greater increase in PM will lose
more life-years than if it were exposed to a smaller
increase in PM. Finally, the life-years lost by the popu-
lation will increase as the relative risk associated with
the increase in PM increases.
Life-years lost are usually reported as averages
over a population of individuals. The population be-
ing averaged over, however, can make a crucial dif-
6 Even in the absence of cumulative effects of exposure, exposure of a population for many years will result in a greater total
number of pollution-related deaths than exposure for only a year or two, because the same relative risk is applied repeatedly, year
after year, to the population, rather than for only a year or two.
D-16
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
ference in the reported average life-years lost, as noted
above. The average life-years lost per exposed indi-
vidual (the ex ante estimate) is just the total life-years
lost by the population of exposed individuals divided
by the number of exposed individuals. This average
will depend on all the factors that the total life-years
lost depends on except the size of the exposed popu-
lation. The average life-years lost by an exposed indi-
vidual is a statistical expectation. It is the average of
the numbers of life-years actually lost by each mem-
ber of the exposed population. Alternatively, it can be
thought of as a weighted average of possible numbers
of years lost, where the weights are the proportions of
the population that lose each number of expected years
of life. Although those individuals who do die prema-
turely from exposure to PM may lose several expected
years of life, most exposed individuals do not actu-
ally die from exposure to PM and therefore lose zero
life-years. The average life-years lost per exposed in-
dividual in a population, alternatively referred to as
the average decrease in life expectancy of the exposed
population, is therefore heavily weighted towards zero.
The average number of life-years lost per individual
who dies of exposure to PM (the ex post measure of
life-years lost) is an average of the numbers of ex-
pected years of life lost by individuals who actually
died prematurely because of PM. Because everyone
who dies prematurely from exposure to PM loses some
positive number of expected years of life, this aver-
age, by definition, does not include zero.
An example of an ex ante measure of life-years
lost is given by a study in the Netherlands (WHO,
1996), which considered a cohort of Dutch males, aged
25-30, and compared the life expectancy of this co-
hort to what it would be in a hypothetical alternative
scenario in which these individuals are continuously
exposed to concentrations of PM25 that are 10 |_ig/m3
lower than in the actual scenario. The life expectancy
of this cohort of 25-30 year old Dutch males was cal-
culated to be 50.21 years in the actual scenario, based
on a 1992 life table from the Netherlands. Assuming
that the relative risk of mortality associated with an
increase of 10 pg/m3 PM2 5 is 1.1 (the average of the
relative risks of 1.14 from Dockery et al., 1993, and
1.07 from Pope et al., 1995), the study authors calcu-
lated death rates in the hypothetical "cleaner" scenario
by dividing the age-specific death rates in the actual
scenario by 1.1. Using these slightly lower death rates,
and assuming that the effect of PM does not begin
until 15 years of exposure, the authors constructed a
life table for the cohort in the hypothetical "cleaner"
scenario. Based on this new life table in a cleaner
world, the life expectancy of the cohort of 25-30 year
old Dutch males was calculated to be 51.32 years in
the hypothetical cleaner scenario. (In calculating life
expectancies in both the "dirty" scenario and the
"clean" scenario, it is assumed that any individual who
does not survive to the next 5-year age group lives
zero more years. For example, a 30 year old individual
either survives to age 35 or dies at age 30.) The change
in life expectancy for this cohort of 25-30 year old
Dutch males, due to a change in PM exposure of 10
(ig/m3 for the rest of their lives (until the age of 90),
was therefore 51.32 years - 50.21 years = 1.11 years.
That is, the average life-years lost by an exposed in-
dividual in this population, under these assumptions,
is 1.11 years.
The estimate of 1.11 years of expected life lost
depends on several things, as mentioned above. If the
study authors had used the relative risk from Pope et
al., 1995, alone, (1.07 instead of 1.1), for example,
the change in life expectancy (the ex ante measure of
life-years lost) for this cohort of25-30 year old Dutch
males would have been 0.80 years. Similarly, chang-
ing the assumption about the duration of exposure also
changes the estimate of ex ante life-years lost. Using
a relative risk of 1.1, but assuming that exposure lasts
only during the first 5 years (i.e., that the death rate in
the first five years, from age 25 through age 30, is
lower but that after that it is the same as in the "dirty"
scenario), the average life-years lost by an exposed
individual in this population is reduced from 1.11 years
to 0.02 years.
By their construction and definitions, the average
life-years lost per exposed individual and the average
life-years lost per affected individual (i.e., per indi-
vidual who dies prematurely from PM) take the same
total number of life-years lost by the exposed popula-
tion and divide them by different denominators. The
average life-years lost per exposed individual divides
the total life-years lost by the total population exposed;
the average life-years lost per affected individual di-
vides the same total life-years lost by only a small
subset of the total population exposed, namely, those
who died from PM. The average per exposed indi-
vidual is therefore much smaller than the average per
affected individual. Because both types of average may
be reported, and both are valid measurements, it is
important to understand that, although the numbers
will be very dissimilar, they are consistent with each
other and are simply different measures of the esti-
mated mortality impact of PM.
D-17
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
To calculate the total (estimated) life-years lost
by a population, it is necessary to follow each age
cohort in the population through their lives in both
scenarios, the "dirty" scenario and the "clean" sce-
nario, and compute the difference in total years lived
between the two scenarios, as WHO (1996) did for
the cohort of Dutch males 25-30 years old. This
method will be referred to as Method 1. In practice,
however, it is not always possible to do this. (Other
changes to the population, such as those from recruit-
ment and immigration, for example, would make such
an exercise difficult.) An alternative method, which
approximates this, is to predict the numbers of indi-
viduals in each age category who will die prematurely
from exposure to PM (i.e., who will die prematurely
in the "dirty" scenario), and multiply each of these
numbers by the corresponding expected number of
years remaining to individuals in that age category,
determined from life expectancy tables. This method
will be referred to as Method 2. Suppose, for example,
that individuals age 25 are expected to live to age 75,
or alternatively, have an expected 50 years of life re-
maining. Suppose that ten 25 year olds are estimated
to die prematurely because of exposure to PM. Their
expected loss of life-years is therefore 50 years each,
or a total of 500 life-years. If the same calculation is
carried out for the individuals dying prematurely in
each age category, the sum is an estimate of the total
life-years lost by the population.
Using Method 1 (and retaining the assumptions
made by WHO, 1996), the average life-years lost per
PM-related death among the cohort of Dutch males is
calculated to be 14.28 years. Using Method 2 it is es-
timated to be 14.43 years.
Although this ex post measure of life-years lost is
much larger than the ex ante measure (1.11 life-years
lost per exposed individual), it only applies to those
individuals who actually die from exposure to PM.
The number of individuals in the age 25-30 Dutch
cohort example who eventually die from exposure to
PM (7,646) is much smaller than the number of indi-
viduals in the age 25-30 Dutch cohort who are ex-
posed to PM (98,177). The total life-years lost can be
calculated either as the number of exposed individu-
als times the expected life-years lost per exposed in-
dividual (98,177* 1.11 = 109,192.1) or as the number
of affected individuals times the expected life-years
lost per affected individual (7,646* 14.28 = 109,192.1).
To further illustrate the different measures of life-
years lost and the effects of various input assump-
tions on these measures, death rates from the 1992
U.S. Statistical Abstract were used to follow a cohort
of 100,000 U.S. males from birth to age 90 in a "dirty"
scenario and a "clean" scenario, under various assump-
tions. Death rates were available for age less than 1,
ages 1-4, and for ten-year age groups thereafter. The
ten-year age groups were divided into five-year age
groups, applying the death rate for the ten-year group
to each of the corresponding five-year age groups. Ex
ante and ex post measures of life-years lost among
those individuals who survive to the 25-29 year old
category were first calculated under the assumptions
in the WHO (1996) study. These assumptions were
that the relative risk of mortality in the "dirty" sce-
nario versus the "clean" scenario is 1.1; that exposure
does not begin until age 25; that the effect of expo-
sure takes fifteen years; that individuals at the begin-
ning of each age grouping either survive to the next
age grouping or live zero more years; and that all in-
dividuals age 85 live exactly five more years. Under
these assumptions, the expected life-years lost per
exposed individual in the 25-29 year old cohort is 1.32
years. There are 96,947 exposed individuals in this
age cohort. The expected life-years lost per affected
individual (i.e., per PM-related death) is 16.44 years
(Method 1). There are 7,804 affected individuals. The
total life-years lost by individuals in this cohort is
128,329.3 (1.32*96,947 = 16.44* 7,804 = 128,329.3).
If the relative risk is changed to 1.07, the expected
life-years lost per exposed individual in the cohort of
25-29 year old U.S. males is reduced from 1.32 to
0.95 years. The expected life-years lost per affected
individual (i.e., per PM-related death) is 16.44 years
(Method 1). Using a relative risk of 1.1 but assuming
no lag (i.e., assuming that exposure starts either at
birth or at age 25 and has an immediate effect), the
expected life-years lost per exposed individual in the
25-29 year old cohort changes from 1.32 to 1.12. The
expected life-years lost per affected individual (i.e.,
per PM-related death) becomes 19.7 years (Method
1).
D-18
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
Estimating Morbidity Effects
In addition to mortality effects, this analysis quan-
tifies effects for a number of non-fatal health end-
points. Several issues arise in implementing the stud-
ies selected for this analysis.
Overlapping Health Effects
Several endpoints reported in the health effects
literature overlap with each other. For example, the
literature reports relationships for hospital admissions
for single respiratory ailments (e.g. pneumonia or
chronic obstructive pulmonary disease) as well as for
all respiratory ailments combined. Similarly, several
studies quantify the occurrence of respiratory symp-
toms where the definitions of symptoms are not unique
(e.g., shortness of breath, upper respiratory symptoms,
and any of 19 symptoms). Measures of restricted ac-
tivity provide a final example of overlapping health
endpoints. Estimates are available for pollution-in-
duced restricted activity days, mild restricted activity
days, activity restriction resulting in work loss. This
analysis models incidence for all endpoints. Double-
counting of benefits is avoided in aggregating eco-
nomic benefits across overlapping endpoints (see
Appendix I).
Studies Requiring Adjustments
Applying concentration-response relationships
reported in the epidemiological literature to the na-
tional scale benefits analysis required by section 812
required a variety of adjustments.
Normalization of coefficients by population. To
be applied nationwide, concentration-response coef-
ficients must reflect the change in risk per person per
unit change in air quality. However, some studies re-
port the concentration-response coefficient, , as the
change in risk for the entire studied population. For
example, Thurston et al. (1994) reported the total num-
ber of respiratory-related hospital admissions/day in
the Toronto, Canada area. To normalize the coeffi-
cient so that it might be applied universally across the
country, it was divided by the population in the geo-
graphical area of study (yielding an estimate of the
change in admissions/person/day due to a change in
pollutant levels).
Within-study meta-analysis. In some cases, stud-
ies reported several estimates of the concentration-
response coefficient, each corresponding to a particu-
lar year or particular study area. For example, Ostro
and Rothschild (1989) report six separate regression
coefficients that correspond to regression models run
for six separate years. This analysis combined the in-
dividual estimates using a fixed coefficient meta-
analysis on the six years of data.
Conversion of coefficients dependent on symptom
status during the previous day. Krupnick et al. (1990)
employed a Markov process to determine the prob-
ability of symptoms that were dependent on symp-
tom status of the previous day. The current analysis
adjusts the regression coefficients produced by the
model in order to eliminate this dependence on previ-
ous day's symptom status.
Concentration-Response Functions:
Health Effects
After selecting studies appropriate for the section
812 analysis, taking into account the considerations
discussed above, the published information was used
to derive a concentration-response function for esti-
mating nationwide benefits for each health effect con-
sidered. In general, these functions combine air qual-
ity changes, the affected population and information
regarding the expected per person change in incidence
per unit change in pollutant level. The following tables
present the functions used in this analysis, incorpo-
rating information needed to apply these functions and
references for information.
Particulate Matter
The concentration-response functions used to
quantify expected changes in health effects associ-
ated with reduced exposure to particulate matter are
summarized in Table D-6. The data profiles selected
for use in this analysis are PM10. In those cases in
which PM10 was not the measure used in a study, this
analysis either converted PM10 air quality data to the
appropriate air quality data (e.g., PM25 or TSP) or,
equivalently, converted the pollutant coefficient from
the study to the corresponding PM10 coefficient, based
on location-specific information whenever possible.
D-19
-------
Tabic D-6. Summary of Concentration-Response Functions for Particulate Matter.
Except where noted otherwise, the functional form is
A cast's cast's * (e^V'A/~'(I - 1)
where "cases" refers to incidence at the first pollution level.
Health
Endpoint
(ICD-9 code)
Baseline
Incidence (per
100,000)
Expos Meas.
from Original
Study
Study Pop.
Applied
Pop.
Functional form3
Uncert & Var.
Sources
mortality
(long-term
exposure)
non-accidental
deaths by
countyb
annual median
pm2,5
50 cities,
all deaths
over age
30
Ppm2.5 = 0.006408
PMI0 data converted to PM2 5 datac
s.e. = 0.00148
Pope et al., 1995
American Cancer
Society cohort
hospital
admissions-
all resp.
illnesses (ICD
460-519)
504d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
same day PM10
65 and older
in New
Haven, CT,
Tacoma, WA
65 and
older
New Haven: 0.00172
Tacoma: 0.00227
average: 0.0020
c.i. = New
Haven: 1.00-1.12
s.e. = 0.00093
Tacoma: 0.97-
1.29
s.e. = 0.00146
Schwartz, 1995
New Haven and
Tacoma
hospital
admissions —
all resp.
illnesses (ICD
460-519)
n/a
mean monthly
PM10
variety of
ages in Salt
Lake Valley,
Utah
all
A cases = p * APM10 * Pop.
where P = 0.8047 monthly admissions / Salt Lake Valley
population (780,000).
= 3.4 x 10"8
(converted from monthly to daily admissions)
s.e. = 0.28
Pope, 1991
Salt Lake Valley
daily
respiratory
admissions
(total)
includes 466,
480,481,482,
485,490, 491,
492, 493
n/a
same-day PM10
Toronto
metro area
all
A cases = P * A PMj0 * Pop
where P = 0.0339 daily admissions / Toronto population
(2.4 million)
= 1.4 x 10"8
(model also includes 03)
s.e. = 0.034/2.4
million
= 1.4 x lO"8
Thurston et al.
1994
Toronto
hospital
admissions
pneumonia
(480-487)
229d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
same day PM10
over 65,
Birmingham
AL
over 65
P = 0.00174
c.i. = 1.07-1.32
s.e. = 0.000536
Schwartz, 1994a
Birmingham
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence (per
100,000)
Expos Meas.
from Original
Study
Study Pop.
Applied
Pop.
Functional form"
Uncert & Var.
Sources
hospital
admissions
COPD
(490-496)
103d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
same day PM10
over 65,
Birmingham
AL
over 65
p = 0.00239
c.i. = 1.08-1.50
s.e. = 0.00084
Schwartz, 1994a
Birmingham
hospital
admissions
pneumonia
(480-487)
229d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
same day
PM10
over 65,
Detroit
over 65
P = 0.00115
s.e. = 0.00039
Schwartz, 1994b
Detroit
hospital
admissions
COPD
(490-496)
103d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
same day
PM10
over 65,
Detroit
over 65
p = 0.00202
s.e. =
0.00059
Schwartz, 1994b
Detroit
hospital
admissions
pneumonia
(480-487)
229d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
same day PM10
65 and over
in Mpls
over 65
p = 0.00157
c.i.= 1.02- 1.33
s.e. = 0.00068
Schwartz, 1994c
Mpls, St. Paul
hospital
admissions
COPD
(490-496)
103d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
current and
previous day
PM.o
65 and over
in Mpls
over 65
P = 0.00451
c.i. = 1.20-2.06
s.e. = 0.00138
Schwartz, 1994c
Mpls, St. Paul
hospital
admissions for
congestive
heart failure
(ICD 428)
231d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
avg same and
previous day
PM,o
65 and older
in Detroit
65 and
older
P = 0.00098
c.i. = 1.012-1.052
s.e. = 0.00031
Schwartz and
Morris, 1995
Detroit
hospital
admissions for
ischemic heart
disease (ICD
410-414)
450d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
24 hr avg PM10
same day
65 and older
in Detroit
65 and
older
P = 0.00056
c.i.= 1.005-1.032
s.e. = 0.00021
Schwartz and
Morris, 1995
Detroit
^3
to
£
3
Si
s
Sr
a
s
o
Q)
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence (per
100,000)
Expos Meas.
from Original
Study
Study Pop.
Applied
Pop.
Functional form"
Uncert & Var.
Sources
hospital
admissions —
all resp.
illnesses (ICD
460-519)
504d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
24 hr avg PM10
over 65,
Spokane
over 65
P = 0.00163
s.e. = 0.00047
Schwartz, 1996,
Spokane
hospital
admissions
COPD
(490-496)
103d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
24 hr avg PM10
over 65,
Spokane
over 65
P = 0.00316
s.e. = 0.00084
Schwartz, 1996,
Spokane
hospital
admissions
pneumonia
(480-487)
229d/year
(incidence in
pop. > 65 years
of total U.S.
pop.)
24 hr avg PM10
over 65,
Spokane
over 65
P = 0.00103
s.e. = 0.00068
Schwartz, 1996,
Spokane
LRS defined
as cough,
chest pain,
phlegm, and
wheeze
not applicable
same day PM10
8-12 yr olds
0-12 yr
olds
Acases =
^ .A*
(1 -Pf() ¦ »
where P0 = the probability of a child in the study pop
suffering from LRS in the base case = 1.45 %
and p = 0.014176
s.e. = 0.0041
Schwartz et al.,
1994d
shortness of
breath, days
not applicable
24 hour avg
PM10
African-
American
asthmatics
between ages
7 and 12
same as
study
pop-
Acases =
(l-Po) * ep,A™70
where P0 = the probability of a child in the study pop.
suffering from shortness of breath in the base case = 5.6
%
and P = 0.008412
s.e. = 0.00363
Ostro et al., 1995
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence (per
100,000)
Expos Meas.
from Original
Study
Study Pop.
Applied
Pop.
Functional form"
Uncert & Var.
Sources
URI
defined as
ranny or
stuffy nose,
wet cough,
burning,
aching, or red
eyes
l,192e
(ages 10-12)
5,307e
(ages <= 12)
same day PM10
10-12 yr old
non-
symptomatic
12 and
under
P = 0.0036
s.e. = 0.0015
Pope et al., 1991
Utah
acute
bronchitis
(ICD 466)
n/a
PM10 annual avg
(converted)
10 to 12 year
olds
18 and
under
P = 0.0330
Acases = *Pop
l_VPo(e(5-A™-,°)
P0 = baseline probability of having bronchitis
=0.065c
s.e. = 0.0216
Dockery et al.,
1989
6 cities
chronic
bronchitis
710/year
(of study pop.)
annual mean
TSP
Seventh Day
Adventists in
California
all
P = 0.00512
convert PM10 to TSP:
Apmio
A TSP~
0.56
where 0.56 is the specific conversion based on region
and initial TSP conc.
not available
Abbey et al., 1993
^3
to
£
3
Si
s
Sr
a
s
o
Q)
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence (per
100,000)
Expos Meas.
from Original
Study
Study Pop.
Applied
Pop.
Functional form*
Uncert & Var.
Sources
chronic
bronchitis
600/year
annual mean
TSP
adults 30-74
years old in
53 U.S.
urban areas
all
A cases/year = (p, - p0) * Pop
where
1^^!:
"On—— ~ $*hPM )
: a ;: :;i>
1 + e 0
where p0 = 0.006 = the probability of developing
physician-diagnosed chronic bronchitis per individual
per year and
P = 0.0012, the PM10 coefficient, converted from the TSP
coefficient, using the relationship:
APM10
A TSP -
0.56
where 0.56 is the specific conversion based on region
and initial TSP conc.
95% CI = (1.02-
1.12) for odds
ratio
corresponding to
a 10 ng/m3
increase in annual
TSP
Schwartz, 1993b
presence of
any of 19
acute
respiratory
symptoms
not applicable
24 hour average
COH in units/
100 ft)8
COH = coeff. of
haze
adult
members of
families of
elementary
school-aged
children in
Glendora-
Covina-
Azusa, CA
adults
18-65
A Symptdays/day = (p, - p0) * Pop
where
1
= ,
"(111";—-— + P"A03)
1 + e P0
and
p0 = the probability of Symptdays per individual for a 24-
hour period in the base case
= 0.19
P = 0.00046s
(Model includes 03, COH, S02)
s.e. = 0.00024"
Krupnick et al.,
1990
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence (per
100,000)
Expos Meas.
from Original
Study
Study Pop.
Applied
Pop.
Functional form"
Uncert & Var.
Sources
moderate or
worse asthma
status
n/a
average PM2 5
during 9:00 am
to 4:00 pm
(|ig/m3)
Denver
asthmatics
between ages
18 and 70
asthmatic
(4%' of
total
pop.)
A asthma status= P[ln(X,/X0)]*Pop
where
X0 = PM10 concentrations with CAA,
X, = PM10 concentrations without CAA, and
P = 0.00038j
(model includes PM2 5 and modeled PM2 5 measures for
periods where PM2 5 measures were missing)
s.e. = 0.00019
Ostro et al., 1991
Denver
Restricted
Activity Days
(RADs)
400,531
days/year11 (of
the total U.S.
pop)
2-wk average
PM2.5 (ng/m3)
All adults
18-65 in
metropolitan
areas in the
U.S.
adults
aged 18-
65
A health effects determined over a 2 wk period
p = 0.0030>'1
s.e. = 0.00018'
Ostro, 1987
respiratory
and
nonrespiratory
conditions
resulting in a
minor
restricted
activity day
(MRAD)
780,000
days/year (cited
as 7.68 days
per person per
year in study)
PM2 5 averaged
over a 2-wk
period
employed
adults across
the U.S.
between the
ages of 18-65
adults
aged 18-
65
number of health effects determined over a 2-week
period
P = 0.00463*1
(Model includes fine particulates and 03)
s.e. = 0.000441
Ostro and
Rothschild, 1989
respiratory
restricted
activity days
(RRADs)
306,000
days/year (cited
as 3.06 days
per person per
year in study)
PM2 5 averaged
over a 2 wk
period
employed
adults across
the U.S.
between the
ages of 18-65
adults
aged 18-
65
number of health effects determined over a 2-wk period
p = 0.00936*1
(Model includes fine particulates and 03)
s.e. = 0.00103'
Ostro and
Rothschild, 1989
Work Loss
Days (WLDs)
150,750'" (of
total U.S. pop)
2-wk average
PM2, (ng/m3)
All adults
18-65 in
metropolitan
areas in the
U.S.
adults
aged 18-
65
A health effects determined over a 2 wk period
P = 0.0029''1
s.e. = 0.00022'
Ostro, 1987
^3
to
£
3
Si
s
Sr
a
s
o
Q)
-------
NOTES:
11 Pollutant coefficients reflect changes in health effects per change in (ig/m3 PM
b Mortality baseline incidence data for each county taken from Vital Statistics of the United States, Vol. II - Mortality, Part B, (U.S. Dept. of Health and Human Services). Incidence
rates were generated for total mortality excluding accidental deaths and adverse effects, suicide, homicide, and other external causes (ICD E800-E999). Rates calculated based on 1990
population.
c PMj0 data converted to PM,5 data by using national urban average PM, 7PM10 ratio = 0.56.
d Centers for Disease Control, 1992. Vital and Health Statistics, Detailed Diagnoses and Procedures, National Hospital Discharge Survey, 1990. Number of 1990 discharges divided by
1990 U.S. population (248,709,873) from City and County Databook, 12th edition, 1994, U.S. Dept. of Commerce, Bureau of the Census, Washington, D.C.
e Pope et al., 1991 NOTE: rates were not available from standard incidence sources and so were calculated from incidence in the study of 10-12 year olds. This may not be entirely
appropriate for older or younger individuals. Children of this age are less likely to have colds than much younger children and may be more representative of the adult population.
f Dockery et al., 1989.
8 Coefficient and standard error are converted from a (3 and s.e. for coefficient of haze (COH) to a (3 and s.e. for PM This was done by using a ratio of COH to TSP of 0.116 from the
study authors (as cited in ESEERCO, 1994) and a ratio of PM10 to TSP of 0.55 (U.S. EPA, 1986).
h Coefficient and standard error incorporate the stationary probabilities as described in Krupnick et al. (1990). To do this, the calculation used the transitional probabilities supplied by
the authors and presented in ESEERCO, 1994.
' U.S. EPA, 1994a.
' B converted from a change in health effects per change in ug/m3 PM,, to a change per ug/m3 PMln using the following relationship: 1 ug/m3 PM, =0.56 ug/m3 PMln (ESEERCO,
1994)
k Number of RADs for all acute conditions from: National Center for Health Statistics. Current Estimates from the National Health Interview Survey: United States, 1990. (Hyattsville,
MD). This number is divided by the U.S. population for 50 states for 1990 (248,709,873) and multiplied by 100,000 (to obtain the incidence per 100,000).
1 Based on fixed-weight meta-analysis of single-year coefficients and standard errors reported in study.
m Number of WLDs of 374,933,000 from: National Center for Health Statistics. Current Estimates from the National Health Interview Survey from 1990. (Hyattsville, MD). Series 10,
No. 181. This number is divided by the U.S. population for 50 states for 1990 (248,709,873) and multiplied by 100,000 (to obtain the incidence per 100,000).
K>
On
n 3 o « 3 S" § ° g
g-^o gt-g < s |-« g| ^ g
O- 8 £T 3 O & 2 acrSB:?
B" S «
^ 3 5 • 8 ^ ^ ^ $
2 S g. a (1 n n
g » " ^ g g I I I §: I % % z
h <-+ ^ p P p Q- 5 • o
O w ^ Oft H rt <1- N c
» s I ag p-o aQ o g
g&-5f°h2o3-§'g.c'1£.fCS'
a" g g.g,
§5 o 3 3 3 3 P s
£3 < Ci- c/3 ST o p ^ r-t- ^
c« ^ ei, n- .«' a o o o
. P O c/3 . i i
-------
Tabic D-7. Summary of Concentration-Response Functions for Ozone.
Except where noted otherwise, the functional form is
Acascs cases * (c v'r" - 1)
where "cases" refers to incidence at the first pollution level.
Health
Endpoint
(ICD-9 code)
Baseline
Incidence
(per
100,000)
Expos Meas
from original
study
Study Pop.
Applied Pop.
Functional form*
Uncert & Var.
Sources
hospital
admissions —
all resp.
illnesses (ICD
460-519)
504/yearb
(incidence
in pop .> 65
years of
total U.S.
pop.)
24 hr avg
(Hg/m3)
65 and older in
New Haven,
CT, Tacoma,
WA
over 65 only
P =
New Haven: 0.0027
Tacoma: 0.007
where
1 ng/m3= 0.510 ppb
(two pollutant model with PM10 and 03)
New Haven:
s.e. = 0.0014
Tacoma:
s.e. = 0.0025
where
lug/m3 = 0.51
PPb
Schwartz, 1995
New Haven and
Tacoma
daily
respiratory
admissions-
includes 466,
480, 481,482,
485,490, 491,
492,493
n/a
1 hour daily max
ozone (ppb)
all
all
for Toronto: P = 0.0388/2.4 million
=1.62xl0"8
A cases/day = (3 * A 03 * pop
(ozone and PM10 model used)
se = 0.0241/2.4
million
= 1.0xl0"8
Thurston et al.,
1994
Toronto
hospital
admissions
pneumonia
(480-487)
229/yearb
(incidence
in pop > 65
years of
total U.S.
pop.)
24-hr avg ppb
over 65,
Birmingham
AL
over 65
P = 0.00262
for 03 alone (single pollutant model only avail.)
s.e. = 0.00196
Schwartz, 1994a
Birmingham
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence
(per
100,000)
Expos Meas
from original
study
Study Pop.
Applied Pop.
Functional form*
Uncert & Var.
Sources
hospital
admissions
COPD
(490-496)
103/yeai*
(incidence
in pop .> 65
years of
total U.S.
pop.)
24-hr avg ppb
over 65,
Birmingham
AL
over 65
P = 0.00314
for 03 only (only single pollutant model avail.)
s.e. = 0.00316
Schwartz, 1994a
Birmingham
hospital
admissions
pneumonia
(480-487)
229/7631*
(incidence
in pop .> 65
years of
total U.S.
pop.)
24-hr avg ppb
over 65, Detroit
over 65
P = 0.00521
(two pollutant model with 03 and PM10)
note: authors suggest a threshold of 25 ppb
s.e. = 0.0013
Schwartz, 1994b
Detroit
hospital
admissions
COPD
(490-496)
103/year11
(incidence
in pop > 65
years of
total U.S.
pop.)
24-hr avg ppb
over 65, Detroit
over 65
P = 0.00549
(two pollutant model with 03 and PM10)
note: authors suggest a threshold of 25 ppb
s.e. = 0.00205
Schwartz, 1994b
Detroit
hospital
admissions
pneumonia
(ICD 480-487)
229/year"
(incidence
in pop .> 65
years of
total U.S.
pop.)
24 hr avg ppb
65 and over in
Mpls
over 65
P = 0.002795
(two pollutant model with 03 and PM10)
s.e. = 0.00172
Schwartz 1994c
Mpls, St. Paul
hospital
admissions —
all resp.
illnesses (ICD
460-519)
504/year11
(incidence
in pop .> 65
years of
total U.S.
pop.)
1 hour daily max
ozone(ppb)
over 65,
Spokane
over 65
P = 0.008562
s.e. = 0.004326
Schwartz, 1996,
Spokane
hospital
admissions
COPD
(490-496)
103/year11
(incidence
in pop .> 65
years of
total U.S.
pop.)
1 hour daily max
ozone(ppb)
over 65,
Spokane
over 65
P = 0.004619
s.e. = 0.007739
Schwartz, 1996,
Spokane
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence
(per
100,000)
Expos Meas
from original
study
Study Pop.
Applied Pop.
Functional form*
Uncert & Var.
Sources
hospital
admissions
pneumonia
(ICD 480-487)
229/year11
(incidence
in pop > 65
years of
total U.S.
pop.)
1 hour daily max
ozone(ppb)
over 65,
Spokane
over 65
P = 0.00965
s.e. = 0.006011
Schwartz, 1996,
Spokane
presence of
any of 19 acute
respiratory
symptoms
n/a
daily one-hour
max. 03 (pphm)
adult members
of families of
elementary
school-aged
children in
Glendora-
Covina-Azusa,
CA
adults 18-65
A Symptdays/day = (p, - p0) * Pop
where
1
Pl =
"On-—2— * M03)
l+«
and
p0 = the probability of having Symptdays per individual for
a 24-hour period
in the base case
= 0.19
P= 1.4 x 10-4c
(Model includes 03, COH, S02)
s.e. 6.7 x 10"5c
Krupnick et al.,
1990
^3
to
£
3
Si
s
Sr
a
s
o
Q)
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence
(per
100,000)
Expos Meas
from original
study
Study Pop.
Applied Pop.
Functional form*
Uncert & Var.
Sources
self-reported
asthma attacks
n/a
1 hour daily
max. oxidants
(ppm)
asthmatics in
Los Angeles
all asthmatics
(4%d of the
total
population)
A asthma attacks/day = (Pi - p0) * Pop
where
1
=
-(In—?— ~ p*A03)
I + e
and
Po = the probability of attacks per asthmatic for a 24-hour
period in the base case,
= 0.027"
P= 1.9 x 10"3f
1.11 = factor to convert measured 03 levels to oxidants
(only model includes oxidants and TSP)
s.e. =
7.2 x 10"4 tE
Whittemore and
Korn, 1980 and
U.S. EPA, 1993b
respiratory and
nonrespira-
tory conditions
resulting in a
minor
restricted
activity day
(MRAD)
780,000/
year1
(of study
pop.)
1 hour daily
max. 03 (ppm)
averaged over 2
weeks
employed
adults across
the U.S.
between the
ages of 18-65
(urban
residents)
all adults
aged 18-65
equation predicts daily change in MRAD
P = 2.2 x 10-31
(Model includes 03 and fine particulates)
s.e. =
6.6 x 10A>
Ostro and
Rothschild, 1989
respiratory
restricted
activity days
(RRADs)
310,000/
year11
(of study
pop.)
1 hour daily max
03 (ppm)
averaged over 2
weeks
employed
adults across
the U.S.
between the
ages of 18-65
(urban
residents)
all adults
aged 18-65
equation predicts daily change in RRAD
p = -0.0054'
(Model includes 03 and fine particulates)
s.e. = 0.0017'
Ostro and
Rothschild, 1989
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence
(per
100,000)
Expos Meas
from original
study
Study Pop.
Applied Pop.
Functional form*
Uncert & Var.
Sources
sinusitis and
hay fever
n/a
hourly 03
averaged over
six years (1974-
1979) inppm
adults in urban
areas surveyed
in the National
Health
Interview
Survey
all
[®(
-------
Health
Endpoint
(ICD-9 code)
Baseline
Incidence
(per
100,000)
Expos Meas
from original
study
Study Pop.
Applied Pop.
Functional form"
Uncert & Var.
Sources
The following two rows should be combined, e.g., cases of DFEV, a 15% for heavy exercisers (using equation based on Avol et al., 1984) should be added to cases of DFEV, a 15%
for moderate exercisers (using equation based on data from Seal et al., 1993)
Decrements in
lung function
as measured by
forced
expiratory
volume in one
second (FEV,)
n/a
Exposure to
ozone for 1.33
hours during
which
individuals were
exercising
continuously for
one hour
(controlled
setting)
Heavily
exercising male
and female
bicyclists
(mean age =
26.4 yrs)
all under age
50"
Acases = a * p * A03 * Pop.
where,
P = 0.00297 for DFEV, a 15%
= 0.00268 for DFEV, a 20%
a = 0.06656°
Avol et al., 1984
Decrements in
lung function
as measured by
FEV,
n/a
Exposure to
ozone for 2.33
hours during
which
individuals were
exercising
intermittently
(total exercise
time = 1 hour)
(controlled
setting)
Moderately
exercising male
and female
college students
(ages 18-35)
all under age
50"
.V -d
In
...
-a
l-X •d.
ill
i
b
Seal, et al., 1993
mm
/ \
x\'d,
In
Wa
• Pop
where,
a = -0.664 for DFEV, 2 15%
= -0.326 for DFEV, > 20%
b = 0.000840 for DFEV, > 15%
= 0.000919 for DFEV, > 20%
d, = 1.06"
d2= 1.00
d3 = 0.70
e, = 0.288"
e2 = 0.224
e3 = 0.640
X0 and X, are ozone concentrations in the CAA and
No-CAA scenarios
-------
NOTES:
s Pollutant coefficients expressed as a change in health effects per change in ppb 03.
b Centers for Disease Control, 1992. Vital and Health Statistics, Detailed Diagnoses and Procedures, National Hospital Discharge Survey, 1990. Number of 1990 discharges divided by
1990 U.S. population (248,709,873) from City and County Databook, 12th edition, 1994, U.S. Dept. of Commerce, Bureau of the Census, Washington, D.C.
c Determined the incremental effect/unit 03 by incorporating stationary probabilities from transitional probabilities. ESEERCO (1994) obtained transitional probabilities for adults from
original study authors.
d U.S. EPA, 1994a.
e Calculated as baseline asthma attack rate (number of attacks per person per year) divided by 365 days per year. Number of attacks per person per year = 9.9 from National Center for
Health Statistics, National Health Interview Survey, 1979 (as cited by Krupnick and Kopp, 1988).
f (3 coefficient and s.e. converted to A in cases/ppb 03 based on the following relationship: 1 ppb 03 = 1.11 ppb oxidants.
g Study did not report a s.e. Thus, the analysis assumed the largest s.e. possible (at p = 0.01, using a two-tailed test of significance)
h Ostro and Rothschild (1989) report average annual MRADs as 7.8 per person, using data from 6 years.
1 (3 is a weighted mean using separate coefficients for six years. Each year's coefficient was weighted by the inverse of the variance for that coefficient.
' Standard error is the square root of the sum of the weights (sqrt[sum(l/var)], where I indicates the individual year).
k Ostro and Rothschild (1989) report average annual RRADs as 3.1 per person, using data from 6 years.
1 Obtained by determining the products of beta coefficients for other independent variables and their mean values and summing these and the constant value.
m Calculated by dividing (3 by asymptotic t-ratio.
" From Table 12 in 1992 Statistical Abstracts, the percent of individuals in the U.S. population under age 50 = 75%.
0 Factor to adjust for differences in concentration among microenvironments and amount of time spent in different microenvironments at heavy exercise rates.
p The values, d., adjust ozone concentrations for various microenvironments (outdoor — near road, outdoor — other, and indoor) using values reported in U.S. EPA, 1993.
q The values, e., adjust the response rates by the percent of time spent in each microenvironment at the relevant exercise rates (i.e., percent of time at a fast rate is used for Avol et al.,
1984, and percent of time at a moderate rate is used for Seal et al., 1993). U.S. EPA (1993) presents information to determine e. values.
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Nitrogen Oxides
Nitrogen dioxide (N02) is the primary focus of health studies on the nitrogen oxides and serves as the basis
for this analysis. The primary pathophysiology of N02 in humans involves the respiratory system and the con-
centration-response function identified for N02 describes the relationships between measures ofN02 and respi-
ratory illness.
A number of epidemiological studies of N02 are available; however, most have either confounded expo-
sures (with other pollutants) or insufficient exposure quantification (e.g., exposure assessment indicates only
absence or presence of a gas stove). Most studies consider N02 generated by gas stoves or other combustion
sources in homes and are therefore not directly usable in concentration-response functions. However, studies by
Melia et al, 1980 and Hasselblad et al, 1992 provide a reasonable basis for development of a concentration
response function. Table D-8 presents the function obtained from their work. The function relates N02 to respi-
ratory illness in children.
D-34
-------
Tabic D-8. Summary of Concentration-Response Functions for NO,
Health
Endpoint
Exposure
Measure from
Original
Studies
Study
Population
Applied
Population
Functional Form'
Uncertainty/
Variab.
Sources
respiratory
illness (as
indicated by
respiratory
symptoms)
N02
measurements
in bedrooms
with Palmes
tubes
(one year time
weighted
average
concentration in
Hg/m3)
children ages
6 to 7
all
(combining
functions for
men and
women)
AResp cases =AProb(Resp)*P op
s.e. = 0.0132
Hasselblad, et
al., 1992.
where:
Prob(resp) = probability of respiratory illness during a one year
period:
Prob(resp)
1
1 + e
log >Jd\
and
log odds Resp =
-0.536 + 0.0275 N02 - 0.0295 gender
gender = 1 for boys and 0 for girls (the term drops out for girls)
NOTES:
s This equation was obtained from two sources. The N02 coefficient was reported in Hasselblad et al., 1992. The background and gender intercepts were obtained via personal
communication with V. Hasselblad 2/28/95 by Abt Associates. The equation was based on an evaluation by Hasselblad et al. of study results obtained by Melia et al. (1980). See text for
further discussion.
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Carbon Monoxide
Three concentration-response relationships are available for estimating the health effects of carbon monox-
ide. The first relates ambient CO levels to hospital admissions for congestive heart failure (Morris et al., 1995).
The second equation (Allred et al., 1989a,b, 1991) relates the CO level in the bloodstream to the relative change
in time of onset of angina pain upon exertion. The third relates the CO level in the bloodstream to the relative
change in time of onset of silent ischemia. Due to the lack of quantitative information relating silent ischemia to
a meaningful physical health effect, this analysis uses only the first two dose-response functions shown in Table
D-9.
D-36
-------
Tabic D-9. Summary of Concentration-Response Functions for Carbon Monoxide.
Health
Baseline
Exposure
Study
Applied
Functional Form
Uncert./
Sources
Endpoint
Incidence
Measure
Population
Population
Variability
Hospital
n/a
average of
Medicare
65 and over
Acases = P * ACO * Pop
s.e. =
Morris et al.,
admiss. for
hourly max
population in
1.9 x 10 s
1995
congestive
CO (ppm)
7 large U.S.
where P = l.lxl 0"7
heart
cities (96% of
7 large U.S.
failure
which are
cities
;>65)
percent
baseline time
CO (in ppm)
men, age 35-
Angina
percent change in time to angina =P*A%COHb
s.e. = 0.81%
Allred, et al.,
change in
to onset of
averaged over
75 years,
patients in
1989a,b, 1991
time to
angina
1 or 8 hours
stable angina,
U.S.=
where:
angina
during
nonsmokers
3,080,000 in
P = -1.89%
treadmill test
(of at least 3
1989"
and
from Allred
months) at
et al. studies
time of study
Frequency
COHb = blood level of carboxyhemoglobin
= 515
of angina
and
seconds at
attacks for
%COHb=
the study
A %COHb = 0.45 * ACOc,
0.63"
population =
where:
4.6 per week
CO = concentration of CO (ppm), for non-
(range = 0 -
smoking adults undertaking light exercise
63)cd
(alveolar ventilation rate of 20L/min) for one
hour at low altitude, with an initial COHb =
0.5%.
OR
A %COHb = 0.12 * ACOe,
where:
conditions are the same as above except that
study individuals are at rest (alveolar
ventilation rate of lOL/min) for 8 hours.
NOTES:
s Calculated as the mean of means from 3 pre-exposure treadmill tests and 1 post-exposure test (control exposure to air) (Allred et al., 1991).
b American Heart Association (1991)
c Allred etal. (1991)
d Multiple daily events are not modeled. Although it is possible that angina attacks may occur more than once per day, the average frequency of attacks was 4.6 per week (< 1 per day).
e Equation calculated from figure in U.S. EPA (1991a), p. 2-7.
O
Q)
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Sulfur Dioxide
This analysis estimated one concentration-response function for S02 using clinical data from two sources
on the responses of exercising asthmatics to S02, as measured by the occurrence of respiratory symptoms in
mild and moderate asthmatics (see Table D-10).
D-38
-------
Tabic D-10. Summary of Concentration-Response Functions for Sulfur Dioxide.
Health
Expos Meas.
Study Pop.
Applied
Functional Form
Uncert & Var.
Sources
Endpoint
from original
Pop.
study
Any
5-minute S02
generally
exercising
logodds Symp = -5.65 + 0.0059 S02 + 1.10 status
s.e. for:
data from Linn
Symptom
concentration,
young
asthmatics -
et al.(1987,
(chest
ppm (using peak
exercising
defined as
const. term=
1988, 1990),
tightness,
to mean ratio
asthmatics
4% of
2.60
Roger et al.
shortness of
from hourly S02
(ventilation
general
where status = asthma status (0 for mild, 1 for moderate)
(1985)
breath, etc.)
concentration of
rate 0.4
population,
for
2:1 to 3:1)
m3/min)
of whom
S02 coeff=
1.7% (range
Prob{symp) = -——
0.0025
0.2% to
j + e -logodds
3.3%) are
for
exercising
status coeff=
during
1.44
waking hours
Cases = Probmjld (effects) ¦ Pop^ +
Proimod Wfects"> ' p°Pm„d
Cases = number of individuals with occurrences of at least moderate effects for
all three measures.
where
Popmiid = exposed population of exercising mild asthmatics (assumed to be 2/3
of asthmatic population);
P°Pmod = exposed population of exercising moderate asthmatics (assumed to be
1/3 of asthmatic population)
^3
to
£
3
Si
3
a
3
o
Q)
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Estimating Welfare Effects of
Exposure
In addition to avoided incidences of adverse hu-
man health effects, the air quality improvements esti-
mated to result from the CAA yield additional ben-
efits, namely welfare benefits. Table D-10 indicates a
variety of benefits expected to have accrued through
the avoidance of air pollution damage to resources.
As indicated, data supporting quantified estimates of
welfare benefits are more limited than those quanti-
fying the relationship between air pollution exposure
and human health. While evidence exists that a vari-
ety of welfare benefits result from air quality improve-
ments, currently available data supports quantifying
only a limited number of potential effects at this time.
The Table lists the effects quantified in the section
812 analysis; each is discussed below.
mate such benefits using reported relationships be-
tween ozone exposure and yields of a variety of com-
modity crops.
It should be noted that the method used to allo-
cate monitor-level ozone concentrations to estimate
crop exposure differed from that used to estimate
ozone health effects. Instead of assigning concentra-
tions from the nearest monitor, the agricultural ben-
efits analysis estimated ozone concentrations for each
county nationwide. This was necessary because of two
factors specific to the agricultural analysis. First, crop
production is reported at the county level, so changes
in crop yields associated with changes in ozone levels
must be estimated for each county. Second, much of
the nation's agricultural production of "commodity
crops" (corn, wheat, soybeans, etc.) occurs at signifi-
cant distances from the location of the population-
oriented ozone monitors. Thus, an algorithm was used
Tabic D-l 1. Selected Welfare Effects of Criteria Pollutants.
Pollutant
Quantified Welfare Effects
IJnqiiiintified Welfare Effects
Ozone
Agriculture - Changes in crop yields
(for 7 crops)
Decreased worker productivity
Changes in other crop yields
Materials damage
Ecological - effects on forests
Ecological - effects on wildlife
l'si rticu lute Mutter/
I SP/ Sulfates
Materials Damage - Household
soiling
Visibility
Other materials damage
Ecological - effects on wildlife
Nitrogen Oxides
Visibility
Crop losses due to acid deposition
Materials damage due to acid
deposition
Effects on fisheries due to acid
deposition
Effects on forest
Sulfii i Dioxide
Visibility
Crop losses due to acid deposition
Materials damage due to acid
deposition
Effects on fisheries due to acid
deposition
Effects on forest
Agricultural Effects
This analysis was able to quantify the benefits to
economic welfare attributable to the increased crop
yields expected from CAA-related air quality improve-
ments. Appendix F describes the method used to esti-
to assign ozone concentrations for the agricultural
analysis for the control and no-control scenarios to
county centroids based on a planar interpolation of
concentrations at the nearest three monitors. Appen-
dix F documents the details of the triangulation of
ozone air quality data.
D-40
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
Materials Damage
Welfare benefits also accrue from avoided air
pollution damage, both aesthetic and structural, to ar-
chitectural materials and to culturally important ar-
ticles. At this time, data limitations preclude the abil-
ity to quantify benefits for all materials whose dete-
rioration may have been promoted and accelerated by
air pollution exposure. However, this analysis does
address one small effect in this category, the soiling
of households by particulate matter. Table D-l 1 docu-
ments the function used to associate nationwide PM-
10 levels with household willingness to pay to avoid
the cleaning costs incurred for each additional (ig/m3
of PM-10.
Visibility
In addition to the health and welfare benefits esti-
mated directly from reduced ambient concentrations
of individual criteria air pollutants, this analysis also
estimates the general visibility improvements attrib-
uted to improved air quality. Visibility effects are
measured in terms of changes in DeciView, a mea-
sure useful for comparing the effects of air quality on
visibility across a range of geographic locations for a
range of time periods. It is directly related to two other
common visibility measures, visual range (measured
in km) and light extinction (measured in km"1); how-
ever, it characterizes visibility in terms of perceptible
changes in haziness independent of baseline condi-
tions.
Visibility conditions under the control and no-
control scenarios were modeled separately for the east-
ern and western U.S. In the east, the Regional Acid
Deposition Model (RADM) generated extinction co-
efficient estimates for each of 1,330 grid cells in the
RADM domain (essentially the eastern half of the
country). The extinction coefficients were translated
to DeciView using the relationship reported in
Pitchford and Malm (1994). In the Western U.S., a
conventional extinction budget approach provided
DeciView estimates for 30 metropolitan areas (SAI,
1994). A linear rollback model provided the corre-
sponding no-control estimates. Visibility estimates for
both portions of the country were generated for the
target years 1975, 1980, 1985, and 1990.
Table D-12 summarizes the methodology used to
predict visibility benefits attributable to the CAA.
Physical benefits for a given year are reported in terms
of the average DeciView change per person in the
modeled population.
Worker Productivity
Available data permits quantification of a final
human welfare endpoint, worker productivity. Crocker
and Horst (1981) and U.S. EPA (1994c) present evi-
dence regarding the inverse relationship between
ozone exposure and productivity in exposed citrus
workers. This analysis applies the worker productiv-
ity relationship (reported as income elasticity with
respect to ozone) to outdoor workers in the U.S. (ap-
proximately one percent of the population). Table D-
12 details the form of the concentration response func-
tion.
Ecological Effects
It is likely that the air pollution reductions
achieved under the CAA resulted in improvements in
the health of aquatic and terrestrial ecosystems. To
the extent that these ecosystems provide a variety of
services (e.g., fishing, timber production, and recre-
ational opportunities), human welfare benefits also
accrued. However, due to a lack of quantified con-
centration-response relationships (or a lack of infor-
mation concerning affected population), ecological
effects were not quantified in this analysis. Appendix
E provides discussion of many of the important eco-
logical benefits which may have accrued due to his-
torical implementation of the CAA.
D-41
-------
Tabic D-12. Summary of Functions Quantifying Welfare Benefits.
Endpoint
Expos Meas.
Applied
Pop.
Functional Form
Uncert & Var.
Sources
Household
Soiling Damage
(change in dollar
valuation)
annual mean
PM.o
all
households
(study based
on
households in
20
metropolitan
areas)
Soiling Damage = P * Pop/PPH * APM10
where
P = $2.52
PPH = people per household (2.68)a
Beta distribution with
mean = $2.52
s.e. = $1.00
interval =
[$1.26-$10.08]
slope parameters:
a= 1.2,
P = 7.3
Manuel et al. (1982);
McClelland, et al.
(1991); Watson and
Jaksch (1982);
ESEERCO (1994)
Visibility
(average change
in DeciView per
person)b'c
Eastern U.S.:
Extinction
coefficient (Ext)
in units of m"1
Western U.S.:
DeciView, dv
(unitless)
all
£ (dVNo-CAA, , " dVCAA, ) X P°P,
A Vis = ^
X>P,
¦. ¦
where,
AVis = avg. change in DeciView per person in modeled population
i = modeled area
dvNo-CAA = DeciView under no control scenario
dvcAA = DeciView under control scenario
Pop; = modeled population in modeled area, i
In the East, Ext (in units of km"1) is converted to dv as follows:
Ext
deciview - 10 In
0.01 km ~l)
not available
Pitchford and Malm
(1994)
-------
Endpoint
Expos Meas.
Applied
Pop.
Functional Form
Uncert & Var.
Sources
worker
hourly 03
individuals in
ai = i*n*(x, - x0)/x0 * Pop * w
not available
Estimated using data
productivity
concentration
occupations
from Crocker and
(resulting in
averaged over a
that require
AI = change in total daily income,
Horst (1981) and
changes in daily
workday or 24-
heavy
r) = income elasticity with respect to 03 conc.,
U.S. EPA, 1994c
wages)
hours (ppm)
outdoor
T| = -0.14 for 24-hour period,
physical
I = total daily income per worker engaged in strenuous outdoor labor
labor
= $73d
W = proportion of outdoor workers in the U.S. population
(study based
= 0.012e
on citrus
X0 = average hourly 03 concentrations with CAA,
workers in S.
X, = average hourly 03 concentrations without CAA
California)
(NOTE: Average number of days worked per year for workers engaged
in strenuous outdoor labor = 213)f
(model includes 03 only)
NOTES:
11 1990 Census
b Visibility is measured in two ways: (1) in terms of extinction coefficient in the eastern U.S. (based on modeling of RADM domain); and (2) as DeciView (dv) in the west (modeling
of 30 western cities) (SAI, 1994).
c DeciView is a haziness index used to characterize visibility through uniform hazes.
d Average daily wage, assuming an 8-hour day, by workers in the job categories listed below, taken from U.S. Bureau of the Census, Earnings by Occupation and Education, 1990.
e Full- and part-time workers (total of 3,100,000) taken from U.S. Bureau of the Census, Earnings by Occupation and Education, 1990. Includes the following job categories: farm
workers; groundskeepers and gardeners, except farm; forestry workers, except logging; timber cutting and logging occupations; brickmasons and stonemasons; brickmason and stonemason
apprentices; roofers; structural metal workers; construction trades, n.e.c.; construction laborers; garbage collectors; and stevedores. Value is divided by total U.S. population.
f Average number of days worked per year, assuming an 8-hour day, by workers in the job categories listed above, taken from U.S. Bureau of the Census, Earnings by Occupation and
Education, 1990.
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Modeling Results
This section summarizes results of the health and
welfare effects modeling. As indicated previously, the
Project Team adopted a Monte Carlo approach in an
effort to capture uncertainty in the benefits analysis.
With respect to estimating avoided incidence of ad-
verse health and welfare effects, two sources of vari-
ability are considered. The first is the statistical un-
certainty associated with each concentration-response
relationship reported in the literature. In addition to
an estimate of a concentration-response function co-
efficient, studies typically report a standard error of
the reported estimate. The second source of uncer-
tainty lies in the choice of studies, where multiple stud-
ies offer estimates for the same endpoint. Different
published results reported in the scientific literature
typically do not report identical findings; in some in-
stances the differences are substantial. This between-
study variability is captured by considering the range
of estimates for a given endpoint.
Table D-13 summarizes health and welfare effects
for each study included in the analysis. The values
presented are mean estimates of the number of cases
of each endpoint avoided due to implementation of
the CAA. A distribution is associated with each mean
estimate, capturing the uncertainty inherent in the es-
timate of the concentration-response coefficient. The
distribution of estimated effects corresponding to a
given study was generated by randomly sampling from
the distribution of coefficients (given by the estimated
coefficient and its standard error reported in the study)
and evaluating the concentration-response function,
yielding an estimate of avoided incidence for the given
effect. This procedure was repeated many times. While
only the central estimates of the resulting distribu-
tions are presented here, the distributions were retained
for use in monetizing and aggregating economic ben-
efits (see Appendix I).7
As shown, for some health endpoints more than
one concentration-response function was used, each
representing a different study. The alternative con-
centration-response functions provide differing mea-
sures of the effect. These can be used to derive a range
of possible results. In the case of lead (Pb), alterna-
tive functions were not used; rather, two analytical
procedures were implemented (labeled the "backward-
looking" and "forward looking" analyses), giving a
range of results for most Pb endpoints (see Appendix
G for discussion of Pb health effects).
The table presents the results of modeling "all U.S.
population" (although, with the exception of Pb, not
all of the 48 state population is modeled, with up to
five percent being excluded in a given year). The re-
sults depict the pattern of health effects incidence
across years. The accuracy of the scale of incidence
is less certain (due to the extrapolation of air quality
data). These results are almost certainly more accu-
rate than the corresponding "50 km" results, but rely
on the assumption that (for a portion of the popula-
tion) distant air quality monitors provide a reason-
able estimate of local air quality conditions. Thus, the
results presented here are somewhat speculative. It is
likely that the estimated health effects are overstated
for that population group (20 to 30 percent of total
population in the case of PM) for which distant moni-
tors are used. (Note, however, that the scaling of
unmonitored county PM concentrations based on re-
gional-scale grid model projections significantly miti-
gates this potential overestimation in the case of PM;
see Appendix C for details). Conversely, there is an
implied zero health impact for that portion of the popu-
lation (three to four percent in the case of PM) ex-
cluded from the analysis altogether, an understatement
of health impacts for that group.
The results indicate the growth of benefits over
the study period, consistent with increasing improve-
ments in air quality between the control and no-con-
trol scenarios from 1970 to 1990.
The mortality effects documented above can be
disaggregated by age. Table D-14 indicates the esti-
mated proportions of premature mortalities for vari-
ous age groups (Pb-induced mortality estimates for
children, men, and women are grouped). Also pre-
sented is the average life expectancy for each group,
indicating the degree of prematurity of PM and Pb-
related mortality.
Table D-15 presents estimated incidence reduc-
tions for several health effects which could be quanti-
fied but not monetized for this analysis.
7 With the exception of visibility, welfare endpoints estimated economic benefits directly and are therefore included in the
monetary benefits results presented in Appendix I.
D-44
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
Tabic D-13. Criteria Pollutants Health Effects ~ Extrapolated to 48 State U.S. Population (Cases per year -
mean estimates).
End point
Study
Pollutant(s) 1975
1980
1985
1990
MORTALITY
Mortality (long-term exposure)
Pope et al., 1995
PM10
58,764
145,884
169,642
183,539
Mortality (Pb exposure) -Male
Average of Backward & Forward
Pb
822
5,281
10,340
12,819
Mortality (Pb exposure) -Female
Average of Backward & Forward
Pb
231
1,474
2,866
3,537
Mortality (Pb exposure) -Infant
Average of Backward & Forward
Pb
456
2,342
3,933
4,944
CHRONIC BRONCHITIS
Chronic Bronchitis
Schwartz, 1993b
PM10
198,973
554,632
720,166
741,775
Abbey et al., 1993
PM10
173,571
454,309
564,753
602,990
OTHER Pb-INDUCED AILMENTS
Lost IQ Points
Average of Backward & Forward
Pb
1,028,492
5,031,157
8,559,426
10,378,268
IQ< 70
Average of Backward & Forward
Pb
3,780
20,074
36,520
45,393
Hypertension-Men
Average of Backward & Forward
Pb
830,299
5,276,999
10,087,115
12,646,876
Cor. Heart Disease
Average of Backward & Forward
Pb
1,313
8,444
16,671
21,069
Atherothrombotic brain infarction - Men
Average of Backward & Forward
Pb
181
1,128
2,165
2,690
Atherothrombotic brain infarction - Women
Average of Backward & Forward
Pb
84
529
1,020
1,255
1 nitial cerebrovascular accident - Men
Average of Backward & Forward
Pb
260
1,635
3,154
3,926
1 nitial cerebrovascular accident - Women
Average of Backward & Forward
Pb
120
758
1,466
1,804
HOSPITAL ADMISSIONS
All Respiratory
Schwartz, 1995, Tacoma
™,„
&03
32,004
77,827
95,435
106,777
Schwartz, 1996, Spokane
PM10
&03
29,393
69,449
93,137
119,290
Pope, 1991, Salt Lake Valley
PM10
30,982
73,093
86,407
95,486
Schwartz, 1995, New Haven
PM,„
&03
23,137
55,096
66,385
73,842
Thurston et al., 1994, Toronto
PM10
&03
13,746
32,383
39,691
46,013
COPD + Pneumonia
Schwartz, 1994c
PM,„
&03
21,898
53,928
64,217
70,528
Schwartz, 1996, Spokane
PM10
&03
19,769
47,294
63,116
80,113
Schwartz, 1994a
PM10
&03
16,942
40,882
49,290
55,227
Schwartz, 1994b
PM10
&03
13,006
30,679
37,434
43,410
Ischemic Heart Disease
Schwartz and Morris, 1995
PM10
6,348
14,709
17,289
19,098
Congestive Heart Failure
Schwartz and Morris, 1995
PM10
5,733
13,365
15,742
17,362
Morris et al., 1995
CO
3,022
8,543
17,028
21,835
OTHER RESPIRATORY-RELATED AILMENTS
-Adults
Any of 19 Acute Symptoms
Krupnick et al., 1990
PM10
&03
41,631,456
98,876,110
117,275,400
129,529,717
- Children
Shortness of breath, days
Ostro et al., 1995
PM,„
20,752,402
50,758,872
58,575,484
68,375,216
Acute Bronchitis
Dockery et al., 1989
PM10
1,936,260
6,255,801
7,644,924
8,541,833
Lower Respiratory Symptoms
Schwartz et al., 1994d
PM10
2,994,048
6,100,276
6,977,680
7,804,860
Upper Respiratory Symptoms
Pope et al., 1991
PM10
500,395
1,292,922
1,557,177
1,683,854
-All Ages
Asthma Attacks
Ostro et al., 1991
PM,„
264,430
548,306
686,953
841,916
Whittemore and Korn, 1980;
03
193
482
816
1,080
EPA ,1983
Increase in Respiratory Illness
Hasselblad, 1992
N02
729,306
2,686,813
6,113,639
9,776,267
Any Symptom
Linnet al. (1987, 1988,1990)
S02
104,896
319,192
282,846
265,650
RESTRICTED ACTIVITY AND WORK LOSS DAYS
RAD
Ostro, 1987
PM10
19,170,337
47,445,314
56,939,271
62,187,720
MRAD
Ostro and Rothschild, 1989
PM10
&03
60,871,610
155,799,151
190,333,140
209,924,785
RRAD
Ostro and Rothschild, 1989
PM10
&03
47,669,732
237,799,482
176,850,171
174,329,691
Work Loss Days
Ostro, 1987
PM10
6,966,775
17,213,581
20,648,906
22,562,752
HUMAN WELFARE
Household Soiling Damage
ESEERCO, 1994
PM10
direct economic valuation
Visibility - East (DeciView chg. per person)
Pitchford and Malm, 1994
DeciView
0.4
1.4
1.9
2.0
Visibility - West (DeciView chg. per person) Pitchford and Malm, 1994
DeciView
2.4
4.9
5.0
6.0
Decreased Worker Productivity
Crocker & Horst, 1981 and EPA, 1994c03
direct economic valuation
Agriculture (Net Surplus)
Minimum Estimate
03
direct economic valuation
Maximum Estimate
03
direct economic valuation
D-45
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic D-14. Mortality Distribution by Age: Proportion of PM- and
Pb-rclatcd Premature Mortalities and Associated Lite Expectancies.
Proportion of Premature Mortalities by Age1
Age Croup
PM"
PI)'
Life Expectancy
Forward (Backward)1
(years)
Infants
33% (20%)
75
5-30
30-34
2%
48
35-39
4%
38
40-44
11% (13%)
45-51
6%
21% (25%)
29
55-C4
13%
22% (27%)
21
65-74
24%
12% (15%)
14
75-W
29%
9
85+
22%
6
100%
100%
Notes:
a Distribution of premature mortalities across ages is fairly consistent across years.
b PM-related mortality incidence estimated only for individuals 30 years and older,
consistent with the population studied by Pope et al., 1995.
c Pb-related mortality incidence was estimated for infants, women aged 45-74. and men in
three age groups (40-54. 55-64. 65-74). each with a distinct concentration-response
relationship.
d Forward (backward) analysis holds other lead sources at constant 1970 (1990) levels -
see Appendix G. Values may not sum to 100% due to rounding.
D-46
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
Tabic D-15. Quantified Benefits Which Could Not Be Monetized - Extrapolated to the
Entire 48 State Population.
ljnJ|pi(Oiiiml
Study
Pollutant
1975
1980
MS
1990
I in its
Pulmonary Fund ion
Decrements
Decreased FEV by 15 %
or more
Avol et al. 1984 & Seal
et al. 1993
03
53
121
196
312
million person-days with
decreased FEV (per year)
Decreased FEV by 20 %
or more
Avol et al. 1984 & Seal
et al. 1993
03
39
87
141
224
million person-days with
decreased FEV (per year)
Chronic Sinusitis and Hay
Fever
Portney and Mullahy,
1990
03
6
8
8
9
million cases/year
Tim e to Onset of Angina
Pain
Allred, et al., 1989a,b,
1991
CO
0.1%
0.4%
0.7%
0.8%
fractional increase in
time to onset of angina
attack
D-47
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Human Health and Welfare
Effects References
Abbey, D.E, F. Petersen, P. K. Mills, and W. L.
Beeson. 1993. "Long-term Ambient Concen-
trations of Total Suspended Particulates,
Ozone, and Sulfur Dioxide and Respiratory
Symptoms in a Nonsmoking Population."
Archives of Environmental Health 48(1): 33-
46.
Allred, E.N., E.R. Bleecker, B.R. Chaitman, T.E.
Dahms, S.O. Gottlieb, J.D. Hackney, M.
Pagano, R.H. Selvester, S.M. Walden, and J.
Warren. 1989a. "Short-term Effects of Car-
bon Monoxide Exposure on the Exercise Per-
formance of Subjects with Coronary Heart
Disease." New England Journal of Medicine
321: 1426-1432.
Allred, E.N., E.R. Bleecker, B.R. Chaitman, T.E.
Dahms, S.O. Gottlieb, J.D. Hackney, D.
Hayes, M. Pagano, R.H. Selvester, S.M.
Walden, and J. Warren. 1989b. Acute Effects
of Carbon Monoxide Exposure on Individu-
als with Coronary Heart Disease. Health Ef-
fects Institute, Cambridge, MA. Research
Report No 25.
Allred, E.N., E.R. Bleecker, B.R. Chaitman, T.E.
Dahms, S.O. Gottlieb, J.D. Hackney, M.
Pagano, R.H. Selvester, S.M. Walden, and J.
Warren. 1991. "Effects of Carbon Monoxide
on Myocardial Ischemia." Environmental
Health Perspectives 91:89-132.
American Heart Association. 1991. 1992 Heart and
Stroke Facts.
Avol, E.L., W.S. Linn, T.G. Venet, D.A. Shamoo, and
J.D. Hackney. 1984. "Comparative Respira-
tory Effects of Ozone and Ambient Oxidant
Pollution Exposure During Heavy Exercise."
Journal of the Air Pollution Control Associa-
tion 34: 804-809.
Centers for Disease Control (CDC). 1992. Vital and
Health Statistics, Detailed Diagnoses and Pro-
cedures, National Hospital Discharge Survey,
1990. (Series 13, Number 113).
Crocker T. D., and R. L. Horst, Jr. 1981. "Hours of
Work, Labor Productivity, and Environmen-
tal Conditions: a Case Study." The Review of
Economics and Statistics 63:361-368.
Dockery, D.W., F.E. Speizer, D.O. Stram, J.H. Ware,
J.D. Spengler, and B.G. Ferris, Jr. 1989. "Ef-
fects of Inhalable Particles on Respiratory
Health of Children." Am. Rev. Respir. Dis.
139: 587-594.
Dockery, D.W., et al. 1993. "An Association Between
Air Pollution and Mortality in Six U.S. Cit-
ies." The New England Journal of Medicine
329 (24): 1753-1759.
Dockery, D.W., J. Schwartz, and J.D. Spengler. 1992.
Air Pollution and Daily Mortality: Associa-
tions with Particulates and Acid Aerosols.
Environ. Res. 59: 362-373.
Empire State Electric Energy Research Corporation
(ESEERCO). 1994. New York State Environ-
mental Externalities Cost Study. Report 2:
Methodology. Prepared by: RCG/Hagler,
Bailly, Inc., November.
Hasselblad, V., D.M. Eddy, and D.J. Kotchmar. 1992.
"Synthesis of Environmental Evidence: Ni-
trogen Dioxide Epidemiology Studies." J. Air
Waste Mgmt. Assoc 42: 662-671.
Health Effects Institute (HEI). 1996. Air Pollution and
Mortality in Philadelphia, 1974-1988. Con-
tributors: Samet, J.M., S.L. Zeger, K. Berhane,
and J. Xu.
Ito, K. and G.D. Thurston. 1996. Daily PM /Mortal-
ity Associations: An Investigation of At-Risk
Sub-Populations. J. Exposure Anal. Environ.
Epidemiol.: in press.
Ito, K, P. Kinney, and G.D. Thurston. 1995. Varia-
tions on PM Concentrations Within Two
Metropolitan Xrcas and Their Implications for
Health Effects Analyses. In: Phalen, R.F. and
D.V. Bates, eds. Proceedings of the Collo-
quium on Particulate Air Pollution and Hu-
man Mortality and Morbidity, Part II; Janu-
ary 1994; Irvine, CA. Inhalation Toxical. 7:
735-745.
D-48
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
Kinney, P.L., K. Ito, andG.D. Thurston. 1995. A Sen-
sitivity Analysis of Mortality/PM Associa-
tions in Los Angeles. In: Phalen1,0 R.F. and
D.V. Bates, eds. Proceedings of the Collo-
quium on Particulate Air Pollution and Hu-
man Mortality and Morbidity; January 1994;
Irvine, CA. Inhalation Toxicol. 7: 59-69.
Krupnick, A.J. and R.J. Kopp. 1988. "The Health and
Agricultural Benefits of Reductions in Am-
bient Ozone in the United States." Resources
for the Future Discussion Paper QE88-10,
Washington, DC. August.
Krupnick A.J., W. Harrington, and B. Ostro. 1990.
"Ambient Ozone and Acute Health Effects:
Evidence from Daily Data "Journal of Envi-
ronmental Economics and Management 18:1-
18.
Lave, L.B. and E.P. Seskin. 1977. Air Pollution and
Human Health. Johns Hopkins University
Press for Resources for the Future. Baltimore
and London.
Manuel, E.H., R.L. Horst, K.M. Brennan, W.N. Lanen,
M.C. Duff, and J.K. Tapiero. 1982. Benefits
Analysis of Alternative Secondary National
Ambient Air Quality Standards for Sulfur Di-
oxide and Total Suspended Particulates, Vol-
umes I-IV. Prepared for U.S. Environmental
Protection Agency, Office of Air Quality
Planning and Standards, Research Triangle
Park, NC. [Cited in ESEERCO, 1994],
McClelland, G., W. Schulze, D. Waldman, J. Irwin,
D. Schenk, T. Stewart, L. Deck, and M.
Thayer. 1991. Valuing Eastern Visibility: a
Field Test ofthe Contingent Valuation Method
(Draft). Prepared for U.S. Environmental Pro-
tection Agency, Washington, D.C.
Melia, R.J.W., C du V Florey, S. Chinn, B.D.
Goldstein, A.G.F. Brooks, H.H. John, D.
Clark, I.B. Craighead, andX. Webster. 1980.
"The Relation Between Indoor Air Pollution
from Nitrogen Dioxide and Respiratory Ill-
ness in Primary Schoolchildren." Clinical
Respiratory Physiology 16:7P-8P.
Linn, W.S., E.L. Avol, R. Peng, D.A. Shamoo, and
J.D. Hackney. 1987. "Replicated Dose-Re-
sponse Study of Sulfur Dioxide Effects in
Normal, Atopic, and Asthmatic Volunteers."
Am Rev Respir Dis. 136: 1127-34.
Linn, W.S., E.L. Avol, D.A. Shamoo, R. Peng, C.E.
Spier, M.N. Smith, and J.D. Hackney. 1988.
"Effect of Metaproterenol Sulfate on Mild
Asthmatics' Response to Sulfur Dioxide Ex-
posure and Exercise." Archives of Environ-
mental Health 43(6): 399-406.
Linn, W.S., D.A. Shamoo, R. Peng, K.W. Clark, E.L.
Avol, and J.D. Hackney. 1990. "Responses
to Sulfur Dioxide and Exercise by Medica-
tion-Dependent Asthmatics: Effect of Vary-
ing Medication Levels."Archives of Environ-
mental Health 45(1): 24-30.
Lipfert, F.W. 1984. "Air Pollution and Mortality:
Specification Searches Using SMSA-based
Data." J. Environ. Econ. Manage. 11: 208-
243.
Melia, R.J.W., C du V Florey, R.W. Morris, B.D.
Goldstein, H.H. John, D. Clark, I.B.
Craighead, and J.C. Mackinlay. 1982. "Child-
hood Respiratory Illness and the Home Envi-
ronment: II. Association Between Respiratory
Illness and Nitrogen Dioxide, Temperature,
and Relative Humidity." International Jour-
nal ofEpidemiology 11: 164-169.
Melia, R.J.W., C du V. Florey, and Y. Sittampalarm.
1983. The Relationship Between Respiratory
Illness in Infants and Gas Cooking in the UK:
a Preliminary Report. Air Quality Vth World
Congress: Proceedings of the International
Union of Air Pollution Prevention Associa-
tions. SEPIC (APPA), Paris France, pp. 263-
269.
Morris, R.D., E.N. Naumova, and R.L. Munasinghe.
1995. "Ambient Air Pollution and Hospital-
ization for Congestive Heart Failure Among
Elderly People in Seven Large U.S. Cities."
American Journal of Public Health 85(10):
1361-1365.
D-49
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Ostro, B.D. 1987. "Air Pollution and Morbidity Re-
visited: a Specification Test." J. Environ.
Econ. Manage. 14: 87-98.
Ostro, B.D. and S. Rothschild. 1989. "Air Pollution
and Acute Respiratory Morbidity: An Obser-
vational Study of Multiple Pollutants." Envi-
ronmental Research 50:238-247.
Ostro, B.D., M.J. Lipsett, M.B. Wiener, and J.C.
Seiner. 1991. "Asthmatic Responses to Air-
borne Acid Aerosols." American Journal of
Public Health 81: 694-702.
Ostro, B.D., J.M. Sanchez, C. Aranda, and G.S.
Eskeland. 1995. "Air Pollution and Mortal-
ity: Results of a Study of Santiago, Chile." J.
Exposure Anal. Environ. Epidemiol., submit-
ted.
Ostro, B.D., J.M. Sanchez, C. Aranda, and G.S.
Eskeland. 1996. Air Pollution and Mortality:
Results from a Study of Santiago, Chile. In:
Lippmann, M. ed. Papers from the ISEA-ISEE
Annual Meeting; September 1994; Research
Triangle Park, NC. J. Exposure Anal. Environ.
Epidemiol.: in press.
Ozkaynak, H., J. Xue, P. Severance, R. Burnett, and
M. Raizenne. 1994. Associations Between
Daily Mortality, Ozone, and Particulate Air
Pollution in Toronto, Canada. Presented at:
Colloquium on Particulate Air Pollution and
Human Mortality and Morbidity: Program
and Abstracts; January; Irvine, CA. Irvine,
CA: University of California Irvine, Air Pol-
lution Health Effects Laboratory; p. PI. 13;
report no. 94-02.
Pitchford, M.L. and W.C. Malm. 1994. "Development
and Applications of a Standard Visual Index."
Atmospheric Environment 28(5): 1049-1054.
Pope, C.A., III. 1991. "Respiratory Hospital Admis-
sions Associated with PM Pollution in Utah,
Salt Lake, and Cache Valleys."Arch. Environ.
Health 46 (2): 90-97.
Pope, C.A., III, and D.W. Dockery. 1992. "Acute
Health Effects of PM Pollution on Symp-
tomatic and Asymptomatic Children." Am.
Rev. Respir. Dis. 145: 1123-1128.
Pope, C.A., III, D.W. Dockery, J.D. Spengler, and
M.E. Raizenne. 1991. "Respiratory Health
and PM Pollution: a Daily Time Series
Analysis!'' Am. Rev. Respir. Dis. 144: 668-
674.
Pope, C.A., III, and L.S. Kalkstein. 1996. Synoptic
Weather Modeling and Estimates of the Ex-
posure-Response Relationship Between Daily
Mortality and Particulate Air Pollution.
Environ. Health Perspect. 104: in press.
Pope, C.A., III, J. Schwartz, and M.R. Ransom. 1992.
Daily Mortality and PM Pollution in Utah
Valley. Arch. Environ, health 47: 211-217.
Pope, C.A., III, M.J. Thun, M.M. Namboodiri, D.W.
Dockery, J.S. Evans, F.E. Speizer, and C.W.
Heath, Jr. 1995. Particulate Air Pollution as a
Predictor of Mortality in a Prospective Study
of U.S. Adults. Am. J. Respir. Crit. Care Med.
151: 669-674.
Portney, P.R. and J. Mullahy. 1990. "Urban Air Qual-
ity and Chronic Respiratory Disease." Re-
gional Science and Urban Economics 20: 407-
418.
Roger, L.J., H.R. Kehrl, M. Hazucha, and D.H.
Horstman. 1985. "Bronchoconstriction in
Asthmatics Exposed to Sulfur Dioxide Dur-
ing Repeated Exercise." J. Appl. Physiol.
59(3): 784-791.
Saldiva, P.H.N., Pope, C.A., III, J. Schwartz, D.W.
Dockery, A.J. Lichtenfels, J.M. Salge, I.
Barone, and G.M. Bohm. 1995. Air Pollution
and Mortality in Elderly People: A Time-Se-
ries Study in Sao Paulo, Brazil. Arch. Environ.
Health 50: 159-163.
Samet, J.M., Lambert, W.E., Skipper, B.J., Cushing,
A.H., Hunt, W.C., Young, S.A., McLaren,
L.C., Schwab, M., and J.D. Spengler. 1993.
"Nitrogen Dioxide and Respiratory Illnesses
in Infants." Am Rev. Respir. Dis. 148: 1258-
1265.
Schwartz, J. 1993a. Air Pollution and Daily Mortal-
ity in Birmingham, Alabama. Am. J.
Epidemiol. 137: 1136-1147.
D-50
-------
Appendix D: Human Health and Welfare Effects of Criteria Pollutants
Schwartz, J. 1993b. "Particulate Air Pollution and
Chronic Respiratory Disease." Environmen-
tal Research 62: 7-13.
Schwartz, J. 1994a. "Air Pollution and Hospital Ad-
missions in Elderly Patients in Birmingham,
Alabama." American Journal of Epidemiol-
ogy 139:589-98.
Schwartz, J. 1994b. "Air Pollution and Hospital Ad-
missions for the Elderly in Detroit, Michigan."
American Journal of Respiratory Care Med
150:648-55.
Schwartz, J. 1994c. "PM , Ozone and Hospital Ad-
missions for the Elderly in Minneapolis-St.
Paul, Minnesota." Archives of Environmen-
tal Health 49(5): 366-374.
Schwartz, J. 1994d. "Acute Effects of Summer Air
Pollution on Respiratory Symptom Report-
ing in Children.",4m JRespir Crit Care Med.
150: 1234-1242.
Schwartz, J. 1995. "Short Term Fluctuations in Air
Pollution and Hospital Admissions of the Eld-
erly for Respiratory Disease." Thorax 50:531-
538.
Schwartz, J. 1996. "Air Pollution and Hospital Ad-
missions for Respiratory Disease." Epidemi-
ology 7(1): 1-9.
Schwartz, J., D.W. Dockery, L.M. Neas. 1996a. Is
Daily Mortality Associated Specifically with
Fine Particles? J. Air Waste Manage. Assoc.:
accepted.
Schwartz, J. and R. Morris. 1995. "Air Pollution and
Cardiovascular Hospital Admissions.'Mm. J.
Epidemiol. 142: 23-35.
Seal, E., W.F. McDonnell, D.E. House, S.A. Salaam,
P.J. Dewitt, S.O. Butler, J. Green, and L.
Raggio. 1993. "The Pulmonary Response of
White and Black Adults to Six Concentrations
of Ozone." Am. Rev. Respir. Dis. 147: 804-
810.
Styer, P., N. McMillan, F. Gao, J. Davis, and J. Sacks.
1995. The Effect of Airborne Particulate
Matter in Daily Death Counts. Environ.
Health Perspect. 103:490-497.
Systems Application International (SAI). 1994. Ret-
rospective Analysis of the Impact of the Clean
Air Act on Urban Visibility in the Southwest-
ern United States. Prepared for the U.S. En-
vironmental Protection Agency, Office of Air
and Radiation. October 31.
Thurston, G., K. Ito, C. Hayes, D. Bates, and M.
Lippmann. 1994. "Respiratory Hospital Ad-
mission and Summertime Haze Air Pollution
in Toronto, Ontario: Consideration ofthe Role
of Acid Aerosols." Environmental Research
65: 271-290.
U.S. Environmental Protection Agency (U.S. EPA).
1985. Costs and Benefits of Reducing Lead
in Gasoline: Final Regulatory Impact Analy-
sis. Office of Policy Analysis. Washington,
DC. EPA-230-05-85-006.
U.S. Environmental Protection Agency (U.S. EPA).
1986. Air Quality Criteria for Particulate
Matter: Updated Assessment of Scientific and
Technical Information Addendum to the 1982
OAQPS Staff Paper. Prepared by the Office
of Air Quality Planning and Standards, Re-
search Triangle Park, North Carolina. EPA
450/05-86-012.
U.S. Environmental Protection Agency (U.S. EPA).
1991a. Acid Rain Benefit Assessment: Draft
Plan for the [Section 812] 1992 Assessment,
Acid Rain Division, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA).
1991 b.Air Quality Criteria for Carbon Mon-
oxide. EPA-600/8-90/045F, U.S. Environ-
mental Protection Agency, Office of Health
and Environmental Assessment, Environmen-
tal Criteria and Assessment Office; Research
Triangle Park, NC.
U.S. Environmental Protection Agency, 1993. Docu-
mentation for Oz-One Computer Model (Ver-
sion 2.0). Office of Air Quality Planning and
Standards. Prepared by: Mathtech, Inc., un-
der EPA Contract No. 68D830094, WA 59.
July.
D-51
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
U.S. Environmental Protection Agency (U.S. EPA).
1993b. External Draft, Air Quality Criteria for
Ozone and Related Photochemical Oxidants.
Volume II. Office of Health and Environmen-
tal Assessment, Environmental Criteria and
Assessment Office, Research Triangle Park,
NC; EPA/600/AP-93/004b. 3v.
U.S. Environmental Protection Agency (U.S. EPA).
1994a. Review of the National Ambient Air
Quality Standards for Sulfur Oxides: Assess-
ment of Scientific and Technical Information.
Supplement to the 1986 OAQPS Staff Paper
Addendum. Air Quality Management Divi-
sion, Office of Air Quality Planning and Stan-
dards. Research Triangle Park, NC. EPA Re-
port No. EPA-452/R-94-013.
World Health Organization (WHO). 1996.Final Con-
sultation on Updating and Revision of the Air
Quality Guidelines for Europe. Bilthoven,
The Netherlands 28-31 October, 1996 ICP
EHH018VD96 2.il.
U.S. Environmental Protection Agency (U.S. EPA).
1994b. Supplement to the Second Addendum
(1986) to Air Quality Criteria for Particulate
Matter and Sulfur Oxides (1982): Assessment
of New Findings on Sulfur Dioxide Acute Ex-
posure Health Effects in Asthmatics (Exter-
nal Review Draft). Environmental Criteria and
Assessment Office. Research Triangle Park,
NC. EPA/600/AP-93/002.
U.S. Environmental Protection Agency (U.S. EPA).
1994c.Documentation for Oz-One Computer
Model (Version 2.0). Office of Air Quality
Planning and Standards. Prepared by:
Mathtech, Inc., under EPA Contract No.
68D30030, WA 1-29. August.
Ware J.H., D.W. Dockery, A. Sprio III, F.E. Speizer,
and B.G. Ferris, Jr. 1984. "Passive Smoking,
Gas Cooking, and Respiratory Health of Chil-
dren Living in Six Cities." American Review
of Respiratory Disease 129:366-374.
Watson, W. and J. Jaksch. 1982. "Air Pollution:
Household Soiling and Consumer Welfare
Losses." Journal of Environmental Econom-
ics and Management. 9: 248-262.
Whittemore, A. S., and E. L. Korn. 1980. "Asthma
and Air Pollution in the Los Angeles Area."
American Journal of Public Health 70:687-
696.
D-52
-------
Appendix E: Ecological Effects of Criteria
Pollutants
Introduction
Benefits to human welfare from air pollution re-
ductions achieved under the CAA can be expected to
arise from likely improvements in the health of aquatic
and terrestrial ecosystems and the myriad of ecologi-
cal services they provide. For example, improvements
in water quality stemming from a reduction in acid
deposition-related air pollutants (e.g., SOx and NOx)
could benefit human welfare through enhancements
in certain consumptive services such as commercial
and recreational fishing, as well as non-consumptive
services such as wildlife viewing, maintenance of
biodiversity, and nutrient cycling. Increased growth
and productivity of U.S. forests could result from re-
duced emissions of ozone-forming precursors, particu-
larly VOCs and NOx, and thus may yield benefits from
increased timber production; greater opportunities for
recreational services such as hunting, camping, wild-
life observation; and nonuse benefits such as nutrient
cycling, temporary C02 sequestration, and existence
value.
In this Appendix, the potential ecological benefits
from CAA pollutant controls are discussed in the con-
text of three types of ecosystems: aquatic, wetland,
and forest. In describing the potential ecological ben-
efits of the CAA, it is clearly recognized that this dis-
cussion is far from being comprehensive in terms of
the types and magnitude of ecological benefits that
may actually have occurred from the implementation
of the CAA. Rather, this discussion reflects current
limitations in understanding and quantifying the link-
ages which exist between air quality and ecological
services, in addition to limitations in the subsequent
valuation of these services in monetary terms. This
discussion also does not cover potential benefits from
improvements in other ecological services, namely ag-
riculture and visibility, which are discussed and quan-
tified in other sections of this report. This appendix
is dedicated to a qualitative evaluation of ecological
benefits. However, where possible, the existing body
of scientific literature is drawn upon in an attempt to
provide insights to the possible magnitude of benefits
that may have resulted from CAA-related improve-
ments of selected ecological services. It is important
to note that the inability to fully value ecological ser-
vices results in a significant undervaluation of the
ecological benefits of air pollution reductions. This
undervaluation should not be interpreted as a devalu-
ation.
Benefits From Avoidance of
Damages to Aquatic Ecosystems
Aquatic ecosystems (lakes, streams, rivers, estu-
aries, coastal areas) provide a diverse range of ser-
vices that benefit the welfare of the human popula-
tion. Commercially, aquatic ecosystems provide a
valuable food source to humans (e.g., commercial fish
and shellfish harvesting), are used for the transporta-
tion of goods and services, serve as important drink-
ing water sources, and are used extensively for irriga-
tion and industrial processes (e.g., cooling water, elec-
trical generation). Recreationally, water bodies pro-
vide important services that include recreational fish-
ing, boating, swimming, and wildlife viewing. They
also provide numerous indirect services such as nu-
trient cycling, and the maintenance of biological di-
versity.
Clearly, these and other services of aquatic eco-
systems would not be expected to be equally respon-
sive to changes in air pollution resulting from the
implementation of the CAA. The available scientific
information suggests that the CAA-regulated pollut-
ants that can be most clearly linked to effects on
aquatic resources include SOx and NOx (through acid
deposition and increases in trace element
bioavailability), NOx (through eutrophication of ni-
trogen-limited water bodies), and mercury (through
changes in atmospheric deposition). Potential ben-
efits from each of these processes (acid deposition,
eutrophication, mercury accumulation in fish) are
described separately in the following sections.
E-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Acid Deposition
Background
Acid deposition refers to the depositing of strong
acids (e.g., H2S04, HN03) and weak acids ((NH^SO^
NH4N03) from the atmosphere to the earth's surface.
Acid deposition can occur in the wet or dry form and
can adversely affect aquatic resources through the
acidification of water bodies and watersheds. Acidi-
fication of aquatic ecosystems is of primary concern
because of the adverse effects of low pH and associ-
ated high aluminum concentrations on fish and other
aquatic organisms. Low pH can produce direct ef-
fects on organisms, through physiological stress and
toxicity processes, and indirect effects, mediated by
population and community changes within aquatic
ecosystems. Acidification can affect many different
aquatic organisms and communities. As pH decreases
to 5.5, species richness in the phytoplankton, zoop-
lankton, and benthic invertebrate communities de-
creases.1 Additional decreases in pH affect species
richness more significantly, and may sometimes af-
fect overall biomass.2 Table E-l presents descrip-
tions of the biological effects of acidification at dif-
ferent pH levels. In evaluating the severity of bio-
logical changes due to acidification, the reversibility
of any changes is an important consideration; biologi-
cal populations and communities may not readily re-
cover from improved water quality under certain cir-
cumstances. Researchers have addressed acidifica-
tion effects through many different experimental pro-
tocols, including laboratory bioassays, particularly
concerning pH, aluminum, and calcium; manipula-
tive whole-system acidification studies in the field;
and comparative, nonmanipulative field studies.
Although acidification affects phytoplankton,
zooplankton, benthic invertebrates, fish, amphibians,
and waterfowl, most acidification research has con-
centrated on fish populations3 Aluminum, which can
1 J. Baker et al., NAPAP SOS/T 13, 1990; Locke, 1993.
2 J. Baker et al., NAPAP SOS/T 13, 1990.
3 NAPAP, 1991.
4 J. Baker et al., NAPAP SOS/T 13, 1990.
5 J. Baker et al., NAPAP SOS/T 13, 1990.
6 Rosseland, 1986.
7 J. Baker et al., NAPAP SOS/T 13, 1990.
8 Mills etal., 1987.
9 NAPAP, 1991.
10 NAPAP, 1991.
be toxic to organisms, is soluble at low pH and is
leached from watershed soils by acidic deposition.4
Acidification may affect fish in several ways. The
direct physiological effects of low pH and high alu-
minum include increased fish mortality, decreased
growth, and decreased reproductive potential. The
mechanism of toxicity involves impaired ion regula-
tion at the gill.5 Population losses occur frequently
because of recruitment failure,6 specifically due to
increased mortality of early life stages.7 Changes at
other trophic levels may affect fish populations by
altering food availability.8 Fish in poorly buffered,
low pH water bodies may accumulate higher levels of
mercury, a toxic metal, than in less acidic water bod-
ies, due to increased mercury bioavailability. The
primary consequence of mercury accumulation ap-
pears to be hazardous levels to humans and wildlife
who consume fish, rather than direct harm to aquatic
organisms (discussed further below).
The CAA-regulated pollutants that are likely to
have the greatest effect on aquatic ecosystems through
acid deposition and acidification are S02 and NOx. In
the atmosphere, S02 and NOx react to form sulfate
and nitrate particulates, which may be dry-deposited;
also the pollutants may react with water and be wet-
deposited as dilute sulfuric and nitric acids. S02 is
considered the primary cause of acidic deposition,
contributing 75 to 95 percent of the acidity in rainfall
in the eastern United States.9
Current Impacts of Acid Deposition
Effects on Water Chemistry
The effects of acid deposition and resulting acidi-
fication of water bodies was intensively studied as part
of a 10-year, congressionally-mandated study of acid
rain problems in the United States.10 Based on the
NAPAP study, it is estimated that 4 percent of the
lakes and 8 percent of the streams in acid-sensitive
E-2
-------
Appendix E: Ecological Effects of Criteria Pollutants
Tabic E-l. Summary of Biological Changes with Surface Water Acidification.
PH
Decrease
Biological K fleets
6.5 to 6.0
Small decrease in species richness ofphytoplankton, zooplankton, and benthic invertebrate communities
resulting from the loss of some acid-sensitive species, but no measurable change in total com munity abundance
or production.
Som e adverse effects (decreased reproductive success) may occur for acid-sensitive fish species (e.g., fathead
minnow, striped bass).
6.0 to 5.5
Loss of sensitive species of minnows and dace, such as blacknosedace and fathead minnow; in some waters
decreased reproductive success of lake trout and walleye.
Distinct decrease in the species richness and change in species composition of the phytoplankton, zooplankton,
and benthic invertebrate communities.
Loss of a number of common invertebrate species from the zooplankton and benthic invertebrate communities,
including zooplankton species such as Diaptomus silicis,Mysis relicta, Epischura lacustris; many species of
snails, clams, mayflies, and amphipods, and some crayfish.
Visual accumulations of filamentous green algae in the littoral zone of many lakes and in some streams.
5.5 to 5.0
Loss of severalimportant sport fish species, including lake trout, walleye, rainbow trout, and smallmouth bass;
as well as additional non-game species such as creek chub.
Continued shift in the species composition and decline in species richness of the phytoplankton, periphyton,
zooplankton, and benthic invertebrate communities; decreases in the total abundance and biom ass of benthic
invertebrates and zooplankton may occur in som e waters.
Loss of several additional invertebrate species common in oligotrophic waters, includin gDtf phn ia galeata
mendotae, Diaphanosoma leuchtenhergiannm,Asplancha priodon ta\ all snails, most species of clams, and many
species of mayflies, stoneflies, and other benthic invertebrates.
Inhibition of nitrification.
Further increase in the extent and abundance of filamentous green algae in lake littoral areas and streams.
5.0 to 4.5
Loss of most fish species, including most important sport fish species such as brook trout and Atlantic salmon.
Measurable decline in the whole-system rates of decomposition of some forms of organic matter, potentially
resulting in decreased rates of nutrient cycling.
Substantial decrease in thenumber of species of zooplankton and benthic invertebrates, including loss of all
clams an dm any insects and crustaceans; measurable decrease in the total community biom ass of zooplankton
and benthic invertebrates in most waters.
Further decline in the species richness of the phytoplankton and periphyton communities.
Reproductive failure of some acid-sensitive species of amphibians such as spotted salamanders, Jefferson
salamandeis, and the leopard frog.
Source: Baker, J. et al. (NAPAP SOS/T 13, 1990), p. 13-173.
regions of the U.S. are chronically acidic due to natu-
ral and anthropogenic causes. NAPAP defines acidic
conditions as occurring when the acid neutralizing
capacity11 (ANC) is below 0 |_ieq/L. Furthermore, ap-
proximately 20 percent of the streams and lakes in
these regions are considered to be extremely suscep-
tible to acidity (defined as ANC <50 j^icq/L) and
slightly more than half show some susceptibility to
acidification (defined as ANC <200 (ieq/L).
In terms of the role of acid deposition as a causal
mechanism for the acidification of water bodies, it is
estimated that 75 percent of the 1,181 acidic lakes
and 47 percent of the 4,668 streams studied under
11 ANC is expressed in units of microequivalents per liter (^eq/L), where an equivalent ANC is the capacity to neutralize one
mole of H+ ions. Generally, waters with an ANC < 0 have corresponding pH values of less than 5.5 (L. Baker et al., NAPAP SOS/T
9, 1990).
E-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
NAPAP receive their dominant source of acid anions
from atmospheric deposition (see Table E-2). On a
regional basis, the importance of acid deposition var-
ies considerably, which is believed to result from re-
gional differences in SOx and NOx emissions and dif-
ferences in the biogeochemistry of individual water-
sheds. For acidic lakes (ANC <0), the regions that
appear most likely to be influenced by acid deposi-
tion include the Adirondacks and Mid-Atlantic High-
land region, with acid deposition cited as the domi-
Florida, where the vast majority (79 percent) are acidic
primarily due to organic acids, rather than acid depo-
sition.
Effects on Fish Habitat Quality
By combining information on relevant water
chemistry parameters (pH, aluminum, calcium), fish
toxicity models, and historical and current distribu-
tions of fish populations in the lakes and streams in-
Table E-2. Comparison of Population of Acidic National Surface Water Survey (NSWS) by
Chemical Category 1
Region
Number of
Deposition
Organic
Acid Mine
Watershed
Acidic
Dominated
Dominated
Drainage
Sulfate
W aters
Dominated
Dominated
(%)
(%)
(%)
(%)
LAKES
New England
173
79
21
—
—
Adirondacks
181
100
—
—
—
Mid-Atlantic Highlands
88
100
—
-
—
Southeastern Highlands
—
—
—
—
—
Florida
477
59
37
—
4
Upper Midwest
247
73
24
-
3
West
15
—
—
—
100
All Lakes
1,181
75
22
—
3
STRIiAMS
Mid-Atlantic Highlands
2,414
56
—
44
—
Mid-Atlantic Coastal Plain
1,334
44
54
-
2
Southeastern Highlands
243
50
—
50
—
Florida
677
21
79
—
—
All Streams
4,668
47
27
26
<1
'Source: NAPAP 1991 (Table 2.2-3,p. 28).
nant source of acidity in 100 percent of the acidic lakes
studied (Table E-2). This is in stark contrast to the
West region, where none of the acidic lakes studied
were dominated by acid deposition (notably, the
sample size of lakes for this region was small to be-
gin with). For acidic streams, the Mid-Atlantic High-
land region contains the greatest proportion of streams
whose acidic inputs are dominated by acid deposition
(56 percent). This contrasts with acidic streams of
12 NAPAP, 1991.
eluded in the National Surface Water Survey (NSWS),
NAPAP investigators estimated the proportion of
water bodies with water chemistry conditions that are
unsuitable for survival of various fish species.12 In
the Adirondack region, where the acidic lakes are
dominated by acid deposition, it is estimated that ten
percent of the lakes are unsuitable for the survival of
acid-tolerant fish species such as brook trout; twenty
percent of the lakes are estimated to be unsuitable for
E-4
-------
Appendix E: Ecological Effects of Criteria Pollutants
the survival of acid-sensitive species such as minnows.
About two percent and six percent of the lakes in the
New England region are estimated to be unsuitable
for acid-tolerant and acid-sensitive fish species, re-
spectively. A greater proportion of streams in the Mid-
Atlantic Highland region are estimated to be unsuit-
able for acid-tolerant and acid-resistant fish species
(18 percent and 30 percent, respectively); however,
about 44 percent of streams surveyed in this region
are thought to be heavily influenced by acid mine
drainage (Table E-2).
Economic Damages to Recreational Fishing
In an effort to assess some of the impacts from
existing levels of acid deposition to public welfare,
NAPAP investigated the current economic damages
associated with acid deposition to trout anglers of New
York, Maine, Vermont, and New Hampshire. The
general approach used consisted of linking the catch
per unit effort (CPUE) for four species of trout at in-
dividual lakes (estimated using participation survey
data) to the relevant water quality conditions at these
lakes (namely, the acid stress index or ASI). Using
historical water quality data, critical water quality
conditions (i.e., the ASI values) were estimated for
lakes in the absence of acid deposition and compared
to current conditions reflecting the presence of acid
deposition. Using two types of travel cost models, the
Random Utility Model (RUM) and Hedonic travel-
cost model (HTCM), estimates of the willingness to
pay (WTP) per trip of sampled trout anglers were ob-
tained. Aggregate estimates of the WTP were obtained
across the populations of trout anglers using statisti-
cal weighting factors. Finally, the difference in total
WTP between the current (acid deposition) scenario
and the historical (acid deposition-free) scenarios was
determined.
The resulting estimates of economic damages to
trout anglers in the four state region are relatively
small. Specifically, damage estimates range from $0.3
million to $1.8 million (in 1989 dollars) for the he-
donic travel-cost and random utility models, respec-
tively. By many accounts, these estimates can be con-
sidered to underestimate actual damages to anglers in
these states. First, data limitations precluded the de-
velopment of meaningful WTP estimates for brook
trout anglers, which may be a significant component
of trout fishing in these areas. Second, resource con-
straints necessitated exclusion of a large population
of trout anglers (i.e., those residing inNew York City).
Third, the economic damage estimates were limited
to trout anglers, thus excluding potentially similar if
not greater economic damages to anglers fishing for
other coldwater or warmwater fish species. In addi-
tion, the NAPAP analysis was performed in the con-
text of recreational fishing in lakes, thereby exclud-
ing potentially important welfare impacts from recre-
ational fishing in streams. Finally, these estimates do
not address non-use values of lakes in this region.
Benefits From Acid Deposition Avoidance Under
the CAA
It is currently estimated that in the absence of
pollution reductions achieved under the Clean Air Act,
total sulfur emissions to the atmosphere would have
increased by nearly sixteen million tons byl990,a40
percent increase above 1990 levels estimated with
CAA controls remaining in place.13 Based on atmo-
spheric transport and deposition modeling, this in-
crease in sulfur emissions corresponds to an approxi-
mate 25 to 35 percent increase in total sulfur deposi-
tion (wet & dry) in large portions of the northeastern
portion of the United States.14 Given sulfur emission
and deposition changes of this magnitude, and the
importance of sulfur emissions in contributing to acid
deposition, one would expect some benefits to human
welfare to be achieved as a result of improved quality
of aquatic ecosystems. To date, however, no formal
benefits assessment of CAA-avoided acid deposition
impacts has been conducted for aquatic ecosystems.
Nevertheless, past benefit assessments involving acid
deposition impacts on aquatic ecosystems provide
some opportunity to gain insights into the relative
magnitude of certain aquatic-based benefits that may
be achieved through pollution reductions under the
CAA.15
Recreational Fishing
NAPAP evaluated the impact of changes in acid
deposition on use values of aquatic ecosystems (i.e.,
recreational fishing)6 In their integrated assessment,
NAPAP valued the impacts of three different sulfur-
13 U.S. EPA, 1995; Table B-2.
14 U.S. EPA 1995, p. 3-10.
15 See, for example, NAPAP, 1991.
16 NAPAP, 1991.
E-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
induced acid deposition scenarios to trout anglers from
NY, VT, NH and ME.17 The three scenarios evalu-
ated were:
1. No change in acid deposition.
2. A 50 percent reduction in acid deposition.
3. A 30 percent increase in acid deposition.
As described above, equations were developed by
NAPAP to estimate the catch per hour for species at
each lake as a function of the ASI value for each lake
and of the technique of the fishers. Baseline and pre-
dicted changes in CPUE were evaluated for all lakes
modeled in the region. Willingness-to-pay estimates
for CPUE per trip were derived for the baseline and
sulfur emission scenarios using two travel-cost mod-
els, a random utility model and a hedonic travel cost
model. These willingness-to-pay estimates were then
combined with the results of a participation model
that predicted the total number of trips taken by trout
anglers. Total welfare changes were determined over
a 50 year period (from 1990 to 2040).
At current levels of acid deposition, NAPAP esti-
mates that trout anglers in these four states will expe-
rience annual losses by the year 2030 of $5.3 or $27.5
million (in 1989 dollars) for the random utility model
and hedonic travel cost model, respectively (see Table
E-3). If acid deposition increases by 30 percent, which
roughly corresponds to the 25 to 35 percent increase
predicted for the northeast U.S. in the absence of CAA
sulfur controls,18 the resulting economic losses to trout
anglers in 2030 would range from $ 10 million to nearly
$100 million annually (in 1989 dollars) for the RUM
and HTCM, respectively. If deposition decreases by
50 percent, annual benefits to recreational anglers are
estimated to be $14.7 million (RUM) or $4.2 million
(HTCM).
While an estimation of CAA-related benefits to
trout anglers based on the 30 percent increase in acid
deposition scenario has some appeal, a strict transfer
of these benefits to the section 812 retrospective analy-
sis is hindered by several factors. First, the NAPAP
benefits estimates are projected for future conditions
(the year 2030). Therefore, the extent to which the
NAPAP benefits reflect conditions and benefits in
1990 (the focus of the section 812 retrospective as-
sessment) is unclear. Second, the NAPAP and CAA
section 812 analyses operate from different baselines
(1990 for the NAPAP study versus 1970-1990 for the
section 812 study). However, the NAPAP estimates
of annual benefits of $10 to $100 million provide a
rough benchmark for assessing the likely magnitude
of the avoided damages to an important and sensitive
recreational fishery in a four-state area most impacted
by surface water acidification from atmospheric depo-
sition.
Tabic E-3. Results from Benefits Assessments of Aquatic
Ecosystem Use Values from Acid Deposition Avoidance.
Study
Use
Viilue
Scenario Modeled
Method
A1111 ii»l Benefits
NAPAP
(1991)
Trout
Fishing
No change in acid
deposition
RUM
HTCM
-$5.3 million
- $27.5 million
(NY,
ME, VT,
NH)
50% decrease in acid
deposition
RUM
HTCM
$14.4 million
$4.2 million
30% increase in acid
deposition
RUM
HTCM
-$10.3 million
-$97.7 million
No new emission
reductions after 1985
RUM
HTCM
-$5.5 million
-$3.5 million
10 million ton reduction
of SCh from 1980 levels
by 2000
RUM
HTCM
$9.7 million
$4.4 million
17 NAPAP, 1991; p. 383-384.
18 U.S. EPA, 1995.
Eutrophication
Eutrophication is the process by
which aquatic systems respond to nu-
trient enrichment. The most common
nutrients involved in eutrophication are
nitrogen and phosphorous (and related
chemical species). When water bod-
ies receive excessive amounts of nu-
trients, adverse impacts on their resi-
dent species and on ecosystem func-
tions can occur from excessive algal
growth and the reduction in dissolved
oxygen caused by decaying algal bio-
mass. Under highly eutrophic condi-
tions, excessive nutrients can cause
depleted oxygen levels that result in
subsequent loss of economically im-
portant benthic organisms (shellfish),
fish kills, and changes in phytoplank-
ton, zooplankton, and fish community
E-6
-------
Appendix E: Ecological Effects of Criteria Pollutants
composition.19 Nuisance algal blooms can have nu-
merous economic and biological costs, including wa-
ter quality deterioration affecting biological resources,
toxicity to vertebrates and higher invertebrates, and
decreased recreational and aesthetic value of waters.20
Although severe eutrophication is likely to adversely
affect organisms, especially fish, a moderate increase
in nutrient levels may also increase fish stocks, by
increasing productivity in the food chain.21
Atmospheric Deposition and Eutrophication
The deposition ofNOx in aquatic systems and their
watersheds is one source of nitrogen that may con-
tribute to eutrophication. The relative importance of
NOx deposition as a contributor to aquatic eutrophi-
cation depends on the extent to which the productiv-
ity of an aquatic ecosystem is limited by nitrogen avail-
ability and the relative importance of nitrogen depo-
sition compared to other internal and external sources
of nitrogen to the aquatic ecosystem. Furthermore,
the vulnerability of aquatic ecosystems to eutrophi-
cation is known to vary seasonally and spatially, al-
though these systems are affected by nutrient deposi-
tion throughout the year. In general, freshwater eco-
systems appear to be more often limited by phospho-
rus, rather than nitrogen, and are not as likely to be
heavily impacted by nitrogen deposition compared to
some estuarine and coastal ecosystems.22 In contrast
to acidification of streams and lakes, eutrophication
from atmospheric deposition of nitrogen is more com-
monly found in coastal and estuarine ecosystems,
which are more frequently nitrogen-limited.23
Unfortunately, there is limited information with
regard to the relative importance of atmospheric depo-
sition as a nitrogen source in many estuarine and ma-
rine ecosystems. Estimates of the importance of at-
mospheric nitrogen deposition are difficult to make
because of uncertainties in estimating deposition, es-
pecially dry deposition, as well as watershed nitrogen
retention.24 Paerl (1993) reviews the importance of
atmospheric nitrogen deposition as a contributor to
eutrophication of coastal ecosystems; he concludes
that 10 to 50 percent of the total nitrogen loading to
coastal waters is from direct and indirect atmospheric
deposition. Estimates for the economically impor-
tant Chesapeake Bay indicate that about 25 to 40 per-
cent of the nitrogen loadings to the bay occur via at-
mospheric deposition.25 Hinga et al. (1991) estimate
that anthropogenic deposition provides 11 percent of
total anthropogenic inputs of nitrogen in Narragansett
Bay, 33 percent for the New York Bight, and 10 per-
cent for New York Bay. Fisher and Oppenheimer
(1991) estimate that atmospheric nitrogen provides
23 percent of total nitrogen loading to Long Island
Sound and 23 percent to the lower Neuse River in
North Carolina. Information on the importance of
atmospheric nitrogen deposition for most other U.S.
coastal ecosystems is not available in the literature.
Episodic atmospheric inputs of nitrogen may be an
important source of nitrogen to nutrient-poor marine
ecosystems, such as the North Atlantic near Bermuda
and the North Sea.26
Valuing Potential Benefits from Eutrophication
Avoidance Under the CAA
It is currently estimated that in the absence of
pollution reductions achieved under the Clean Air Act,
total nitrogen emissions to the atmosphere would have
increased by nearly 90 million tons by 1990, a two-
fold increase above 1990 levels estimated with CAA
controls remaining in place.27 However, the ability
to determine the potential economic benefit from such
a reduction in nitrogen emissions is heavily con-
strained by gaps in our current biological and eco-
nomic knowledge base of aquatic ecosystems.
One water body that has received much study in
the area of nitrogen-induced eutrophication is Chesa-
peake Bay. As previously discussed, it is estimated
that atmospheric deposition of nitrogen contributes
approximately 25 percent to the total nitrogen load-
19 Paerl, 1993.
20 Paerl, 1988.
21 Hansson and Rudstam, 1990; Rosenberg et al., 1990; Paerl, 1993.
22 Hecky andKilham, 1988; Vitousek and Howarth, 1991.
23 U.S. EPA, 1993; Paerl, 1993.
24 U.S. EPA, 1993.
25 U.S. EPA, 1994.
26 Owens et al., 1992.
27 U.S. EPA, 1995; Table B-3.
E-7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
ings to the bay.28 In deposition terms, an estimated
15 to more than 25 percent increase in total nitrogen
deposition has been forecast in the Chesapeake Bay
watershed by 1990 in the absence of CAA pollution
controls.29 These results are based on an estimated
40,000 tons of atmospherically deposited nitrogen (as
nitrate and ammonia) to Chesapeake Bay in 1985,30
which means a 20 percent increase in atmospheric
deposition would amount to approximately 8,000 ad-
ditional tons.
One indirect method available to gauge the po-
tential economic relevance of avoidance of such at-
mospheric nitrogen loadings to Chesapeake Bay is
through the avoidance cost of nitrogen controls. How-
ever, such an assessment is difficult due to the site,
facility, and treatment-specific variation in treatment
costs. For example, Camacho (1993) reviewed nitro-
gen treatment costs for chemical treatment of water
from important point sources (mostly public owned
treatment works) and found that costs ranged from
$9,600 to $20,600 per ton (annual costs, 1990 dol-
lars), depending on the facility evaluated. Biological
treatment of nitrogen from point sources was far more
expensive, varying from $4,000 to $36,000 per ton.
For control of non-point source loading, values of ni-
trogen removal practices ranged from $1,000 to
$285,000 per ton.31 Taking chemical addition as one
possible example, the avoided costs of treatment of
8,000 tons of nitrogen would range from about $75
million to about $170 million annually (in 1990 dol-
lars).
Mercury
Mercury, in the form of methyl mercury, is a neu-
rotoxin of concern and can accumulate in tissue of
fish to levels that are hazardous to humans and aquatic-
feeding wildlife in the U.S. In relation to the section
812 CAA retrospective analysis, mercury is of inter-
est for two reasons. First, potential benefits to human
welfare may have occurred as a result of mercury
emission controls implemented under EPA's National
Emission Standards for Hazardous Air Pollutants
(NESHAP). Second, experimental and observational
evidence suggests that acidification of water bodies
enhances mercury accumulation in fish tissues.32
Therefore, CAA-mandated reductions in sulfur and
nitrogen oxide emissions and subsequent acid depo-
sition may have resulted in indirect benefits from a
reduction in mercury accumulation in fish and subse-
quent improvements to human health and welfare.
The accumulation of mercury to hazardous levels
in fish has become a pervasive problem in the U.S.
and Canada. A rapid increase in advisories occurred
during the 1980s, including a blanket advisory affect-
ing 11,000 lakes in Michigan.33 The Ontario Minis-
tries of Environment and Natural Resources (1990)
recommend fish consumption restrictions for 90 per-
cent of the walleye populations, 80 percent of small-
mouth bass populations, and 60 percent of lake trout
populations in 1,218 Ontario lakes because of mer-
cury accumulation. In many instances, mercury has
accumulated to hazardous levels in fish in highly re-
mote water bodies that are free from direct aqueous
discharges of mercury.34 Mass balance studies have
shown that atmospheric deposition of mercury can
account for the accumulation of mercury in fish to
high levels in lakes of these remote regions.35 The
potential impacts of mercury on the health of humans
and fish-eating (piscivorous) wildlife has lead EPA
to recently establish water quality criteria to protect
piscivorous species in the Great Lakes.36
Although mercury accumulation in fish via atmo-
spheric deposition is now widely recognized as a po-
tential hazard to human health and certain wildlife
species, studies establishing quantitative linkages be-
tween sources of mercury emissions, atmospheric
deposition of mercury, and subsequent accumulation
in fish are lacking. Thus at the present time, we are
unable to quantify potential benefits from CAA-
avoided mercury accumulation in fish of U.S. water
28 U.S. EPA, 1993.
29 U.S. EPA 1995, Figure C-6.
30 NERA, 1994.
31 Shuyler, 1992.
32 Bloom et al., 1991; Watras and Bloom, 1992; Miskimmin et al., 1992; Spry and Wiener, 1991; Wiener et al., 1990.
33 Watras et al., 1994.
34 Glass et al., 1990; Sorenson et al., 1990; Grieb et al. 1990; Schofield et al. 1994.
35 Fitzgerald et al. ,1991.
36 U.S. EPA, 1995.
E-8
-------
Appendix E: Ecological Effects of Criteria Pollutants
bodies. Given the pervasiveness of the mercury prob-
lem with CAA-pollution controls, potential benefits
to human health and welfare from avoidance of fur-
ther mercury related damages to aquatic ecosystems
could be substantial.
It should also be noted that atmospheric deposi-
tion is a major contributor to surface water loads of
other toxic pollutants as well. For example, scientists
believe that about 35 to 50 percent of the annual load-
ings of a variety of toxic chemicals to the Great Lakes
may be from the air; for lead, atmospheric deposition
currently accounts for an estimated 95 percent of the
total load in the Great Lakes.37 CAA-related reduc-
tions in air emissions of toxic pollutants (such as lead)
undoubtedly reduced the loading of these chemicals
to the Great Lakes and other water bodies; the magni-
tude of the benefits of reducing these exposures to
humans and wildlife is not known.
Benefits from A voided Damages
to Wetland Ecosystems
Introduction
This review addresses the effects of air pollutants
on wetland ecosystems; the focus is on acidification
and nutrient loading. Valuable service flows of wet-
land ecosystems include flood control, water quality
protection and improvement, wildlife and fish habi-
tat, and biodiversity. The limited scientific evidence
suggests that air pollutants may most affect
biodiversity, in particular because of nutrient loading
through nitrogen deposition.
Wetlands are broadly characterized as transitional
areas between terrestrial and aquatic systems in which
the water table is at or near the surface or the land is
periodically covered by shallow water.38 Types of
wetlands include swamps (forested wetlands), marshes
(herbaceous vegetation), and peatlands, which are
wetlands that accumulate partially decayed vegeta-
tive matter due to limited decomposition.39 Peatlands
include bogs and fens. Bogs receive water solely from
precipitation, are generally dominated by Sphagnum
moss, and are low in nutrients. Fens receive water
from groundwater and precipitation, contain more
marsh-like vegetation, and have higher pH and nutri-
ent levels than bogs.40 Most of the limited work on
the effects of atmospheric deposition on wetlands has
been done in peatlands, specifically in Europe, where
levels of atmospheric deposition are generally much
higher than in the U.S.
The air pollutants of greatest concern with respect
to effects on wetland ecosystems are oxides of nitro-
gen (NOx) and oxides of sulfur (SOx), primarily sul-
fur dioxide (S02). Air pollutants may affect wetland
ecosystems by acidification of vulnerable wetlands
and by increasing nutrient levels. Acidification in
vulnerable wetlands may affect vegetation adversely,
as appears to have occurred in Europe. In wetlands
where nitrogen levels are low, increased nitrogen
deposition may alter the dynamics of competition
between plant species. Species adapted to low-nitro-
gen levels, including many endangered species, may
decrease in abundance.41
Effects of Acidification
Limited evidence suggests that acidic deposition
and decreased pH may harm certain wetland plants,
alter competitive relations between wetland plants and
cause changes in wetland drainage and water reten-
tion.
Work concerning the possible acidification of
peatlands is inconclusive. Acidic deposition is un-
likely to result in displacement of base cations from
cation exchange sites in bogs, and therefore it will not
cause a drop in pH.42 Peatland sediments are low in
Al3+, so mobilization of toxic aluminum is not a con-
cern as it is in forest soils and aquatic ecosystems.43
Acidification might affect certain fen ecosystems.
Gorham et al. (1984) have hypothesized that acidic
deposition could leach base cations from mineral-poor
fens and decrease pH levels. This could result in a
37U.S. EPA, 1994.
38 Cowardin et al., 1979.
39 Mitsch and Gosselink, 1986.
40 Mitsch and Gosselink, 1986.
41 U.S. EPA, 1993.
42 Gorham etal., 1984.
43 Turner et al., NAPAP SOS/T 10, 1990.
E^9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
transition to bog vegetation such as Sphagnum and
away from sedge meadow vegetation. At this time,
this remains a hypothesis; however, pH did not de-
crease in a mineral-poor Ontario fen during a four-
year period in which researchers experimentally in-
creased acidic deposition.44
In European wetlands affected by high levels of
deposition for many years, acidic deposition has seri-
ously affected wetland vegetation. Roelofs (1986)
reports that acidification of heath pools in the Nether-
lands has caused a change in species composition with
Sphagnum and rushes replacing the original vegeta-
tion. Likewise, significant declines in Sphagnum in
British bogs have occurred in areas affected by 200
years of atmospheric pollution, including nitrogen
deposition.45 It is unclear how such changes have af-
fected wetland service flows apart from the effects on
biodiversity; however, water retention has decreased
and significant erosion has occurred in seriously per-
turbed British bogs near Manchester and Liverpool.46
Effects of Nutrient Loading
Atmospheric deposition may affect wetlands by
increasing the level of nutrients, particularly nitrogen,
in wetlands. Sulfur is not a limiting nutrient in
peatlands,47 but nitrogen commonly limits plant
growth.48 The effects of increased nitrogen levels in
wetlands include an increased threat to endangered
plant species and possible large-scale changes in plant
populations and community structure. Endangered
and threatened plant species are common in wetlands,
with wetland species representing 17 percent of the
endangered plant species in the U.S. (U.S. EPA, 1993).
These plants are often specifically adapted to low ni-
trogen levels; examples include isoetids49 and insec-
tivorous plants.50 In eastern Canadian wetlands, na-
tionally rare species are most common in infertile
sites.51 When nitrogen levels increase, other species
44 Rochefort et al., 1990.
45 Lee et al., 1986.
46 Lee et al., 1986.
47 Turner et al., NAPAP SOS/T 10, 1990.
48 U.S. EPA, 1993.
49 Boston, 1986.
50 Moore et al., 1989.
51 Moore et al., 1989; Wisheu and Keddy, 1989.
52 Lee & Woodin 1988, Aerts et al., 1992.
53 Ferguson et al., 1984; Lee et al., 1986.
adapted to higher levels of nitrogen may competitively
displace these species. Thus, NOx emissions that in-
crease nitrogen levels in nitrogen-poor wetlands may
increase the danger of extinction for threatened and
endangered species.
By changing competitive relations between plant
species, increased nitrogen deposition may broadly
affect community structure in certain wetlands. Com-
mon species that thrive in nitrogen-poor wetlands may
become less abundant. Many nitrogen-poor bogs in
the northern U.S. are dominated by Sphagnum spe-
cies. These species capture low levels of nitrogen from
precipitation. Increased nitrogen levels may directly
harm Sphagnum and cause increased nitrogen to be
available to vascular plants that may out compete Sph-
agnum.52 Studies in Great Britain have documented
large declines in Sphagnum moss because of atmo-
spheric pollution;53 nitrogen loading may play an im-
portant role in these declines. However, Rochefort et
al. (1990) document limited effects of fertilization
from experimentally-increased N03" and S042 depo-
sition on an Ontario mineral-poor fen over a four-year
period, apart from initially increased Sphagnum
growth. Thus, increased nitrogen loading might ad-
versely or beneficially affect wetland plants depend-
ing on baseline nitrogen concentrations in the wet-
land, atmospheric nitrogen loading, and species re-
quirements for and sensitivity to nitrogen.
Increases in nitrogen levels due to NOx emissions
will have the greatest effect on wetlands that are ex-
tremely nitrogen-limited and that receive small
amounts of nitrogen naturally. Since bogs, including
Sphagnum bogs, receive little surface water runoff,
they get most of their nutrient and water loadings
through precipitation. These bogs may receive a total
of approximately 10 kg nitrogen per hectare per year
(kg N/ha/yr), which is one to two orders of magnitude
less nitrogen than other freshwater wetlands and
E-10
-------
Appendix E: Ecological Effects of Criteria Pollutants
saltmarshes receive.54 As atmospheric deposition of
nitrogen has been estimated to be at least 5.5 to 11.7
kg N/ha/yr,55 changes in NOx emissions would most
likely affect these bogs. The results of a model by
Logofet and Alexandrov (1984) suggest that a tree-
less, nutrient-poor bog may undergo succession to a
forested bog because of the input of greater than 7 kg
N/ha/yr.
As in freshwater wetlands, significantly increased
nitrogen deposition to coastal wetlands will increase
productivity and alter the competitive relationships
between species.56 However, studies showing this
increased productivity have used 100 to 3000 kg N/
ha/yr.57 Therefore, limited changes in NOx emissions
may not affect coastal wetland productivity.
Summary of Wetland Ecosystem Effects
The effects of air pollutants on wetlands have re-
ceived little attention, in contrast to the large body of
work on the effects of acid rain on aquatic and forest
ecosystems. Little evidence exists suggesting that
acidification due to atmospheric deposition is a ma-
jor threat to wetlands. In particular, peatlands are
naturally acidic, although mineral-poor fens may be
at risk from acidification. Nitrogen loading may alter
community composition in wetlands naturally low in
nutrients, such as bogs. Nitrogen loading may threaten
rare species adapted to low nitrogen levels. In Britain
and The Netherlands, heavy atmospheric deposition
over a long period appears to have caused serious de-
clines in Sphagnum in peatlands.
Air pollutants appear to most seriously threaten
rare and endangered species, biodiversity, and com-
munity composition in wetlands, particularly bogs.
These changes are difficult to associate with changes
in economic value; even the qualitative nature of the
effects is uncertain. Air pollutants may not signifi-
cantly affect such important wetland service flows as
flood control, water quality protection, and wildlife
habitat in most wetlands, so the impacts on the more
readily monetized aspects of the economic value of
wetlands may be limited.
Benefits from A voided Damages
to Forests
Introduction
Forests occupy 33 percent of the land mass in the
U.S. (some 738 million acres) and provide a wealth
of services to the U.S. population.58 Notable services
provided by forests include timber production, recre-
ational opportunities such as hunting, camping, hik-
ing, and wildlife observation, water quality protec-
tion, nutrient removal and cycling, flood control, ero-
sion control, temporary carbon sequestration, preser-
vation of diversity, and existence values. In 1991,
hunting participation alone accounted for 236 million
recreation days that included 214 million person trips
with estimated expenditures valued at $12.3 billion.59
The Clean Air Act-regulated pollutants of great-
est concern with respect to effects on forest ecosys-
tems are oxides of sulfur (SOx), primarily sulfur di-
oxide (S02), oxides of nitrogen (NOx), and volatile
organic compounds (VOCs). While extremely high
ambient concentrations of S02 and NOx may directly
affect vegetation, such effects are uncommon in the
U.S.;60 the indirect effects of these pollutants are of
greater concern. Specifically, emissions of S02 and
NOx are known to contribute to acid deposition in
portions of the United States, with S02 contributing
75 to 95 percent of the acidity in rainfall in the east-
ern U.S.61 Acid deposition is of concern to forests
primarily from the acidification of soils (i.e., by re-
ducing seed germination, altering nutrient and heavy
metal availability). Direct foliar damage can occur
from precipitation with extremely low pH levels (i.e.,
3.0-3.6 and below), although these levels are lower
than ambient levels in the U.S.62 VOCs and NO are
54 U.S. EPA, 1993.
55 U.S. EPA, 1993.
56 U.S. EPA, 1993.
57 U.S. EPA, 1993.
58 Powell etal. 1993.
59 U.S. DOI, 1993.
60 Shriner et al., NAPAP SOS/T 18, 1990.
61 NAPAP, 1991.
62 Shriner et al., NAPAP SOS/T 18, 1990.
E-ll
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
important precursors to ozone formation, which can
affect leaf photosynthesis and senescence and decrease
cold hardiness, thereby causing deleterious impacts
on tree growth, survival and reproduction. Deposi-
tion of NOx may also alter the nutrient balance of for-
est soils, which in turn might alter the competitive
relationships between tree species and affect species
composition and diversity.63
Current Air Pollutant Effects on Forests
Acid Deposition Impacts
In 1985, NAPAP organized the Forest Response
Program (FRP) to evaluate the significance of forest
damage caused by acidic deposition, the causal rela-
tionships between air pollutants and forest damage,
and the dynamics of these relationships regionally.
Research was focussed on four forest regions: East-
ern Spruce-Fir, Southern Commercial Forests, East-
ern Hardwoods, and Western Conifers. With the ex-
ception of high-elevation spruce-fir forests, the avail-
able evidence suggests that acidic deposition does not
currently affect these forests and that observed de-
clines in sugar maple and southern pines are not due
to acidic deposition.64
Circumstantial evidence suggests that acidic depo-
sition may affect high-elevation spruce-fir forests in
the northeastern U.S. These forests have extensive
contact with acidic cloud water.65 Experimental evi-
dence suggests that acidic deposition may affect cold
hardiness in red spruce, an important component of
the spruce-fir forest. Significant declines in red spruce
growth and in its importance in the forest have oc-
curred in New York and northern New England. The
proximate cause of death of red spruce in the region
is pathogens and insects; acidic deposition may inter-
act with these biological stresses and with weather-
induced stress to produce adverse effects in red spruce.
Ozone may also play a role in red spruce decline in
this region.66 Available evidence suggests that soil
aluminum and soil pH levels have not affected red
spruce adversely.67
63 U.S. EPA, 1993.
64 Barnard et al., NAPAP SOS/T 16, 1990; NAPAP, 1991.
65 Barnard et al., NAPAP SOS/T 16, 1990.
66 Shriner et al., NAPAP SOS/T 18, 1990.
67 Barnard et al., NAPAP SOS/T 16, 1990.
68 U.S. EPA, 1996a.
69 Hogsettetal. , 1995.
Ozone Impacts
Experimental Evidence
For practical reasons, the majority of experimen-
tal evidence linking ozone exposure to damage to tree
species has been derived from studies of individual
plants, especially seedling and branch studies.68 Re-
sults from these studies suggest that ozone exposure
can reduce photosynthesis and increase senescence in
leaves. Subsequently, such effects from ozone may
alter the carbohydrate allocation to plant tissues such
as roots, which may affect plant growth and cold har-
diness. Decreases in cold tolerance may be particu-
larly important for trees in northern latitudes and high
elevations. Recent work on quantifying the relation-
ship between ozone exposure and plant responses sug-
gest that seedlings of aspen, ponderosa pine, black
cherry, tulip poplar, sugar maple, and eastern white
pine seedlings may experience biomass reductions of
approximately 10 percent at or near ambient ozone
exposures.69 Because trees are perennials, the effect
of even a 1-2 percent per year loss in seedling biom-
ass (versus 10 to 20 percent yield loss in crops), if
compounded over multiple years under natural field
conditions of competition for resources, could be se-
vere.
Although indicative of short-term relative re-
sponse to ozone exposure, results from these experi-
ments are unable to provide reliable information on
the long-term effects of ozone on forests. This limi-
tation arises because the effects of ozone on forests
will depend on both the response of individual plants
to ozone exposure and the response of populations of
plants, which interact with their environment. Popu-
lation response will be altered by the varying intraspe-
cific genetic susceptibility to ozone. Individual plant
response will also be affected by many environmen-
tal factors, including insect pests, pathogens, plant
symbionts, competing plants, moisture, temperature,
light, and other pollutants. Consistent evidence on
the interaction of ozone with other environmental fac-
tors is lacking. Furthermore, most experimental stud-
E-12
-------
Appendix E: Ecological Effects of Criteria Pollutants
ies have only studied exposure for one growing sea-
son; effects on forest species may occur over de-
cades.70 Therefore, considerable uncertainties occur
in scaling across individuals of different ages, from
individuals to populations and communities, and
across time.
Observational Evidence
Studies of the forests of the San Bernardino Moun-
tains provide the strongest case for linking ozone ex-
posure to damages to an entire forest ecosystem. These
forests have been exposed to extremely high ambient
ozone levels over the past 50 years due to their prox-
imity to the Los Angeles area. The area has been ex-
tensively studied regarding the effects of ozone, as
described in U.S. EPA (1996a). The ecosystem has
been seriously affected by ozone pollution, with the
climax-dominant, but ozone-sensitive ponderosa pine
and Jeffrey pine declining in abundance, replaced by
more ozone-tolerant species. These sensitive species
have experienced decreased growth, survival, and re-
production, and susceptibility to insects. The effects
of ozone on these species have resulted in other eco-
system effects, including the buildup of a large litter
layer, due to increased needle senescence. The de-
cline of the fire-tolerant ponderosa and Jeffrey pines
may seriously affect the fire ecology of the ecosys-
tem, with fire-sensitive species becoming more com-
mon. Ozone concentrations have been declining in
recent decades, and crown injury of ponderosa and
Jeffrey pine has decreased. However, the two species
have continued to decline in abundance, as measured
by total basal area, compared with other species over
the period 1974 to 1988.71 The nature of community
dynamics, particularly in mixed species, uneven aged
stands, indicates that subtle long-term forest responses
(e.g., shifts in species composition) to elevated levels
of a chronic stress like exposure to ozone are more
likely than wide-spread community degradation.72
Limited field studies have been completed in other
forest ecosystems. Foliar injury has been observed in
the Jefferson and George Washington National For-
ests and throughout the Blue Ridge Mountains, in-
cluding areas of the Shenandoah National Park.73 In
the Great Smoky Mountains National Park, surveys
made in the summers from 1987 through 1990 found
95 plant species exhibited foliar injury symptoms con-
sistent with those thought to be caused by ozone.74
Foliar ozone injury has also been documented in Na-
tional Parks and Forests in the Sierra Nevada moun-
tains.75
Growth and productivity of seedlings have been
reported to be affected by ozone for numerous spe-
cies in the Blue Ridge Mountains of Virginia. In the
Shenandoah National Park, Duchelle et al. (1982,
1983) found that tulip poplar, green ash, sweet gum,
black locust, as well as several evergreen species (e.g.,
Eastern hemlock, Table Mountain pine, pitch pine,
and Virginia pine), common milkweed, and common
blackberry all demonstrated growth suppression of
seedlings. Except for the last two species mentioned,
almost no visible injury symptoms accompanied the
growth reductions. Studies of mature trees in the
Appalachian Mountains also indicate that injury as-
sociated with exposure to ozone and other oxidants
has been occurring for many years.76 Researchers have
also found that major decreases in growth occurred
for both symptomatic and asymptomatic trees during
the 1950s and 1960s in the Western U.S.77 The ad-
verse response of a number of fruit and nut trees to
ozone exposure has been reported.78
Monitoring by the USDA Forest Service shows
that growth rates of yellow pine in the Southeast have
been decreasing over the past two decades in natural
stands but not in pine plantations.79 Solid evidence
linking this growth reduction to air pollutants is lack-
70 U.S. EPA, 1996a.
71 Miller et al., 1989 and Miller et al., 1991.
72 Shaver et al., 1994
73 Hayes and Skelly, 1977; Skelly et al., 1984
74 Neufeld, et al., 1992
75 Peterson and Arbaugh, 1992
76 Benoit et al., 1982
77 Peterson et al., 1987; Peterson and Arbaugh, 1988, 1992; Peterson et al., 1991
78 McCool and Musselman, 1990; Retzlaff et al., 1991, 1992a, b
79 NAPAP, 1991.
E-13
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
ing, although ozone, in particular, may be a factor.80
Ambient ozone levels in the region are high enough
to damage sensitive tree species, including pine seed-
lings during experimental exposure .81 Due to the com-
mercial importance of yellow pine, the economic im-
pacts of ozone on forest ecosystems in this area could
be significant if ozone is affecting growth.
Although the ecosystem effects occurring in the
San Bernardino forest ecosystem have occurred at very
high ozone exposures, lower ozone exposure else-
where in the U.S. may still affect forests. The EPA
Ozone Staff Paper82 assessed the risk to vegetation,
including forests, under current ambient air quality.
Using a GIS approach, it was found that under the
base year (1990) air quality, a large portion of Cali-
fornia and a few localized areas in North Carolina and
Georgia have seasonal ozone levels above those which
have been reported to produce greater than 17 percent
biomass loss in 50 percent of studied tree seedling
species. A broader multistate region in the east is
estimated to have air quality sufficient to cause 17
percent biomass loss in seedlings, while at least a third
of the country, again mostly in the eastern U.S., most
likely has seasonal exposure levels which could al-
low up to 10 percent yield loss in 50 percent of stud-
ied seedlings. The Staff Paper did not present mon-
etized benefits because of lack of exposure-response
functions.83
Even small changes in the health of ozone-sensi-
tive species may affect competition between sensi-
tive and tolerant species, changing forest stand dy-
namics.84 Depending on the sensitivities of individual
competing species, this could affect timber produc-
tion either positively or negatively, and affect com-
munity composition and, possibly, ecosystem pro-
cesses.
Endangered species
Ozone effects may also reduce the ability of af-
fected areas to provide habitats to endangered spe-
cies. For example, two listed endangered plant spe-
cies, the spreading aven and Roan Mountain bluet,
are currently found at a small number of sites in east-
ern Tennessee and western North Carolina—forested
areas where ozone-related injury is of concern.85 In
addition, ozone-related effects on individual ozone-
sensitive species that provide unique support to other
species can have broader impacts. For example, one
such species is the common milkweed, long known
for its sensitivity to ozone and usefulness as an indi-
cator species of elevated ozone levels, as well as be-
ing the sole food of the monarch butterfly larvae.
Thus, a major risk associated with of the loss of milk-
weed foliage for a season is that it might have signifi-
cant indirect effects on the monarch butterfly popula-
tion. A large number of studies have shown that
ozone-sensitive vegetation exists over much of the
U.S., with many native species located in forests and
Class I areas, which are federally mandated to pre-
serve certain air quality related values.
Valuation of Benefits From CAA-
Avoided Damages to Forests
Background
To quantitatively assess the economic benefits of
avoided damages of relevant CAA pollutants to for-
ests, it is necessary to link estimated changes in air
pollution to measures of forest health and conditions
that can be readily quantified in economic terms. For
commercial timber production, this would require
quantifying the relationship between atmospheric
deposition and measures of forest productivity such
as timber yield. For assessing recreational benefits,
linkages would have to be drawn between air pollu-
tion and vulnerable factors that influence forest-based
recreation (e.g., site-characteristics such as canopy
density, type of tree species, degree of visible tree
damage, etc.). While important strides have been
made in establishing these linkages (e.g., NAPAP
modeling of air pollution effects on forest soil chem-
istry and tree branch physiology), critical gaps in our
ability to predict whole tree and forest responses to
air pollution changes have precluded the establish-
ment of such quantitative linkages.86 Critical knowl-
80 NAPAP, 1991.
81 NAPAP, 1991.
82 U.S. EPA, 1996b
83 U.S. EPA, 1996b.
84 U.S. EPA, 1996a.
85 U.S. EPA, 1996b
86 NAPAP, 1991.
E-14
-------
Appendix E: Ecological Effects of Criteria Pollutants
edge gaps exist in our ability to extrapolate experi-
mental results from seedling and branch studies to
whole tree and forest responses, to account for key
growth processes of mature trees, to integrate various
mechanisms by which air pollution can affect trees
(e.g., soil acidification, nitrification, and direct foliar
damage, winter stress, etc.), and to account for the
interaction of other stressors on forest health and dy-
namics (susceptibility to insect damage, drought, dis-
ease, fire, nutrient and light competition, etc.).
Despite these constraints to quantifying economic
benefits from air pollution reductions on forest eco-
systems, relevant studies that have attempted to value
air pollution damages on forests are reviewed and
summarized below. In some cases, the relationship
between air pollution and forest response is estimated
using expert judgement (e.g., for NAP AP assessment
from various growth scenarios). In other cases, dam-
age estimates reflect current impacts of air pollution
on forests, and the dose-response relationship is ab-
sent. In the aggregate, this summary provides some
insight into possible CAA-related benefits from
avoided damages to a select and narrowly focussed
group of forest services, but, because of severe data
constraints, does not provide an estimate of the over-
all range of forest-based benefits possible under the
CAA.
Commercial Timber Harvesting
The economic impact of hypothetical growth re-
ductions in northeastern and southeastern trees (both
hardwood and softwood species) was intensively stud-
ied under NAPAP.87 Growth reductions ranging from
5 to 10 percent over a 5 to 10 year period, depending
on the species and location, were assumed to occur as
a result of all forms of air pollution based on expert
opinion derived from a survey by deSteigner and Pye
(1988). Timber market responses to these hypoth-
esized growth declines were modeled until the year
2040 using a revised version of the Timber Assess-
ment Market Model (TAMM90) and the Aggregate
Timberland Assessment System (ATLAS), which was
used to simulate timber inventories on private tim-
berland in the United States. Economic welfare out-
puts included changes in consumer and producer sur-
plus and changes in revenue to southeast stumpage
owners. Results indicate that annualized reductions
in consumer and producer surplus would total $0.5
billion by the year 2000 and $3 billion by the year
2040 (in 1967 dollars). Simulated effects on stump-
age owners' revenues were minimal ($10 to $20 mil-
lion).
In an attempt to estimate the net economic dam-
ages from ozone effects on selected U.S. forests,
NAPAP studied the effect of various assumed reduc-
tions in growth rates of commercial southeastern pine
forests (both natural and planted).88 For both planted
and natural plus planted pines, the following changes
in growth rates were assumed to occur: a two percent
increase, no change, a two percent decrease, a five
percent decrease, and a ten percent decrease. The two
to five percent growth reductions were considered as
possible outcomes from current ozone induced dam-
age to southeastern forests, although no quantitative
linkage between ozone exposure and damages was
established. The ten percent growth reduction sce-
nario was primarily included for evaluating model
sensitivity to severe changes in growth and was con-
sidered out of the range of likely ozone damage esti-
mates. The TAMM and ATLAS models were again
used to simulate timber market responses under
baseline and hypothesized growth change scenarios
from 1985 to 2040. Results indicate that annual
changes in total economic surplus (i.e., the sum of
consumer and producer surplus and timber owner rev-
enues in 1989 dollars) would range from an increase
of $40 million (for the two percent increase in growth
scenario) to a decrease of $110 million (for the ten
percent decrease in growth scenario) for planted and
natural pine model simulations.
In the context of estimated benefits from avoid-
ance of other damages in the absence of the Clean Air
Act from 1970 to 1990,89 the magnitude of economic
damages estimated to the commercial timber indus-
try are comparatively small. For example, economic
damage estimates range up to $3 billion annually for
five to ten percent growth rate reductions in northeast
and southeast forests, and just $110 million for south-
eastern pines. However, in the context of damages to
forest-based services as a whole, the NAPAP-derived
commercial timber damage estimates should be
viewed as representing a lower bound estimate for a
variety of reasons. First, these damage estimates
exclude other categories of possible forest-based ben-
87 Haynes and Kaiser, NAPAP SOS/T 27 Section B, 1990.
88 NAPAP, 1991.
89 Most notably avoided human health effects, which are estimated on the order of $300 to $800 billion annually.
E-15
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
efits, including recreational and non-use values. Sec-
ond, even within the context of timber-related dam-
ages, the NAPAP forest-damage studies focused on a
portion of U.S. forests (northeastern and southeastern
U.S.); amuch greater geographic range of forests could
become susceptible to timber-related damages in the
absence of CAA controls. Finally, the NAPAP dam-
age estimates consider only two types of tree species:
planted and naturally grown pines, although these spe-
cies are economically important. Damages to other
commercially harvested tree species, such as mixed
pine and hardwood forests, are therefore excluded.
Non-marketed Forest Services
In an effort to address the potential benefits re-
sulting from avoidance of acid deposition-induced
damages to non-marketed forest-based services (e.g.,
recreation use, existence value), an extensive review
of the economic literature was conducted under the
auspices of NAPAP.90 From their review, NAPAP
could not identify any single study or model that could
be reliably used to quantify economic benefits from
avoided acid deposition-caused damages to non-mar-
keted forest services (such as recreational use) on a
regional or national basis. The primary limitation in
many of the studies reviewed was the absence of a
quantitative linkage between the value of a recreational
user day and important site characteristics which could
be tied to air pollution effects. In addition, most stud-
ies were narrowly focused geographically to specific
sites and did not attempt to value system-wide (larger
scale) damages that could result from acid deposition
over an entire region. Since the availability of nearby
substitution sites will affect the recreational value for
a given site, the benefits from such site-specific stud-
ies may not reflect actual economic damages incurred
from wide-scale air pollution impacts on forests. The
inability of studies to consider additional crowding at
unaffected sites in addition to changes in recreational
participation rates as a function of air pollution dam-
ages was also recognized as an important limitation.
Despite not being able to quantitatively assess the
benefits from avoided acid deposition-induced dam-
ages to nonmarket forest services, several important
concepts emerge from NAPAP's review of recre-
ational benefits, that bear relevance to the section 812
retrospective analysis. First, several studies were iden-
tified that established a relationship between key for-
est site characteristics and the value of recreational
participation. For example, Brown et al. (1989) used
90 Rosenthal, NAPAP SOS/T 27 Section B, 1990.
contingent valuation to evaluate the relationship be-
tween scenic beauty ratings and willingness of
recreationalists to pay at pictured sites. Based on their
interviews with over 1400 recreationalists at ten dif-
ferent sites in Arizona, positive correlations were es-
tablished between scenic beauty rankings determined
from one group of recreationalists and willingness to
pay to recreate determined by a separate group of
recreationalists (r2 ranged from 0.27 to 0.98 depend-
ing on ranking). In another study, Walsh et al. (1989)
developed a functional relationship between reduc-
tion of recreational benefits and tree density changes
that reflected varying levels of insect damage at six
campgrounds in the Front Range of the Colorado
Rockies. By using both contingent valuation and travel
cost models, Walsh et al. (1989) were able to show
that 10 percent, 20 percent, and 30 percent decreases
in tree densities reduces the total recreational benefits
at their sites by 7 percent, 15 percent and 24 percent,
respectively. Although results from these studies are
limited to the sites from which they were derived, they
do support the intuition that the degree of visible dam-
age to forests is to some extent correlated with the
magnitude of damages to forest-based recreation ex-
pected. This finding supports the notion that the avoid-
ance of damages to forest ecosystems from CAA-in-
duced pollution controls (albeit currently unquantified)
have likely benefited forest-based recreation in the
U.S.
In addition to establishing relationships between
recreational value and visible damage to forest sites,
there is evidence linking air pollution (ozone) effects
on forests to economic damages to non-use values of
forests. For example, D.C. Peterson et al. (1987) val-
ued ozone-induced damages to forests surrounding the
Los Angeles area. Using contingent valuation meth-
ods, D.C. Peterson et al. (1987) surveyed
recreationalists (a random survey of households in the
San Bernardino, Los Angles and Orange counties) and
residents (a sample of property owners within the San
Bernardino and Angeles national forests) for their
willingness to pay to prevent forest scenes from de-
grading one step on a "forest quality ladder" depict-
ing various levels of ozone-induced damages. The
mean willingness to pay to protect further degrada-
tion was $37.61 and $119.48 per household for
recreationalists and residents, respectively. Annual
damages to Los Angeles area residences from a one-
step drop on the forest quality ladder were estimated
between $27 million and $147 million.
E-16
-------
Appendix E: Ecological Effects of Criteria Pollutants
These estimates cannot be directly translated into
a rough estimate of the potential non-use values of
avoided forest damages. Considering the limited size
of the population generating the estimated benefits of
forest degradation, however, they do provide evidence
that the recreational and non-use benefits may sub-
stantially exceed the commercial timber values.
E-17
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Ecosystem Effects References
Aerts, R., B. Wallen, andN. Maimer. 1992. Growth-
limiting nutrients in Sphagnum-dominated
bogs subject to low and high atmospheric ni-
trogen supply. Journal of Ecology 80: 131-
140.
Baker, L.A., P.R. Kaufmann, A.T. Herlihy, J.M.
Eilers, D.F. Brakke, M.E. Mitch, R.J. Olson,
R.B. Cook, B.M. Ross-Todd, J.J. Beauchamp,
C.B. Johnson, D.D. Brown, and D.J. Blick.
1990. Current Status of Surface Water Acid-
Base Chemistry. NAPAP SOS/T Report 9,
In: Acidic Deposition: State of Science and
Technology, Volume II, National Acid Pre-
cipitation Assessment Program, 722 Jackson
Place NW, Washington, D.C. 20503.
Baker, J.P., D.P. Bernard, S.W. Christensen, M.J. Sale,
J. Freda, K. Heltcher, D. Marmorek, L. Rowe,
P. Scanlon, G. Suter, W. Warren-Hicks, and
P. Welbourn. 1990. Biological effects of
changes in surface water acid-base chemis-
try. NAPAP SOS/T Report 13, In: Acidic
Deposition: State of Science and Technology,
Volume II, National Acid Precipitation As-
sessment Program, 722 Jackson Place NW,
Washington, D.C. 20503.
Barnard, J.E., A.A. Lucier, R.T. Brooks, A.H. Johnson,
P.H. Dunn, and D.F. Karnosky. 1990.
Changes in forest health and productivity in
the United States and Canada. NAPAP SOS/
T Report 16, In: Acidic Deposition: State of
Science and Technology, Volume III, Na-
tional Acid Precipitation Assessment Pro-
gram, 722 Jackson Place NW, Washington,
D.C.20503.
Benoit, L. F., J. M. Skelly, L. D. Moore, and L. S.
Dochinger. 1982. Radial growth reductions
in Pinus strobus L. correlated with foliar
ozone sensitivity as an indicator of ozone-in-
duced losses in eastern forests. Can. J. For.
Res. 12:673-678.
Bloom N. S., C. J. Watras, and J. P. Hurley. 1991.
Impact of acidification on the methylmercury
cycle of remote seepage lakes. Water Air Soil
Pollution 56:477-491.
Boston, H.L. 1986. A discussion of the adaptations
for carbon acquisition in relation to the growth
strategy of aquatic isoetids. Aquatic Botany
26: 259-270.
Brown, T.C., M.T. Richards, and T.C. Daniel. 1989.
Scenic Beauty and Recreation Value: Assess-
ing the Relationship. In: J. Vining, ed., So-
cial Science and Natural Resources Recre-
ation Management, Westview Press, Boulder,
Colorado.
Camacho, R. 1993. Financial Cost Effectiveness of
Point and Nonpoint Source Nutrient Reduc-
tion Technologies in the Chesapeake Bay
Basin. Report No. 8 of the Chesapeake Bay
Program Nutrient Reductio Strategy Reevalu-
ation. Washington D C.: U.S. Environmental
Protection Agency, February.
Cowardin, L.M., V. Carter, F.C. Golet, and E.T.
LaRoe. 1979. Classification of wetlands and
deepwater habitats ofthe United States. U.S.
Fish & Wildlife Service Pub. FWS/OBS-79/
31, Washington, D.C., 103 pp.
deSteigner, J. E. and J. M. Pye. 1988. Using scientific
opinion to conduct forestry air pollution eco-
nomic analysis. In: A. Jobstl, ed., Proceed-
ings of the Symposium on the Economic As-
sessment of Damage Caused to Forests by Air
Pollutants. IUFRO Working Party S4.04-02,
September 13-17, 1988. Gmunden, Austria.
Duchelle, S. F., J. M. Skelly, T. L. Sharick, B. I.
Chevone, Y. Yang, and J. E. Nellessen. 1983.
Effects of ozone on the productivity of natu-
ral vegetation in a high meadow of the
Shenandoah National Park of Virginia. J of
Env. Manage. 17:299-308.
Duchelle, S. F., J. M. Skelly, and B. I. Chevone. 1982.
Oxidant effects on forest tree seedling growth
in the Appalachian Mountains. J Water, Air,
Soil Pollut. 18:363-373.
Ferguson, P., R.N. Robinson, M.C. Press, and J.A.
Lee. 1984. Element concentrations in five
Sphagnum species in relation to atmospheric
pollution. Journal of Bryology 13: 107-114.
E-18
-------
Appendix E: Ecological Effects of Criteria Pollutants
Fisher, D.C. and M. Oppenheimer. 1991. Atmo-
spheric nitrogen deposition and the Chesa-
peake Bay estuary. Ambio 20:102-108.
Fitzgerald, W.F., R.P. Mason, and G.M. Vandal. 1991.
Atmospheric Cycling and Air-Water Ex-
change of Mercury Over Mid-Continental
Lacustrine Regions. Water, Soil, Air & Soil
Poll. 56:745-767.
Glass, G. E., J.A. Sorenson, K.W. Schmidt, and G.R.
Rapp, Jr. 1990. New Source Identification of
Mercury Contamination in the Great Lakes.
Environ. Sci. Technol. 24: 1059-1069.
Gorham, E., S.E. Bayley, and D.W. Schindler. 1984.
Ecological effects of acid deposition upon
peatlands: A neglected field in "acid-rain"
research. Can. J. Fish. Aquat. Sci. 41: 1256-
1268.
Grieb, T.M., C.T. Driscoll, S.T. Gloss, C.L. Schofield,
G.L. Bowie, and D.B. Porcella. 1990. Fac-
tors affecting mercury accumulation in fish
in the upper Michigan peninsula. Environ.
Toxicol. Chem. 9:9191-930.
Hansson, S. and L.G. Rudstam. 1990. Eutrophica-
tion and Baltic fish communities. Ambio
19:123-125.
Hayes, E. M., and J. M. Skelly. 1977. Transport of
ozone from the northeast U.S. into Virginia
and its effect on eastern white pines. Plant
Dis. Rep. 61: 778-782.
Haynes, R.W. and H.F. Kaiser. 1990. Forests: Meth-
ods for Valuing Acidic Deposition/Air Pol-
lution Effects. NAPAP SOS/T Report 27,
Section B2, In: Acidic Deposition: State of
Science and Technology, Volume IV, Na-
tional Acid Precipitation Assessment Pro-
gram, 722 Jackson Place NW, Washington,
D.C.20503.
Hecky, R.E. and P. Kilham. 1988. Nutrient limita-
tion of phytoplankton in freshwater and ma-
rine environments: a review of recent evi-
dence on the effects of enrichment. Limnol-
ogy and Oceanography 33:796-822.
Hinga, K.R., A.A. Keller, and C.A. Oviatt. 1991.
Atmospheric deposition and nitrogen inputs
to coastal waters. Ambio 20:256-260.
Hogsett, W. E., A. A. Herstom, J. A. Laurence, E. H.
Lee, J. E. Weber, and D. T. Tingey. 1995.
Risk characterization of tropospheric ozone
to forests. In: Comparative Risk Analysis and
Priority Setting for Air Pollution Issues. Pro-
ceedings of the 4th U.S.-Dutch International
Symposium. Pittsburgh, PA. Air and Waste
Management Association. 119-145.
Lee, J.A. and S.J. Woodin. 1988. Vegetation struc-
ture and the interception of acidic deposition
by ombrotrophic mires. In: Vegetation Struc-
ture in Relation to Carbon and Nutrient
Economy, J.T.A. Verhoeven, G.W. Heil, and
M.J.A. Werger, Eds. SPB Academic Publish-
ing bv, The Hague, pp. 137-148.
Lee, J.A., M.C. Press, and S.J. Woodin. 1986. Ef-
fects of NO on aquatic ecosystems. In: En-
vironment and Quality of Life: Study on the
Need for an NO Long-term Limit Value for
the Protection ofTerrestrial and Aquatic Eco-
systems. Commission of the European Com-
munities, Luxembourg, pp. 99-116.
Locke, A. 1993. Factors influencing community
structure along stress gradients: zooplankton
responses to acidification. Ecology 73:903-
909.
Logofet, D.O. and G.A. Alexandrov. 1984. Model-
ling of matter cycle in a mesotropic bog eco-
system. II. Dynamic model and ecological
succession. Ecol. Modell. 21:259-276
McCool, P. M., and R. C. Musselman. 1990. Impact
of ozone on growth of peach, apricot, and al-
mond. Hortscience 25: 1384-1385.
Miller, P.R., J.R. McBride, S.L. Schilling, and A.P.
Gomez. 1989. Trend of ozone damage to
conifer forests between 1974 and 1988 in the
San Bernardino mountains of southern Cali-
fornia. In: Effects of Air Pollution on West-
ern Forests. R.K. Olson and A.S. Lefohn, Eds.
Pittsburgh, PA: Air and Waste Management
Association, Pittsburgh, PA, pp. 309-324
(Transaction series no. 16).
E-19
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Miller, P.R., J.R. McBride, and S.L. Schilling. 1991.
Chronic ozone injury and associated stresses
affect relative competitive capacity of species
comprising the California mixed conifer for-
est type. In: Memorias del Primer Simposial
National; Agricultura Sostenible: Una Option
para Desarollo sin Deterioro Ambiental.
Comision de Estudios Ambientales, Colegio
de Postgraduados, Montecillo, Edo. Mexico,
Mexico, pp. 161-172.
Mills, K.H., S.M. Chalanchuk, L.C. Mohr, and I.J.
Davies. 1987. Responses of fish populations
in Lake 223 to 8 years of experimental acidi-
fication. Can. J. Fish. Aquat. Sci. 44(Suppl.
1): 114-125.
Miskimmin, B.M., J.W.M. Rudd, and C.A. Kelly.
1992. Influence of dissolved organic carbon,
pH, and microbial respiration rates on mer-
cury methylation and demethylation in lake
water. Can. J. Fish. Aquat. Sci. 49:17-22
Mitsch, W.J. and J.G. Gosselink. 1986. Wetlands.
Van Nostrand reinhold. New York.
Moore, D.R.J., P A. Keddy, C.L. Gaudet, and I.C.
Wisheu. 1989. Conservation of wetlands:
do infertile wetlands deserve a higher prior-
ity? Biological Conservation 47: 203-217.
National Acid Precipitation Assessment Program
(NAPAP). 1991. 1990 Integrated assessment
report. National Acid Precipitation Assess-
ment Program, 722 Jackson Place NW, Wash-
ington, D.C.20503.
National Economics Research Associates (NERA),
Inc. 1994. The Benefits of Reducing Emis-
sions of Nitrogen Oxides under Phase I of
Title IV of the 1990 Clean Air Act Amend-
ments.
Neufeld, H. S., J. R. Renfro, W. D. Hacker, and D.
Silsbee. 1992. Ozone in Great Smoky Moun-
tains National Park: dynamics and effects on
plants. In: Tropospheric ozone and the envi-
ronment II - effects, modeling and control:
papers from an international specialty con-
ference; November; Atlanta, GA, Pittsburgh,
PA: Air & Waste Management Association;
pp. 594-617. (A&WMA transactions series:
TR-20).
Ontario Ministry of the Environment and Ministry of
Natural Resources. 1990. Guide to eating
Ontario sport fish. Public Information Cen-
tre, Environment Ontario, Toronto.
Owens, N.J.P., J.N. Galloway, and R.A. Duce. 1992.
Episodic atmospheric nitrogen deposition to
oligotrophic oceans. Nature 357:397-399.
Paerl, H.W. 1988. Nuisance phytoplankton blooms
in coastal, estuarine, and inland waters.
Limnol. Oceanogr. 33:823-847.
Paerl, H.W. 1993. Emerging role of atmospheric ni-
trogen deposition in coastal eutrophication:
biogeochemical and trophic perspectives.
Can. J. Fish. Aquat. Sci. 50:2254-2269.
Peterson D.C. et al. 1987. Improving accuracy and
reducing costs of environmental benefit as-
sessments. Office of Policy, Planning and
Evaluation, U.S. Environmental Protection
Agency, Washington D.C.
Peterson, D. L., and M. J. Arbaugh. 1988. An evalu-
ation of the effects of ozone injury on radial
growth of ponderosa pine (Pinus ponderosa)
in the southern Sierra Nevada. JAPCA 38:
921-927.
Peterson, D. L., and M. J. Arbaugh. 1992. Conifer-
ous forests of the Colorado front range. Part
B: ponderosa pine second-growth stands. In:
Olson, R. K.; Binkley, D.; Boehm, M., eds.
The response of western forests to air pollu-
tion. New York, NY: Springer-Verlag; pp.
365 and 385-401. (Billings, W. D.; Golley,
F.; Lange, O. L.; Olson, J. S.; Remmert, H.
Ecological studies; analysis and synthesis v.
97).
Peterson, D.L., M. J. Arbaugh, and J. R. Linday. 1991.
Regional growth changes in ozone-stressed
ponderosa pine (Pinus ponderosa) in the Si-
erra Nevada, California, USA. Holocene 1:50-
61.
Peterson, D. L., M. J. Arbaugh, V. A. Wakefield, and
P. R. Miller. 1987. Evidence of growth re-
duction in ozone-injured Jeffrey pine (Pinus
jeffreyi Grev. and Balf.) in Sequoia and Kings
Canyon National Parks. JAPCA 37: 906-912.
E-20
-------
Appendix E: Ecological Effects of Criteria Pollutants
Powell, D. S., et al. 1993. Forest Resources of the
United States, 1992. USDA-Forest Service,
Fort Collins, CO. General Technical Report
RM-234.
Retzlaff, W. A., T. M. DeJong, and L. E. Williams.
1992a. Photosynthesis and growth response
of almond to increased atmospheric ozone
partial pressures. J. Environ. Qual. 21:
208-216.
Retzlaff, W. A., L. E. Williams, and T. M. Dejong.
1992b. Photosynthesis, growth, and yield
response of 'Casselman' plum to various
ozone partial pressures during orchard estab-
lishment. J. Am. Soc. Hortic. Sci. 117:
703-710.
Retzlaff, W. A., L. E. Williams, and T. M. DeJong.
1991. The effect of different atmospheric
ozone partial pressures on photosynthesis and
growth of nine fruit and nut tree species. Tree
Physiol. 8: 93-105.
Rochefort, L., D. Vitt, and S. Bayley. 1990. Growth,
production, and decomposition dynamics of
Sphagnum under natural and experimentally
acidified conditions. Ecology 71(5): 1986-
2000.
Roelofs, J.G.M. 1986. The effect of airborne sulphur
and nitrogen deposition on aquatic and ter-
restrial heathland vegetation. Experientia
42:372-377.
Rosenthal, D. 1990. Forest Recreation. NAPAP SOS/
T Report 27, Section B2.4, In: Acidic Depo-
sition: State of Science and Technology, Vol-
ume IV, National Acid Precipitation Assess-
ment
Rosenberg, R., R. Elmgren, S. Fleischer, P. Jonsson,
G. Persson, and H. Dahlin. 1990. Marine
eutrophication case studies in Sweden.
Ambio 19:102-108.
Rosseland, B.O. 1986. Ecological effects of acidifi-
cation on tertiary consumers: fish population
responses. Water Air Soil Pollution 30:451-
460.
Schofield, C.L., C.T. Driscoll, R. K. Munson, C. Yan,
and J.G. Holsapple. 1994. The Mercury Cycle
and fish in the Adirondack Lakes. Environ.
Science & Tech. 28:3:136-143.
Shaver, C. L., K. A. Tonnessen, and T. G. Maniero.
1994. Clearing the air at Great Smoky Moun-
tains National Park. Ecol. App. 4: 690-701.
Shriner D.S., W.W. Heck, S.B. McLaughlin, D.W.
Johnson, P.M. Irving, J.D. Joslin, and C.E.
Peterson. 1990. Response of vegetation to
atmospheric deposition and air pollution.
NAPAP SOS/T Report 18, In: Acidic Depo-
sition: State of Science and Technology, Vol-
ume III, National Acid Precipitation Assess-
ment Program, 722 Jackson Place NW, Wash-
ington, D.C.20503.
Shuyler L.R. 1992. "Cost Analysis for Nonpoint
Source Control Strategies in the Chesapeake
Basin." Annapolis MD: U.S. Environmental
Protection Agency Chesapeake Bay Program,
March.
Skelly, J. M., Y. S. Yang, B. I. Chevone, S. J. Long, J.
E.Nellessen, andW.E. Winner. 1984. Ozone
concentrations and their influence on forest
species in the Blue Ridge Mountains of Vir-
ginia. In: Davis, D.D.; Millen, A.A.;
Dochinger, L., eds. Air pollution and the pro-
ductivity of the forest: proceedings of the
symposium; October 1983; Washington, DC.
Arlington, VA; Izaak Walton League of
America Endowment; pp. 143-159.
Sorenson, J.A., G. E. Glass, K. W. Schmidt, and G.R.
Rapp, Jr. 1990. Airborne Mercury Deposition
and Watershed Characteristics in Relation to
Mercury Concentrations in Water, Sediment,
Plankton and Fish of Eighty Northern Min-
nesota Lakes. Environ. Sci. Technol. 24:
1716-1727.
Spry, D.J. and J.G. Wiener. 1991. Metal
bioavailability and toxicity to fish in low-al-
kalinity lakes: a critical review. Environmen-
tal Pollution 71:243-304.
E-21
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Turner, R.S., R.B. Cook, H. Van Miegroet, D.W.
Johnson, J.W. Elwood, O.P. Bricker, S.E.
Lindberg, and G. M. Hornberger. 1990.
Watershed and Lake Processes Affecting Sur-
face Water Acit-Base Chemistry. NAPAP
SOS/T Report 10, In: Acidic Deposition:
State of Science and Technology, Volume II,
National Acid Precipitation Assessment Pro-
gram, 722 Jackson Place NW, Washington,
D.C.20503.
U.S. DOI, 1993. Fish and Wildlife Service and U.S.
Department of Commerce, 1991 National
Survey of Fishing, Hunting, and Wildlife-
Associated Recreation, U.S. Government
Printing Office, Washington D.C.
U.S. EPA. 1993. Air Quality Criteria for Oxides of
Nitrogen. Office of Health and Environmen-
tal Assessment, Environmental Criteria and
Assessment Office, Research Triangle Park,
NC; EPA report no. EPA600/8-91/049bF. 3v.
U.S. EPA. 1994. Deposition of Air Pollutants to the
Great Waters: First Report to Congress. Of-
fice of Air Quality Planning and Standards.
Research Triangle Park, NC: EPA Report
Number EPA-453/R-93-055.
U.S. EPA. 1995. The Benefits and Costs of the Clean
Air Act 1970 to 1990 — Report to Congress.
U.S. EPA. 1996a. External Draft, Air Quality Crite-
ria for Ozone and Related Photochemical
Oxidants. Volume II. Office of Health and
Environmental Assessment, Environmental
Criteria and Assessment Office, Research
Triangle Park, NC; EPA report no. EPA/600/
AP-93/004af-cf.
U.S. EPA. 1996b. Review of National Ambient Air
Quality Standards for Ozone: Assessment of
Scientific and Technical Information: OAQPS
Staff Paper. Office of Air Quality Planning
and Standards. Research Triangle Park, NC;
EPA report no. EPA-452YR-96-007. June.
Vitousek, P.M. and R.W. Howarth. 1991. Nitrogen
limitation on land and in the sea: How can it
occur? Biogeochemistry 13:87-115.
Walsh, R.G., F.A. Ward, and J.P. Olienyk. 1989. Rec-
reational demand for trees in national forests.
J. Environ. Manage. 28:255-268.
Watras, C.J. and N.S. Bloom. 1992. Mercury and
methylmercury in individual zooplankton:
implications for bioaccumulation. Limnol-
ogy and Oceanography 37:1313-1318.
Watras, C.J., N.S. Bloom, R.J.M. Hudson, S. Gherini,
R. Munson, S.A. Claas, K.A. Morrison, J.
Hurley, J.G. Wiener, W.F. Fitzgerald, R.
Mason, G. Vandal, D. Powell, R. Rada, L.
Rislov, M. Winfrey, J. Elder, D. Krabbenhoft,
A.W. Andren, C. Babiarz, D.B. Porcella, and
J.W. Huckabee. 1994. Sources and fates of
mercury and methylmercury in Wisconsin
lakes. In: Mercury Pollution: Integration and
Synthesis, C.J. Watras and J.W. Huckabee,
Eds. Lewis Publishers, Boca Raton, Florida,
pp.153-180.
Wiener, J.G., R.E. Martini, T.B. Sheffy, and G.E.
Glass. 1990. Factors influencing mercury
concentrations in walleyes in northern Wis-
consin lakes. Trans. Am. Fish. Soc. 119:862-
870.
Wisheu, I.C. and P.A. Keddy. 1989. The conserva-
tion and management of a threatened coastal
plain plant community in eastern North
America (Nova Scotia, Canada). Biological
Conservation 48: 229-238.
E-22
-------
Appendix F: Effects of Criteria Pollutants on
Agriculture
Introduction
One potential impact of air pollutants on economic
welfare is their effect on agricultural crops, including
annual and perennial species. Pollutants may affect
processes within individual plants that affect growth
and reproduction, thereby affecting yields of agricul-
tural crops. Possible physiological effects of pollut-
ants include the following: decreased photosynthesis;
changes in carbohydrate allocation; increased foliar
leaching; decreased nutrient uptake; increased sensi-
tivity to climatic stress, pests, and pathogens; de-
creased competitive ability; and decreased reproduc-
tive efficiency. These physiological effects, in con-
junction with environmental factors and intraspecies
differences in susceptibility, may affect crop yields.
Primary air pollutants that might damage plants
include S02, NOx, and volatile organic compounds
(VOCs). These pollutants may have direct effects on
crops, or they may damage crops indirectly by con-
tributing to tropospheric (ground-level) ozone,
peroxyacetyl nitrate (PAN), and/or acid deposition,
all of which damage plants. Tropospheric ozone is
formed by photochemical reactions involving VOCs
and NOx, while S02 and NOx cause acidic deposition.
While all of these air pollutants may inflict incre-
mental stresses on crop plants, in most cases air pol-
lutants other than ozone are not a significant danger
to crops. Based primarily on EPA's National Acid
Precipitation Assessment Program (NAPAP) conclu-
sions,1 this analysis considers ozone to be the primary
pollutant affecting agricultural production.
This analysis estimates the economic value of the
difference in agricultural production that has resulted
due to the existence ofthe CAA since 1970. The analy-
sis is restricted to a subset of agricultural commodi-
ties, and excludes those commodity crops for which
ozone response data are not available. Fruits, veg-
etables, ornamentals, and specialty crops are also ex-
cluded from this analysis. To estimate the economic
value of ozone reductions under the CAA, agricul-
tural production levels expected from control scenario
ozone conditions are first compared with those ex-
pected to be associated with ozone levels predicted
under the no-control scenario. Estimated changes in
economic welfare are then calculated based on a com-
parison of estimated economic benefits associated with
each level of production.
Ozone Concentration Data
To estimate the nationwide crop damages as a
result of ozone exposure, the first step is to estimate
the nationwide ozone concentrations under the con-
trol and no-control scenarios. This section describes
the methodology used to estimate ozone concentra-
tions for each county in each of these two scenarios.
First, historical ozone concentration data at the
monitor level were compiled from EPA's AIRS sys-
tem. Differences between the modeled control and no-
control scenario ozone concentrations were then used
to scale historical data to derive no-control scenario
ozone air quality profiles.2 Next, the ozone index used
in the exposure response evaluation was calculated
and applied at the monitor level. For this analysis, the
W126 index, a peak-weighted average of cumulative
ozone concentrations, was selected to conform with
the index currently being used by EPA in ozone
NAAQS benefits analysis. The W126 index is one of
several cumulative statistics, and may correlate more
closely to crop damage than do unweighted indices.3
EPA has not yet made a final determination of the
appropriate index to use in agricultural benefits analy-
1 Shriner et al., 1990; NAPAP, 1991.
2 Derivation of these ozone air quality profiles for the control and no-control scenario is summarized in the following section and
described in detail in Appendix C.
3 Lefohn et al., 1988.
FA
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
sis; thus this analysis should be viewed only as an
indicator of the magnitude of potential benefits.
The third step in ozone concentration estimation
involved the use of triangulation and planar interpo-
lation to arrive at a W126 statistic at the county, rather
than at the monitor, level. For each county centroid,
the closest surrounding triangle of monitors is located
and the W126 is calculated for that county using a
distance-weighted average of the ozone concentration
at each of these monitors.
Control and No-control Scenario Ozone
Concentration Data
The initial estimation of ozone concentrations in
the control and no-control scenarios was performed
by Systems Applications International (SAI). To cre-
ate the control scenario, SAI compiled ozone data from
the EPA's Aerometric Information and Retrieval Sys-
tem (AIRS).4 SAI summarized these data by fitting
gamma distributions to them and providing the alpha
and the beta parameters to these distributions. Each
of these distributions describes a set of ozone con-
centration levels, and the distributions are categorized
by year, season, and averaging time. SAI defines six
distinct "seasons," each composed of a two month
period in the year. This analysis uses those distribu-
tions which describe 1-hour average ozone concen-
trations taken from 7 AM to 7 PM and separated into
seasons. The analysis utilizes only those monitor
records that were modeled in both the control and no-
control scenarios.
To determine the ozone concentrations forthe no-
control scenario, SAI utilized the Ozone Isopleth Plot-
ting with Optional Mechanisms-IV (OZIPM4) model.
The input data required for OZIPM4 includes air qual-
ity data, surface and upper-air meteorological data,
and estimates of anthropogenic and biogenic emis-
sions of volatile organic compounds, NOx and CO.5
To create these inputs, SAI used (among other sources)
outputs from the Regional Acid Deposition Model
(RADM) and the SJVAQS/AUSPEX Regional Mod-
eling Adaptation project (SARMAP). Additional de-
tail concerning the development of ozone concentra-
tion data is available in Appendix C and in the SAI
report to EPA.6
Calculation of the W126 Statistic
Using the SAI ozone concentration distributions,
we calculated a sigmoidally weighted ozone index for
each monitor. The generalized sigmoidal weighting
function used in calculating such indices is presented
in Lefohn and Runeckles (1987) as:
where:
w = 1 /| 1 +M'c\p(-A - i)\ (1)
w. = weighting factor for concentration
(unitless)
c. = concentration (ppm)
M= an arbitrary constant
A = an arbitrary constant
The constants M and A are chosen to give different
weights to higher or lower concentrations. The index
used in this analysis is the W126 statistic, which is
calculated as follows:7
w = 1 / 11 + 4403 • exp( -126 • c )| (2)
and
fYl26=Jjv. (3)
Missing values are accounted for by multiplying the
resulting W126 statistic by the ratio of the number of
potential observations to the number of actual obser-
vations (i.e., total hours in period/hours of data in pe-
riod).
To calculate W126 indices from the monitor level
gamma distributions, we used an inverse cumulative
density function to calculate a separate representative
air concentration for each hour in the two month sea-
son. These values are then used in the above equation
to obtain a monitor-level W126 statistic.
To ensure that the interpretation of the gamma
distributions in this manner does not generate errors,
we tested our gamma-derived control-scenario W 126s
4 SAI, ICF Kaiser, 1995.
5 SAI, ICF Kaiser, 1995.
6 SAI, ICF Kaiser, 1995.
7 Lefohn et al.. 1988.
-------
Appendix F: Effects of Criteria Pollutants on Agriculture
against W126s calculated directly from the AIRS da-
tabase. We found that insignificant error resulted from
the utilization of the gamma distributions to create
W126 statistics.
Aggregating Ozone Data to the County
Level
Because crop production data are available at the
county level, the lowest level of aggregation that could
be used for ozone indices is also the county level.
Therefore, monitor level data needed to be aggregated
to a county level. For each county, we first located
the monitors from which we would be interpolating
data. To identify these monitors, we searched for the
three monitors which formed the closest triangle
around the centroid of the county.8 The closest tri-
angle was defined as that triangle in which the sum of
the distances from the three monitors to the county
centroid was the least. The algorithm stopped search-
ing for closest triangles of monitors when it had
searched all monitors within 500 km of a given county
centroid (an arbitrary distance, selected to reduce com-
putational requirements).
For coastal counties and some rural counties in
some years, monitor triangles around the county cen-
troid do not exist. We assigned the W126 value from
the monitor closest to the centroid to these counties.
Approximately 15 percent of all county-years (36,973
of248,880 records) were assigned W126s in this man-
ner.
For the remaining 85 percent, after the closest tri-
angle of monitors was found, a "planar interpolation"
was used to calculate the W126 at that county for that
year. One way to picture this process is to plot each of
the three monitors as a point in space. For each moni-
tor, the x axis represents longitude, the y axis repre-
sents latitude and the z axis represents the W126 sta-
tistic. A plane can then be drawn through these three
points in space. Furthermore, using the equation for
the plane, and given the x and y values (latitude and
longitude) for the county centroid, the county
centroid's z value (W126 statistic) can be calculated.
In essence, this procedure calculates a distance-
weighted average of three monitors' W126 index val-
ues and assigns this value to the county centroid.
The result of this data manipulation is a monthly
W126 statistic for each county in the continental
United States for the years 1971-1990. From these
data, yield change estimates were generated, and eco-
nomic impacts were estimated.
Yield Change Estimates
There are several steps involved in generating
yield change estimates. The first is the selection of
relevant ozone exposure-response functions (mini-
mum and maximum) for each crop in the analysis.
Ozone data, triangulated to the county level, are trans-
formed into an index suitable for use in the selected
function(s) to estimate county level predicted yield
losses for both the control and no-control scenarios.
In the next step, the proportion of each county to the
national production of each crop is calculated to per-
mit national aggregation of estimated yield losses.
Finally, the control scenario percentage relative yield
loss (PRYL) is compared to the minimum and maxi-
mum PRYL for the no-control scenario. Each step is
discussed in more detail below.
Exposure-Response Functions
To estimate yield impacts from ozone, exposure-
response functions are required for each crop to be
analyzed. This analysis was restricted to estimating
changes in yields for those commodity crops for which
consistent exposure-response functions are available
and that are included in national agricultural sector
models. To maintain consistency with the current
ozone NAAQS benefits analysis being conducted by
OAQPS, NCLAN-based exposure-response functions
using a Weibull functional form and a 12-hour W126
ozone index were used.
Several crops included in the NCLAN research
program were not evaluated in this analysis. Non-com-
modity crops that are not modeled in national agri-
cultural sector models were not included in this analy-
sis: lettuce, tomatoes, potatoes, alfalfa, tobacco, tur-
nips, and kidney beans. In addition, one commodity
crop, spring wheat, was excluded because the NCLAN
exposure-response function was only developed for
winter wheat.
8 The vast majority of monitors had latitude and longitude data available through AIRS. 1,528 of 1,536 monitors were located in
this manner. For the remaining 8 monitors, if in a given year of monitor data another monitor exists in the county of the unfound
monitor, we discarded the unlocated monitor's data. Otherwise, we located that monitor at the county's centroid. We located 5 of the
remaining 8 monitors in this fashion.
F-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Minimum/Maximum Exposure-Response
Functions
Estimated responsiveness of a given crop to ozone
varies within the NCLAN data. This range of response
is partially explained by the program's evaluation of
several cultivars for some crops; ozone sensitivity
varies across cultivars. In addition, the conditions for
different experiments varied due to variations in lo-
cation, year, and additional treatments included in
some experiments. No one exposure-response func-
tion can be assumed to be representative of all culti-
vars in use, or of all environmental conditions for crop
production. To develop a range of benefits estimates
that reflects this variation in responsiveness, a mini-
mum responsiveness and a maximum responsiveness
function were selected for each crop. In actuality, a
number of different cultivars are planted by produc-
ers, and so ozone response will be a weighted average
of the responsiveness of each cultivarto its ozone con-
dition and its proportion of total acreage. It is impor-
tant to note that these values do not necessarily bound
the analysis, since the number of cultivars evaluated
by NCLAN is small relative to the number grown for
many crops.
For the crops used
in this study, CERL
conducted an analysis
to identify the ozone
concentration required
to reduce yields by 10
percent for each crop
cultivar using its 12-
hour W126 exposure-
response function. For
each crop, the function
demonstrating the low-
est ozone concentration
at a 10 percent yield
loss represents the
maximum response,
and the function with
the highest concentra-
tion at 10 percent yield
loss represents the mini-
mum response. Table F-
1 reports the minimum
and maximum expo-
sure-response functions
for each crop. Two crops, peanuts and sorghum, did
not have multiple NCLAN experiments on which to
base a comparison of the responsiveness of different
cultivars or the variation in response with different
experimental conditions.
Calculation of Ozone Indices
Each NCLAN ozone exposure-response experi-
ment exposed each studied crop over a portion of the
crop's growing season. The duration of the NCLAN
experiments was provided by CERL and was rounded
to the nearest month. The W126 index is cumulative,
and so is sensitive both to the duration over which it
is calculated and to the specific month(s) within a
growing season that are included in it. Because crop-
ping seasons vary across the U.S., the ozone index
used to calculate county-level changes in yield due to
ozone must reflect the local season for each crop. To
determine which portion of the growing season a par-
ticular exposure period pertains to (in order to calcu-
late an exposure index), we developed state-specific
growing seasons based on planting and harvesting data
developed by USDA.9 The W126 index was calcu-
Table F-1. Agriculture Exposure-Response Functions.
Crop
Cultivar
Equation
Type
Yield Function
(PRYL, ppm)
Du ration
(days)
Barley
CM-7 2
Both
l-exp(-(W126/6998.5)'388
95
Com-Kield
PAG 397
Min
l-exp(-(W126/94.2)4176
83
Com-Kield
Pioneer 3780
Max
l-exp(-(W126/92.7)2585
83
Cotton
McNair235
Min
l-exp(-(W126/l 13.3)1397
125
Cotton
Acala SJ2
Max
l-exp(-(W126/74.6)1066
98
Grain
Sorghum
DeKalb A28+
Both
1 -exp (-(W126/205.3)1957
85
Peanuts
NC-6
Both
l-exp(-(W126/96.8)1890
112
Soy beans
Corsoy-79
Min
l-exp(-(W126/476.7)'113
93
Sov beans
Davis
Max
l-exp(-(W126/l 30.1)1 000
93
Wheal
ART
Min
l-exp(-(W126/76.8)2031
54
Wheal
VONA
Max
l-exp(-(W126/24.7)100)
61
Source: EPA/CERL (unpublished) for all functions.
9 USDA, 1984. Some states did not have explicit growing seasons reported for certain crops due to the low production in these
states. In these cases a proxy state growing season was used. In most of these cases the proxy growing season was taken from a state
with an adjoining boundary within the same geographic region.
F-4
-------
Appendix F: Effects of Criteria Pollutants on Agriculture
lated using the county level ozone data developed in
the prior section, summed for the number of months
of NCLAN experimental duration, with the exposure
period anchored on the usual harvest month for each
crop.10
Calculations of County Weights
Because the benefits analysis did not require a
regional level of disaggregation and to minimize com-
putational burdens the economic analysis was con-
ducted at a national level. Ozone data and estimated
yield responses, however, were developed at a county
level. To conduct a national analysis, the county level
yield change estimates were weighted to develop a
single national percent relative yield loss for each crop
relative to the control scenario, for both the minimum
and the maximum yield responses. As a part of cal-
culating a percent change in yield at the national level,
weights for each county and crop were created for
1975, 1980, 1985, and 1990. The weights for these
four years were used to represent the year itself and
the four years immediately following it (e.g., 1975
weights were also used for 1976, 1977, 1978, and
1979). Although weather and other conditions may
change the proportion of counties' production to the
total national production in each year, five year
weights should reflect stable periods of agricultural
policy between each Farm Bill, and are sufficient for
the level of precision needed for this analysis. The
weights were calculated by dividing the production
level of a crop in a county11 by the sum of all states'
reported production for that crop.12 These county
weights were applied to the percent relative yield loss
results for each county, as discussed below.
Calculation of Percent Change in Yield
Ozone exposure-response functions are expressed
in terms of percent relative yield loss (PRYL); the
ozone level being analyzed is compared with "clean"
(charcoal filtered/zero ozone) air. To calculate a per-
cent change in yield between the control and no-con-
trol scenarios, we first calculate a PRYL based on the
county-level control scenario W126 ozone index, and
a PRYL based on the no-control scenario W126 in-
dex. Next, the county weights are applied to the
PRYLs. The change in yield, measured relative to the
hypothetical zero-ozone crop production, is then:
(PRYLC-PRYLJ (4)
To obtain the change in terms of our (non-zero)
baseline yield, we divide by that yield, and get:
(PRYLC- PRYL,,) / (100- PRY I.,) (5)
To create the national percent change in yield for
each crop, the results of this equation are summed for
each scenario (maximum and minimum) and for each
year. Tables F-2 and F-3 present the resulting percent
yield changes that were used as inputs to the economic
model.
Economic Impact Estimates
To estimate the economic benefits of controls on
ozone precursor pollutants under the Clean Air Act,
changes in yields due to those controls need to be
evaluated in terms of their effect on agricultural mar-
kets. To do this, yield changes can be incorporated
into an economic model capable of estimating the as-
sociated changes in economic surpluses within the
agricultural economy, preferably one that reflects
changes in producers' production decisions and de-
mand substitution between crops. This type of dy-
namic analysis is needed because even small changes
in yield or price expectations can cause large shifts in
the acreage allocated to specific crops, and the degree
to which alternative crops will be substituted (particu-
larly for feed uses).
Agricultural Simulation Model (AGSIM)
The modeling approach used in this analysis is an
econometric model of the agricultural sector, which
estimates demand and supply under different produc-
tion technologies and policy conditions. The
AGricultural Simulation Model (AGSIM) has been
10 This analysis required "rounding" some months: if a harvest date was specified to be from the 15th to the end of a month, the
W126 index was calculated using that month's data; if the harvest date was specified to be from the first to the 14th of a month, the
W126 index was calculated using the prior month's data as the final month in the exposure period.
11 USDA, 1995.
12 The national total does not include USDA areas designated "other counties". These areas are groups of counties that for one
reason or another (disclosure rules, low amount of production, etc.) are not individually listed. Because we did not have ozone values
for these groups, we did not use their production levels in the calculation of the total national production.
F-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic F-2. Relative No-control to Control Percent Yield Change (harvested acres) for the Minimum
Scenario.
Crop
Year
ISarlev
Corn
Cotton
Peanuts
Soybeans
Sorjih um
Winter Wheat
1975
-0.000020
-0.000171
-0.011936
-0.006635
-0.001166
-0.000717
-0.005631
1976
-0.000013
-0.000329
-0.017505
-0.024048
-0.002171
-0.001841
-0.004841
1977
-0.000013
-0.000169
-0.013114
-0.015150
-0.001562
-0.001118
-0.005464
1978
-0.000019
-0.000291
-0.018692
-0.017606
-0.002480
-0.001844
-0.005894
1979
-0.000027
-0.000100
-0.017217
-0.013067
-0.001898
-0.001389
-0.004998
1980
-0.000019
-0.000200
-0.021315
-0.022761
-0.002397
-0.002222
-0.005385
1981
-0.000016
-0.000071
-0.018552
-0.014269
-0.001951
-0.000802
-0.003964
1982
-0.000020
-0.000070
-0.017295
-0.014200
-0.002107
-0.001050
-0.004773
1983
-0.000023
-0.000617
-0.020842
-0.028601
-0.003901
-0.002366
-0.005904
1984
-0.000027
-0.000111
-0.023552
-0.019225
-0.002919
-0.002881
-0.006121
1985
-0.000025
-0.000132
-0.020947
-0.017965
-0.002645
-0.001726
-0.007316
1986
-0.000029
-0.000158
-0.027 968
-0.031605
-0.002899
-0.001564
-0.007597
1987
-0.000033
-0.000358
-0.034 5 84
-0.043854
-0.003776
-0.001812
-0.009669
1988
-0.000027
-0.000662
-0.03 5 069
-0.038085
-0.004563
-0.002922
-0.019873
1989
-0.000024
-0.000150
-0.031245
-0.022094
-0.003769
-0.001359
-0.007605
1990
-0.000024
-0.000210
-0.03 7 9 88
-0.047395
-0.003819
-0.001567
-0.006449
Note: There is only one scenario for barley, peanuts, and sorghum, because there was only one exposure-response function. •
Table F-3. Relative No-control to Control Percent Yield Change (harvested acres) for the Maximum
Scenario.
Crop
Year
Barley
Com
Cotton
Peanuts
Soybeans
Sorghum
Winter Wheat
1975
-0.000020
-0.001139
-0.021059
-0.006635
-0.005808
-0.000717
-0.034803
1976
-0.000013
-0.002281
-0.032063
-0.024048
-0.010298
-0.001841
-0.040303
1977
-0.000013
-0.001232
-0.025773
-0.015150
-0.007764
-0.001118
-0.049636
1978
-0.000019
-0.002015
-0.033075
-0.017606
-0.011803
-0.001844
-0.050308
1979
-0.000027
-0.001052
-0.031433
-0.013067
-0.009592
-0.001389
-0.052211
1980
-0.000019
-0.001537
-0.037278
-0.022761
-0.011845
-0.002222
-0.054128
1981
-0.000016
-0.000923
-0.035058
-0.014269
-0.009902
-0.000802
-0.053470
1982
-0.000020
-0.000974
-0.034101
-0.014200
-0.010815
-0.001050
-0.058409
1983
-0.000023
-0.003888
-0.040405
-0.028601
-0.018597
-0.002366
-0.063556
1984
-0.000027
-0.001443
-0.043890
-0.019225
-0.014502
-0.002881
-0.067612
1985
-0.000025
-0.001377
-0.040845
-0.017965
-0.013384
-0.001726
-0.072177
1986
-0.000029
-0.001451
-0.052426
-0.031605
-0.014754
-0.001564
-0.081225
1987
-0.000033
-0.002565
-0.061295
-0.043854
-0.018578
-0.001812
-0.089042
1988
-0.000027
-0.004318
-0.061660
-0.038085
-0.021767
-0.002922
-0.120703
1989
-0.000024
-0.001987
-0.059573
-0.022094
-0.018739
-0.001359
-0.086958
1990
-0.000024
-0.002056
-0.071659
-0.047395
-0.018670
-0.001567
-0.082309
Note: There is only one scenario for bariey, peanuts, and sorghum, because there was only one exposure-response function.
F-6
-------
Appendix F: Effects of Criteria Pollutants on Agriculture
used extensively to evaluate air pollution impacts, as
well as a number of other environmental policy analy-
ses. AGSIM is an econometric-simulation model that
is based on a large set of statistically estimated de-
mand and supply equations for agricultural commodi-
ties produced in the United States. The model is ca-
pable of estimating how farmers will adjust their crop
acreages between commodities when relative profit-
ability changes as a result of crop yield and produc-
tion cost changes. Acreage and yield changes from
various scenarios will affect total production of crops,
which then affects commodity prices and consump-
tion. The commodity price changes, in turn, affect
profitability and cropping patterns in subsequent years.
Federal farm program and conservation reserve ef-
fects are also incorporated into the model.
The initial version of AGSIM (which went
through various acronym revisions) was developed
in 1977.13 The model was developed to permit esti-
mation of aggregate impacts associated with relatively
small changes in crop yields or production costs, which
might result from various policy conditions such as
changes in pesticide input availability, or in this case,
changes in crop exposure to ozone. Subsequent revi-
sions to the model as well as the current specification
are described in Taylor (1993a).14 Several policy ap-
plications of AGSIM were tested and reported in Tay-
lor (1993b)15 to provide a comparison to the results
of several alternative agricultural sector models. These
tests included an expansion of Conservation Reserve
acreage, reduced target prices, elimination of agricul-
tural programs in the U.S. other than the Conserva-
tion Reserve Program (CRP), and a tax on nitrogenous
fertilizer use in the U.S. The model has been used to
evaluate the effects of changes to the CRP,16 changes
in agricultural price support programs,17 and evalua-
tions of policies concerning pesticide availability.18
AGSIM is designed to estimate changes in the
agricultural sector resulting from policies that affect
either the yields or the costs of crop production.
Changes in economic variables are computed by com-
paring a policy simulation of the model with a baseline
simulation of the model. For this retrospective evalu-
ation, the baseline reflects actual farm programs,
prices, and other parameters since 1970. The model's
author, Dr. C. Robert Taylor, modified AGSIM for
this analysis to reflect production conditions and poli-
cies as they changed through the 20-year span of the
Clean Air Act, from 1970 to 1990. During this pe-
riod, U.S. farm policy parameters changed every five
years with the enactment of each Farm Bill, and pro-
ducer participation varied significantly over the pe-
riod. Over this time, due to policy, weather, techno-
logical development, and other variations, production
levels and prices have varied, as have production tech-
nologies, costs of production, and relevant cultivars.
To reflect these changes, Dr. Taylor re-estimated de-
mand relationships forthree periods (1975-79; 1980-
84; and 1985-89) based on the farm policies in effect
in each period, and structured the model to run on a
national level rather than a regional level. The period
from 1970-1975 was not modeled because of data limi-
tations and because there was limited impact from the
CAA on ozone levels during that period.
The AGSIM baseline production and price data
serve as the control scenario baseline. Percent rela-
tive yield losses (PRYLs) between the control and no-
control scenarios are the relevant input parameter for
this analysis, from which AGSIM calculates new yield
per planted acre values. Based on these values (as well
as on lagged price data, ending stocks from the previ-
13 Taylor, C.R., R.D. Lacewell, andH. Talpaz. 1979. Use of Extraneous Information with the Econometric Model to Evaluate
Impacts of Pesticide Withdrawals. Western J. of Ag. Econ. 4:1-8.
"Taylor, C.R. 1993a. AGSIM: An Econometric-Simulation Model of Regional Crop and National Livestock Production in the
United States. In: C.R. Taylor, K.H. Reichelderfer, and S.R. Johnson (Eds) Agricultural Sector Models for the United States:
Descriptions and Selected Policy Applications. Ames Iowa: Iowa State University Press.
15 Taylor, C.R. 1993b. Policy Evaluation Exercises with AGSIM. In: C.R. Taylor, K.H. Reichelderfer, and S.R. Johnson (Eds)
Agricultural Sector Models for the United States: Descriptions and Selected Policy Applications. Ames Iowa: Iowa State University
Press.
16 Taylor, C.R. 1990. Supply Control Aspects of the Conservation Reserve. In: T.L. Napier (Ed) Implementing the Conservation
Title of the Food Security Act of 1985. Ankeny, Iowa: Soil and Water Conservation Society; Taylor, C.R., H.A. Smith, J.B. Johnson,
and T.R. Clark. 1994. Aggregate Economic Effects of CRP Land Returning to Production. J. of Soil and Water Conservation 49:325-
328.
17Talyor, C.R. 1994. Deterministic vs. Stochastic Evaluation of the Aggregate Effects of Price Support Programs. Agricultural
Systems 44:461-474.
18 Taylor, C.R., G.A. Carlson, F.T. Cooke, K.H. Reichelderfer, and I.R. Starbird. Aggregate Economic Effects of Alternative Boll
Weevil Management Strategies. Agricultural Econ. Res. 35:19-19;Taylor, C.R., J.B. Penson Jr., E.G. Smith, and R.D. Knutson. 1991.
Impacts of Chemical Use Reduction in the South. S. J. Of Ag. Econ. 23:15-23.
F-7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
ous year, and other variables), AGSIM predicts acre-
age, production, supply, and price parameters for each
crop for each year, as well as calculating yield per
harvested acre. From these results and the demand
relationships embedded in the model, AGSIM calcu-
lates the utilization of each crop (i.e., exports, feed
use, other domestic use, etc.), as well as the change in
consumer surplus, net crop income, deficiency pay-
ments and other government support payments. Net
surplus is calculated as net crop income plus consumer
surplus, less government payments. The first year of
results is 1976 because AGSIM must have one year
(1975) of lagged data.
Table F-4 presents the net changes in economic
surpluses (in 1990 dollars) annually and as a cumula-
tive present value (discounted at 5 percent) over the
period 1976-1990 due to the Clean Air Act. The posi-
tive surpluses exhibited in almost all years are a re-
sult of the increase in yields associated with lower
ozone levels than those predicted to occur under the
no-control scenario. The present value of the estimated
agricultural benefits of the CAA ranges between $7.8
billion in the minimum response case to approximately
$37 billion in the maximum response case. This range
represents the impacts that would occur if all of the
acreage planted to a given crop had an ozone response
function similar to either the minimum available re-
sponse function or the maximum available response
function. The available response functions do not nec-
essarily bracket the true range of potential crop re-
sponses, and it is unrealistic to anticipate that all acre-
age will be planted in cultivars with a uniform response
to ozone exposure. These considerations notwithstand-
ing, these values do indicate the likely magnitude of
agricultural benefits associated with control of ozone
precursors under the CAA, but not the precise value
of those benefits. In addition to estimating the present
value of net surplus at a discount rate of five percent,
two alternative discount rates were used. At a three
percent discount rate, the range of net surplus is be-
tween $6.7 billion and $32 billion; at seven percent
discount rate, the range is between $9 billion and $43
billion. For more detail on AGSIM intermediate model
outputs, see Abt Associates (1996).
Change in
Change in
Change in
Change in
Farm Program Payments
Net Crop Income
Consumer Surplus
Net Surplus
Year
Minimum
Maximum
Minimum
Maximum
Minimum
Maximum
Minimum
Maximum
1976/77
0
0
243
486
236
993
477
1,479
1977/78
0
0
-97
-259
349
1,557
253
1.297
1978/79
43
345
30
298
392
1,646
379
1.597
1979/8(1
0
0
-140
-406
449
2,000
309
1,594
1980/81
0
0
8
-178
392
2,049
400
1,870
1981/82
112
518
-99
-406
440
2,594
231
1,670
1982/83
168
981
64
107
377
2,730
273
1,856
1983/84
153
1,009
231
958
316
1,969
395
1,917
1984/85
-182
808
82
560
-279
1,686
-14
1,437
1985/86
289
1,291
181
879
616
2,054
509
1,644
1986/87
270
1,356
230
966
462
2,265
422
1,875
1987/88
469
2,033
320
1,405
708
2,990
558
2,361
1988/89
557
2,023
316
1,508
796
2,943
556
2,428
1989/9(1
329
1,401
161
614
527
2,572
358
1,785
1990/!) 1
414
1,927
180
473
618
3,047
384
1,593
Cumulative Present Value of Net Surplus at 5 percent ($ 1990)
7,763
37.225
F-8
Table F-4. Change in Farm Program Payments. Net Crop Income. Consumer Surplus, and Net
Surplus Due to the CAA (millions 1990 $).
-------
Appendix F: Effects of Criteria Pollutants on Agriculture
Conclusions
Agricultural benefits associated with control of
ozone precursors under the Clean Air Act are likely
to be fairly large. Because it is possible that overtime
producers have adopted more ozone-resistant culti-
vars, it may be appropriate to consider the lower end
of the range of predicted benefits to be more indica-
tive of the likely total benefits. The estimates devel-
oped in this analysis, however, do not represent all of
the likely benefits accruing to agriculture, in that many
high-value and/or ozone sensitive crops could not be
included in the analysis due to either exposure-re-
sponse data limitations or agricultural sector model-
ing limitations. The second consideration implies that
benefits will likely be larger than estimated. The mini-
mum case may be the most appropriate starting point,
however, due to the first consideration: the current
crop mix is probably biased toward lower ozone re-
sponsiveness. Therefore, we anticipate that cumula-
tive total agricultural benefits from the Clean Air Act
are on the order of ten billion dollars (real terms).
F-9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Agricultural Effects References
Abt Associates. 1996. Section 812 Retrospective
Analysis: Quantifying Health and Welfare
Benefits (Draft). Prepared by Abt Associates
under Contract No. 68-W4-0029. U.S. EPA,
Office of Policy, Planning, and Evaluation.
Lefohn, Allen S. et. al. 1988. A comparison of indi-
ces that describe the relationship between
exposure to ozone and reduction in the yield
of agricultural crops. Atmospheric Environ-
ment 22: 1229-1240.
Lee, E. Henry et. al. 1994. Attainment and effects is-
sues regarding alternative secondary ozone air
quality standards. J. Environ. Qual. 23:1129-
1140.
National Acid Precipitation Assessment Program
(NAPAP). 1991. 1990 Integrated assessment
report. National Acid Precipitation Assess-
ment Program, 722 Jackson Place NW, Wash-
ington, D.C.20503.
SAI, ICF Kaiser. 1995. Retrospective Analysis of
ozone air quality in the United States: final
report. Prepared by Systems Applications In-
ternational under contract 68-D4-0103. U.S.
EPA, Office of Policy Analysis and Review.
Shriner, D.S., W.W. Heck, S.B. McLaughlin, D.W.
Johnson, P.M. Irving, J.D. Joslin, and C.E.
Peterson. 1990. Response of vegetation to at-
mospheric deposition and air pollution.
NAPAP SOS/T Report 18, In: Acidic Depo-
sition: State of Science and Technology, Vol-
ume III, National Acid Precipitation Assess-
ment Program, 722 Jackson Place NW, Wash-
ington, D.C.20503.
Taylor, C.R. 1990. Supply Control Aspects of the
Conservation Reserve. In: T.L. Napier (Ed)
Implementing the Conservation Title of the
Food Security Act of 1985. Ankeny, Iowa:
Soil and Water Conservation Society.
Taylor, C.R. 1993a. AGSIM: An Econometric-Simu-
lation Model of Regional Crop and National
Livestock Production in the United States. In:
C.R. Taylor, K.H. Reichelderfer, and S.R.
Johnson (Eds) Agricultural Sector Models for
the United States: Descriptions and Selected
Policy Applications. Ames Iowa: Iowa State
University Press.
Taylor, C.R. 1993b. Policy Evaluation Exercises with
AGSIM. In: C.R. Taylor, K.H. Reichelderfer,
and S.R. Johnson (Eds) Agricultural Sector
Models for the United States: Descriptions
and Selected Policy Applications. Ames Iowa:
Iowa State University Press.
Taylor, C.R. 1994. Deterministic vs. Stochastic Evalu-
ation of the Aggregate Effects of Price Sup-
port Programs. Agricultural Systems 44:461-
474.
Taylor, C.R., G.A. Carlson, F T. Cooke, K.H.
Reichelderfer, and I.R. Starbird. Aggregate
Economic Effects of Alternative Boll Weevil
Management Strategies. Agricultural Econ.
Res. 35:19-19.
Taylor, C.R., R.D. Lacewell, andH. Talpaz. 1979. Use
of Extraneous Information with the Econo-
metric Model to Evaluate Impacts of Pesti-
cide Withdrawals. Western J. of Ag. Econ.
4:1-8.
Taylor, C.R., J.B. Penson Jr., E.G. Smith, and R.D.
Knutson. 1991. Impacts of Chemical Use
Reduction in the South. S.J. Of Ag. Econ.
23:15-23.
1994. Aggregate Economic Effects of CRP Land Re-
turning to Production. J. of Soil and Water
Conservation 49:325-328.
USDA. 1984. Usual Planting and Harvesting Dates
for U.S. Field Crops. Statistical Reporting
Service Agricultural Handbook No. 628.
USDA. 1995. Crops County Data. National Agricul-
tural Statistics Service Dataset (Electronic
File) 93100A and 93100B.
F-10
-------
Appendix G: Lead Benefits Analysis
Introduction
The scientific understanding of the relationship
between lead and human health is rapidly expanding.
This expansion is documented in numerous EPA stud-
ies on the health effects associated with lead expo-
sure. In a pioneering study, Schwartz etal. (U.S. EPA,
1985) quantified a number of health benefits that
would result from reductions in the lead content of
gasoline. The work was extended by EPA's analysis
of lead in drinking water (U.S. EPA, 1986a) and by
an EPA-funded study of alternative lead National
Ambient Air Quality Standards (U.S. EPA, 1987).
Despite this substantial research, much uncertainty
remains. While the health effects of very high levels
of blood lead (PbB) are quite severe (including con-
vulsions, coma and death from lead toxicity) and have
been known for many years, the effects of lower lead
doses continue to be the subject of intensive scien-
tific investigation. Dose-response functions are avail-
able for only a handful of health endpoints associated
with elevated blood lead levels. Other known or
strongly suspected health endpoints cannot be quan-
tified due to a lack of information on the relationship
between dose and effect. Table G-l presents the health
effects that are quantified in this analysis, as well as
important known health effects that are not quanti-
fied.
Some of the health effects that are quantified in
this analysis have not been estimated in previous EPA
analyses. This is largely due to more recent informa-
tion about the dose-response functions that makes it
possible to expand the health effect coverage beyond
what was done previously. Recent information is avail-
able for previously unquantified health effects, and
new information on previously estimated dose-re-
sponse functions is also available.
Tabic G-l. Quantified and Unquantified Health Effects of Lead.
Population Group
Quantified Health Kffeet
linqiiantified Health Kffeet
Adult Mule
For men in specified age ranges:
Hypertension
Non-fatal coronary heart disease
Non-fatal Strokes
Mortality
Quantified health effects for men in other age
ranges
Other cardiovascular diseases
Neurobehavioral function
Adult Feniiile
For women in specified age ranges:
Non-fatal coronary heart disease
Non-fatal stroke
Mortality
Quantified health effects for women in other age
ranges
Other cardiovascular diseases
Reproductive effects
Neurobehavioral function
Children
IQ loss effect on lifetime earnings
IQ loss effects on special educationalneeds
Neonatal mortality due to low birth weight
caused by maternal exposure to lead
Fetal effects from maternal exposure (including
diminished IQ)
Other neurobehavioral and physiological effects
Delinquent and anti-social behavior
G-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Methods Used to Measure and
Value Health Effects
The following sections present relevant dose-re-
sponse relationships for three population groups: chil-
dren, men, and women. These sections also discuss
data sources used for the dose-response relationships,
although an extensive review of the literature is not
given.1 In addition, each section includes the meth-
ods used to value the changes in health effects deter-
mined using these dose-response relationships.
Health Benefits to Children
Changes in IQ
Elevated Pb levels may induce a number of ef-
fects on the human nervous system. Generally, these
neurobehavioral effects are more serious for children
than for adults because of children's rapid rate of de-
velopment. It is believed that neurobehavioral defi-
cits in children may result from both pre-natal and
post-natal exposure. These nervous system effects may
include hyperactivity, behavioral and attentional dif-
ficulties, delayed mental development, and motor and
perceptual skill deficits. Quantification of certain
manifestations of these effects is possible because
sufficient data exist to estimate a dose-response rela-
tionship and IQ loss. The relationship used in the
analysis is discussed below.
Quantifying the Relationship Between Blood
Lead Levels and IQ
1991), 70 percent of the data were below 10 (jg/dL;
thus, the Bellinger data were linearized in the 5 to 15
(ig/dL range. For the studies that did not transform
lead concentrations, the regression coefficients were
used directly. Given the typical uncertainty within
individual studies, the variation in the regression co-
efficients among studies was not more than would be
expected. The relationship determined by Schwartz
(1993) suggests that for a 1 pg/dL increase in lead, a
decrease of 0.25 IQ points can be expected. The p-
value (< 0.0001) indicates that this relationship is
highly significant.
To obtain the total change in number of IQ points
for a population of children, the 0.25 points lost per
(ig/dL change in blood lead is multiplied by the aver-
age blood lead level for that population. The average
blood lead level modeled in this analysis is a geomet-
ric mean, not an arithmetic mean. To adjust for this,
we use a relationship between the expected value and
the geometric mean of a lognormally distributed ran-
dom variable:
l<:\x\ = exp In
(GM)
(In (GSD))2
(1)
where E(X) is the expected value (mean) of the distri-
bution, GM is the geometric mean, and GSD is the
geometric standard deviation. Taking the natural loga-
rithm of Equation 1 and rearranging gives the ratio
between the expected value and the GM:
In (li(Xj) - In (GM) =
(In (GSD))2
(2)
A dose-response relationship for IQ decrements
has been estimated by a meta-analysis of seven re-
search studies.2 Regression coefficients for each study
were used to determine a weighted average linear re-
gression coefficient for the relationship between lead
and IQ. Each regression coefficient was weighted by
the inverse of the variance of each estimate. To deter-
mine an overall coefficient, the regression coefficients
for studies that used natural logarithms of lead as the
exposure index were linearized. In general, the coef-
ficient was linearized in the blood lead range of 10 to
20 (ig/dL. However, in one study (Bellinger et al.,
In
K(X)
GM
(In (GSD))2
K(X)
GM
exp
(In (GSD)):
(3)
(4)
For a GSD of 1.6 (the assumed GSD of children's
blood lead levels3), the resulting ratio between E(X)
and GM is 1.117. This ratio is used in equation 5.
1 For a detailed review of this literature see U.S. Environmental Protection Agency, (1986b) Air Quality Criteria Document for
Lead, and 1989 Addendum. Environmental Criteria and Assessment Office, Office of Research and Development, March.
2 Schwartz, 1993.
3 Suggested value for sub-populations provided by IEUBK guidance manual (U.S. EPA, 1994).
G-2
-------
Appendix G: Lead Benefits Analysis
The total lost IQ points for each group was estimated
as:
(TOTAL LOST 10)k = AG.\lt x 1.117 x 0.25 x (Pop)l/ 7 (5)
where (Pop)k represents the number of children (up to
age six) around a given industrial source (in the case
of estimating benefits from reduced industrial emis-
sions) or the total U.S. population of children (in the
case of estimating benefits from reductions in gaso-
line lead emissions).
As shown in equation 5, the population of chil-
dren up to age six is divided by seven to avoid double
counting. If we assume that children are evenly dis-
tributed by age, this division applies this equation to
only children age 0-1. If we did not divide, this equa-
tion would count a child who is age zero in the first
year of the analysis and count that same child 6 more
times in successive years. Dividing by seven does cre-
ate some undercounting because in the first year of
the analysis children from age 1 to 6 are not accounted
for, while presumably they are affected by the lead
exposure.
The analysis assumes a permanent loss oflQ based
on blood lead levels estimated for children six years
and younger. Recent studies4 provide concrete evi-
dence of long-term effects from childhood lead expo-
sure.
Valuing Changes in Children's Intelligence
Available economic research provides little em-
pirical data for society's willingness to pay (WTP) to
avoid a decrease in an infant's IQ. Some research,
however, has addressed monetization of a subset of
the effects of decreased IQ. These effects would rep-
resent components of society's WTP to avoid IQ de-
creases. Employed alone, these monetized effects
should underestimate society's WTP. Nevertheless,
for the purpose of this analysis, these effects are used
to approximate the WTP to avoid IQ decrements.
IQ deficits incurred through lead exposure are
assumed to persist throughout the exposed infant's
lifetime. Two consequences of this IQ decrement are
4 For example, Bellinger (1992).
then considered: the decreased present value of ex-
pected lifetime earnings for the infant, and the in-
creased educational resources expended for a infant
who becomes mentally handicapped or is in need of
compensatory education as a consequence of lead
exposure. The value of foregone earnings is addressed
in this section.
The reduction in IQ has a direct and indirect ef-
fect on earnings. The direct effect is straightforward:
lower IQs decrease job attainment and performance.
Reduced IQ also results in reduced educational attain-
ment, which, in turn, affects earnings and labor force
participation. These effects on earnings are additive
since the studies used for this analysis have controlled
for these effects separately.5 If personal decisions
about the total amount of education and labor force
participation were based entirely on each individual
maximizing the expected present value of lifetime
income, the magnitude of the indirect effect on in-
come of a small change in educational attainment
would be close to zero,6 and certainly less than the
magnitude of the direct effect. However, individuals
make educational decisions based on a number of
considerations in addition to the effect on the present
value of lifetime earnings, such as satisfaction (util-
ity) derived from formal education, non-compensa-
tion aspects of alternative career opportunities, the
ability to pay educational costs, etc. Such consider-
ations could lead to either a positive or negative mar-
ginal return to education. Studies7 of educational at-
tainment and lifetime earnings have generally identi-
fied a positive marginal return to education, suggest-
ing that the educational attainment decision may not
be based simply on expected earnings.
This analysis uses two sets of estimates of the ef-
fects of IQ on earnings. The first estimate, used by
Abt Associates in a previous analysis, is based on sev-
eral older studies. The second estimate is based on
Salkever (1995).
Older Estimate of the Effect of IQ on Earnings:
The Direct Effect of IQ on Wage Rate
Henry Aaron, Zvi Griliches, and Paul Taubman
have reviewed the literature examining the relation-
5 IQ is also correlated with other socio-economic factors which have not been quantified in this analysis.
6 This is a straightforward result of the "envelope theorem" in economics. In this context, the envelope theorem shows that if
individuals select the level of education that maximizes expected income, then the marginal benefit of additional education (i.e., the
partial derivative of income with respect to education) will be zero at that optimal education level.
7 Including Chamberlain and Griliches (1977), Ashenfelter and Ham (1979), and Salkever (1995)
G-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
ship between IQ and lifetime earnings.8 They found
that the direct effect, (schooling held constant) of IQ
on wage rates ranged from 0.2 percent to 0.75 percent
per IQ point. Perhaps the best of these studies is
Griliches (1977).9 He reported the direct effect of IQ
on wage rates to be slightly more than 0.5 percent per
IQ point. Because this is roughly the median estimate
of the U.S. EPA review of the literature, this estimate
is used.
Older Estimate of the Effect of IQ on Earnings:
The Indirect Effect of IQ on Earnings
From Needleman et al. (1990) it is possible to
estimate the change in years of schooling attained per
one IQ point change. The study's regression coeffi-
cients for the effect of tooth lead on achieved grade
provide an estimate of current grade achieved. How-
ever, many of these children were in college at the
time and are expected to achieve a higher grade level.
Following Schwartz (1990), after adjusting the pub-
lished results for the fact that a higher percentage of
children with low tooth lead were attending college, a
0.59 year difference in expected maximum grade
achieved between the high and low exposure groups
was estimated. It is assumed that educational attain-
ment relates with blood lead levels in proportion to
IQ. The difference in IQ score between the high and
low exposure group was 4.5 points (from Needleman
et al. (1990)). Dividing 0.59/4.5 = 0.131 suggests that
the increase in lead exposure which reduces IQ by
one point may also reduce years of schooling by 0.131
years.
Studies that estimate the relationship between
educational attainment and wage rates (while control-
ling for IQ and other factors) are less common. Cham-
berlain and Griliches (1977) estimate that a one year
increase in schooling would increase wages by 6.4
percent. In a longitudinal study of 799 subjects over 8
years, Ashenfelter and Ham (1979) reported that an
extra year of education increased the average wage
rate over the period by 8.8 percent. We use the aver-
age of these two estimates (7.6 percent) to calculate
the indirect effect of increased schooling on the present
value of lifetime income. Increased wages per IQ point
are calculated using: (7.6 percent wage increase/school
year) x (0.131 school years/IQ) = 1.0 percent increase
in earnings per IQ point.
8 U.S. EPA, 1984.
There is one final indirect effect on earnings.
Changes in IQ affect labor force participation. Fail-
ure to graduate high school, for example, correlates
with participation in the labor force, principally
through higher unemployment rates and earlier retire-
ment ages. Lead is also a strong correlate with atten-
tion span deficits, which likely reduce labor force par-
ticipation. The results of Needleman et al. (1990) re-
lating lead to failure to graduate high school can be
used to estimate changes in earnings due to labor force
participation. Using the odds ratio from Needleman
et al., it was estimated that a one IQ point deficit would
also result in a 4.5 percent increase in the risk of fail-
ing to graduate. Krupnick and Cropper (1989) pro-
vide estimates of labor force participation between
high school graduates and non-graduates, controlling
for age, marital status, children, race, region, and other
socioeconomic status factors. Based on their data,
average participation in the labor force is reduced by
10.6 percent for persons failing to graduate from high
school. Because labor force participation is only one
component of lifetime earnings (i.e., earnings = wage
rate X years of work), this indirect effect of schooling
is additive to the effect on wage rates. Combining this
estimate with the Needleman result of 4.5 percent in-
crease in the risk of failing to graduate high school
per IQ point, indicates that the mean impact of one IQ
point loss is a (10.6 percent x 4.5 percent) = 0.477
percent decrease in expected earnings from reduced
labor force participation.
Combining the direct effect of 0.5 percent with
the two indirect effects (1.0 percent for less schooling
and 0.477 percent for reduced labor force participa-
tion) yields a total of 1.97 percent decrease in earn-
ings for every loss of one IQ point.
Newer Estimate of the Effect oflO on Earnings:
Salkever (1995)
One of the most recent studies of the effects of IQ
on earnings is Salkever (1995). Such an analysis with
more recent data is valuable because the labor market
has undergone many changes over the quarter cen-
tury in which earlier studies have appeared. Like the
analysis of the effect of IQ on earnings presented
above, Salkever (1995) estimates this as the sum of
direct and indirect effects. The direct effect is the sum
of effects of IQ test scores on employment and earn-
9 Griliches used a structural equations model to estimate the impact of multiple variables on an outcome of interest. This method
has conceptual advantages over other empirical estimates used in the literature because it successfully controls for the many con-
founding variables that can affect earnings.
G-4
-------
Appendix G: Lead Benefits Analysis
ings for employed persons, holding years of school-
ing constant. The indirect effect works through the
effect of IQ test scores on years of schooling attained,
and the subsequent effect of years of schooling on the
probability of employment, and on earnings for em-
ployed persons.
Salkever (1995) provides updated estimates all of
the necessary parameters using the most recent avail-
able data set, the National Longitudinal Survey of
Youth (NLSY). Three regression equations provide
these parameters. The years of schooling regression
shows the association between IQ scores and highest
grade achieved, holding background variables con-
stant. The employment regression shows the associa-
tion between IQ test scores, highest grade, and back-
ground variables on the probability of receiving earned
income. This regression thus provides an estimate of
the effect of IQ score on employment, holding school-
ing constant, and the effect of years of schooling on
employment, holding IQ constant. The earnings re-
gression shows the association between IQ test scores,
highest grade, and background variables on earnings,
for those with earned income.
These regressions provide parameters needed to
estimate the total effect of a loss of an IQ point on
earnings. The direct effects of IQ on employment and
earnings for employed persons, holding schooling
constant, come from the employment and earnings
regressions. The indirect effect of IQ on employment
through schooling is the product of the effect of IQ on
years of schooling, from the years of schooling re-
gression, and the effect of highest grade on employ-
ment, from the employment regression. The indirect
effects of IQ on earnings for employed persons through
schooling is the product of the effect of IQ on years
of schooling, from the years of schooling regression,
and the effect of highest grade on earnings for em-
ployed persons, from the earnings regression.
The total estimated effect of a loss of an IQ point
on earnings is larger than the previous estimate of 1.97
percent. Based on the Salkever study, the most recent
estimate of the effect of an IQ point loss is now a
reduction in earnings of 1.93 percent for men and 3.22
percent for women, which is a participation-weighted
average of 2.39 percent.
Value of Foregone Earnings
In the next step to monetize intelligence effects,
the percent earnings loss estimate must be combined
with an estimate of the present value of expected life-
10 U.S. Department of Commerce, 1993
time earnings. Data on earnings for employed per-
sons and employment rates as a function of educa-
tional attainment, age, and gender were reported for
the U.S. population in 1992 by the Bureau of the Cen-
sus.10 Assuming this distribution of earnings for em-
ployed persons and labor force participation rates re-
mains constant overtime, and further assuming a trend
rate of real wage growth (productivity effect), an an-
nual discount factor, and year-to-year survival prob-
abilities, the current Census data on earnings can be
used to calculate the mean present value of lifetime
earnings of a person born today. This analysis assumed
a person received earned income from age 18 to age
64, and assumed a real wage growth rate of one per-
cent and an annual discount rate of five percent. Men
tend to earn more than women because of higher wage
rates and higher labor force participation. However,
for both men and women, expected lifetime earnings
increase greatly with education.
While the Census data are most likely the best
available basis for projecting lifetime earnings, a num-
ber of uncertainties deserve mention. Labor force par-
ticipation rates of women, the elderly, and other groups
will most likely continue to change over the next de-
cades. Real earnings of women will probably continue
to rise relative to real earnings of men. Unpredictable
fluctuations in the economy's growth rate will prob-
ably affect labor force participation rates and real wage
growth of all groups. Medical advances will probably
raise survival probabilities.
One problem that was addressed was the fact that
the current educational distribution for older persons
today is an especially poor predictor of educational
attainment for those born today, since educational at-
tainment has risen over time. In fact, if one simply
projected educational attainment for a person born
today using this method, this person would lose years
of schooling with age (starting between ages 40 and
50), since average years of schooling declines with
age in a one-time snapshot of the current population.
To address this issue, the analysis assumes education
levels cannot fall as a person ages.
Note that use of earnings is an incomplete mea-
sure of an individual's value to society. Those indi-
viduals who choose not to participate in the labor force
for all of their working years must be accounted for,
since the lost value of their productive services may
not be accurately measured by wage rates. The larg-
est group are those who remain at home doing house-
work and child rearing. Also, volunteer work contrib-
G-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
utes significantly to social welfare and rates of
volunteerism tend to increase with educational attain-
ment and income.11 If the opportunity cost of non-
wage compensated work is assumed to be the average
wage earned by persons of the same sex, age, and edu-
cation, the average lifetime earnings estimates would
be significantly higher and could be approximated by
recalculating the tables using full employment rates
for all age/sex/education groups. To be conservative,
only the value of lost wages is considered in this analy-
sis.
The adjusted value of expected lifetime earnings
obtained above is a present value for an individual
entering the labor force at age 18 and working until
age 64. Given a five percent discount rate, the other
assumptions mentioned, and current survival prob-
abilities,12 the present value of lifetime earnings of a
person born today would be $170,169.
Costs of Additional Education
The increase in lifetime earnings from additional
education is the gross return to education. The gross
return to education, however, does not reflect the cost
of the additional education. The cost of the marginal
education must be subtracted from the gross return in
order to obtain the net increase per IQ point from ad-
ditional education. There are two components of the
cost of marginal education; the direct cost of the edu-
cation, and the opportunity cost of lost income during
the education. An estimate of the educational cost
component is obtained from the U.S. Department of
Education.13 The marginal cost of education used in
this analysis is assumed to be $5,500 per year. This
figure is derived from the Department of Education's
reported ($5,532) average per-student annual expen-
diture (current plus capital expenditures) in public
primary and secondary schools in 1989-'90. For com-
parison, the reported annual cost of college education
(tuition, room and board) in 4 year public institutions
is $4,975, and $12,284 for private institutions.
The estimated cost of an additional 0.131 years
of education per IQ point (from the older estimate of
IQ effects) is $721 (i.e., 0.131 x $5,500). Because this
marginal cost occurs at the end of formal education,
it must be discounted to the time the exposure and
damage is modeled to occur (age zero). The average
level of educational attainment in the population over
age 25 is 12.9 years.14 Therefore, the marginal educa-
tional cost is assumed to occur at age 19, resulting in
a discounted present value cost of $285.
The other component of the marginal cost of edu-
cation is the opportunity cost of lost income while in
school. Income loss is frequently cited as a major fac-
tor in the decision to terminate education, and must
be subtracted from the gross returns to education. An
estimate of the loss of income is derived assuming
that people in school are employed part time, but
people out of school are employed full time. The op-
portunity cost of lost income is the difference between
full-time and part-time earnings. The median annual
income of people ages 18-24 employed full-time is
$16,501, and $5,576 for part-time employment.15 The
lost income associated with being in school an addi-
tional 0.131 years is $1,431, which has a present dis-
counted value at age zero of $566.
Salkever found a smaller effect of IQ on educa-
tional attainment (0.1007 years per IQ point, versus
0.131 years), which results in smaller estimated costs.
Using the same method and data described above, the
estimated present value of educational cost per IQ
point is $219, and the income opportunity cost is $435.
Final Estimate of the Effect of IQ on Earnings.
Combining the value of lifetime earnings with the
two estimates of percent wage loss per IQ point yields
a low estimate of $170,169 x 1.97 percent = $3,000
per lost IQ point, and a higher estimate of $4,064 based
on Salkever (1995). Subtracting the education and
opportunity costs reduces these values to $2,505 and
$3,410 per IQ point, respectively. This analysis uses
the midpoint of these two estimates, which is $2,957.
Of course, changing the discount rate would change
this estimate. With an assumed discount rate of seven
percent, the final estimate is only $1,311. With an
assumed discount rate of three percent, the final esti-
mate rises to $6,879.
11 U.S. Department of Commerce, 1986. Table No. 651, p. 383.
12 Special education costs for children who do not survive to age 18 are not counted, which results in some underestimation of
benefits. However, most child mortality occurs before the age of 7, when the special education begins, so this under-counting is not
substantive.
13 "Digest of Education Statistics". U.S. Dept. of Education, 1993.
14 "Digest of Education Statistics". U.S. Dept. of Education, 1993.
15 "Money Income of Households, Families, and Persons in the United States: 1992". U.S. Department of Commerce, 1993.
G-6
-------
Appendix G: Lead Benefits Analysis
Children with IQs Less Than 70
Quantifying the Number of Children with IQs
Less than 70
In addition to the total IQ point decrements that
can be predicted to occur in a population of children
having a specified blood lead distribution, increases
are also expected to occur in the incidence of children
having very low IQ scores as the mean blood lead
level for that population increases. IQ scores are
normalized to have a mean of 100 and a standard de-
viation of fifteen. An IQ score of 70, which is two
standard deviations below the mean, is generally re-
garded as the point below which children require spe-
cial compensatory education tailored to the mentally
handicapped.
The relationship presented here for estimating
changes in the incidence of IQ < 70 was developed to
make use of the most current IQ point decrement func-
tion provided by Schwartz (1993). It is assumed that
for a baseline set of conditions where a population of
children has a blood lead distribution defined by some
geometric mean and geometric standard deviation, that
population also has a normalized IQ point distribu-
tion with a mean of 100 and a standard deviation of
15. For this baseline condition, the proportion of the
population expected to have IQ < 70 is determined
from the standard normal distribution function:
The integral in the standard normal distribution
function does not have a closed form solution. There-
fore, values forO(z) are usually obtained readily from
software with basic statistical functions or from tables
typically provided in statistics texts. The solution for
O(z) where z = -2 is 0.02275. That is, for the normal-
ized IQ score distribution with mean of 100 and stan-
dard deviation of 15, it is expected that approximately
2.3 percent of children will have IQ scores below 70.
To estimate changes in the proportion of children
with IQ scores below 70 associated with changes in
mean blood lead levels for a population of children,
the following two key assumptions are made:
1. The mean IQ score will change as a result of
changes in the mean blood lead level as:
AIQ = -0.25 xAPbB
where
A/<2 and APbB
are the changes in the mean IQ score and in
the mean blood lead levels, respectively, be-
tween the no-control and control scenarios.
This relationship relies on Schwartz' estimate
(1993) of adecrease of 0.25 IQ points for each
(ig/dL increase in blood lead. Note that the
mean blood lead level referred to here is the
arithmetic mean (or expected value) for the
distribution obtained as described previously
from the GM and GSD.
P(1()<1 0) = O(z)
(6)
2. The standard deviation for the IQ distribution
remains at 15.
where:
P(IQ <70)
O(z)
Probability of IQ scores less than
70
standard normal variate; com-
puted for an IQ score of 70, with
mean IQ score of 100 and stan-
dard deviation of 15 as:
70- 100 _ o
15
(7)
Standard normal distribution
function:
^/2n
du
(8)
Using these assumptions, the change in the pro-
portion of children having IQ <70 can then be deter-
mined for a given change in mean blood lead from:
mio
-------
Appendix G: Lead Benefits Analysis
where:
APr(CA
v ™
DBP
DBP,
= change in 2 year probability
of cerebrovascular accident in
women aged 45-74;
mean diastolic blood pressure in
the control scenario; and
mean diastolic blood pressure in
the no-control scenario.
The predicted incidences of avoided BI and CA
were multiplied by 70 percent to estimate only non-
fatal strokes (Shurtleff, 1974).
Valuing reductions in strokes
The value of avoiding an initial cerebrovascular
accident or an initial atherothrombotic brain infarc-
tion for women was calculated in the same way as for
men (see above). Of women in the United States be-
tween the ages of 45 and 74 (the age group for which
lead-related stroke was predicted), 38.2 percent are
ages 45-54 and the remaining 61.8 percent are ages
55-74. Using these percentages, and the gender- and
age-specific values in Taylor et al. (1996) the average
value among women ages 45-74 of avoiding either
type of stroke was estimated to be about $150,000.
Changes in Premature Mortality
Quantifying the relationship between blood
pressure and premature mortality
The risk of premature mortality in women can be
estimated by the following equation:
A PrfMOliT
.¦j |
1
1
_|_ £,5.40374 - 0.01511 *DBPi) | + g5.40374-0.01511*DBP2)
(25)
Quantifying Uncertainty
Characterizing Uncertainty Surrounding the Dose-
Response Relationships
The dose-response functions described for each
health endpoint considered above generally quantify
the adverse health effects expected due to increased
exposure to lead. For children, these relationships are
described directly in terms of changes in blood lead.
For adults, effects are estimated in terms of changes
in blood pressure (which are related to changes in
blood lead levels). As with any estimate, uncertainty
is associated with the dose-response relationships.
Consistent with the approach outlined in Appen-
dix D for the non-lead criteria air pollutants, this analy-
sis attempts to capture the uncertainty associated with
these relationships. This is accomplished by estimat-
ing a distribution associated with each dose-response
coefficient using the information reported in the lit-
erature. This analysis assumes these distributions to
be normal. For each of the coefficients used to relate
adverse health effects to lead exposure, Table G-2
summarizes the means and standard deviations of the
normal distributions used in this analysis.
Characterizing Uncertainty Surrounding the
Valuation Estimates
The procedure for quantifying uncertainty asso-
ciated with the valuation estimates is similar to that
used to characterize the dose-response coefficient es-
timates. The valuation distributions for health effects
considered in the lead analysis are documented in
Appendix I.
where:
APr(MORT
DBP
DBP,
) = the change in 2 year prob-
n/ o J i
ability of death for women aged
45-74;
mean diastolic blood pressure in
the control scenario; and
mean diastolic blood pressure in
the no-control scenario.
Valuing reductions in premature mortality
The value of reducing premature mortality for
women is assumed to be equal to that estimated for
all premature mortality, $4.8 million per incident (see
Appendix I).
G-15
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic G-2. Uncertainty Analysis: Distributions Associated With Dosc-
Rcsponsc Coefficients Used to Estimate Lead Health Effects.
Parameters of Normal
distributions describing Dose-
Response Coefficients
Health Effect
Mean
Standard
Deviation
Blood Lead-Blood Pressure Coefficient
1.44
0.85
(Adults)
Adult Males
Mortality (ages 40-54)
0.03516
0.16596
Mortality (ages 55-64)
0.01866
0.00533
Mortality (ages 65-74)
0.00547
0.00667
Chronic Heart Disease (ages 40-59)
0.030365
0.003586
Chronic Heart Disease (ages 60-64)
0.02351
0.028
Chronic Heart Disease (ages 65-74)
0.02031
0.00901
Cerebrovascular Accidents
0.04066
0.00711
Athcrollvrombolic Brain Infarctions
0.0484
0.00938
Hypertension
0.793
not available
Adult Females
Mortality (ages 45-74)
0.01511
0.00419
Chronic Heart Disease
0.03072
0.00385
Cerebrovascular Accidents
0.04287
0.00637
Atherothrombotic Brain Infarctions
0.0544
0.00754
Children
Infant Mortality
0.0001
not available
Lost IQ Points
0.245
0.039
IQ<70 (cases)
relies on Lost IQ Point distribution
Methods Used to
Determine Changes in
Lead Emissions from
Industrial Processes
from 1970 to 1990
This analysis used several
sources to determine the
changes in facility-specific
emissions of lead from indus-
trial processes. To summarize,
the analysis extracted 1990 fa-
cility-specific lead emissions
data from the Toxics Release
Inventory (TRI), which pro-
vides recent emissions data for
over 20,000 U.S. manufactur-
ing facilities. This study then
adjusted these data by the rela-
tive changes in lead emissions
between 1970 and 1990; these
relative changes were derived
from several data sources de-
scribed below. This method
yielded facility-specific emis-
sions for five year intervals be-
tween 1970 and 1990 for both
the controlled and uncontrolled
scenarios. The five-year values
were interpolated to derive an-
nual changes for each year be-
tween 1970 and 1990. Specific
details on this approach are
given below.
TRI Data
Industrial Processes and Boilers
and Electric Utilities
This section describes the methods and data
sources used to estimate changes in blood lead levels
due to changes in lead emissions from industrial pro-
cesses and boilers between 1970 and 1990 and from
electric utilities between 1975 and 1990. The estimates
of the changes in health effects resulting from changes
in lead emissions due to the CAA are also presented.
The Toxics Release Inven-
tory (TRI) is mandated by the
Superfund Amendment Reau-
thorization Act (SARA) Title
III Section 313 and requires that U.S. manufacturing
facilities with more than 10 employees file annual re-
ports documenting multimedia environmental releases
and off-site transfers for over 300 chemicals. Facili-
ties report both stack and fugitive releases to air. Re-
ported releases are generally estimates rather than
precise quantifications. Emissions data can be pre-
sented as numerical point estimates, or, if releases are
below 1,000 pounds, as an estimated range of emis-
sions.
G-16
-------
Appendix G: Lead Benefits Analysis
From the TRI data base, this analysis extracted
data from the reporting year 1990 for all facilities re-
porting emissions of lead to air, as either stack or fu-
gitive emissions. Data were reported as annual emis-
sions (in pounds per year). Where emissions are re-
ported as a range, this analysis used the upper bound
of the range to represent the emissions.28 TRI facili-
ties also report their location by latitude and longi-
tude. In order to later match facilities emitting lead
with Census data on surrounding exposed populations,
this analysis uses the latitudes and longitudes of lead-
emitting facilities.
Derivation of Industrial Process Emissions
Differentials 1970-1990
The TRI database is the Agency's single best
source of consistently reported release data; however,
the database does not include information for most of
the years modeled in this analysis. Furthermore, this
analysis required estimates of hypothetical emissions
in the absence of the CAA. Therefore, estimates were
created for the emissions of lead from industrial
sources under the CAA, and in the absence of the CAA,
for the years 1970, 1975, 1980, 1985, and 1990. The
percent changes, or differentials, reflected by these
estimates were then applied to the 1990 TRI data to
obtain facility-level release estimates for the years of
interest for the control and no-control scenarios.
The method for creating these differentials cap-
tured the two potential causes of the differences be-
tween emissions from industrial sources regulated by
the CAA and emissions from those same sources in
the absence of the CAA. The first cause of the differ-
ence in emissions is a change in overall industrial
output, resulting from the macroeconomic impact of
the CAA. The second element is a change in emis-
sions per unit of output, which results from the adop-
tion of cleaner processes and the application of emis-
sions control technology mandated by the CAA. The
methods used to project the effects of these two causes,
described below, were designed to be as consistent as
possible with other emissions projection methods for
other segments of the CAA retrospective analysis.
Data sources
Data for the differentials estimates were taken
from the following sources:
the lorgenson/Wilcoxen (I/W) model projec-
tions, conducted as part of the section 812
analysis. This data source addresses the first
cause of changes in emissions: the macroeco-
nomic changes that resulted from the imple-
mentation of the 1970 CAA. The I/W model
calculated the change in economic output for
each of thirty-five industrial sectors, roughly
analogous to two-digit standard industrial
classification (SIC) codes, that resulted from
the CAA's implementation. The specific out-
put used from the I/W model in this analysis
was the percentage change in economic out-
put for the various industrial sectors, rather
than any absolute measure of economic ac-
tivity.
the 1991 OAQPS Trends database. This data
base is an emissions projection system that
was used to produce the report, "The National
Air Pollutant Emission Estimates, 1940-
1990." It contains information on economic
activity, national level emissions and emis-
sion controls, by industrial process, from 1970
through 1990. Three different elements were
extracted from the Trends database: the emis-
sions of lead per unit economic output for
various industrial processes for the years
1970-1990; annual economic output data for
these industrial processes; and the emission
calculation formula.
the National Energy Accounts (NEA), com-
piled by the Bureau of Economic Analysis.
This database records the historical levels of
industrial energy consumption, disaggregated
by fuel type at the approximately three-digit
SIC code level.
The manner in which these data were combined
to derive lead emissions estimates is described be-
low.
Estimates of industrial process emissions in the
control scenario
Emissions data for industrial processes were esti-
mated for the years 1970, 1975,1980, 1985,and 1990.
For each of these years, this analysis extracted an
emission factor and a control efficiency for each lead-
28 Ranges are infrequently reported and are either reported as 0-500 lbs. or 500-1000 lbs. The infrequency of the incidence of a
facility reporting a range and the relatively small quantities of lead released by those facilities means any overestimation of benefits
that results from using the upper limit of the range is extremely minor.
G-17
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
emitting industrial process in the Trends database.
Emissions factors are expressed as amount of lead
emitted per unit of economic activity, and control ef-
ficiencies are reported as the percent that emissions
are reduced through the application of pollution con-
trol technology to the process. The year-specific emis-
sion factors and control efficiencies were multiplied
by the economic activity data for that year, for that
process, as reported in the Trends database, using the
following equation found in the Trends report:
Emissions = (Economic Activity) X
(Emission Factor) X (1 - Control Efficiency) (26)
This calculation yielded the estimated control sce-
nario emissions, by industrial process. Industrial pro-
cesses were then assigned to an NEA code. Finally,
all processes assigned to a given NEA code were
summed to give a total emissions estimate for that
NEA code.
Estimates of industrial process emissions in the
no-control scenario
The results from the J/W model were used to es-
timate process emissions in the no-control scenario.
As stated above, the J/W model provides percent
changes in economic outputs by industrial sector. To
use these values, lead-emitting industrial processes
(in the Trends database) were assigned to a J/W sec-
tor. The percent change for that sector from the J/W
model was then used to adjust the economic activity
data for that process from the Trends database. These
adjusted economic output figures were used together
with 1970 emission factors and control efficiencies to
derive the estimated lead emissions for each indus-
trial process in the no-control scenario. The 1970
emission factors and control efficiencies were used
for all years in the analysis (1970, 1975, 1980, 1985
and 1990) in the no-control scenario; this assumes that
emissions per unit economic output and control effi-
ciencies would have been constant over time in the
absence of the CAA. This is the same approach that
was used to project the changes in emissions from
industrial processes for other criteria pollutants in other
portions of the CAA retrospective analysis. The pro-
cess-level emissions were then aggregated to the NEA-
code level, as in the controlled scenario.
Matching TRI Data to Industrial Process
Emissions Differentials
The methods described in the preceding section
yielded emissions estimates from industrial processes
in the control and no-control scenarios, by NEA code.
We used these estimates to derive percent changes in
emissions between control and no-control scenarios,
by NEA code, for application to the TRI emissions
data. However, since TRI data are reported by SIC
code, we first mapped" NEA codes to the appropriate
SIC codes, and used the percent change for each NEA
code to represent the percent change for all SIC codes
covered by that NEA code.
It should be noted that the Trends data base cov-
ers only the most important sources of lead in air, not
all sources; as a result, not all SIC codes reporting
lead emissions in TRI correspond to an NEA code for
which emission differentials have been estimated.
However, we assume that the TRI emissions sources
that have a match are the most important sources of
lead air emissions. In fact, although only 48 out of
519 legitimate SIC codes reporting lead emissions in
TRI have matching differentials, these SIC codes ac-
count for over 69 percent of the lead emissions re-
ported in TRI. The remaining 31 percent of the emis-
sions are distributed relatively evenly among the re-
maining 471 SIC codes, each of which contributes a
small amount to total emissions.
For the 31 percent of the emissions without dif-
ferentials, this analysis has no information regarding
the change in the lead emissions overtime or between
the control and no-control scenarios; therefore, we are
unable to predict benefits attributable to the CAA for
these emission sources. Although excluding these
sources may lead us to underestimate total benefits,
we believe these sources are unlikely to contribute
significantly to the difference between control and no-
control scenarios. The Trends data focus on the point
sources of lead emissions of greatest concern to the
Project Team and of greatest regulatory activity. If a
process within an SIC code does not appear in the
Trends, it is unlikely to have had specific CAA con-
trols instituted over the past 20 years. A lack of con-
trol efficiencies for smaller sources prevents them
from being included.
It should also be noted that the total industrial
process emissions of lead estimated in the 1990 Trends
report actually exceeds the reported lead emissions in
G-18
-------
Appendix G: Lead Benefits Analysis
TRI, despite the fact that TRI covers more SIC codes.
This is probably attributable in part to the fact that
TRI covers only a subset of the facilities contributing
to economic output in an SIC code. TRI reporting rules
only require facilities with greater than 10 employees
and who use certain amounts of lead in their processes
to submit information to TRI, while the Trends report
attempted to estimate emissions from all sources con-
tributing to the economic output for the industrial sec-
tor, regardless of size. However, the components of
the Trends data base used in this analysis (i.e., emis-
sions factors, economic output data) represent typical
conditions at average facilities; they do not allow for
the representation of the distribution of emissions
across particular facilities. In contrast, a major strength
of the TRI is its match of emissions data with geo-
graphical information. Because the distribution of
emissions geographically determines the size of ex-
posed populations, this analysis used the TRI data,
rather than Trends data, to characterize lead release
quantities, and used the Trends figures only to char-
acterize relative emissions and changes over time,
rather than to estimate total quantities.
Because the Trends data are intended only as an
estimate of emissions using typical conditions at av-
erage facilities, and do not capture the differences in
facility-level emissions, the data do not provide suffi-
cient information to make specific quantitative adjust-
ments to the TRI-based benefits estimates to account
for the overall higher emissions estimates in Trends.
However, since Trends does generally suggest that
there are many more sources than are accounted for
by TRI, it is possible that our benefits calculations
may be underestimated.
Some additional assumptions were necessary
when matching the TRI lead release data and the dif-
ferentials from the Trends data. Ideally, we would like
to know whether the facilities present at a given loca-
tion, as reported in the 1990 TRI, were present and
operating in earlier years; whether facilities operat-
ing in 1970 have ceased to operate; and whether new
facilities would have been constructed in the no-con-
trol situation. Unfortunately, data do not exist in an
accessible form at this level of detail for the years
1970 through 1990. Therefore, forthe purposes ofthis
exercise, we have assumed that the locations and num-
bers of the 1990 sources are the same as they were in
1970.
Methods Used to Determine Changes in
Lead Emissions from Industrial Boilers
from 1970 to 1990
Several sources were used to determine the change
in lead emissions from industrial boilers. TRI
locational data, Trends database national fuel con-
sumption levels and emissions factors, and NEA and
SIC codes were used to derive the emissions for the
control and no-control scenarios.
TRI Data
The TRI does not appear generally to contain com-
bustion emissions data. In general, the emissions data
are from process sources. We reached this conclusion
based on two pieces of information:
(1) TRI reporting requirements: TRI has three
reporting requirements: (a) the facility must fall in
SIC codes 20-39; (b) the facility must employ more
than 10 persons; and (c) the facility must manufac-
ture or process more than 25,000 pounds of a TRI
chemical, or otherwise use more than 10,000 pounds
Firms must submit reports only for the chemical that
exceeds the thresholds given in item (c), but they must
report all releases of that chemical, including releases
from uses that would not qualify alone. If the TRI
chemical is part of a blended substance and the quan-
tity of the TRI chemical in the blend exceeds the
threshold, it must be reported. For industrial boilers,
if the amount of lead in the fuel were to exceed the
10,000 pounds threshold, then the firm would be re-
quired to report all emissions of lead from combus-
tion of fuel. There is an exemption, however, for in-
gredients present in small proportions. If the amount
of lead in the oil were less than 0.1 percent (1,000
ppm), then the firm would not be required to report
the emissions.
The conclusion from the above information is that
most firms burning used oil are probably not report-
ing lead combustion emissions to TRI because these
releases fall outside the TRI reporting requirements.
The concentration at which lead is typically found is
used oil (100 ppm) (NRDC, 1991) is much less than
the minimum concentration required for reporting
(1,000 ppm).
(2) Use data from the TRI data base: The hypoth-
esis that firms do not report lead combustion was con-
firmed by an analysis of the data submitted by the
G-19
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
firms reporting lead use to TRI. On the TRI submis-
sion forms, firms must indicate how the chemical is
used. Our analysis of category codes submitted by
firms reporting lead emissions showed the following
four use category reports: as a formulation compo-
nent; as a reactant; as an article component; and re-
packaging only. None of these category codes sug-
gest that the source of the reported lead release is com-
bustion. Therefore, we may conclude that all of the
lead emissions reported in TRI are process emissions.
Based on these analyses, the Project Team could
not use the TRI release data to evaluate releases of
lead from industrial combustion. However, the TRI
geographical information was used to locate indus-
trial facilities by longitude and latitude in order to
combine combustion data with population informa-
tion. For combustion emissions, the calculations in-
cluded all TRI reporting facilities, not just those who
reported lead emissions. The assignment of combus-
tion emissions to these facilities is described below.
Derivation of Industrial Combustion Emissions
1970-1990
As with industrial process emissions, estimates
were created for the emissions of lead from industrial
combustion under the CAA, and in the absence of the
CAA, for the years 1970, 1975, 1980, 1985, and 1990.
These emissions estimates were used, in combination
with the TRI data base geographic information, to
obtain facility-level release estimates for the years of
interest for the control and no-control scenarios. The
method for deriving these emissions estimates in-
cluded both the macroeconomic impact of the CAA
and the change in emissions per unit of output that
resulted from specific pollution control mandates of
the CAA. The same data sources were used to derive
combustion differentials as were used to derive pro-
cess differentials. The particular data elements and
the methods by which these data were combined to
derive lead emissions estimates from industrial com-
bustion are described below.
Estimates of combustion emissions under the
control scenario
The Trends database contains a national aggre-
gate industrial fuel consumption estimate, by fuel type
(coal, natural gas, oil). For each fuel type, the fuel
consumption estimate was disaggregated by the share
of that fuel used by each NEA industrial category,
using the NEA data base. It should be noted that the
NEA includes data only for the years 1970 through
1985. For 1990, the 1985 figures were used to disag-
gregate the national-level consumption figure into
NEA industrial categories.
The Trends database also contains emissions fac-
tors for industrial fuel use, by fuel type, as well as
control efficiencies. The lead emissions from indus-
trial combustion for each NEA category was derived
by multiplying the fuel-specific combustion estimate
for each NEA category by the emission factor and
control efficiency for that fuel type. The result was
emissions of lead by NEA code and by fuel type.
Emissions from all fuel types were then summed by
NEA code. By using the NEA data to disaggregate
the industrial fuel consumption figures, the analysis
assumes that the industrial combustion emissions are
the same among all industries covered by a given NEA
code, an assumption which may bias the analysis.
Estimates of combustion emissions under the
no-control scenario
As in the control scenario, the national aggregate
industrial fuel consumption estimate, by fuel type
(coal, natural gas, oil), was disaggregated by the share
of that fuel used by each NEA industrial category.
The fuel use was then adjusted by one of two factors:
(1) seven of the NEA codes were specifically mod-
eled by the Industrial Combustion Emissions (ICE)
model — for these sectors, the ICE modeled percent
changes were used instead of J/W percent changes; or
(2) the remaining NEA codes were matched to J/W
sectors — the J/W percent changes were then applied
to those matched NEA codes. These fuel use estimates
were then combined with the 1970 emission factors
and control efficiencies for industrial combustion by
fuel type from the Trends database to obtain combus-
tion-related lead emissions from industrial boilers in
the no-control scenario, by NEA code.
The process-specific data in the Trends database,
and the energy use data in the NEA, are much more
disaggregated than the J/W sectoral projections. For
the purpose of the analysis, it was assumed that all of
the specific industrial processes in the Trends data-
base and industrial categories in the NEA data set as-
signed to a given J/W sector changed at the same rate
as the entire J/W sector. For example, if the economic
activity in the J/W Sector 20, "Primary Metals,"
changed by one percent between the control and no-
control scenarios, then the analysis assumed that eco-
nomic activity in each industrial process assigned to
G-20
-------
Appendix G: Lead Benefits Analysis
the Primary Metals sector also increased by one per-
cent. This approach assumes that the economic ac-
tivities of specific industries within a sector are equally
affected by the imposition of the CAA. This assump-
tion is consistent with the projection of the change in
emissions from industrial processes for the other cri-
teria air pollutants, which were calculated using a simi-
lar process.
Matching TRI Data to Industrial Combustion
Emissions Data
Because of the structure of the TRI reporting re-
quirements, it does not appear that TRI generally con-
tains releases from combustion sources. Although TRI
may incidentally contain lead combustion emissions,
TRI would contain data on such releases only if the
reporting facility also used more than 10,000 pounds
of lead per year for manufacturing or processing. As
a result, the combustion releases, estimated using the
methods described above, do not have corresponding
data in the TRI data base. Therefore, we devised a
different method for estimating benefits from changes
in combustion releases.
The first step in the method was to divide the es-
timates of total releases of lead from industrial com-
bustion, by NEA code, by an estimate of the number
of facilities in each NEA code. The number of facili-
ties in each NEA category was estimated using the
1987 Census of Manufactures. This Census, conducted
by the U.S. Department of Commerce, tallies the num-
ber of facilities by four-digit SIC code; these SIC codes
were matched to the NEA codes.
Dividing total lead emissions emitted by number
of facilities yielded the average yearly lead emissions
from industrial combustion for each SIC code. We
then assigned this average value to all reporting TRI
facilities in the SIC code. The consequence of this
approach is that the modeling of combustion from
industrial facilities includes substantially more sources
than the modeling of industrial process emissions;
combustion emissions are assigned to essentially all
facilities reporting to TRI, while the process emis-
sions are only evaluated for facilities actually report-
ing lead air emissions from processes.
One unavoidable drawback to this approach is that
it cannot capture differences in release quantities
among facilities within an SIC code. Furthermore, this
approach does not capture all combustion emissions
because we assign average emissions only to facili-
ties that report to TRI. TRI facilities account for be-
tween two percent and 50 percent of all facilities listed
in the Census of Manufacturers, depending on the SIC
code. Because of the inability to place the remaining
facilities geographically, this analysis excludes the
consideration of emissions from non-TRI facilities.
Methods Used to Determine Changes in
Lead Emissions from Electric Utilities
from 1975 to 1990
The estimation of lead emissions from electric
utilities required data from three different sources.
Energy use data for the control and no-control sce-
narios were obtained from the national coal use esti-
mates prepared for the section 812 analysis by ICF
Incorporated. The OAQPS Trends Database provided
emissions factors and control efficiencies. Individual
plant latitudes, longitudes, and stack information were
collected from the EPA Interim Emissions Inventory.
This analysis combines these three sets of data and
estimates annual lead emissions at the plant level for
coal burning electric utilities in the control and no-
control scenarios. This section describes the sources
and the methods used to create the final data set.
Coal-Use Data
The energy use data obtained from national coal-
use estimates provide plant level energy consumption
information for 822 electric utilities. The data set were
separated into four distinct sets for the years 1975,
1980, 1985, and 1990. Each set of data provided the
state where the plants are located, the plant names,
and the amount of coal consumed, for both the con-
trol and no-control scenarios. The four data sets were
combined into one comprehensive set by matching
the plants' names and states.
The EPA Interim Emissions Inventory
The EPA Office of Air Quality Planning and Stan-
dards Technical Support Division provided the 1991
EPA Interim Emissions Inventory. The Interim Inven-
tory contains data for all electric utility and industrial
plants in the United States including latitude, longi-
tude, stack height, stack diameter, stack velocity, and
stack temperature. The additional stack parameter data
allowed the use of plant-specific parameters in the air
modeling for electric utilities rather than average pa-
rameters for all facilities as was done for industrial
emissions.
G-21
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Matching the Coal-Use Data to the Interim
Emissions Inventory
The combination of the Interim Emissions Inven-
tory and the coal-use data required two steps. First,
the Interim Emissions Inventory had to be pared down
to include only electric utility data, and to narrow the
information provided for each utility. Second, the two
databases had to be combined. One difficulty in com-
bining them was the lack of a common data field that
would allow a quick and complete matching process.
Electric utility plants were identified in the In-
terim Emissions Inventory by SIC code (code 4911).
The associated stack information file, which lists the
size of every stack on every plant, was reduced to in-
clude only the tallest stack for each plant. This pro-
vides a reasonable estimate of the stack height at which
most emissions occur. The air modeling assumes that
each electric utility releases its emissions from the
largest stack that exists at that plant.
Next, the procedure matched the abridged Interim
Emissions Inventory file with the coal use data. Due
to the lack of a common data field between the two
sets, this process required several phases. Both data
sets had name fields, but these fields utilized differ-
ent naming conventions for the plants. Therefore the
name fields were matched directly, with individual
words in the names, and then with abridged words
from the names. Abridged word matches were double
checked by ensuring that the names were indeed simi-
lar and by verifying that the state fields matched. Fi-
nally some matches were made by hand.
Only 27 unmatched plants with positive coal use
remained. There were 493 matched plants with posi-
tive coal usage and these were included in the final
data set.29 To eliminate under-counting of emissions,
the emissions from the 27 unmatched plants were al-
located to matched plants within the states where the
unmatched plants were located. Allocations were
weighted according to the emission level for each
matched plant within that state in the year in which
the allocation was being made.
Emissions Factors and Control Efficiencies
At this stage, the electric utilities data set con-
tained coal consumption by plant by year in the con-
trol and no-control cases as well as air modeling pa-
rameters. Using emission factors for lead and control
efficiencies for electric utilities, estimates of lead emis-
sions per plant per year could now be calculated. As
in the industrial source analysis, the emission factors
and control efficiencies come from the 1991 OAQPS
Trends database.
Control efficiencies are available for coal-fired
electric utilities in each year between 1975 and 1990.
As in the industrial source analysis, it is assumed that
pollution control on coal-burning power plants with-
out the CAA would be the same as the pollution con-
trol level in 1970. Therefore, the control efficiency
from 1970 is used in the no-control analysis.
The emission factor obtained from the Trends
database is expressed in terms of lead emitted per ton
of coal burned (6,050 grams per 1,000 tons).30 The
combined data set, though, contains quantity of coal
burned per plant per year in energy units (trillions of
BTUs). To reconcile this difference, a conversion fac-
tor was obtained from a 1992 DOE report titled Cost
and Quality of Fuels for Electric Utility Plants 1991.
The conversion factor used (20.93 million BTUs per
ton of coal) is the average BTU per pound of coal
burned for all domestic electric utility plants in 1990.
Data for a small subset of other years were also pro-
vided in the DOE report, but they did not differ sig-
nificantly from the 1990 number. Therefore, the 1990
conversion factor (637.3 pounds of lead per trillion
BTU) is assumed valid over the entire study period.
The final equation for lead emissions looks quite simi-
lar to the equation used in the industrial source analy-
sis.31 The only change is that "Economic Activity"
has been replaced by "Coal Consumed" for this par-
ticular analysis:
Emissions = (Coal Consumed) X
(Emission Factor) X (1 - Control Efficiency) (26)
This equation produces estimates of the emissions
per plant per year in both the control and the no-con-
trol scenarios.
29 Plants with zero coal usage were not immediately excluded from the analysis due to the possibility of analyzing lead emissions
from oil combustion at these plants. However, OAQPS has suggested that oil combustion comprises under two percent of the total
lead emitted from electric utilities. For this reason, the electric utility analysis focused entirely on coal.
30 The actual figure cited is 12.1 metric pounds per 1,000 tons. A metric pound is one two-thousandth of a metric ton.
31 U.S. EPA, 1991a
G-22
-------
Appendix G: Lead Benefits Analysis
Use of Air Dispersion Modeling to
Estimate Ambient Air Lead Levels
To link estimates of lead emissions to blood lead
levels of populations living in the vicinity of a facil-
ity, the lead benefits model first uses air dispersion
modeling to estimate air lead concentrations surround-
ing facilities that emit lead into the air. The air con-
centrations are then linked to blood lead levels.
This analysis uses the Industrial Source Complex
Long Term (ISCLT) air dispersion model, a steady-
state Gaussian plume model, to estimate long-term
lead concentrations downwind of a source. The con-
centration is modeled as a function of site parameters
(stack height, stack velocity).32 The general form of
the concentration equation from a point source at a
distance r greater than one meter away is as follows:33
r (28)
a,v'jt ^2n r& ucz
where,
C|ir = concentration at distance r ((.ig/ni3).
Q = pollutant emission rate (g/sec),
f = frequency of occurrence of wind speed
and direction,
0 = sector width (radians),
S = smoothing function used to smooth
discontinuities at sector boundaries,
u = mean wind speed (m/sec),
c = standard deviation of vertical concentra-
z
tion distribution (m),
V = vertical term (m),
K = scaling coefficient for unit agreement.
For each facility modeled in the lead benefits
model, a 21 by 21 kilometer grid around the source is
specified. The model stores data in 1 km by 1 km cells
and calculates the air lead concentrations for each of
the 441 cells surrounding a given facility. Fugitive
sources are modeled similarly, the only difference
being a modified form of Equation 28.
For facility-specific weather data, the model used
Stability Array (STAR) data. The STAR data contain
information on typical wind speed and direction for
thousands of weather stations in the U.S. For each
facility, the model accesses the STAR data for the
weather station nearest the source. Standard default
parameters are used for the other parameters because
facility-specific data are not available for them (ex-
cept for utilities). Table G-3 lists default parameters
for the ISCLT, and summarizes sources for other pa-
rameters.
Industrial process emissions were modeled as ei-
ther point or fugitive sources, depending on how they
were reported in TRI. All industrial combustion emis-
sions were modeled as "fugitive" emissions. This is a
more appropriate model scenario for boiler emissions
than a 10 meter stack scenario. All electric utility
sources were modeled as point sources.
The model tracks all lead emissions to a given
grid cell. That is, if the plumes of two or more sources
overlap in a given cell, the air concentration in the
given cell is determined from the sum of all of the
contributing sources.
Determination of Blood Lead Levels
from Air Lead Concentrations
Once the air lead concentrations surrounding a
given plant are estimated, the model estimates blood
lead levels for children and adults living in those ar-
eas. This section describes the methods and data
sources used to derive blood lead levels from esti-
mated air lead concentrations.
Relationship Between Air Lead Concentrations
and Blood Lead Levels
The rates at which lead is absorbed from air de-
pend on the age of the exposed individual, distance
from the facility, the initial concentration of blood
lead, and other factors. In addition, rates determined
from empirical data may differ depending on whether
or not the analyses from which rates are derived have
controlled for factors such as lead exposure through
deposition on dust and soil (i.e., "indirect exposure").
Especially when children constitute the exposed group,
the inclusion of indirect exposure results in higher air
lead to blood lead slopes. In both cases, the slope re-
32 Ideally, reported stack and fugitive air releases would be modeled using site-specific data (such as source area or stack height).
However, since TRI does not contain such facility-specific information, default values are used to model TRI facilities.
33 This equation is from U.S. EPA (1992). The equation is for a specific wind speed, direction, and category (ijk). Each facility
has several combinations of these that must be added to arrive at a total concentration at that point. The equation for area sources is
similar.
G-23
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic G-3. Air Modeling Parameters.
Pa nun etc r
Industrial
Source
Value
Klectric
Utility
Value
Source/
Comment
Stack height
10m
site-specific or 115.0
m*
Industrial - U.S. EPA (1992) Utilities — U.S.
EPA (1991b)
Kxit velocity
0.01 m/s
site-specific or 22.5
m/s*
Industrial - U.S. EPA (1992) Utilities — U.S.
EPA (1991b)
Stack dhiinctil
1 m
site-specific or 5.15
m*
Industrial - U.S. EPA (1992) Utilities — U.S.
EPA (1991b)
l.xit <51 s tem peril ture
293° K
site-specific or
427.5 *
Industrial - U.S. EPA (1992) Utilities — U.S.
EPA (1991b)
Area source size
10m2
10 m2
U.S. EPA (1992)
Area source height
3m
3m
U.S. EPA (1992)
Lend emission rate
site-specific
site-specific
Industrial - TRIS (lbs/yr)
Utilities — SAI & OAQPS (lbs/yr)
Frequency of wind speed
and direction
site-specific
site-specific
STAR data
Secto rwidtli
22.5°
22.5°
360° divided by 16 wind directions
Wind speed
site-specific
site-specific
ST AR data (m/sec)
Smoothing function
calculated
calculated
Vertical term
calculated
calculated
* average value for electric utilities, utilized for utilities without this information
lationship is expressed as the change in blood lead
((.ig/dL) per change in air concentration (pg/m3).
In performing this analysis, a choice had to be
made between the use of air lead:blood lead relation-
ships that account for inhalation exposure ("direct"
slopes) and those that account for exposure to lead
deposited from air onto soil and dust ("indirect"
slopes). The choice of which slopes to use considered
both the effects on the estimate of benefits over time
(from 1970 to 1990) and the estimate of the differ-
ence in benefits between the control and no-control
scenarios. The indirect slope is more comprehensive
in its coverage of the types of exposures that will re-
sult from air releases, and thus captures more of the
health effects predicted to occur from lead exposures,
especially to children. For this reason, indirect slopes
are preferred to direct slopes, especially when com-
paring the control and no-control scenarios: using only
the direct slope would underestimate the benefits of
avoiding deposition that controls confer. However,
indirect slopes may capture effects from exposure to
soil and dust lead deposited from both current air re-
leases and historic air releases. Since lead's dissipa-
tion from soil is slow relative to its removal from air,
the reservoir of lead in soil and dust is unlikely to
change at the same rate as the reductions in air lead
concentrations. Therefore, using indirect slopes to
represent a change in blood lead over time due to re-
duced air emissions may overestimate the change in
blood lead, and thus overestimate the benefits of re-
ductions overtime, to the extent that the indirect slope
captures exposure to the total reservoir of soil and dust
lead, rather than only recently deposited lead.
Given that the focus of this analysis is the differ-
ence between the control and no-control scenarios, it
is important to capture both the benefits from reduced
lead deposition that result from the CAA, and the di-
rect benefits from reduced air concentrations. There-
fore, this analysis modeled changes in blood lead lev-
els using indirect slopes. It should be kept in mind
G-24
-------
Appendix G: Lead Benefits Analysis
that this choice may overestimate blood lead changes
overtime for both the control and no-control scenarios.
The relationship between concentrations of lead
in ambient air and blood lead concentrations has been
evaluated by a variety of methods. These include ex-
perimental studies of adult volunteers, as well as epi-
demiological studies of different populations of chil-
dren and adults. The discussion below describes the
slopes used in this analysis for children and adults,
and for individuals with blood lead values greater than
30 (Jg/dL.
Children
U.S. EPA (1986b) reports that slopes which in-
clude both direct (inhalation) and indirect (via soil,
dust, etc.) air lead contributions vary widely, but typi-
cally range from three to five (ig/dL increment in
children's blood lead per (.ig/ni3 increment in air lead
concentration (roughly double the slope due to inhaled
air lead alone). Since hand dust levels can play a sig-
nificant role in blood lead levels (U.S. EPA, 1986b),
this higher slope may be due to mouthing behavior of
children that brings them into contact with dust and
soil.
Specific values for estimating contribution of air
lead to blood lead, including indirect pathways, are
cited in U.S. EPA (1986b); slope values (ranging from
-2.63 to 31.2) and data sources for these values are
presented in Table 11-36 of U.S. EPA (1986b). The
median of these values is 4.0 pg/dL per (Jg/m3, which
matches the midpoint of the range of typical slope
values. This analysis used this value to represent the
relationship between air lead concentrations and blood
lead concentrations for children living in the vicinity
of point sources of lead emissions.
The use of this slope assumes that indirect expo-
sure" principally measures indirect effects of lead
emissions to air (through deposition to dust and soil).
However, it is possible that these slopes include other
exposures not related to air lead. In many cases re-
searchers have measured other possible exposures,
such as water and food, and have confirmed that the
most significant contribution comes from soil and dust
lead, which is assumed to result from air deposition
of lead. Those studies that measured lead in tap water
showed that mean levels were generally low or not
significantly related to blood lead. Landrigan et al.
(1975) measured lead in pottery and food; lead in pot-
tery was found in only 2.8 percent of homes, and food
and water made no more than a negligible contribu-
tion to lead uptake. Lead in paint was measured in
some studies.34 Landrigan and Baker (1981) measured
lead in paint at levels greater than one percent in about
one fourth to one third of the houses in each area stud-
ied. Brunekreef et al. (1981) measured high levels of
paint in some houses, but excluded these data points
from the analysis.
Despite the possibility of confounding factors, this
analysis uses the median value determined above (4.0
(ig/dL per (ig/m3) as the appropriate slope for chil-
dren living within five kilometers of the point source.
Five kilometers is chosen as the cut off point because
the data from most of the studies cited collected the
majority of their data points near lead smelters.35 Fur-
thermore, these slopes, although measured primarily
in the vicinity of smelters, are assumed applicable to
all point sources that emit lead into the air.
Adults
For adult males and females, the air lead/ blood
lead slopes that include indirect effects due to soil and
dust differ very little from slopes that include only
direct effects. This result is expected since the higher
indirect slope values estimated for children are as-
sumed to be as a result of mouthing behavior typical
of young children.
U.S. EPA (1986b) describes several population
studies that estimate indirect slopes for men; these
slopes range from -0.1 to 3.1 (ig/dL per (ig/m3.36 Snee
(1981) determined a weighted average of these stud-
ies and one other study.37 The average slope, weighted
by the inverse of each study's variance, is 1.0 pg/dL
per (ig/m3. However, the Azar study measured the di-
rect relationship between air lead and blood lead. Ex-
cluding the Azar study from the weighted average,
the average slope is 1.1 (jg/m3. Excluding the highest
and lowest slopes from this group (from Goldsmith,
34 Landrigan and Baker, 1981; Brunekreef et al., 1981.
35 U.S. EPA, 1986b, Table 11-36.
36 Johnson et al., 1976; Nordman, 1975; Goldsmith, 1974; Tsuchiya et al., 1975; Fugas et al., 1973.
37 Azar et al., 1975.
G-25
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
1974 and Tsuchiya et al., 1975), both of which had
difficulties,38 the resulting slope is 1.4 (ig/dL per jjg/
m3.
Slopes for females range from 0.6 to 2.4 for gen-
eral atmospheric conditions.39 Snee determined an
average slope for women of 0.9 (ig/dL per (ig/m3,
weighted by the inverse of the variances of the stud-
ies. Excluding the slope for women from Goldsmith
(1974), the resulting slope for women is 1.0 pg/dL
per (ig/m3.
These values are adjusted by a factor of 1.3 to
account for the resorption of lead from bone tissue
(according to Chamberlain, 1983), thus deriving an
adjusted slope estimate of 1.8 (jg/dL blood lead per
(ig/m3 increment in air lead concentration for men and
1.3 for women. These are the slope estimates used in
this analysis.
Individuals with initial blood lead levels of 30
jig/dL and greater
For individuals with high blood lead levels, the
air lead to blood lead uptake slopes have been shown
to be much shallower, as described by U.S. EPA
(1986b). An appropriate change in blood lead per
change in air lead is 0.5 pg/dL per (jg/m3 for indi-
viduals that have initial blood lead levels in the range
of 30 to 40 (jg/dL. This value is based on cross-sec-
tional and experimental studies.40 For individuals with
initial blood lead levels greater than 40 (jg/dL, an ap-
propriate range of slopes is 0.03 to 0.2, as determined
by occupational studies listed in Table 11-37 of U.S.
EPA (1986b). The median value of these studies is
0.07. These two slopes (0.5 for the population with
blood lead levels between 30 and 40 (jg/dL and 0.07
for blood lead levels greater than 40 pg/dL) are used
for both children and adults in this analysis. These
relationships are summarized in Table G-4.
Estimates of Initial Blood Lead Concentrations
The benefits model requires an initial distribution
of blood lead levels in the exposed populations to
model health benefits of reducing lead air emissions.
The model estimates the new distribution of blood lead
levels that would exist after a given change in air con-
centrations using the slopes described above. Finally,
the model estimates the difference between the two
distributions. This analysis begins with an initial 1970,
no-control scenario blood lead distribution from which
all subsequent changes are modeled. This approach
requires an estimate of the blood lead distributions in
the U.S. population in 1970. Unfortunately, there are
no actual national blood lead distribution estimates
for 1970. Although the first NHANES study covered
1970, blood lead data were not collected in this study41
Nonetheless, a 1970 distribution of blood lead was
estimated using NHANES II data (from 1976-1980),
combined with estimates of typical changes in blood
lead levels from 1970-1976 observed in localized
screening studies.
Tabic G-4. Estimated Indirect Intake Slopes: Increment of Blood Lead Concentration (in ng/dL) per
Unit of Air Lead Concentration (ng/m ')•
Individuals with blood lead
levels <30 ng/d 1,
Individualswith blood lead
levels 30-40 ng/d I
Individuals with blood
lead levels > 40 fig/dL
Adult Males
1.8
0.5
0.07
Adult Females
1.3
0.5
0.07
Children
4.0
0.5
0.07
38 Goldsmith (1974) refrigerated (rather than froze) the blood samples, and did not analyze the samples until 8 or 9 months after
they were taken, and restricted the analysis to one determination for each blood sample. Tsuchiya et al. (1975) measured air lead
concentrations after blood samples were taken; blood was drawn in August and September of 1971, whereas air samples were taken
during the 13 month period from September 1971 to September 1972.
39 Tepper and Levin, 1975; Johnson et al., 1976; Nordman, 1975; Goldsmith, 1974; Daines et al., 1972.
40 U.S. EPA, 1986b.
41 NCHS, 1993a.
G-26
-------
Appendix G: Lead Benefits Analysis
A major drawback to this approach is the uncer-
tainty in deriving the 1970 estimates. Another draw-
back to beginning with the 1970 level and modeling
changes from that point is the analysis only represents
changes in lead exposure from air; reductions from
other sources of lead exposure are not accounted for.
The purpose of this analysis is to identify changes at-
tributable to the CAA mandates; changes from other
sources of lead exposure should not be considered.
However, due to nonlinear nature of the lead concen-
tration-response functions (see above), the overall
exposure context in which the air lead exposure re-
ductions take place will influence the estimate of ben-
efits from those reductions. Specifically, at higher
blood lead levels, the slope of the concentration-re-
sponse curve is shallower than at lower levels. As a
result, a given change in the mean blood lead level
may result in a smaller change in the health effect if
the change occurs from a relatively high starting level.
On the other hand, if one accounts for the fact that
other sources of lead exposure are reduced at the same
time that the given air reductions occur, then those air
emissions reductions may result in greater changes in
health risk.
This issue is of concern even though the analysis
focuses on the difference between the control and no-
control scenarios, since the health benefit implications
of the emissions differentials between the scenarios
will depend on the point on the blood lead distribu-
tion curve at which the differences are considered.
That is, a difference between a mean blood lead of 25
(ig/dL and one of 20 (jg/dL may have different health
implications than a difference between 15 pg/dL and
10 (jg/dL, even though the absolute value of the dif-
ference is the same (5 pg/dL).
An alternative method is to "start" with a 1990
blood lead level and to "back-calculate" benefits by
representing the differentials as increases over the
1990 levels, rather than decreases from 1970 levels.
The advantage of this approach is that it accounts for
reductions in lead exposure from other sources, as rep-
resented by current blood lead levels. Its disadvan-
tage is that it holds other sources constant to (lower)
1990 levels, and thus the modeling may underesti-
mate actual blood lead distributions in earlier years,
and thereby overestimate benefits from controls dur-
ing those years. This analysis presents the results of
both approaches, indicated as "forward-looking" and
"backward-looking".
Combination of Air Concentration
Estimates with Population Data
The modeled air lead concentrations at various
distances from the sources were combined with popu-
lation data from the Census Bureau to arrive at an
estimate of the number of cases of health effects for
each of the years from 1970 to 1990 in both the con-
trol and no-control scenarios. The primary census in-
formation was accessed from the Graphical Exposure
Modeling System Database (GEMS), an EPA main-
frame database system. The following data were ob-
tained from GEMS forthe years 1970,1980, and 1990:
total population for each Block Group/Enumeration
District (BG/ED); state and county FIPS codes asso-
ciated with each BG/ED; latitude and longitude of each
BG/ED; and population of males under 5 and females
under 5 for each BG/ED. The intervening five year
intervals (1975 and 1985) were estimated using the
Intercensal County Estimates from the Census, which
estimate annual populations on a county by county
basis. The decennial Census data and the Intercensal
County Estimates data sets were related by county
FIPS codes; the population in each BG/ED was as-
sumed to grow or shrink at the same rate as the county
population as a whole.
Since the concentration-response data are particu-
lar to specific sex and adult age groups, additional
population data were also required to determine the
sizes of affected subpopulations. For 1990 age and
sex, the U.S. Census, 1992 was used, with age groups
tallied as necessary. For 1980 age and sex, the U.S.
Census, 1982 was used, with age groups also tallied
as necessary. The 1970 age and sex breakdowns were
obtained through personal communication with the
Census Bureau.42 The age and sex percentages were
interpolated for intervening years.
Pregnant women are often a subpopulation of in-
terest for lead effects. Although pregnant women
themselves may be harmed by exposure to lead, this
analysis was concerned with pregnant women because
of possible effects on their fetuses who will be born
42 Personal communication, Karl Kuellmer, Abt Associates and the Bureau of Census, Population, Age and Sex telephone staff,
March, 1994.
G-27
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
and evince effects as young children. To estimate the number of exposed fetuses who were born during the years
of interest,43 birth rates for 1970, 1980 and 1990 were obtained from the Census Bureau.44 These birth rates
were used to interpolate for years between 1970 and 1980, and for the years between 1980 and 1990.
Results
For both the control and no-control scenarios, Table G-5 shows estimated lead emissions from electric
utilities, industrial processes, and industrial combustion. Tables G-6 and G-7 show the differences in health
impacts between the two scenarios (for industrial processes, industrial combustion and electric utilities only) for
the "forward-looking" and "backward-looking" analyses. The modeled population for each year is also pre-
sented.
Tabic G-5. Estimated Lead Emissions from Electric Utilities.Industrial Processes and
industrial Combustion fin Tons).
1970
1975
1980
1985
1990
Electric Utilities11
Control Sceimrio
1,351
636
175
190
Electric Utilities"
No-control Scrim rio
2,309
3,143
3,670
3,864
Industrial Processes Control
Sccna rio
7,789
3,317
1,032
670
658
In tl lis ti-i:i 1 Processes
No-control Scen;i rio
7,789
7,124
6,550
5,696
5,305
1 n (1 lis tl-hi 1 ('o in Ihi st i on
Control Sceniirio
4,329
4,354
1,880
190
187
1 n d us tr i:i 1 Co in bu st i on
No-control Scen;i rio
4,329
4,457
4,653
4,584
4,596
' Appropriate dataon electric utilities do not exist for yearsprior to 1975.
43 Note that we do not record the number of pregnancies, since the valuation only applies if the child is born and lives to exhibit
the effect. Neither are we concerned with whether the births are single or multiple births, since each fetus is at risk, whether a
pregnant women carries one or more fetuses.
44 Personal communication, Karl Kuellmer, Abt Associates and the Bureau of Census, Population, Fertility/Births telephone
staff.
G-28
-------
Appendix G: Lead Benefits Analysis
Table G-6. Yearly Differences in Number of Health Effects Between the Control and No-
control Scenarios: Industrial Processes, Boilers, and Electric Utilities (Holding Other Lead
Sources at Constant 1970 Levels).
Health Effect
1975
1980
1985
1990
Mortality
Men (40-54)
0.1
1.5
2.5
2.7
Men (55-64)
0.0
1.1
1.8
1.8
Men (65-74)
0.0
0.4
0.7
0.8
Women (45-74)
0
0.8
1.3
1.4
I nfant s
0
0 001
0 00?
0 00?
Total
0.1
3.9
6.3
6.7
Coronary Heart Disease
Men (40-54)
0.1
1.8
3.0
3.3
Men (55-64)
0.0
0.7
1.2
1.2
Men (65-74)
0.0
1.0
1.6
1.7
Women (45-74)
n_L
LI
2_L
2_L
Total
0.2
4.8
8.0
8.3
Strokes
Cerebrovascular Accident (men 45-74)
0.1
1.1
1.8
1.8
Cerebrovascular Accident (women 45-74)
0
0.5
0.9
0.9
Brain Infarction (men 45-74)
0
0.7
1.1
1.1
Brain Infarction (women 45-74)
ill
Oil
il£
Total
0.1
2.7
4.4
4.4
Hvpcrtcnsion (men 20-74)
149
3.790
6.350
6.670
IQ Decrement.
Lost IQ Points
630
14.300
22,700
23.900
IQ<70 (cases)
3
60
120
125
Population Exposed (millions)
188
197
207
217
G-29
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic G-7. Yearly Differences in Number of Health Effects Between the Control and No-control
Scenarios: Industrial Processes. Boilers, and Electric Utilities (Holding Other Lead Sources at
Constant 1990 Levels).
Health Effect
1975
1980
1985
1990
Mortality
Men (40-54)
0.3
6.9
11.5
12.5
Men (55-64)
0.2
5.1
8.3
8.2
Men (65-74)
0.1
2.0
3.5
3.9
Women (45-74)
0.2
3.9
6.4
6.4
I nfants
a
0 001
0 00?
0 002
Total
0.8
17.9
29.7
31.0
Coronarv Heart Disease
Men (40-54)
0.4
8.3
13.8
15.0
Men (55-64)
0.1
3.4
5.6
5.6
Men (65-74)
0.2
4.4
7.6
8.0
Women (45-74)
iL2
5JI
2A
21
Total
0.9
22.1
36.6
38.3
Strokes
Cerebrovascular Accident (men 45-74)
0.2
5.0
8.1
8.2
Cerebrovascular Accident
0.1
2.6
4.1
4.2
(women 45-74)
Brain Infarction (men 45-74)
0.1
2.8
4.6
4.7
Brain Infarction (women 45-74)
ILL
1A
11
11
Total
0.5
12.0
19.5
19.8
Hypertension (men 20-74)
422
10.800
18.100
19,000
IQ Decrement
Lost IQ Points
630
14.300
22,700
23,900
IQ<70 (cases)
0
31
50
61
Population Exposed (millions)
188
197
207
217
G-30
-------
Appendix G: Lead Benefits Analysis
Reduction in Health Effects
Attributable to Gasoline Lead
Reductions
Estimating Changes in Amount of Lead
in Gasoline from 1970 to 1990
The relationship between the national mean blood
lead level and lead in gasoline is calculated as a func-
tion of the amount of lead in gasoline consumed. Thus,
to calculate the health benefits from gasoline lead re-
ductions, necessary inputs are estimates of lead in
gasoline consumed over the period 1970 to 1990 and
the amount of lead in gasoline that would have been
consumed in the absence of the Clean Air Act. These
values are calculated using the quantity of both leaded
and unleaded gasoline sold each year and the con-
centration of lead in leaded and unleaded gasoline for
each year in the period of interest. For each year, the
relationship is expressed as:
Fraction of Total Sales Comprised of Leaded Gaso-
line (FRACpJ: For the control scenario, this analysis
used information reported by Kolb and Longo (1991)
for the fraction of the gasoline sales represented by
leaded gasoline for the years 1970 through 1988. For
1989 and 1990, data were taken from DOE (1990 and
1991, respectively). For the no-control scenario, all of
the gasoline sold was assumed to be leaded for all years.
Lead Content of Gasoline (Pb, , , and Pb , , ):
J 1 Leaded unleaded.
Argonne National Laboratory in Argonne, Illinois was
the source for the data on the lead content of leaded
and unleaded gasoline for the period 1974-1990.
Argonne compiled these data from historical sales data
submitted to EPA, from Clean Air Act regulations on
lead content, and from recent Motor Vehicle Manu-
facturers Association (MVMA) surveys. For 1970
through 1973, this analysis assumed the lead content
of gasoline to be at the 1974 level. For the no-control
scenario, this analysis used the 1974 lead content in
leaded gasoline as the lead content in all gasoline for
each year.
LEAD =
SOLD
365 days
Y l^r^x/'^+d-FRACj:^PHMJ (29)
where:
LEAD =
SOLD =
FRACpb =
Pb
Pb
average lead per day in gasoline
sold in a given year (metric tons/
day),
total quantity of gasoline sold
(million gal/yr),
fraction of total gasoline sales
represented by leaded gasoline
(dimensionless),
lead content of leaded gasoline
(g/gal), and
lead content of unleaded gasoline
(g/gal).
Gasoline Sales (SOLD) : Data on annual gasoline
sales were taken from a report by Argonne National
Laboratories (1993) which presented gasoline sales
for each state in five year intervals over the period
1970-1990. This analysis used linear interpolation to
estimate the gasoline sales for years between the re-
ported years. These data were summed to obtain na-
tional sales figures.
Estimating the Change in Blood Lead
Levels from the Change in the Amount of
Lead in Gasoline
Several studies have found positive correlations
between gasoline lead content and blood lead levels.45
Data from the National Health and Nutrition Examina-
tion Survey (NHANES II) have been used by other re-
searchers who determined similar positive correlations
between gasoline lead and blood lead levels.46
The current analysis used a direct relationship be-
tween consumption of lead in gasoline and blood lead
levels to estimate changes in blood lead levels result-
ing from Clean Air Act regulation of the lead content
of gasoline. This relationship was based on regression
analyses of the reduction of leaded gasoline presented
in the 1985 Regulatory Impact Analysis (RIA).47 Sev-
eral multiple regressions were performed in the RIA to
relate gasoline usage with individuals' blood lead lev-
45 U.S. EPA, 1985; Billick et al., 1979; Billick et al., 1982.
46 Janney, 1982; Annest et al., 1983; Centers for Disease Control, 1993; National Center for Health Statistics, 1993b.
47 U.S. EPA, 1985.
G-31
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
els, which were taken from NHANES II. These re-
gressions of blood lead on gasoline usage controlled
for such variables as age, sex, degree of urbanization,
alcohol consumption, smoking, occupational expo-
sure, dietary factors, region of the country, educational
attainment, and income. The regressions suggested
that a decrease of 100 metric tons per day (MTD) of
lead used in gasoline is associated with a decrease in
mean blood lead concentration of 2.14 (ig/dL for
whites and 2.04 (ig/dL for blacks. In both of these
regressions, gasoline use was found to be a highly sig-
nificant predictor of blood lead (p < 0.0001).48
To determine a single gasoline usage-blood lead
slope for the entire population of the U.S., this analy-
sis used the average of the slopes for blacks and for
whites, weighted by the percentage of blacks and
whites in the U.S. during the time period of the analy-
sis.49 The resulting relationship is 2.13 pg/dL blood
lead per 100 metric tons of lead in gasoline consumed
per day. The same relationship was used to model
changes in both children's and adults' blood lead lev-
els. The U.S. EPA (1985) analyzed data from a study
of black children in Chicago during the time period
1976 to 1980 and determined a slope of 2.08 (ig/dL
per 100 MTD. This slope for children is very similar
to the one used in this analysis.
1970-Forward and 1990-Backward Approaches
As with the industrial processes and boilers analy-
sis, this analysis used two different approaches to de-
termine mean blood lead levels based on changes in
lead concentrations in gasoline. In the 1970-forward
approach, the calculations began with the estimated
blood lead level for 1970. The change in blood lead
level from one year to the next was based upon the
change in the amount of lead in gasoline sold, as dis-
cussed above, for both the control and no-control sce-
narios. For example, to calculate the blood lead level
for 1971, the calculated change in blood lead from
1970 to 1971 was added to the 1970 value. This pro-
cess was repeated for each succeeding year up to 1990.
The 1990-backward approach began with a mean
blood lead level in 1990 for the control scenario. For
the no-control scenario, the starting blood lead was
estimated from the 1990 level used in the control sce-
nario, plus an additional blood lead increment result-
ing from the difference between the 1990 consump-
tion of lead in gasoline under the two scenarios. Again,
the difference in mean blood lead levels from one year
to the next was based on the change in gasoline lead
for the corresponding years. For example, the differ-
ence in blood lead levels between 1990 and 1989 was
subtracted from the 1990 level to determine the 1989
level. The process was continued for each year back
to 1970.
Relating Blood Lead Levels to Population Health
Effects
The mean blood lead levels calculated using the
methods described above were used in the dose-re-
sponse functions for various health effects (e.g., hy-
pertension, chronic heart disease, mortality). This in-
formation was then combined with data on the resi-
dent population of the 48 conterminous states in each
year to determine the total incidence of these health
effects attributable to lead in gasoline. A Department
of Commerce Publication (1991) was used to obtain
the total population in 1970, 1980, and 1983-1990,
while a different publication was the source of the
1975 population values.50 Linear interpolation was
used to estimate the populations in years for which
specific data were not available.
For certain health effects, it was necessary to know
the size of various age groups within the population.
Two different sources were used to estimate the pro-
portions of the population in the age groups of inter-
est. A U.S. Census summary (U.S. Dept. of Com-
merce, 1990) was used for information for 1990 for
children and adults and for 1980 for adults, and Cen-
sus Telephone Staff (U.S. Dept. of Commerce, 1994)
provided information for 1980 for children and 1970
for children and adults. The populations for the inter-
vening years were estimated by linear interpolation.
Changes in Leaded Gasoline Emissions and
Resulting Decreased Blood Lead Levels and Health
Effects
Table G-8 shows the estimated quantity of lead
burned in gasoline in the five year intervals from 1970
to 1990. Tables G-9 and G-10 show the difference in
48 U.S. EPA, 1985.
49 U.S. Department of Commerce, 1992. Although the percentages of blacks and whites changed slightly over this time period
(1970-1990), the change did not affect the value of the weighted slope.
50 U.S. Dept. of Commerce, 1976.
G-32
-------
Appendix G: Lead Benefits Analysis
health impacts between the two scenarios (for lead in
gasoline only) for the "forward-looking" and "back-
ward-looking" analyses. In general, health effect ben-
efits resulting from gasoline lead reductions exceed
those predicted from lead reductions at the point
sources examined (i.e., industrial processes and boil-
ers and electric utilities) by three orders of magni-
tude.
Table G-8. Lead Burned in Gasoline (in tons).
1970
1975
1980
1985
1990
Control Scenario
No-control Scenario
176,100
176,100
179,200
202,600
86,400
206,900
22,000
214,400
2,300
222,900
G-33
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic G-9. Yearly Differences in Number of Health Effects Between the Control and No-control
Scenarios: Lead in Gasoline only (Holding Other Lead Sources at Constant 1970 Levels).
Health Effect
1975
1980
1985
1990
Mortality
Men (40-54)
309
1.820
3.340
4,150
Men (55-64)
220
1,340
2.380
2,700
Men (65-74)
81
520
999
1,260
Women (45-74)
155
939
1.710
2,060
1 nfants
7 340
3 930
4 940
Total
1,220
6,960
12.400
15,100
Coronary Heart Disease
Men (40-54)
230
1.360
2.540
3,280
Men (55-64)
92
563
1.030
1,220
Men (65-74)
113
723
1.380
1,750
Women (45-74)
71
Ml
805
965
Total
508
3.090
5,760
7,210
Strokes
Cerebrovascular Accident (men 45-74)
147
884
1,610
1,960
Cerebrovascular Accident
73
442
805
965
(women 45-74)
Brain Infarction (men 45-74)
85
508
927
1,130
Brain Infarction (women 45-74)
42
786
521
£21
Total
352
2.120
3.862
4,679
Hypertension (men 20-74)
677,000
4.200.000
7.840.000
9,740,000
IQ Decrement
Lost IQ Points
1,030,000
5.020.000
8.580.000
10,400,000
IQ<70 (cases)
3,780
20.100
36.500
45,300
Population Exposed (millions)
214
225
237
247
G-34
-------
Appendix G: Lead Benefits Analysis
Tabic G-10. Yearly Differences in Number of Health Effects Between the Control and No-
control Scenarios: Lead in Gasoline only (Holding Other Lead Sources at Constant 1990
Levels).
Health Effect
1975
1980
1985
1990
Mortality
Men (40-54)
476
3.040
6.140
7,950
Men (55-64)
342
2,250
4.430
5,240
Men (65-74)
128
886
1.880
2,480
Women (45-74)
242
1.590
3.210
4,030
I nfanls
7 340
3 930
4 940
Total
1,640
10.100
19.600
24,600
Coronary Heart Disease
Men (40-54)
356
2.280
4.690
6,310
Men (55-64)
142
945
1.910
2,370
Men (65-74)
176
1.220
2,570
3,380
Women (45-74)
in
1411
1 490
1 860
Total
787
5.180
10.700
13,900
Strokes
Cerebrovascular Accident (men 45-74)
225
1.460
2,940
3,720
Cerebrovascular Accident
113
740
1.490
1,860
(women 45-74)
Brain Infarction (men 45-74)
129
837
1,680
2,120
Brain Infarction (women 45-74)
23.
422
955
1 190
Total
540
3.514
7,065
8,890
Hypertension (men 20-74)
984,000
6.350.000
12.300.000
15,600,000
IQ Decrement
Lost IQ Points
1,030,000
5.030.000
8.580.000
10,400,000
IQ<70 (cases)
3,790
20.100
36.500
45,300
Population Exposed (millions)
214
225
237
247
G-35
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Lead Benefits Analysis
References
Abt Associates, Inc. 1992. The Medical Costs of Five
Illnesses Related to Exposure to Pollutants.
Prepared for: Nicholas Bouwes, Regulatory
Impacts Branch, Economics and Technology
Division, Office of Pollution Prevention and
Toxics, U.S. Environmental Protection
Agency, Washington, D.C.
Abt Associates, Inc. 1995. The Impact of the Clean
Air Act on Lead Pollution: Emissions Reduc-
tions, Health Effects, and Economic Benefits
From 1970 to 1990, Draft. Prepared for Eco-
nomic Analysis and Innovations Division,
Office of Policy Planning and Evaluation,
U.S. EPA. January 19.
Annest, J.L., J.L. Pirkle, D. Makuc, J.W. Neese, D.D.
Bayse, and M.G. Kovar. 1983. "Chronologi-
cal Trend in Blood Lead Levels Between 1976
and 1980." New England Journal of Medi-
cine 308: 1373-1377.
Argonne National Laboratories (Argonne). 1993.
National Gasoline Sales Data, 1970-1990.
Ashenfelter, O. and J. Ham. 1979. "Education, Un-
employment and Earnings." J. Political
Economy 87(5): S99-S131.
Azar, R.D., et al. 1975. An Epidemiologic Approach
to Community Air Lead Exposure Using Per-
sonal Air Samplers. In: Griffin, T.B. and
Knelson, J.H., eds. Lead. Stuttgart, West Ger-
many: Georg Thieme Publishers; pp.254-290.
(Coulston, F. and Korte, F., eds. Environmen-
tal Quality and Safety: Supplement V. 2).
Bellinger, D., J. Sloman, A. Leviton, M. Rabinowitz,
H.L. Needleman, and C. Waternaux. 1991.
"Low-level Lead Exposure and Children's
Cognitive Function in the Preschool Years."
Pediatrics 87(2): 219-227.
Bellinger, D.C. 1992. "Lead Exposure, Intelligence
and Academic Achievement. Pediatrics
90(6): 855.
Billick, I.H., A.S. Curran, and D.R. Shier. 1979.
"Analysis of Pediatric Blood Lead Levels in
New York City for \910-1916 "Environmen-
tal Health Perspectives 31: 183-190.
Billick, I.H., et al. 1982. Predictions of Pediatric Blood
Lead Levels from Gasoline Consumption. U.S.
Department of Housing and Urban Develop-
ment. [Cited in U.S. EPA, 1985.]
Brunekreef, B.D., et al. 1981. "The Arnhem Lead
Study: 1. Lead Uptake by 1- to 3-year-old
Children Living in the Vicinity of a Second-
ary Lead Smelter in Arnhem, the Nether-
lands." Environ. Res. 25: 441-448.
Centers for Disease Control (CDC). 1985. Prevent-
ing Lead Poisoning in Young Children. U.S.
Department of Health and Human Services,
Public Health Service, Centers for Disease
Control, Atlanta, GA.
Centers for Disease Control (CDC). 1991a. Strategic
Plan for Elimination of Childhood Lead Poi-
soning. U.S. Department of Health and Hu-
man Services, Public Health Service, Centers
for Disease Control, Atlanta, GA. February.
Centers for Disease Control (CDC). 1991b. Prevent-
ing Lead Poisoning in Young Children. U.S.
Department of Health and Human Services,
Public Health Service, Centers for Disease
Control, Atlanta, GA. October.
Centers for Disease Control (CDC). 1993. Personal
communication between Abt Associates and
Jim Pirkle. November 16.
Chamberlain, A.C. 1983. "Effect of Airborne Lead
on Blood Lead."Atmos. Environ. 17: 677-692.
Daines, R.H., et al. 1972. "Air Levels of Lead Inside
and Outside of Homes." Ind. Med. Surg. 41:
26-28.
Dietrich, K.N., K.M. Krafft, R. Shukla, R.L.
Bornschein, and P.A. Succop. 1987. The
Neurobehavioral Effects of Prenatal and Early
Postnatal Lead Exposure. In: Toxic Sub-
stances and Mental Retardation:
Neurobehavioral Toxicology and Teratology,
G-36
-------
Appendix G: Lead Benefits Analysis
S.R. Schroeder, Ed. American Association of
Mental Deficiency, Washington, DC, pp. 71-
95 (Monograph No. 8).
Elixhauser, A., R. M. Andrews, and S. Fox. 1993.
Clinical Classifications for Health Policy
Research: Discharge Statistics by Principal
Diagnosis and Procedure. Agency for Health
Care Policy and Research, Center for Gen-
eral Health Services Intramural Research,
U.S. Department of Health and Human Ser-
vices.
Environmental Law Institute (ELI). 1992. Projecting
With and Without Clean Air Act Emissions
for the Section 812 Retrospective Analysis:
A Methodology Based upon the Projection
System Used in the OAQPS "National Air
Pollutant Emission Estimates: Reports."
[Jorgenson/Wilcoxen Model Projections], Jim
Lockhart.
Fugas, M., et al. 1973. "Concentration Levels and
Particle Size Distribution of Lead in the Air
of an Urban and an Industrial Area as a Basis
for the Calculation of Population Exposure."
In: Barth, D., et al. eds. Environmental Health
Aspects of Lead: Proceedings, International
Symposium; October 1972; Amsterdam, The
Netherlands. Luxembourg: Commission of
the European Communities, pp. 961-968.
Goldsmith, J.R. 1974. Food Chain and Health Impli-
cations of Airborne Lead. Sacramento, CA:
State of California, Air Resources Board;
Report No. ARB-R-102-74-36.
Griliches, Zvi. 1977. "Estimating the Returns to
Schooling: Some Econometric Problems."
Econometrica 45:1-22.
Hasselblad, V. 1995. Personal Communication be-
tween V. Hasselblad and Abt Associates, Feb-
ruary 28, 1995.
Janney, A. 1982. The Relationship Between Gasoline
Lead Emissions and Blood Poisoning in
Americans. Prepared for U.S. EPA, Office of
Policy Analysis. [Cited in U.S. EPA, 1985.]
Johnson, D.E., et al. 1976. Base Line Levels of Plati-
num and Palladium in Human Tissue. U.S.
EPA, Health Effects Research Laboratory,
Research Triangle Park, N.C. EPA-600/1-76-
019.
Kakalik, J., et al. 1981. The Cost of Special Educa-
tion. Rand Corporation Report N-1791-ED.
Kolb, J. and K. Longo. 1991. Memorandum to Joel
Schwartz, U.S. EPA, Washington, DC, No-
vember 5.
Krupnick, A.J. and M.L. Cropper. 1989. Valuing
Chronic Morbidity Damages: Medical Costs
and Labor Market Effects. Draft Final Report
to U.S. Environmental Protection Agency,
Office of Policy Planning and Evaluation.
June 26.
Landrigan, P.J., et al. 1975. "Epidemic Lead Absorp-
tion near an Ore Smelter: the Role of Particu-
late Lead." N. Engl. J. Med. 292: 123-129.
Landrigan, P.J. and E.L. Baker. 1981. "Exposure of
Children to Heavy Metals from Smelter: Epi-
demiology and Toxic Consequences."
Environ. Res. 25: 204-224.
McGee and Gordon. 1976. The Results of the
Framingham Study Applied to Four Other
U.S.-based Epidemiologic Studies of Coro-
nary Heart Disease. The Framingham Study:
An Epidemiological Investigation of Cardio-
vascular Disease. Section 31, April.
National Center for Health Statistics (NCHS). 1993a.
Facsimile received by Abt Associates from
Margaret McDowell regarding the types of
laboratory tests conducted during NHANES
I. December 14.
National Center for Health Statistics (NCHS). 1993b.
Personal communication between Abt Asso-
ciates and NCHS Public Information Special-
ist. November 3.
National Energy Accounts, Bureau of Economic
Analysis.
G-37
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Needleman, H.L., et al. 1990. "The Long Term Ef-
fects of Exposure to Low Doses of Lead in
Children. "New England Journal of Medicine
322(2): 83-88.
NHANES, National Health and Nutrition Examina-
tion Survey.
NHANES II, National Health and Nutrition Exami-
nation Survey, 1976-1980.
Nordman, C.H. 1975. Environmental Lead Exposure
in Finland: a Study on Selected Population
Groups [dissertation]. Helsinki, Finland: Uni-
versity of Helsinki.
Oliver, T. 1911. "Lead Poisoning and the Race" Brit-
ish Medical Journal 1(2628): 1096-1098.
[Cited in USEPA (1990).]
Piomellietal. 1984. "Management of Childhood Lead
Poisoning." Pediatrics 4: 105.
Pirkle, J.L., J. Schwartz, J.R. Landis, and W.R. Harlan.
1985. "The Relationship Between Blood Lead
Levels and Blood Pressure and its Cardiovas-
cular Risk Implications." American Journal
of Epidemiology 121: 246-258.
Pirkle, J. L., et al. 1994. "Decline in Blood Lead Lev-
els in the United States, the National Health
and Nutrition Examination Survey
(NHANES)." JAMA 272(4): 284.
Pooling Project Research Group. 1978. "Relationship
of Blood Pressure, Serum Cholesterol, Smok-
ing Habit, Relative Weight and ECG Abnor-
malities to Incidence of Major Coronary
Events: Final Report of the Pooling Project."
Journal of Chronic Disease. Vol. 31.
Rabinowitz, M., D. Bellinger, A. Leviton, H.
Needleman, and S. Schoenbaum. 1987. "Preg-
nancy Hypertension, Blood Pressure During
Labor, and Blood Lead Levels." Hyperten-
sion 10(4): October.
Salkever, D.S. 1995. "Updated Estimates of Earnings
Benefits from Reduced Exposure of Children
to Environmental Lead." Environmental Re-
search 70: 1-6.
Schwartz, J. 1988. "The Relationship Between Blood
Lead and Blood Pressure in the Nhanes II
Survey "EnvironmentalHealth Perspectives.
78: 15-22.
Schwartz, J. 1990. "Lead, Blood Pressure, and Car-
diovascular Disease in Men and Women."
Environmental Health Perspectives, in press.
Schwartz, J. 1992a. "Blood Lead and Blood Pressure:
a Meta-analysis." Presented at the Annual
Meeting of Collegium Ramazzini. November.
Schwartz, J. 1992b. "Chapter 13: Lead, Blood Pres-
sure and Cardiovascular Disease." In: Human
Lead Exposure, H. L. Needleman, Ed. CRC
Press.
Schwartz, J. 1993. "Beyond LOEL's, p Values, and
Vote Counting: Methods for Looking at the
Shapes and Strengths of Associations."
Neurotoxicology 14(2/3): October.
Shurtleff, D. 1974. Some Characteristics Related to
the Incidence of Cardiovascular Disease and
Death. The Framingham Study: An Epidemio-
logical Investigation of Cardiovascular Dis-
ease. Section 30, February.
Silbergeld, E.K., J. Schwartz, and K. Mahaffey. 1988.
"Lead and Osteoporosis: Mobilization of Lead
from Bone in Postmenopausal Women."En-
vironmental Research 47: 79-94.
Snee, R.D. 1981. "Evaluation of Studies of the Rela-
tionship Between Blood Lead and Air Lead."
Int. Arch. Occup. Environ. Health 48: 219-
242.
Taylor, T.N., P H. Davis, J.C. Torner, J. Holmes, J.W.
Meyer, and M. F. Jacobson. 1996. "Lifetime
Cost of Stroke in the United States." Stroke
27(9): 1459-1466.
Tepper, L.B. and L.S. Levin. 1975. "A Survey of Air
and Population Lead Levels in Selected
American Communities." In: Griffin, T.B.;
Knelson, J.H., eds. Lead. Stuttgart, West Ger-
many: Georg Thieme Publishers; pp. 152-196.
(Coulston, F.; Korte, f., eds. Environmental
Quality and Safety: Supplement v. 2).
G-38
-------
Appendix G: Lead Benefits Analysis
Tsuchiya, K., et al. 1975. "Study of Lead Concentra-
tions in Atmosphere and Population in Japan."
In: Griffin, T.B. andKnelson, J.H.. eds. Lead.
Stuttgart, West Germany: Georg Thieme Pub-
lishers; pp.95-145. (Coulston, F.; Korte, F.,
eds/ Environmental Quality and Safety:
Supplement v. 2)
U.S. Census. 1982. United States Summary, General
Population Characteristics, Table 41: Single
Years of Age by Race, Spanish Origin, and
Sex: 1980.
U.S. Census. 1992. United States Summary, General
Population Characteristics, Table 13: Single
Years by Sex, Race, and Hispanic Origin:
1990.
U.S. Department of Commerce. 1976. Statistical Ab-
stract of the United States: 95th Edition. Bu-
reau of the Census. Washington, DC.
U.S. Department of Commerce. 1980. U.S. Census,
United States Summary, General Population
Characteristics.
U.S. Department of Commerce. 1986. Statistical Ab-
stract of the United States: 105th Edition.
Bureau of the Census. Washington, DC.
U.S. Department of Commerce. 1987. Census of
Manufacturers.
U.S. Department of Commerce. 1990. Earnings by
Occupation and Education: 1990. Bureau of
the Census. Washington, DC.
U.S. Department of Commerce. 1990. U.S. Census,
United States Summary, General Population
Characteristics.
U.S. Department of Commerce. 1991. Statistical Ab-
stract of the United States: 111th Edition.
Bureau of the Census. Washington, DC.
U.S. Department of Commerce. 1992. Statistical Ab-
stract of the United States: 112th Edition.
Bureau of the Census. Washington, DC.
U.S. Department of Commerce. 1993. Money Income
of Households, Families, and Persons in the
United States: 1992. Bureau of the Census,
Series P60-184.
U.S. Department of Commerce. 1993. Personal Com-
munication between Bureau of Census, Popu-
lation, Age and Sex Telephone Staff and Karl
Kuellmer of Abt Associates on December 8,
1993.
U.S. Department of Commerce. 1994.City and County
Databook: 1994. Bureau of the Census. Wash-
ington, DC.
U.S. Department of Commerce. 1994. Personal Com-
munication between Bureau of Census, Popu-
lation, Age and Sex Telephone Staff and Karl
Kuellmer of Abt Associates on February 7,
1994.
U.S. Department of Education. 1993. Digest of Edu-
cational Statistics, 1993. National Center for
Educaton Statistics, Office of Educational
Research and Improvement. D.Ed, publica-
tion number NCES 93-292.
U.S. Department of Energy (DOE). 1990. Petroleum
Supply Annual, 1989, Volume 1. DOE publi-
cation number EIA-0340(89)/l
U.S. Department of Energy (DOE). 1991. Petroleum
Supply Annual, 1990, Volume 1. DOE publi-
cation number EIA-0340(90)/l
U.S. Department of Energy (DOE). 1992. Cost and
Quality of Fuels for Electric Utility Plants
1991. DOE/EIA-0191 (91) Energy Informa-
tion Administration, August.
U.S. Environmental Protection Agency (U.S. EPA).
1984. A Survey of the Literature Regarding
the Relationship Between Measures oflQ and
Income. Prepared by ICF, Inc. Report to U.S.
Environmental Protection Agency, Office of
Policy Analysis, June.
U.S. Environmental Protection Agency (U.S. EPA).
1985. Costs and Benefits of Reducing Lead
in Gasoline: Final Regulatory Impact Analy-
sis. Prepared by U.S. Environmental Protec-
tion Agency, Office of Policy Analysis, Eco-
nomic Analysis Division. February.
G-39
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
U.S. Environmental Protection Agency (U.S. EPA).
1986a. Reducing Lead in Drinking Water: A
Benefit Analysis. Prepared by U.S. Environ-
mental Protection Agency, Office of Policy
Planning and Evaluation, Draft Final Report.
December.
U.S. Environmental Protection Agency (U.S. EPA).
1986b. Air Quality Criteria for Lead: Volume
III. Environmental Criteria and Assessment
Office, Research Triangle Park, NC. EPA-
600/8-83/028cF. June.
U.S. Environmental Protection Agency (U.S. EPA).
1987. Methodology for Valuing Health Risks
of Ambient Lead Exposure. Prepared by
Mathtech, Inc. for U.S. Environmental Pro-
tection Agency, Office of Air Quality Plan-
ning and Standards, Ambient Standards
Branch, Contract No. 68-02-4323.
U.S. Environmental Protection Agency (U.S. EPA).
199()a. AIRS Facility Subsystem Source Clas-
sification Codes and Emission Factor List-
ing for Criteria Air Pollutants. Office of Air
Quality Planning and Standards, Research
Triangle Park, NC. EPA-450/4-90-003.
March.
U.S. Environmental Protection Agency (U.S. EPA).
1990b. National Air Pollutant Emission Esti-
mates 1940-1988. Office of Air Quality Plan-
ning and Standards, Technical Support Divi-
sion, National Air Data Branch. Research
Triangle Park, NC. EPA-450/4-90-001.
U.S. Environmental Protection Agency (U.S. EPA).
1990c. Review of the National Ambient Air
Quality Standards for Lead: Assessment of
Scientific and Technical Information. OAQPS
Staff Paper, Air Quality Management Divi-
sion, Research Triangle Park, NC. December.
U.S. Environmental Protection Agency (U.S. EPA).
1991b. The Interim Emissions Inventory. Of-
fice of Air Quality Planning and Standards,
Technical Support Division, Source Recep-
tor Analysis Branch. Research Triangle Park,
NC.
U.S. Environmental Protection Agency (U.S. EPA).
1992.1990 Toxics Release Inventory. Office
of Pollution Prevention and Toxics, Washing-
ton, DC. EPA-700-S-92-002.
U.S. Environmental Protection Agency (U.S. EPA).
1994. Guidance Manualfor the Integrated Ex-
posure Uptake Biokinetic Model for Lead in
Children. February. EPA 540-R-93-081.
U.S. Environmental Protection Agency (U.S. EPA)
database. Graphical Exposure Modeling Sys-
tem Database (GEMS).
Wallsten and Whitfield. 1986. Assessing the Risks to
Young Children of Three Effects Associated
with Elevated Blood Lead Levels. Argonne
National Laboratory. December.
Wittels, E.H., J.W. Hay, and A.M. Gotto, Jr. 1990.
"Medical Costs of Coronary Artery Disease
in the United States," The American Journal
of Cardiology 65: 432-440.
U.S. Environmental Protection Agency (U.S. EPA).
1991a. National Air Quality and Emissions
Trends Report, 1989. Office of Air Quality
Planning and Standards. Research Triangle
Park, NC. EPA-450/4-91-003.
G-40
-------
Appendix H: Air Toxics
Introduction
Air toxics are defined as air pollutants other than
those six criteria pollutants for which EPA sets ac-
ceptable concentrations in ambient air. The SARA 313
Toxic Release Inventory (TRI), covering 328 of the
approximately 3000 potentially hazardous compounds
detected in air, estimated that approximately 1.2 mil-
lion tons of air toxics were released to the atmosphere
in 1987 from U.S. stationary sources alone. While the
TRI estimate tends to understate emissions of toxics
for a number of reasons, it does show that large quan-
tities of toxics are emitted into the atmosphere annu-
ally.
Effects of air toxics emissions are divided into
three categories for study and assessment: cancer;
"noncancer" effects, e.g. a wide variety of serious
health effects such as abnormal development, birth
defects, neurological impairment, or reproductive
impairment, etc.; and ecological effects. Each year,
these air toxics emissions contribute to significant
adverse effects on human health, human welfare, and
ecosystems. In EPA's 1987 Unfinished Business Re-
port1 cancer and noncancer air toxics risk estimates
were considered sufficiently high, relative to risks
addressed by other EPA programs, that the air toxics
program area was among the few rated "high risk".
Limited Scope of this
Assessment
The effects of air toxics emissions are difficult to
quantify. The adverse health effects of toxics are of-
ten irreversible, not mitigated or eliminated by reduc-
tion in ongoing exposure, and involve particularly
painful and/or protracted disease. Therefore these ef-
fects are not readily studied and quantified in human
clinical studies, in contrast to, for example, ambient
ozone. In addition, epidemiological studies of effects
in exposed populations are often confounded by si-
multaneous exposure of subjects to a variety of pol-
lutants. Therefore, the effects of air toxics are often
quantified by extrapolating data from animal studies
to human exposure and expressed as risk per unit of
exposure. Incidence of noncancer effects, for example,
often are difficult to translate into monetized benefits.
Similarly, the quantification of ecological effects
due to emissions of air toxics is hampered by lack of
sufficient information regarding contribution of
sources to exposure, associations between exposure
to mixtures of toxics and various ecological endpoints,
and economic valuation for ecological endpoints.
The air toxics portion of this study is, of neces-
sity, separate and more qualitative in nature than the
benefit analysis conducted for the criteria air pollut-
ants. Limitations in the quantitative analyses of air
toxics effects led the Project Team to decide to ex-
clude the available quantitative results from the pri-
mary analysis of CAA costs and benefits. Table H-l
presents the range of potential human health and eco-
logical effects that can occur due to air toxics expo-
sure. As indicated, this appendix presents quantita-
tive estimates of benefits of CAA air toxics control
for the cancer mortality endpoint for only nonutility
stationary source and mobile source categories.
Noncancer effects and ecological effects are described
qualitatively.
1 U.S. EPA. Office of Policy Planning and Evaluation. Unfinished Business: A Comparative Assessment of Environmental
Problems. February 1987.
H-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Tabic H-l. Health and Welfare Effects of Hazardous Air Pollutants.
Effect Category
Q u a n ti tie d Effec ts
Unquantificd Effects
Other Possible Effects
Human Health
Cancer Mortality
- nonutility stationary
source
- mobile source
Cancer Mortality
- utility source
- area source
Noncancer effects
- neurological
- respiratory
- reproductive
- hematopoetic
- developmental
- immunological
- organ toxicity
Human Welfare
Decreased income and
recreation
opportunities due to
fish advisories
Odors
Decreased income resulting
from decreased physical
performance
Ecological
Effects on wildlife
Effects on plants
Ecosystem effects
Loss ofbiological
diversity
Effects on global climate
Other Welfare
Visibility
Materials Damage
History of Air Toxics Standards
under the Clean Air Act of 1970
The 1970 Clean Air Act required the EPA to list a
chemical as a hazardous air pollutant if it met the leg-
islative definition provided:
"The term 'hazardous air pollutant' means
an air pollutant to which no ambient air
quality standard is applicable and which in
the judgment of the Administrator may cause,
or contribute to, an increase in mortality or
an increase in serious irreversible, or
incapacitating reversible, illness. "2
Once a HAP was listed, the EPA Administrator
was required to:
2 42 U.S.C. § 1857(a)(1).
3 42 U.S.C. § 1857(b).
"establish any such standard at the level
which in his judgment provides an ample
margin of safety to protect the public health
from such hazardous air pollutant. "3
In other words the EPA had to first determine that
a chemical was a HAP, and then regulate the emis-
sions of each HAP based solely on human health ef-
fects and with an ample margin of safety. This regu-
latory mandate proved extremely difficult for EPA to
fulfill, for reasons discussed below, and the result was
that only seven HAPs were regulated over a period of
20 years.
Listing chemicals became a difficult task because
of debates within and outside of the EPA surrounding
issues of how much data are needed and which meth-
H-2
-------
Appendix H: Air Toxics
odologies should be used to list a chemical as a HAP.
An even more difficult issue was how to define the
Congressional mandate to provide an "ample margin
of safety." For carcinogens, there is generally no
threshold of exposure considered to be without risk.
What level of risk, then, is acceptable, and how should
it be calculated? The EPA struggled to provide an-
swers to these questions, and was challenged in court.
The end result was a 1987 ruling by the D.C. Circuit
Court that provided the EPA with a legal framework
with which to determine an "ample margin of safety."
This framework was interpreted and used by the EPA
in its 1989 benzene regulations.
Quantifiable Stationary Source
Air Toxics Benefits
One might be tempted to presume that the few
federal HAP standards set would have achieved rela-
tively substantial reductions in quantifiable risk. While
some standards set under section 112 of the Clean Air
Act appear to have achieved significant reductions in
cancer incidence, the coverage, quantification, and
monetization of the full range of potential adverse
effects remains severely limited. This fact serves to
highlight the inadequacy of current methods of evalu-
ating HAP control benefits. This limited ability to es-
timate the total human health and ecological benefits
of HAP reductions is an important area for future re-
search. Thus the quantifiable benefits for CAA air
toxics control presented here are limited in scope.
There are three sources of information that pro-
vide a picture of potential stationary source air toxics
benefits of the CAA. EPA's Cancer Risk studies at-
tempted to broadly assess the magnitude and nature
of the air toxics problem by developing quantitative
estimates of cancer risks posed by selected air toxics
and their sources. Secondly, risk assessments con-
ducted in conjunction with the promulgation of Na-
tional Emissions Standards for Hazardous Air Pollut-
ants (NESHAPs) offer a snapshot of potential mon-
etized cancer mortality benefits. Finally, the Project
Team attempted to estimate historical non-utility sta-
tionary source HAP-related direct inhalation cancer
incidence reductions. Results from each of these stud-
ies are presented below.
EPA Analyses of Cancer Risks from
Selected Air Toxic Pollutants
The Agency conducted two efforts to broadly as-
sess the magnitude and nature of the air toxics prob-
lem. The 1985 report entitled, "The Air Toxics Prob-
lem in the United States: An Analysis of Cancer Risks
for Selected Pollutants"4 otherwise known as the "Six
Month Study," was intended to serve as a "scoping"
study to provide a quick assessment of the air toxics
problem utilizing only readily available data on com-
pound potencies, emissions, and ambient pollutant
concentrations. The Agency updated this analysis of
cancer risks in the 1990 report entitled "Cancer Risk
from Outdoor Exposure to Air Toxics" referred to here
as the "1990 Cancer Risk study."5
For the pollutant and source categories examined,
the 1990 Cancer Risk study estimated the total na-
tionwide cancer incidence due to outdoor concentra-
tions of air toxics to range from 1,700 to as many as
2,700 excess cancer cases peryear, with 14 compounds
accounting for approximately 95 percent of the an-
nual cancer cases. Additionally, point sources con-
tribute 25 percent of annual cases and area sources
contribute 75 percent of annual cases. Mobile sources
account for 56 percent of the nationwide total.6
The Six Month study indicates that the criteria air
pollutant programs appear to have done more to re-
duce air toxics levels during the 1970 to 1990 period
than have regulatory actions aimed at specific toxic
compounds promulgated during the same period.
Metals and polynuclear compounds usually are emit-
ted as particulate matter and most of the volatile or-
ganic compounds are ozone precursors. As such, they
are regulated under State Implementation Plan (SIP)
and New Source Performance Standard (NSPS) pro-
grams and Title II motor vehicle regulations. A num-
ber of reports cited indicate significant reductions in
air toxics emissions attributable to actions taken un-
4 U.S. EPA. Office of Air Quality Planning and Standards. The Air Toxics Problem in the United States: An Analysis of Cancer
Risks for Selected Pollutants. May 1985. EPA-450/1-85-001.
5 U.S. EPA. Office of Air Quality Planning and Standards. Cancer Risk from Outdoor Exposure to Air Toxics. September 1990.
EPA-450/1-90-004a.
6 The 1990 Cancer Risk study reported approximately 500 - 900 more cancer cases per year than the Six Month Study due
primarily to the inclusion of more pollutants, better accounting of emissions sources, and, in some cases, increases in unit risk
estimates.
H-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
der SIP, NSPS and mobile source programs. Addi-
tionally, EPA conducted a comparison of air quality
and emissions data for 1970 with the estimates of can-
cer incidence for 1980.7 Methods, assumptions and
pollutants included were held constant over the pe-
riod. The analysis showed a significant decrease in
incidence during the decade due to improvements in
air quality, presumably related to general regulatory
programs. For the 16 pollutants studied, estimated
nationwide cancer incidence decreased from 3600 in
1970 to 1600 in 1980. The 1990 Cancer Risk Study
did not attempt to update this analysis.
Although it is difficult to draw quantitative con-
clusions from these two studies regarding the ben-
efits of CAA air toxics control, it is apparent that the
pollutant-specific and source category-specific
NESHAPs were not structured to reduce significant
air toxic emissions from area and mobile sources. In
fact, the 1990 Cancer Risk Study indicates that con-
siderable cancer risk remained prior to passage of the
1990 CAA Amendments: as many as 2,700 excess
cancer cases annually. Some studies indicate that the
criteria air pollutant program played a critical role
during the 1970 to 1990 period in achieving air toxic
emission reductions and therefore decreasing cancer
risk.
Cancer Risk Estimates from NESHAP
Risk Assessments
In looking back at the estimated effects of the HAP
standards, EPA found that the effects of the NESHAPs
were not quantified completely. These estimates oc-
curred at a time when emission estimation and risk
assessment methodologies for HAPs were first being
developed. One consequence is that because emissions
were not fully characterized, air toxics exposures could
not be completely assessed. Additionally, most assess-
ments only focused on the specific HAP being listed
under the CAA and did not assess the reduction of
other pollutants, which are currently considered HAPs.
For example, while the vinyl chloride standard reduces
emissions of ethylene dichloride, these emission re-
ductions were not assessed in the risk assessment. In
a different context, reductions of HAP may also
achieve reductions of VOC and PM. The benefits of
such reductions generally were also not evaluated. In
addition, EPA generally did not assess the potential
exposure to high, short-term concentrations of HAP
and therefore did not know whether toxic effects from
acute exposures would have been predicted and pos-
sibly addressed by the HAP standards.
In addition, people living near emission sources
of concern are often exposed to a mix of pollutants at
once. Some pollutants have been shown to act syner-
gistically together to create a health risk greater than
the risk that would be expected by simply adding the
two exposure levels together. More research is needed
to understand the effects of multiple-pollutant expo-
sures. Finally, HAP risks tend to be distributed un-
evenly across exposed populations, with particularly
high exposures occurring closest to emission sources.
It should be noted that HAP exposure to specific popu-
lations may tend to fall disproportionately among the
poor and minorities, who are more likely to live in
close proximity to emitting facilities.
With the above caveats in mind, Table H-2 pro-
vides information about maximum individual risk
taken from the Federal Register notices for the
NESHAPs promulgated before the 1990 amendments
to the Clean Air Act. The benefits are calculated by
multiplying the estimated annual incidence reduction
by the $4.8 million valuation per statistical life (1990
dollars). These benefit estimates provide a snapshot
of potential monetized benefits for the year in which
each NESHAP was promulgated. Of course these es-
timates do not include air toxics benefits for other
health and ecological benefit categories, or air toxics
benefits from co-control of criteria air pollutants. All
uncertainties associated with the original estimates
remain.
Non-utility Stationary Source
Cancer Incidence Reductions
The Project Team commissioned two studies to
estimate reductions in cancer incidence due to pre-
1990 NESHAPs: the PES Study and the ICF Re-analy-
sis. The methodology used for most air pollutant evalu-
ations involved a "back calculation" for the estima-
tion of incidence reductions. However, the EPA has
elected not to rely on the results of this analysis given
critical methodological flaws. Despite the Project
Team's concerns, the methodology and results of the
two studies are presented below in the interest of full
disclosure and to guide efforts to develop a more valid
7 Hunt, W.F., Faoro, R.B. and Outran, T.C., "Estimation of Cancer Incidence Cases and Rates for Selected Toxic Air Pollutants
Using Ambient Air Pollution Data, 1970 vs. 1980". U.S. EPA. April 1985.
H-4
-------
Appendix H: Air Toxics
Table H-2. Cancer Incidence Reductions and Monetized Benefits for NESHAPs.
Pollutant
Source
Cil tOgO I V
Year
Prom ulgated
l'rc- Reg
M iixim u in
Individ ual
Risk
Post-Reg
IVI ii xi in u in
Individual
Risk
Red uction
in Cancer
Incidence
(per year)
Benefits in
Siii ill ion per
ye a r
(1990S)
l)C 11 zone
1985
1.5xl0"3
4.5xl04
.31
1.5
ho n /one
coke by-
product
1984
7xl0"3
2xl04
1.95
9.4
1)011/one
storage
vessels
1982
4.5xl04
3xl0"5
0.01 to 0.06
0.05 to
0.3
hen/.one
waste
operations
1986
2xl0"3
5xl0"5
0.55
2.6
1)011/0110
transfer
operations
1987
6xl0"3
4x10"5
0.98
4.7
sir sonic
primary
copper
1986
1.3xl0"3 to
5xl0'6
1.2xl0"3
to 3x10"6
0.09
0.4
sir sonic
glass manuf.
1986
7xl0"4 to
3xl0"5
1.7xl04
to 6x10"6
0.117 to
0.0034
0.02 to 0.6
asbestos
demolition
1973
100
480
vinyl
chloride
PVC
production
1975
10.5
50.4
and reliable analysis of the health-related benefits of
HAP reductions in the upcoming section 812 Prospec-
tive studies.
PES Study
Methodology
The first attempt to estimate, for this study, his-
torical non-utility stationary source HAP-related di-
rect inhalation cancer incidence reductions was con-
ducted by Pacific Environmental Services (PES). The
basic approach used in the PES study was to adjust
the cancer incidence estimates developed for EPA's
1990 Cancer Risk study to reflect the changes in emis-
sions of, and exposures to, 14 key HAPs: arsenic, as-
bestos, benzene, 1,3-butadiene, carbon tetrachloride,
chloroform, hexavalent chromium, dioxin, ethylene
dichloride, ethylene dibromide, formaldehyde, gaso-
line vapors, products of incomplete combustion
(PICs), and vinyl chloride.
The first step was to compile baseline incidence
levels, defined as cancer cases per million population,
for each of the 14 pollutants. The point estimates of
incidence from the 1990 Cancer Risk study were used
for this purpose. For some source categories, the "best
point estimate" from the 1990 Cancer Risk study was
used, for others a mid-point was selected.8 These
baseline incidence levels were based on measured
ambient concentrations of the pollutant, modeled con-
centrations, or both.
The second step involved allocating baseline in-
cidence levels to the individual source categories
known to emit the relevant pollutant. In some cases,
8 For some of the source categories, the original NESHAP/Air Toxic Exposure and Risk Information System (NESHAP/
ATERIS) estimates of incidence were not available, in which case the baseline incidence was obtained from the 1989 National Air
Toxics Information Clearinghouse( NATICH) Database Report. (See PES, "Draft Summary of Methodology Used for Cancer from
Stationary Sources," memorandum from Ken Meardon, PES to Vasu Kilaru, US EPA, March 22, 1993, p. 2.)
H-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
adjustments were made to reflect differences among
the vintages of source category-specific data.9 All
baseline incidence estimates were ultimately ex-
pressed relative to a 1985 base year.10 The assump-
tion was then made that source-category incidence
rates were proportional to the level of emissions from
that source category.
Next, levels of control for each source category-
specific incidence rate were estimated for each of the
target years of the present analysis (i.e., 1970, 1975,
1980, 1985, and 1990).11 Source category-specific
activity level indicators were then established and
linked to changes in corresponding activity indica-
tors provided by the J/W macroeconomic modeling
results. Activity levels were estimated for each source
category, for each of the target years, and for each of
the two scenarios.
Finally, source category/pollutant combination
incidence levels for both the control and no-control
scenarios were developed. These incidence levels were
developed based on the baseline incidence levels, the
activity indicators, and the control levels for each year.
Both of these latter two factors varied between the
control and no-control scenarios. The activity levels
differed based on the specific levels of related sector
economic activity predicted by the J/W model for the
control and no-control scenario. The control levels
prevailing in each of the target years were used for
the control scenario, and the 1970 control level was
applied throughout the 1970 to 1990 period for the
no-control scenario.12 The formula used for these cal-
culations was as follows:13
A, p, (l - r)
A. = x x l7rH x lyfrry1 (1)
by by ^ by'
where:
I =
cancer incidence for a source category-
pollutant combination
A =
activity level for a source category
P =
population
C =
control level for a source category-pol-
lutant combination
ty =
target year (1970 ... 1990)
by =
base year
Findings
The PES analysis concluded that substantial re-
ductions in HAP-related cancer cases were achieved
during the reference period of the present study. The
vast majority of these estimated reductions were at-
tributable to reduced exposures to asbestos, particu-
larly from manufacturing and fabricating sources.14
In fact, roughly 75 percent of the total reduction in
cancer cases averaged over the 1970 to 1990 period
were attributed to asbestos control.15 Figure H-l sum-
marizes the PES study overall cancer incidence re-
ductions and the relative contribution of asbestos-re-
lated reductions over the study period.
The Project Team had several concerns about the
PES results. First and foremost, the reductions in as-
bestos-related cancer cases appeared to be substan-
tially higher than expected, particularly in the earlier
target years. Second, the control scenario activity level
indicators for several sources with which Project Team
members were familiar did not appear to be even re-
motely consistent with actual historical activity pat-
terns.16 Finally, the level of documentation of the ana-
lytical methodologies, assumptions, and results was
insufficient to ascertain the validity and reliability of
9 For example, six discrete sources for vinyl chloride were identified in the Six-Month Study Update. Point estimate incidences
for each of these source categories came from separate references with databases corresponding to different years. (See PES, "retro-
spective analysis for section 812(a) Benefits Study," September 30, 1992, p. 8.)
10 See PES, March 22, 1993 memorandum, p. 3.
11 Control level estimates were based on one of the following: control efficiencies for related criteria pollutants defined in the
criteria pollutant analysis, reference documents such as Control Technology Guidelines (CTGs) or Background Information Docu-
ments (BIDs), preambles for related regulations, or EPA experts. (See PES, March 22, 1993 memorandum, p. 3.)
12 More detailed descriptions of the methodology and associated uncertainties are provided in "Retrospective Analysis for section
812(a) Benefits Study," a September 30, 1992 memorandum from Ken Meardon, PES to Vasu Kilaru, US EPA.
13 See PES, March 22, 1993 memorandum, p. 4.
14 PES, "Cancer Risk Estimates from Stationary Sources," memorandum from Ken Meardon, PES to Vasu Kilaru, US EPA,
March 5, 1993.
15ICF, "Direct Inhalation Incidence Benefits," Draft Report, November 11, 1994, p. 10.
16 For example, the activity indicators for Municipal Waste Combustors (MWCs) incorporated in the PES analysis decline
dramatically throughout the 1975 to 1990 period. (See PES, March 5, 1993 memorandum to Vasu Kilaru, p. 10). In reality, overall
MWC capacity and throughput increased significantly over this period.
H-6
-------
Appendix H: Air Toxics
Figure H-l. PES Estimated Reductions in HAP-Related
Cancer Cases.
6
5
-T3
o
3
-o 4
-o
c
£ w
« 3 J
O
s- E—1
O L
§
o
l
0
the results. Ultimately, the Project Team determined
that it was necessary to conduct a formal review and
re-analysis of the cancer incidence reductions associ-
ated with non-utility stationary source HAP controls.
The results of the PES analysis remain a relevant part
of the record of the present study, however, since they
provided a substantial basis for the subsequent re-
analysis by ICF Incorporated.
ICF Re-analysis
Methodology
The purposes of the ICF Re-analysis were to ex-
amine the methodology and results of the PES study,
particularly to address the aforementioned concerns
of the Project Team, and to develop a revised set of
estimates. Due to significant constraints on the re-
sources remaining for HAP analysis in the section 812
study, however, only a few key HAPs could be inves-
tigated in depth and many important issues could not
be addressed.17 Furthermore, the effects of two early
and potentially important HAP standards -the Beryl-
lium and Mercury NESHAPs- could not be evalu-
ated. Nevertheless, the ICF Re-analysis clarified some
potential sources of uncertainty in the PES re-
sults and provided revised cancer incidence re-
duction estimates for several HAPs.
A key uncertainty in the PES results was
associated with the use of a "back-calculation"
technique to estimate incidence reductions for
some HAPs. The back-calculation technique
estimates uncontrolled incidence by dividing
residual incidence by the assumed control effi-
ciency. This approach means uncontrolled inci-
dence, and therefore incidence reductions, are
highly sensitive to small changes in assumed
control efficiency.18 In some cases, the PES
analysis may have used control efficiencies
which were too high, resulting in overestima-
tion of uncontrolled incidence and therefore in-
cidence reductions attributable to the CAA.19
The vinyl chloride incidence reduction estimates ap-
pear to be significantly influenced by the use of this
back-calculation technique. Another important source
of uncertainty identified by ICF involved the poten-
tial overestimation of incidence totals when source
apportionment is based on measured ambient concen-
trations.20 ICF was unable, however, to perform an
extensive evaluation of the activity level indicators
used in the PES study.21
The first step undertaken in the re-analysis was to
conduct a screening test to identify the HAPs which
accounted for the most significant estimated incidence
reductions. Based on this screening analysis, ICF
eliminated 1,3-butadiene, carbon tetrachloride, chlo-
roform, gasoline vapors, chromium, formaldehyde,
and PICs from the detailed re-analysis effort.
Detailed reviews were then conducted for the re-
maining HAPs: vinyl chloride, dioxins, ethylene
dibromide (EDB), ethylene dichloride (EDC), ben-
zene, asbestos, and arsenic. In the re-analysis of these
HAPs, ICF determined whether a forward- or back-
calculation technique was used for the relevant source
categories of a given HAP, reviewed the regulatory
DOther HAPs
BAsbestos
1975 1980 1985 1990
Year
17 For example, the Project Team sought to develop and apply a methodology for estimating a central tendency estimate for the
total carcinogenic risk imposed by all the HAPs examined. The intent was to address concerns about potential overestimation of
aggregate risk measures when combining upper bound risk estimates of multiple HAPs. Unfortunately, resources were insufficient to
continue development of this methodology.
18 An example of this back-calculation technique illustrating the sensitivity to the assumed control efficiency is presented on page
12 of the draft ICF report.
19 See ICF Draft Report, p. 12.
20 See ICF Draft Report, p. 9.
21 See ICF Draft Report, p. 13.
H-7
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
history of the relevant source categories to re-evalu-
ate the assumed control efficiencies, and reviewed
the upper-bound unit risk factor for each HAP.
Revised total incidence reduction estimates for each
HAP and for each target year were then calculated
using the same basic calculation procedure used
by PES. Finally, ICF identified a number of residual
deficiencies in the analysis which could only be
addressed through additional research and analy-
sis.22
Findings
The ICF Re-analysis largely affirmed the origi-
nal results obtained by PES; primarily because the
PES analysis itself served as the basis for the re-
analysis and only minor adjustments were adopted
for many critical variables. In particular, most
Project Team concerns regarding the PES method-
ology could not be resolved, including uncertain-
ties associated with activity levels, assumed con-
trol efficiencies, and the unexpectedly high esti-
mated incidence reductions associated with asbes-
tos. In fact, the ICF Re-analysis produced a revised
upper bound estimate for vinyl chloride-related in-
cidence reductions which were even higher than
the asbestos benefits.
Several sets of results were developed by ICF
and presented in either the November 1994 draft
report or in briefing materials prepared for the Sci-
ence Advisory Board Clean Air Act Compliance
Analysis Council Physical Effects Subcommittee
(SAB/ACCACAPERS) in May 1995. The first set
of results is based on the assumption of 100 per-
cent source compliance with HAP control require-
ments. An alternative set of results was developed as-
suming an 80 percent compliance rate with applicable
standards. Given the linear effect of changes in com-
pliance rates, these results were precisely 20 percent
lower than the first set of estimates. At the May 1995
ACCACAPERS briefing, estimates based on the 100
percent compliance estimates were presented. For as-
bestos, the revised incidence reductions were pre-
sented and characterized as upper bound. The asbes-
tos estimates were then combined with upper and
lower bound estimates for vinyl chloride and for "all
other compounds." Figure H-2 presents the total can-
cer incidence reductions derived from the ICF Re-
analysis, using the asbestos estimates combined with
the lower bound estimates for non-asbestos HAPs.
Figure H-2. ICF Estimated Reductions in Total HAP-
Related Cancer Cases Using Upper Bound Asbestos
Incidence and Lower Bound Non-Asbestos HAP Inci-
dence.
3
~0
W 72
O m
a
c3
o
~Other HAPs
nAsbestos
1975 1980 1985 1990
Year
Figure H-3. ICF Estimated Reduction in Total HAP-
Related Cancer Cases Using Upper Bound Incidence for
All HAPs.
12
11 -
10
^ e
£
W 72
72 -3
s- E—1
o
a
03
O
I
in
~ Other HAPs
nAsbestos
1975 1980 1985 1990
Year
Figure H-3 presents a comparable compilation reflect-
ing the upper bound estimates for all HAPs.
The Project Team remains concerned about these
incidence reduction estimates, particularly given the
doubts raised by the SAB/ACCACAPERS at the May
1995 presentation of these results. For instance, sev-
eral critical assumptions are needed to make this analy-
sis valid when applied to EPA's NESHAPs. The flaws
in these assumptions are described below.
(1) The risk estimates described in the 1990 Can-
cer Risk study, which served as the baseline for deter-
mining risk reductions, were accepted without ques-
tion. There are myriad uncertainties in these estimates
22 Additional details of the ICF Re-analysis methodology can be found in ICF,
Report, November 11, 1994.
'Direct Inhalation Incidence Benefits," Draft
H-8
-------
Appendix H: Air Toxics
that must be recognized, as the study was designed
only to generate rough order-of-magnitude estimates
of the extent of the air toxics cancer problem.
(2) The percent control efficiency for emission
reductions, which is calculated in each standard, would
have to apply across every facility. Typically, the
emissions reductions vary between facilities; using a
single average reduction could skew the results.
(3) There is a direct correlation between the num-
ber of tons of emissions reduced and incidence re-
duced by a specific regulation. Given the assumption
of a linear, non-threshold dose-response curve (as is
typically done for cancer), this is theoretically cor-
rect.
(4) Finally, the back calculation approach assumes
that there is 100 percent compliance with the regula-
tion.
EPA staff reviewed the "back calculation" ap-
proach for one of the more controversial aspects of
the vinyl chloride (VC) NESHAP. The PES study es-
timates benefits at 426 cases reduced in 1990. The
ICF Re-analysis resulted in an even higher estimate,
between 1,000 and 7,000 cases annually. An analysis
by EPA staff indicated that these vinyl chloride risk
estimates are highly suspect given historical cancer
incidence data for hepatic angiocarcinoma, a specific
cancer that has been linked to vinyl chloride (Koppikar
and Fegley, 1995). The following analysis demon-
strates the inadequacies of the assumptions in the 1993
study.
(1) In the actual standard, no control technology
was required for emissions from oxychlorination vents
at ethylene dichloride (EDC)/VC plants. Applying
"back calculation" for these emissions is inappropri-
ate.
(2) In 1985, there were an estimated 8,000 fabri-
cation plants which processed resins produced by PVC
plants, thus resulting in VC emissions, which were
exempt from the VC NESHAP. They emit very small
quantities of VC and back calculation is not appropri-
ate.
(3) The 1993 study uses a baseline estimate of 18
residual cases from the NESHAP/ATERIS database.
There is no evidence that these cases resulted only
from emissions from PVC and EDC/VC plants.
(4) The risk analysis performed for the October
21, 1976 final VC regulation projected an incidence
reduction of 11 cases per year.
In contrast, the PES study, using the "back calcu-
lation" method derived the following annual incidence
reductions:
1980 - 250 cases
1985 - 360 cases
1990 - 430 cases
The subsequent back calculation conducted in the
ICF Re-analysis resulted in incidence reductions as
much as an order of magnitude higher than these.
Even considering the slightly different industrial
output assumptions imposed by macroeconomic mod-
eling, such a stark contrast is difficult to explain ex-
cept for a critically flawed approach. Growth in ac-
tivity and population nor other factors explain the dif-
ference in these two estimates. Given that the same
general methodology was used for all of the air toxic
pollutant assessments as was used for the VC
NESHAP evaluation, there is reason to believe that
cancer incidence results for the other air toxic pollut-
ants are also flawed.
Mobile Source HAP Exposure
Reductions
EPA's Cancer Risk report estimated that approxi-
mately 60 percent of the total carcinogenic risk posed
by HAPs was attributed to mobile sources, with sta-
tionary sources contributing 15 percent and area
sources contributing the remaining 25 percent.23 The
relative importance of mobile sources to total HAP
exposure was a significant motivation behind EPA's
subsequent effort to examine exposures and risks from
mobile source HAPs.24 Although available analytical
resources were severely limited, the Project Team
nevertheless decided it was necessary to perform at
least an initial screening analysis to estimate the dif-
ferences in mobile source HAP exposures between
the control and no-control scenarios configured for
the present study.
23 Cancer Risk report, Page ES-12.
24 See US EPA/OAR/OMS, "Motor Vehicle-Related Air Toxics Study," EPA 420-R-93-005. April 1993.
H-9
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Methodology
The approach used by ICF/SAI in conducting the
mobile source HAP analysis closely followed the ap-
proach used in the EPA Motor Vehicle-Related Air
Toxics Study (MVATS).25 Recognizing the dearth of
HAP ambient concentration and exposure data, both
studies use carbon monoxide (CO) concentrations as
the basis for estimating mobile source HAP concen-
trations and exposures. An important difference be-
tween the two studies, however, is that the ICF/SAI
study adjusted the estimated change in ambient CO
concentrations to take account of background26 and
non-mobile source27 CO emissions. The HAP expo-
sure function used in the ICF/SAI analysis is summa-
rized by the following equation:
]•=((('x A)-B)xSxMx
(I 'OCx HAP)
~(T)
(2)
where :
E
C
B
S
M
VOC
HAP
CO
exposure to motor vehicle-emitted
HAP
annual ambient CO concentration to
annual CO exposure concentration
conversion factor
county-level annual average ambient
CO concentration
background CO concentration
no-control to control scenario CO
concentration adjustment factor
(equals 1 for the control scenario)
total CO exposure to mobile source
CO exposure conversion factor
VOC emissions by year, county, and
scenario
VOC speciation factor by mobile
source HAP
CO emissions by year, county, and
scenario
Details of the derivation of each of the variables
applied in the above equation are provided in the ICF/
SAI report. However, in essence, the calculation in-
volves the following basic steps.
First, annual average county-level CO ambient
monitoring data are compiled from the EPA
Aerometric Information Retrieval System (AIRS)
database. After adjusting for background and non-
mobile source contributions, these annual average
ambient CO concentrations are converted to annual
average CO exposure concentrations. As in the EPA
MVATS, this conversion is made based on the Haz-
ardous Air Pollutant Exposure Model - Mobile Sources
(HAPEM-MS) population exposure model, which
takes account of time spent in five indoor and out-
door microenvironments: indoors at home, other in-
door, in-vehicle, outdoors near roadway, and other
outdoor.28 After adjusting for CO exposures attribut-
able to non-mobile sources of CO, the CO exposures
are converted to exposures for each of the mobile
source HAPs based on available VOC speciation data
and the ratio of co-located VOC and CO emissions.29
These calculations are repeated for the no-control sce-
nario after adjusting for differences in CO ambient
concentrations for each target year and for differences
in fuel composition.
Results
By 1990, CAA controls resulted in significant
reductions in exposure to motor vehicle HAPs. Fig-
ure H-4 summarizes the nationwide annual average
exposure levels, in micrograms per cubic meter, for
each of the five HAPs analyzed under the control and
no-control scenarios. Additional detailed results, in-
cluding breakdown by urban versus rural environ-
ments and comparisons with the EPA MVATS esti-
mates, are provided in the ICF/SAI report.
Analytical resources to carry forward these expo-
sure estimates to derive estimates of the changes in
motor vehicle HAP-related adverse effects attribut-
able to historical CAA programs were not available.
25 ICF/SAI, "Retrospective Analysis of Inhalation Exposure to Hazardous Air Pollutants from Motor Vehicles," October 1995, p. 4.
26 Background CO is produced by the oxidation of biogenic hydrocarbons. See ICF/SAI, p. 7.
27 The EPA MVATS attributed all measured CO to motor vehicles, resulting in an overestimation of motor-vehicle contributions to
CO concentration changes. See ICF/SAI, p. 8. The MVATS assumption would also lead to a subsequent overestimation of changes in
HAP exposures.
28 See ICF/SAI, p. 3.
29 The same HAP emission fractions used in the EPA MVATS were used herein, except for diesel PM which is not proportional to
VOC emissions. Instead, diesel PM emission factors were developed using year-specific PART5 diesel PM emission factors and VMT
estimates for diesel-powered vehicles.
H-10
-------
Appendix H: Air Toxics
Figure H-4. National Annual Average Motor Vehicle
HAP Exposures (|_ig/m3).
Benzene Acetaldehyde DieselPM
Formaldehyde 13-Butadiene
Non-Cancer Health Effects
Broad gaps exist in the current state of knowl-
edge about the quantifiable effects of air toxics expo-
sure. This is particularly true for a wide range of health
effects such as tumors, abnormal development, birth
defects, neurological impairment, or reproductive
impairment, etc. For example, the EPA's Non-Can-
cer Study30 found that ambient concentrations for a
substantial number of monitored and modeled HAPs
exceeded one or more health benchmarks.31 However
no accepted methodology exists to quantify the ef-
fects of such exceedences. More data on health ef-
fects is needed for a broad range of chemicals.
Ecological Effects
Through the 1970s and 1980s, the adverse effects
of toxic pollution on the Great Lakes became clear
and undeniable. Over the same time period, scientists
began collecting a convincing body of evidence that
toxic chemicals released to the air can travel long dis-
tances and be deposited on land or water far from the
original sources. An example of this evidence is the
presence of such contaminants as PCBs, toxaphene,
and other pesticides in fish in Lake Siskiwit, a lake on
an island on upper Lake Superior, which has no water-
borne sources of pollution. Toxaphene, a pesti-
cide used primarily in the southeastern U.S. cot-
ton belt, has been found as far away as the Arc-
tic, with a decreasing air concentration gradient
from the southeast toward the Great Lakes and
the north Atlantic regions.
Similarly, a growing body of evidence
~Control showed that pollutants that were persistent (do
iNo-Controi not easily break down) and bioaccumulating (not
significantly eliminated from the body) were
magnifying up the food chain, such that top
predator fish contained levels up to millions of
times greater than the harmless levels in the
water. As such, those who ate those large fish,
such as humans, eagles, mink, and beluga whales
could receive very high exposures to the pollut-
ants. Wildlife were beginning to show adverse
effects in the wild, that could be duplicated in the lab.
In the Great Lakes, such chemicals as PCBs, mercury,
dieldrin, hexachlorobenzene, Lindane, lead com-
pounds, cadmium compounds, DDT/DDE, and oth-
ers are of significant concern. In other places in the
country, similar effects are being experienced, espe-
cially with mercury, which is transported primarily
by air, but exposure to which is primarily through con-
taminated fish. It was this kind of information about
DDT and toxaphene that led to their being banned in
the U.S. under FIFRA.
While ecological and economical sciences are not
yet sufficiently advanced to support the kind of com-
prehensive, quantitative evaluation of benefits needed
forthe present study, selected local and regional scale
adverse ecological effects of HAPs, and their adverse
consequences for human health and welfare, can and
have been surveyed. In May 1994, the EPA issued its
first "Report to Congress on Deposition of Air Pollut-
ants to the Great Waters."32 The Great Waters Report
examined the pollutants contributing to adverse eco-
logical effects, the potential significance of the con-
tribution to pollutant loadings from deposition of air-
borne pollutants, and the potential adverse effects as-
sociated with these pollutant loadings. Key HAPs iden-
tified in the Great Waters Report include PCBs, mer-
cury, dioxins, and other heavy metals and toxic or-
ganics.
30 U.S. Environmental Protection Agency, "Toxic Air Pollutants and Noncancer Risks: Screening Studies," External Review
Draft, September, 1990.
31 Relevant benchmarks include Acceptable Daily Intake (ADI), the estimate of daily exposure at which adverse health effects are
unlikely; and Lowest Observed Actual Effect Level (LOAEL), which is the lowest exposure level at which significant adverse health
effects are observed.
1994.
! USEPA/OAR/OAQPS, "Deposition of Air Pollutants to the Great Waters, First Report to Congress," EPA-453/R-93-055, May
u-Ti
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Of particular relevance to the present assessment,
the Great Waters Report demonstrated the significance
of transport and transformation of HAPs through food
webs, leading to increased toxicity and
biomagnification. A prime example of adverse trans-
port and transformation is mercury. Transformation
from inorganic to methylated forms significantly in-
creases the toxic effects of mercury in ecosystems. A
prime example of biomagnification is PCBs. As noted
in the Great Waters Report:
"Pollutants of concern [such as PCBs]
accumulate in body tissues and magnify up
the food web, with each level accumulating
the toxics from its diet and passing the burden
along to the animal in the next level of the
food web. Top consumers in the food web,
usually consumers of large fish, may
accumulate chemical concentrations many
millions of times greater than the
concentrations present in the water...High risk
groups...include breast-feeding mothers
because breast-fed babies continue to
accumulate [pollutants] from their mothers
after birth. For example, they can have PCB
levels four times higher than their mothers
after six to nine months of breast feeding. "33
Because of the risk of significant exposure to in-
fants and other high-risk groups, such as "sport an-
glers, Native Americans, and the urban poor,"34 a sub-
stantial number of fish consumption advisories have
been issued in recent years. Current fish advisories
for the Great Lakes alone include widespread adviso-
ries for PCB's, chlordane, mercury and others, cau-
tioning that nursing mothers, pregnant women, women
who anticipate bearing children, female children of
any age and male children age 15 and under not eat
certain high-food chain fish species. It should be noted
as well that 40 states have issued mercury advisories
in some freshwater bodies, and nine states have is-
sued mercury advisories for every freshwater
waterbody in the state (these states are Maine, New
Hampshire, Vermont, Massachusetts, New York, New
Jersey, Missouri, Michigan, and Florida).
There is little evidence indicating that the CAA
had much beneficial effect on air toxic deposition to
water bodies. Since the early NESHAPs were based
on direct inhalation, primarily cancer effects close to
33 EPA-453/R-93-055, May 1994, p. ix.
34 EPA-453/R-93-055, May 1994, p. x.
a plant, they did not address the issue of cumulative
effects of persistent pollutants far from the source. It
was for this reason that section 112(m) was included
in the 1990 CAA Amendments, with requirements to
study and document the atmospheric contribution of
water pollutants, the adverse human health and envi-
ronmental effects resulting and the sources that should
be controlled to prevent adverse effects, and addition-
ally, to promulgate regulations to prevent adverse ef-
fects.
Conclusions — Research Needs
As has been demonstrated, there are broad gaps
in the current state of knowledge about the quantifi-
able effects of air toxics exposure for a wide range of
both human health and environmental effects. The
following discussion outlines areas in which further
research is needed in order to adequately quantify the
benefits of air toxics control.
Health Effects
Develop health effects data on pollutants for
which limited or no data currently exists. Such
studies should be focused on pollutants with
a relatively high probability of exposure and/
or potential adverse health effects.
Understand mechanism of action of pollut-
ants, for example through pharmacokinetic
modeling. This will allow for a more accu-
rate assessment of the effects of these pollut-
ants on humans.
Conduct research on factors that affect varia-
tions in susceptibility of human populations
and determine the distribution of these fac-
tors in the U.S.
Conduct research to better understand inter-
active effects of multiple pollutant exposures.
Develop methodologies to derive alternative
estimates of human cancer risk from existing
upper-bound methods.
Acquire data and develop dose-response re-
lationships for critical noncancer effects such
as developmental, neurotoxic, mutagenic, res-
H-12
-------
Appendix H: Air Toxics
piratory and other effects. In particular, de-
sign methodology to quantify effects of ex-
posures above health benchmarks.
Acquire data and develop methods to estimate
effects from acute exposure.
Exposure Assessment
Expand data collection efforts: pre- and post-
control emissions; HAP speciation; facilities
location; facility parameters (stack heights,
distances from stacks to fencelines, etc.).
Develop more comprehensive exposure mod-
els which incorporate activity patterns, indi-
rect exposures, total body burden, ratios of
time spent indoors to outdoors.
Continue to refine uncertainty analysis meth-
ods.
Ecosystem Effects
Reliable estimates/measures of the levels of
persistent bioaccumulating toxics in different
media (air, water column, soils and sediments)
Work to correlate levels of persistent
bioaccumulating toxics with exposures, biota
concentrations/accumulation, and adverse
effects, especially subtle effects such as wast-
ing, behavioral effects, and developmental
effects.
Criteria for effects, such as a wildlife corre-
late to a RfD or dose-response curve. This
work should be done to complement the mass
balance efforts now being completed, which
will model source emissions to water column
concentrations, then design research to pre-
dict effects on living resources given those
predicted levels.
Work to determine the effects of mixtures of
persistent bioaccumulating toxic pollutants,
and to determine cause-effect relationships of
exposures over long periods of time.
Studies to evaluate toxic effects in less well
understood terrestrial systems such as: soil
organisms/invertebrates, food web effects,
amphibian effects, effects on endangered spe-
cies and phytotoxic effects.
Work to improve understanding of effects of
toxic air pollutants on wetland species and
wetland functions.
Economic Valuation
Develop valuation estimates for endpoints for
which inadequate estimates currently exist.
These valuation estimates must be consistent
with the kinds of damages expected.
Initiate broad-scope economic valuation of air
toxics program using survey techniques.
H-13
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Air Toxics References
Hunt, W.F., R.B. Faoro, and T.C. Curran, "Estima-
tion of Cancer Incidence Cases and Rates for
Selected Toxic Air pollutants Using Ambi-
ent Air Pollution Data, 1970 vs. 1980," U.S.
Environmental Protection Agency, April
1985.
ICF Kaiser, "Direct Inhalation Incidence Benefits,"
Draft Report, November 11, 1994.
ICF Kaiser and Systems Applications International,
"Retrospective Analysis of Inhalation Expo-
sure to Hazardous Air Pollutants From Mo-
tor Vehicles," October 1995.
Koppikar, Aparna and Robert Fegley. 1995. "Analy-
sis of 'Reasonableness' of Cancer Risk As-
sessment Findings for Asbestos and Vinyl
Chloride in section 812 Retrospective Cost-
Benefit Analysis," Memorandum to Jim
DeMocker, Office of Policy Analysis and
Review, Office of Air and Radiation, U.S.
Environmental Protection Agency. Novem-
ber 2, 1995.
Pacific Environmental Services, "Cancer Risk Esti-
mates From Stationary Services," Memoran-
dum to Vasu Kilaru, U.S. EPA, March 5,
1993.
Pacific Environmental Services, "Draft Summary of
Methodology Used For Cancer From Station-
ary Services," Memorandum to Vasu Kilaru,
U.S. EPA, March 22, 1993.
Pacific Environmental Services, "Retrospective
Analysis for Section 812 (a) Benefits Study,"
September 30, 1992.
U.S. Environmental Protection Agency, The Air Toxics
Problem in the United States: An Analysis of
Cancer Risks for Selected Pollutants, Office
of Air Quality Planning and Standards, EPA-
450/1-85-001, May 1985.
U.S. Environmental Protection Agency, Cancer Risk
From Outdoor Exposure to Air Toxics, Of-
fice of Air Quality Planning and Standards,
EPA-450/l-90-004a, September 1990.
U.S. Environmental Protection Agency, Deposition
of Air Pollutants to the Great Waters, First
Report to Congress, Office of Air Quality
Planning and Standards, EPA-453/R-93-055,
May 1994.
U.S. Environmental Protection Agency, Motor Ve-
hicle-Related Air Toxics Study, Office of
Mobile Sources, EPA-420/12-93-005, April
1993.
U.S. Environmental protection Agency, "Toxic Air
Pollutants and Noncancer Risks: Screening
Studies," External Review Draft, September
1990.
U.S. Environmental protection Agency, Unfinished
Business: A Comparative Assessment of En-
vironmental Problems, Office of Policy, Plan-
ning, and Evaluation, February 1987.
H-14
-------
Appendix I: Valuation of Human Health and
Welfare Effects of Criteria Pollutants
This appendix describes the derivations of the
economic valuations for health and welfare endpoints
considered in the benefits analysis. Valuation esti-
mates were obtained from the literature and reported
in dollars per case avoided for health effects, and dol-
lars per unit of avoided damage for welfare effects.
This appendix first introduces the method for mon-
etizing improvements in health and welfare, followed
by a summary of dollar estimates used to value ben-
efits and detailed descriptions of the derivation of each
estimate. These economic valuations are given both
in terms of a central (point) estimate as well as a prob-
ability distribution which characterizes the uncertainty
about the central estimate. All dollar values are
rounded and are in 1990 dollars. Next, results of the
economic benefits analysis are presented. Finally, un-
certainties in valuing the benefits attributable to the
CAA are explored.
Methods Used to Value Health
and Welfare Effects
Willingness to pay (WTP) and willingness to ac-
cept (WTA) are the two measures commonly used to
quantify the value an individual places on something,
whether it is something that can be purchased in a
market or not. Both WTP and WTA are measures of
the amount of money such that the individual would
be indifferent between having the good (or service)
and having the money. Whether WTP or WTA is the
appropriate measure depends largely on whether an
increase or a decrease of the good is at issue. WTP is
the amount of money an individual would be willing
to pay to have a good (or a specific increase in the
amount of the good) — i.e., the amount such that the
individual would be indifferent between having the
money and having the good (or having the specific
increase in the good). WTA is the amount of money
the individual would have to be compensated in order
to be indifferent to the loss of the good (or a specific
decrease in the amount of the good). WTP is the ap-
propriate measure if the baseline case is that the indi-
vidual does not have the good or when an increase in
the amount of the good is at issue; WTA is the appro-
priate measure if the baseline case is that the indi-
vidual has the good or when a decrease in the amount
of the good is at issue. An important difference be-
tween WTP and WTA is that, in theory, WTP is lim-
ited by the individual's budget, whereas WTA is not.
Nevertheless, while the underlying economic valua-
tion literature is based on studies which elicited ex-
pressions of WTP and/or WTA, the remainder of this
report refers to all valuation coefficients as WTP esti-
mates. In some cases (e.g., stroke-related hospital
admissions), neither WTA nor WTP estimates are
available and WTP is approximated by cost of illness
(COI) estimates, a clear underestimate of the true
welfare change since important value components
(e.g., pain and suffering associated with the stroke)
are not reflected in the out-of-pocket costs for the
hospital stay.
For both market and non-market goods, WTP re-
flects individuals' preferences. Because preferences
are likely to vary from one individual to another, WTP
for both market and non-market goods (e.g., health-
related improvements in environmental quality) is
likely to vary from one individual to another. In con-
trast to market goods, however, non-market goods
such as environmental quality improvements are pub-
lic goods, whose benefits are shared by many indi-
viduals. The individuals who benefit from the envi-
ronmental quality improvement may have different
WTPs for this non-market good. The total social value
of the good is the sum of the WTPs of all individuals
who "consume" (i.e., benefit from) the good.
In the case of health improvements related to pol-
lution reduction, it is not certain specifically who will
receive particular benefits of reduced pollution. For
example, the analysis may predict 100 days of cough
avoided in a given year resulting from CAA reduc-
tions of pollutant concentrations, but the analysis does
not estimate which individuals will be spared those
days of coughing. The health benefits conferred on
individuals by a reduction in pollution concentrations
are, then, actually reductions in the probabilities of
having to endure certain health problems. These ben-
efits (reductions in probabilities) may not be the same
for all individuals (and could be zero for some indi-
viduals). Likewise, the WTP for a given benefit is
likely to vary from one individual to another. In theory,
the total social value associated with the decrease in
risk of a given health problem resulting from a given
1-1
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
reduction in pollution concentrations is
N
(1)
/=l
where Bi is the benefit (i.e., the reduction in prob-
ability of having to endure the health problem) con-
ferred on the ith individual (out of a total of N) by the
reduction in pollution concentrations, and WTP^ET)
is the ith individual's WTP for that benefit. If a re-
duction in pollution concentrations affects the risks
of several health endpoints, the total health-related
social value of the reduction in pollution concentra-
tions is
N J
TP.(B.J (2)
2=1 ;=l
where B;j is the benefit related to the jth health
endpoint (i.e., the reduction in probability of having
to endure the jth health problem) conferred on the ith
individual by the reduction in pollution concentrations,
and WTP^Rj) is the ith individual's WTP for that
benefit.
The reduction in probability of each health prob-
lem for each individual is not known, nor is each
individual's WTP for each possible benefit he or she
might receive known. Therefore, in practice, benefits
analysis estimates the value of a.vto//.s7/6'a/health prob-
lem avoided. For example, although a reduction in
pollutant concentrations may save actual lives (i.e.,
avoid premature mortality), whose lives will be saved
cannot be known ex ante. What is known is that the
reduction in air pollutant concentrations results in a
reduction in mortality risk. It is this reduction in mor-
tality risk that is valued in a monetized benefit analy-
sis. Individual WTPs for small reductions in mortal-
ity risk are summed over enough individuals to infer
the value of a statistical life saved. This is different
from the value of a particular, identified life saved.
Rather than "WTP to avoid a death," then, it is more
accurate to use the term "WTP to avoid a statistical
death," or, equivalently, "the value of a statistical life."
Suppose, for example, that a given reduction in
PM concentrations results in a decrease in mortality
risk of 1/10,000. Then for every 10,000 individuals,
one individual would be expected to die in the ab-
sence of the reduction in PM concentrations (who
would not die in the presence of the reduction in PM
concentrations). IfWTP forthis 1/10,000 decrease in
mortality risk is $500 (assuming, for now, that all in-
dividuals' WTPs are the same), then the value of a
statistical life is 10,000 x $500, or $5 million.
A given reduction in PM concentrations is un-
likely, however, to confer the same risk reduction (e.g.,
mortality risk reduction) on all exposed individuals
in the population. (In terms of the expressions above,
Bi is not necessarily equal to B , for i j). In addition,
different individuals may not be willing to pay the
same amount for the same risk reduction. The above
expression for the total social value associated with
the decrease in risk of a given health problem result-
ing from a given reduction in pollution concentrations
may be rewritten to more accurately convey this. Us-
ing mortality risk as an example, for a given unit risk
reduction (e.g., 1/1,000,000), the total mortality-re-
lated benefit of a given pollution reduction can be
written as
N
number of units of risk reduction).
2=i x (WTPper unit risk reduction). (3)
where (number of units of risk reduction^ is the
number of units of risk reduction conferred on the ith
exposed individual as a result of the pollution reduc-
tion, (WTP per unit risk reduction^ is the ith
individual's willingness to pay for a unit risk reduc-
tion, and N is the number of exposed individuals.
If different subgroups of the population have sub-
stantially different WTPs for a unit risk reduction and
substantially different numbers of units of risk reduc-
tion conferred on them, then estimating the total so-
cial benefit by multiplying the population mean WTP
to save a statistical life times the predicted number of
statistical lives saved could yield a biased result. Sup-
pose, for example, that older individuals' WTP per
unit risk reduction is less than that of younger indi-
viduals (e.g., because they have fewer years of ex-
pected life to lose). Then the total benefit will be less
than it would be if everyone's WTP were the same. In
addition, if each older individual has a larger number
of units of risk reduction conferred on him (because a
given pollution reduction results in a greater absolute
reduction in risk for older individuals than for younger
individuals), this, in combination with smaller WTPs
of older individuals, would further reduce the total
benefit.
While the estimation of WTP for a market good
(i.e., the estimation of a demand schedule) is not a
simple matter, the estimation of WTP for a non-mar-
ket good, such as a decrease in the risk of having a
particular health problem, is substantially more diffi-
cult. Estimation of WTP for decreases in very spe-
cific health risks (e.g., WTP to decrease the risk of a
day of coughing or WTP to decrease the risk of ad-
mission to the hospital for respiratory illness) is fur-
ther limited by a paucity of information. Derivation
of the dollar value estimates discussed below was of-
ten limited by available information.
1-2
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Valuation of Specific Heaith Endpoints
Valuation of Premature Mortality Avoided
As noted above, it is actually reductions in mor-
tality risk that are valued in a monetized benefit
analysis. Individual WTPs for small reductions in
mortality risk are summed over enough individuals
to infer the value of a statistical life saved. This is
different from the value of a particular, identified
life saved. The "value of a premature death
avoided," then, should be understood as shorthand
for "the value of a statistical premature death
avoided."
The value of a premature death avoided is based
on an analysis of 26 policy-relevant value-of-life
studies (see Table 1-1). Five of the 26 studies are
contingent valuation (CV) studies, which directly
solicit WTP information from subjects; the rest are
wage-risk studies, which base WTP estimates on
estimates of the additional compensation demanded
in the labor market for riskier jobs. Each of the 26
studies provided an estimate of the mean WTP to
avoid a statistical premature death. Several plau-
sible standard distributions were fit to the 26 esti-
mates of mean WTP. A Weibull distribution, with
a mean of $4.8 million and standard deviation of
$3.24 million, provided the best fit to the 26 esti-
mates. The central tendency estimate of the WTP
to avoid a statistical premature death is the mean of
this distribution, $4.8 million. The considerable un-
certainty associated with this approach is discussed
in detail below, in the subsection titled "The Eco-
nomic Benefits Associated with Mortality," within
the section titled "Uncertainties."
Life-years lost is a possible alternative measure
of the mortality-related effect of pollution, as dis-
cussed in Appendix D. If life-years lost is the mea-
sure used, then the value of a statistical life-year lost,
rather than the value of a statistical life lost would be
needed. Moore and Viscusi (1988) suggest one ap-
proach for determining the value of a statistical life-
year lost. They assume that the willingness to pay to
save a statistical life is the value of a single year of
life times the expected number of years of life remain-
ing for an individual. They suggest that a typical re-
spondent in a mortal risk study may have a life ex-
pectancy of an additional 35 years. Using a mean es-
timate of $4.8 million to save a statistical life, their
approach would yield an estimate of $ 137,000 per life-
year lost or saved. If an individual discounts future
additional years using a standard discounting proce-
dure, the value of each life-year lost must be greater
than the value assuming no discounting. Using a 35
year life expectancy, a $4.8 million value of a statisti-
cal life, and a 5 percent discount rate, the implied value
Tabic 1-1. Summary of Mortality Valuation Estimates
(millions of 1990 dollars).
Sttausllj
TTyperf
Ksliin ate
Valuation
(in illions
!Ȥ$))
Kneisner and Leeth (1991) (US)
Labor Market
0.6
Smith and Gilbert (1984)
Labor Market
0.7
Dillingham (1985)
Labor Market
0.9
Butler (1983)
Labor Market
1.1
Miller and Guria (1991)
Cont. Value
1.2
Moore and Viscusi (1988a)
Labor Market
2.5
Viscusi, Magat, and Huber (1991b)
Cont. Value
2.7
Gegax et al. (1985)
Cont. Value
3.3
Marin and Psacharopoulos (1982)
Labor Market
2.8
Kneisner and Leeth (1991)
(Australia)
Labor Market
3.3
Gerking, de Haan, and Schulze
(1988)
Cont. Value
3.4
Cousineau, Lacroix, and Girard
(1988)
Labor Market
3.6
Jones-Lee (1989)
Cont. Value
3.8
Dillingham (1985)
Labor Market
3.9
Viscusi (1978, 1979)
Labor Market
4.1
R.S. Smith (1976)
Labor Market
4.6
V.K. Smith (1976)
Labor Market
4.7
Olson (1981)
Labor Market
5.2
Viscusi (1981)
Labor Market
6.5
R.S. Smith (1974)
Labor Market
7.2
Moore and Viscusi (1988a)
Labor Market
7.3
Kneisner and Leeth (1991) (Japan)
Labor Market
7.6
Herzog and Schlottman (1987)
Labor Market
9.1
Leigh and Folson (1984)
Labor Market
9.7
Leigh (1987)
Labor Market
10.4
Gaten (1988)
Labor Market
13.5
SOURCE: Viscusi, 1992
of each life-year lost is $293,000. The Moore and
Viscusi procedure is identical to this approach, but
uses a zero discount rate.
Using the value of a life-year lost and the expected
number of years remaining (obtained from life ex-
pectancy tables), and assuming a given discount rate,
values of age-specific premature mortality can be de-
rived. Examples of valuations of pollution-related
mortality using the life-years lost approach are given
below, in the subsection titled "The Economic Ben-
efits Associated with Mortality," within the section
titled "Uncertainties."
Valuation of Hospital Admissions Avoided
In the case of hospital admissions, cost-of-illness
(COI), or "costs avoided," estimates were used in lieu
of WTP because of the lack of other information re-
1-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
garding willingness to pay to avoid illnesses that ne-
cessitate hospital admissions. For those hospital ad-
missions which were specified to be the initial hospi-
tal admission (in particular, hospital admissions for
coronary heart disease (CHD) events and stroke), COI
estimates include, where possible, all costs of the ill-
ness, including the present discounted value of the
stream of medical expenditures related to the illness,
as well as the present discounted value of the stream
of lost earnings related to the illness. (While an esti-
mate of present discounted value of both medical ex-
penditures and lost earnings was available for stroke,
the best available estimate for CHD did not include
lost earnings. The derivations of the COI estimates
for CHD and stroke, both lead-induced effects, are
discussed in Appendix G.)
In those cases for which it is unspecified whether
the hospital admission is the initial one or not (that is,
for all hospital admissions endpoints other than CHD
and stroke), it is unclear what portion of medical ex-
penditures and lost earnings after hospital discharge
can reasonably be attributed to pollution exposure and
what portion might have resulted from an individual's
pre-existing condition even in the absence of a par-
ticular pollution-related hospital admission. In such
cases, the COI estimates include only those costs as-
sociated with the hospital stay, including the hospital
charge, the associated physician charge, and the lost
earnings while in the hospital. The derivations of these
costs are discussed in Abt Associates Inc., 1996.
These COI estimates are likely to substantially
understate total WTP to avoid an illness that began
with a hospital admission or to avoid a particular hos-
pital admission itself. First, most of the COI estimates
fall short of being full COI estimates either because
of insufficient information or because of ambiguities
concerning what portion of post-hospital costs should
be attributed to pollution exposure. Even full COI es-
timates will understate total WTP, however, because
they do not include the value of avoiding the pain and
suffering associated with the illness for which the in-
dividual entered the hospital.
Valuation of Chronic Bronchitis Avoided
Although the severity of cases of chronic bron-
chitis valued in some studies approaches that of
chronic obstructive pulmonary disease, to maintain
consistency with the existing literature we do not treat
those cases separately for the purposes of this analy-
sis. Chronic bronchitis is one of the only morbidity
endpoints that may be expected to last from the initial
onset of the illness throughout the rest of the
individual's life. WTP to avoid chronic bronchitis
would therefore be expected to incorporate the present
discounted value of a potentially long stream of costs
(e.g., medical expenditures and lost earnings) associ-
ated with the illness. Two studies, Viscusietal. (1991)
and Krupnick and Cropper (1992) provide estimates
of WTP to avoid a case of chronic bronchitis. The
study by Viscusi et al., however, uses a sample that is
larger and more representative of the general popula-
tion than the study by Krupnick and Cropper (which
selects people who have a relative with the disease).
The valuation of chronic bronchitis in this analysis is
therefore based on the distribution of WTP responses
from Viscusi et al. (1991).
Both Viscusi et al. (1991) and Krupnick and Crop-
per (1992), however, defined a case of severe chronic
bronchitis. It is unclear what proportion of the cases
of chronic bronchitis predicted to be associated with
exposure to pollution would turn out to be severe cases.
The incidence of pollution-related chronic bronchitis
was based on Abbey et al. (1993), which considered
only new cases of the illness.1 While a new case may
not start out being severe, chronic bronchitis is a
chronic illness which may progress in severity from
onset throughout the rest of the individual's life. It is
the chronic illness which is being valued, rather than
the illness at onset.
The WTP to avoid a case of pollution-related
chronic bronchitis (CB) is derived by starting with
the WTP to avoid a severe case of chronic bronchitis,
as described by Viscusi et al. (1991), and adjusting it
downward to reflect (1) the decrease in severity of a
case of pollution-related CB relative to the severe case
described in the Viscusi study, and (2) the elasticity
of WTP with respect to severity. Because elasticity is
a marginal concept and because it is a function of se-
verity (as estimated from Krupnick and Cropper,
1992), WTP adjustments were made incrementally,
in one percent steps. At each step, given a WTP to
avoid a case of CB of severity level sev, the WTP to
avoid a case of severity level 0.99*sev was derived.
This procedure was iterated until the desired severity
level was reached and the corresponding WTP was
derived. Because the elasticity of WTP with respect
to severity is a function of severity, this elasticity
changes at each iteration. If, for example, it is believed
that a pollution-related case of CB is of average se-
1 It is important that only new chronic bronchitis be considered in this analysis because WTP estimates reflect lifetime expendi-
tures and/or losses associated with this chronic illness, and incidences are predicted separately for each year during the period 1970-
1990. If the total prevalence of chronic bronchitis, rather than the incidence of only new chronic bronchitis were predicted each year,
valuation estimates reflecting lifetime expenditures could be repeatedly applied to the same individual for many years, resulting in a
severe overestimation of the value of avoiding pollution-related chronic bronchitis.
1-4
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
verity, that is, a 50 percent reduction in severity from
the case described in the Viscusi study, then the itera-
tive procedure would proceed until the severity level
was half of what it started out to be.
The derivation of the WTP to avoid a case of pol-
lution-related chronic bronchitis is based on three com-
ponents, each of which is uncertain: (1) the WTP to
avoid a case of severe CB, as described in the Viscusi
study, (2) the severity level of an average pollution-
related case of CB (relative to that of the case de-
scribed by Viscusi), and (3) the elasticity of WTP with
respect to severity of the illness. Because of these three
sources of uncertainty, the WTP is uncertain. Based
on assumptions about the distributions of each of the
three uncertain components, a distribution of WTP to
avoid a pollution-related case of CB was derived by
Monte Carlo methods. The mean of this distribution,
which was about $260,000, is taken as the central ten-
dency estimate of WTP to avoid a pollution-related
case of CB. Each of the three underlying distributions
is described briefly below.
The distribution of WTP to avoid a severe case of
CB was based on the distribution of WTP responses
in the Viscusi study. Viscusi et al. derived respon-
dents' implicit WTP to avoid a statistical case of
chronic bronchitis from their WTP for a specified re-
duction in risk. The mean response implied a WTP of
about $1,000,000 (1990 dollars)2; the median response
implied a WTP of about $530,000 (1990 dollars).
However, the extreme tails of distributions of WTP
responses are usually considered unreliable. Because
the mean is much more sensitive to extreme values,
the median of WTP responses is often used rather than
the mean. Viscusi et al. report not only the mean and
median of their distribution of WTP responses, how-
ever, but the decile points as well. The distribution of
reliable WTP responses from the Viscusi study could
therefore be approximated by a discrete uniform dis-
tribution giving a probability of one ninth to each of
the first nine decile points. This omits the first five
and the last five percent of the responses (the extreme
tails, considered unreliable). This trimmed distribu-
tion of WTP responses from the Viscusi study was
assumed to be the distribution of WTPs to avoid a
severe case of CB. The mean of this distribution is
about $720,000 (1990 dollars).
The distribution of the severity level of an aver-
age case of pollution-related CB was modeled as a
triangular distribution centered at 6.5, with endpoints
at 1.0 and 12.0. These severity levels are based on the
severity levels used in Krupnick and Cropper, 1992,
which estimated with relationship between ln(WTP)
and severity level, from which the elasticity is derived.
The most severe case of CB in that study is assigned a
severity level of 13. The mean of the triangular distri-
bution is 6.5. This represents a 50 percent reduction
in severity from a severe case.
The elasticity of WTP to avoid a case of CB with
respect to the severity of that case of CB is a constant
times the severity level. This constant was estimated
by Krupnick and Cropper, 1992, in the regression of
ln(WTP) on severity, discussed above. This estimated
constant (regression coefficient) is normally distrib-
uted with mean = 0.18 and standard deviation = 0.0669
(obtained from Krupnick and Cropper, 1992).
The distribution of WTP to avoid a case of pollu-
tion-related CB was generated by Monte Carlo meth-
ods, drawing from the three distributions described
above. On each of 16,000 iterations (1) a value was
selected from each distribution, and (2) a value for
WTP was generated by the iterative procedure de-
scribed above, in which the severity level was de-
creased by one percent on each iteration, and the cor-
responding WTP was derived. The mean of the re-
sulting distribution of WTP to avoid a case of pollu-
tion-related CB was $260,000.
This WTP estimate is reasonably consistent with
full COI estimates derived for chronic bronchitis, us-
ing average annual lost earnings and average annual
medical expenditures reported by Cropper and
Krupnick, 1990. Using a 5 percent discount rate and
assuming that (1) lost earnings continue until age 65,
(2) medical expenditures are incurred until death, and
(3) life expectancy is unchanged by chronic bronchi-
tis, the present discounted value of the stream of medi-
cal expenditures and lost earnings associated with an
average case of chronic bronchitis is estimated to be
about $77,000 for a 30 year old, about $58,000 for a
40 year old, about $60,000 for a 50 year old, and about
$41,000 for a 60 year old. A WTP estimate would be
expected to be greater than a full COI estimate, re-
flecting the willingness to pay to avoid the pain and
suffering associated with the illness. The WTP esti-
mate of $260,000 is from 3.4 times the full COI esti-
mate (for 30 year olds) to 6.3 times the full COI esti-
mate (for 60 year olds).
2 There is an indication in the Viscusi paper that the dollar values in the paper are in 1987 dollars. Under this assumption, the
dollar values were converted to 1990 dollars.
1-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Valuation of Other Morbidity Endpoints Avoided
WTP to avoid a day of specific morbidity end-
points, such as coughing or shortness of breath, has
been estimated by only a small number of studies (two
or three studies, for some endpoints; only one study
for other endpoints). The estimates for health end-
points involving these morbidity endpoints are there-
fore similarly based on only a few studies. However,
it is worth noting that the total benefit associated with
any reduction in pollutant concentrations is determined
largely by the benefit associated with the correspond-
ing reduction in mortality risk because the dollar value
associated with mortality is significantly greater than
any other valuation estimate. More detailed explana-
tions for valuation of specific morbidity endpoints is
given in Table 1-2.
Estimates of WTP may be understated for a couple
of reasons. First, if exposure to pollution has any cu-
mulative or lagged effects, then a given reduction in
pollution concentrations in one year may confer ben-
efits not only in that year but in future years as well.
Benefits achieved in later years are not included. Sec-
ond, the possible effects of altruism are not consid-
ered in any of the economic value derivations. Indi-
viduals' WTP for reductions in health risks for others
are implicitly assumed to be zero.
Table 1-2 summarizes the derivations of the eco-
nomic values used in the analysis. More detailed de-
scriptions of the derivations of lead-related endpoints
(hospital admissions for CHD and stroke, Lost IQ
points, IQ below 70, and hypertension) are discussed
in Appendix G.
Valuation of Welfare Effects
With the exception of agricultural benefits, eco-
nomic valuations for welfare effects quantified in the
analysis (i.e., household soiling damage, visibility and
worker productivity) are documented in Table 1-2. For
agricultural benefits, estimated changes in crop yields
were evaluated with an agricultural sector model,
AGSIM. This model incorporates agricultural price,
farm policy, and other data for each year. Based on
expected yields, the model estimates the production
levels for each crop, the economic benefits to con-
sumers, and the economic benefits to producers asso-
ciated with these production levels. To the extent that
alternative exposure-response relationships were
available, a range of potential benefits was calculated.
Appendix F documents the derivation of the monetary
benefits associated with improved agricultural pro-
duction. The derivation of the residential visibility
valuation estimate is discussed further below.
Visibility Valuation
Residential visibility has historically been valued
through the use of contingent valuation studies, which
employ surveys and questionnaires to determine the
economic value respondents place on specified
changes in visibility. A number of such studies have
been published in the economics literature since the
late 1970s, and have reported a wide range of result-
ing values for visibility, expressed as household will-
ingness to pay (WTP) for a hypothesized improve-
ment in visibility. Those studies were carefully re-
viewed for their applicability to the retrospective
analysis.
One limitation of many existing contingent valu-
ation studies of visibility is that they are local or re-
gional in scope, soliciting values for visibility from
residents of only one or two cities in a single region
of the country. Studies of visibility values from west-
ern cities, the most recent of which was published in
1981, have reported somewhat lower values than those
from eastern cities, raising the question of whether
eastern and western visibility are different commodi-
ties and should be valued differently in this analysis.
While the different visibility values reported in
the literature may appear to imply that visibility is
not valued equally by survey respondents in the east-
ern and western U.S., other evidence suggests that
eastern and western visibility are not fundamentally
different commodities, and that the retrospective ben-
efits calculations should not be based on separate east-
ern and western visibility values. For example,
NAPAP data indicate that California's South Coast
Air Basin, which encompasses Los Angeles and ex-
tends northward to the vicinity of San Francisco, has
median baseline visibility more characteristic of the
eastern U.S. than of other areas of the west (NAPAP
1991; IEc 1992, 1993a), reflecting the influence of
the higher humidity typical of coastal areas. While
inland areas of the west will tend to have lower hu-
midity, and hence greater baseline visibility than ei-
ther the eastern region or the western coastal zone,
such baseline visibility differences are accounted for
in the conversion from the visual range metric to
DeciView.
1-6
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Perhaps the most compelling rationale for employ-
ing a single nationwide visibility valuation strategy
in the retrospective benefits analysis, however, relates
to the air quality modeling output used to calculate
the control and no-control scenario visibility profiles,
and its implications for the valuation of visibility as a
commodity. The RADM model and linear scaling
technique used for the retrospective analysis model
visibility improvements nationwide as changes in re-
gional atmospheric haze. In other words, although the
magnitude of visibility effects may vary between re-
gions, the model output does not distinguish between
a change in eastern visibility and a change in western
visibility as distinct phenomena. Thus, there is no clear
reason to value those same visibility changes differ-
ently in calculating the benefits of visibility improve-
ments. Consequently, a single, consistent valuation
basis has been applied to residential visibility improve-
ments nationwide for this analysis.
In light of advances in the state of the art of con-
tingent valuation over the last decade, the age of many
of the existing studies raised questions regarding their
suitability to serve as the primary basis for the vis-
ibility benefits estimates. A review of the survey and
data analysis methods used in the available studies
indicated that a study conducted for EPA by
McClelland etal. (1991) addressed many of the meth-
odological flaws of earlier studies, employing survey
methods and analytical techniques designed to mini-
mize potential biases (IEc 1992). Although this study
is unpublished, given its methodological improve-
ments over earlier studies it was chosen as the basis
for the central tendency of the visibility benefits esti-
mate, yielding an estimated value of $ 14 per unit im-
provement in DeciView as the annual household WTP
for visibility improvements (IEc 1997), as specified
in Table 1-2.
The difficulty of accurately defining the expected
statistical distribution of WTP values for visibility
improvements on the basis of published studies of
uneven reliability, along with the considerable varia-
tion in reported visibility values, led to the selection
of a hypothesized triangular distribution of values to
characterize the uncertainty in the visibility benefits
estimate. Reliance on any single study to estimate the
uncertainty range would be unlikely to adequately
characterize variations in visibility values that might
exist across cities, and in any case would fail to cap-
ture the full variability of visibility values reported in
the literature. Therefore, to ensure that the retrospec-
tive study characterizes the full range of uncertainty
in visibility values nationwide, the upper and lower
bounds of the triangular distribution were derived by
combining results from appropriate eastern and west-
ern residential visibility valuation studies.
Most of the existing residential visibility valua-
tion studies were found to suffer from part-whole bias,
which results from the failure to differentiate values
for visibility from those for other air quality ameni-
ties, such as reductions in adverse health effects. Of
the studies reviewed for this analysis, only the
McClelland study and Brookshire et al. (1979) have
attempted to obtain bids explicitly for visibility im-
provements (IEc 1992). Since part-whole bias will tend
to produce overstated values for visibility, reported
values from all studies that do not correct for part-
whole bias were adjusted prior to calculating the lower
bound of the uncertainty range. The upper bound of
the uncertainty range was calculated using the unad-
justed values from all studies, which is equivalent to
assuming that the entire value of respondents' stated
WTP for improved air quality can be attributed to in-
creased visibility.
The uncertainty range specified in Table 1-2 cal-
culated using a consensus function derived from a
regression analysis, incorporates a 25 percent adjust-
ment for part-whole bias (i.e., reported values were
multiplied by 0.25) in calculating the lower bound.
This represents an approximate midpoint of the range
defined by the McClelland study's finding that respon-
dents allocated, on average, 18.6 percent of their total
WTP to improvements in visibility, and Chestnut and
Rowe's (1989) conclusion that visibility improvement
accounted for 34 percent of the total WTP reported in
the Brookshire et al. study. Similarly, the "Denver
Brown Cloud" study results indicate that respondents
allocated 27.2 percent of their total WTP to visibility
improvements (Irwin et al. 1990). Therefore, the ap-
plication of a 25 percent adjustment for part-whole
bias to all but the McClelland and Brookshire values
would appear to be supported by the recent literature,
with the resulting consensus value representing a plau-
sible lower bound for the uncertainty range of visibil-
ity values. The consensus function approach, incor-
porating the part-whole bias adjustment, yields esti-
mated upper and lower bound values of $21 and $8,
respectively, for annual household WTP per unit im-
provement in DeciView.
1-7
-------
Tabic 1-2. Unit Values for Economically Valuing Health and Welfare Endpoints.
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
Mortality
$4.8 million per
statistical life
Weibull distribution,
mean = $4.8 million
std. dev. = $3.24 million
Central Est: $ value is the mean of value-of-statistical-life estimates
from 26 studies (5 contingent valuation and 21 labor market studies).
Uncertainty: Best-fit distribution to the 26 sample means. The Weibull
distribution prevents selection of negative WTP values.
$293,000 per
statistical life-
year
Weibull distribution,
mean = $293,000
std. dev. = $198,000
Central Est: $ value is the mean of the distribution of the value of a
statistical life-year, derived from the distribution of the value of a
statistical life (see below).
Uncertainty: Assuming the discount rate is five percent, and assuming
an expected 35 yrs. remaining to the avg. worker in the wage-risk
studies (see above), the value of a statistical life-year is just a constant,
0.061, multiplied by the value of a statistical life. The distribution of
the value of a life-year is derived from the distribution of the value of a
statistical life. Given that this is a Weibull distribution, as indicated
above, the value of a statistical life-year is also a Weibull distribution,
with mean equal to 0.061 multiplied by the mean of the original
Weibull distribution (0.061x$4.8 million = $293,000) and standard
deviation equal to 0.061 multiplied by the standard deviation of the
original distribution (0.061 x $3.24 = $198,000). (If the discount rate
were considered to also be uncertain, then the distribution of a
statistical life-year would depend on this distribution as well and would
have to be generated by Monte Carlo methods.)
-------
Health or Welfare
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Endpoint
Central Estimate
Uncertainty Distribution
Chronic Bronchitis (CB)
$260,000
A Monte Carlo-generated
distribution, based on three
underlying distributions, as
described more fully under
"Derivation of Estimates"
and in the text.
Central Est: $ value is the mean of a Monte Carlo distribution of WTP
to avoid a case of pollution-related CB. WTP to avoid a case of
pollution-related CB is derived by adjusting WTP to avoid a severe
case of CB (as described in Viscusi et al., 1991) for the difference in
severity and taking into account the elasticity of WTP with respect to
severity of CB. The mean of the resulting distribution is $260,000.
Uncertainty: The distribution of WTP to avoid a case of pollution-
related CB was generated by Monte Carlo methods, drawing from each
of three distributions: (1) WTP to avoid a severe case of CB is assigned
a 1/9 probability of being each of the first nine deciles of the
distribution of WTP responses in Viscusi et al., 1991; (2) the severity of
a pollution-related case of CB (relative to the case described in the
Viscusi study) is assumed to have a triangular distribution, centered at
severity level 6.5 with endpoints at 1.0 and 12.0 (see text for further
explanation); and (3) the constant in the elasticity of WTP with respect
to severity is normally distribution with mean = 0.18 and standard
deviation = 0.0669 (from Krupnick and Cropper, 1992). See text for
further explanation.
IQ Changes
1. Lost IQ Points
$3,000 per lost IQ
point
none available
Central Est: $ value is the mean of estimates based on results of 2
studies. With an assumed 5% discount rate, the results in Schwartz
(1994) yield an estimate of $2,500 per IQ point; the results of Salkever
(1995) yield an estimate of $3,400. These estimates include the
combined effects on lifetime earnings: (1) directly based on IQ
decrement, and (2) indirectly based on lower educational attainment
and reduced labor force participation (subtracting from indirect benefits
the costs of additional education and associated opportunity cost).
2. Incidence of IQ < 70
$42,000
none available
Central Est: $ value measures reduction in education costs in terms of
special needs for lower IQ students (in mainstream schools).
3
B-
x'
s"
a
S'
s
§
£
a
3
o
^0
-------
Health or Welfare
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Endpoint
Central Estimate
Uncertainty Distribution
Hypertension
$680 per case per
year
none available
Central Est: $ value quantifies costs associated with phvsician care,
medications, and hospital charges, in addition to opportunity cost of
lost work time due to the disability.
Hospital Admissions
1. Strokes
- initial
cerebrovascular
accidents (ICD code
436)
- initial
atherothrombotic
brain infarctions
(ICD code 434)
$200,000 for
males;
$150,000 for
females
none available
Central Est: $ values for males and females are based on ase- and
gender-specific estimates of lifetime cost of stroke from Taylor et al.,
1996. Estimates include both direct costs (medical expenditures) and
indirect costs (reduced productivity) and assume a five percent discount
rate.
Uncertainty: Although there is uncertainty surrounding the central
estimates presented, there is insufficient information to characterize this
uncertainty.
2. Coronary Heart Disease
(CHD)
$52,000
A Monte Carlo-generated
distribution, based on the
uncertainty about what
proportion of pollution-
related CHD events is acute
myocardial infarction, what
proportion is angina
pectoris, and what
proportion is unstable
angina pectoris (see
"Derivation of Estimates").
Central Est: $ value is the mean of the Monte Carlo-generated
distribution of WTP to avoid a pollution-related case of CHD,
described below.
Uncertainty: The distribution was based on the estimates of the total
medical costs within 5 years of diagnosis of each of three types of CHD
events examined in the Framingham Study, including acute myocardial
infarction, angina pectoris, and unstable angina pectoris (Wittels et al.,
1990). It is unknown what proportion of pollution-related CHD events
are of each type. On each iteration, three proportions were drawn from
three continuous uniform distributions, such that the three proportions
summed to 1.0. The $ value for an iteration is the weighted average of
the $ values for the three types of CHD event (from Wittels et al.,
1990), weighted by the corresponding proportions selected.
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
3. "Respiratory Illness"
$6,100
Normal distribution,
mean = $6,100
std. dev. = $55
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients of all ages, and the cost of associated
physician care, with the opportunity cost of time spent in the hospital.
Source of hospital charge estimate: Elixhauser et al., 1993. Source of
physician charge estimates: Abt Associates Inc., 1992.
Uncertainty: variation about the central estimate based on the standard
error reported for the hospital charge component (greater than the other
two components by an order of magnitude).
4. COPD
(ICD codes 490-496)
$8,100
Normal distribution, with
mean = $8,100
std. dev. = $190
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients 65 and older, and the cost of
associated physician care, with the opportunity cost of time spent in the
hospital. Source of cost-of-illness estimates: Abt Associates Inc., 1992.
Uncertainty: variation about the central estimate derived from a
standard error estimated for the hospital charge component measured
by another study (Elixhauser et al., 1993). The reported standard error
for hospital charge was applied to the combined cost-of-illness and
opportunity cost estimate by assuming that relative variabilities
surrounding the respective means were similar (i.e., coefficients of
variation are equal). The hospital charge represents the vast majority of
the total value to avoid a hospital admission for COPD.
5. Pneumonia
(ICD codes 480-487)
$7,900
Normal distribution, with
mean = $7,900
std. dev. = $110
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients of all ages, and the cost of associated
physician care, with the opportunity cost of time spent in the hospital.
Source of hospital charge estimate: Elixhauser et al., 1993. Source of
physician charge estimates: Abt Associates Inc., 1992.
Uncertainty: Applied the standard error associated with the hospital
charge component to the central estimate of $7,900. The hospital
charge represents the vast majority of the total value to avoid a hospital
admission for pneumonia.
3
B-
x'
s"
a
S'
s
§
£
a
3
o
^0
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
6. Congestive Heart Failure
(ICD code 428)
$8,300
Normal distribution, with
mean = $8,300
std. dev. = $120
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients of all ages, and the cost of associated
physician care, with the opportunity cost of time spent in the hospital.
Source of hospital charge estimate: Elixhauser et al., 1993. Source of
physician charge estimates: Abt Associates Inc., 1992.
Uncertainty: Applied the standard error associated with the hospital
charge component to the central estimate of $8,300. The hospital
charge represents the vast majority of the total value to avoid a hospital
admission for congestive heart failure.
7. Ischemic Heart Disease
(ICD codes 410-414)
$10,300
Normal distribution, with
mean = $10,300
std. dev. = $88
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients of all ages, and the cost of associated
physician care, with the opportunity cost of time spent in the hospital.
Source of hospital charge estimate: Elixhauser et al., 1993. Source of
physician charge estimates: Abt Associates Inc., 1992.
Uncertainty: Applied the standard error associated with the hospital
charge component to the central estimate of $10,300. The hospital
charge represents the vast majority of the total value to avoid a hospital
admission for ischemic heart disease.
Respiratory Ailments Not Requiring Hospitalization
1. Upper Resp. Symptoms
(URS)
(defined as one or
more of the
following: runny or
stuffy nose, wet
cough, burning,
aching, or red eyes)
$19
Continuous uniform
distribution over the
interval [$7, $33]
Central Est: Combinations of the 3 svmptoms for which WTP
estimates are available that closely match those listed by Pope et al.
result in 7 different "symptom clusters," each describing a "type" of
URS. A $ value was derived for each type of URS, using IEc mid-
range estimates of WTP to avoid each symptom in the cluster and
assuming additivity of WTPs. The $ value for URS is the average of
the $ values for the 7 different types of URS.
Uncertainty: taken to be a continuous uniform distribution across the
range of values described by the 7 URS types.
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
2. Lower Resp. Symptoms
(LRS)
(defined in the study
as two or more of the
following: cough,
chest pain, phlegm,
and wheeze.)
$12
Continuous uniform
distribution over the
interval [$5, $19]
Central Est: Combinations of the 4 svmptoms for which WTP
estimates are available that closely match those listed by Schwartz et
al. result in 11 different "symptom clusters," each describing a "type"
of LRS. A $ value was derived for each type of LRS, using IEc mid-
range estimates of WTP to avoid each symptom in the cluster and
assuming additivity of WTPs. The $ value for LRS is the average of
the $ values for the 11 different types of LRS.
Uncertainty: taken to be a continuous uniform distribution across the
range of values described by the 11 LRS types.
3. Acute Bronchitis
$45
Continuous uniform
distribution over the
interval [$13, $77]
Central Est: Average of low and hieh values recommended by TF.C for
use in section 812 analysis (Neumann et al., 1994).
Uncertainty: continuous distribution between low and hiph values
(Neumann et al., 1994) assigns equal likelihood of occurrence of any
value within the range.
4. Acute Respiratory
Symptoms and Illnesses
- Presence of any of
19 acute respiratory
symptoms
- Any Resp.
Symptom
- Increase in Resp.
Illness
$18
1. URS, probability = 40%
LRS, probability = 40%
URS+LRS, prob. = 20%
2. If URS, use URS $ dist.
If LRS, use LRS $ dist.
If URS+LRS, randomly
select one value each from
URS and LRS $
distributions; sum the two
Central Est: Assuming that respiratorv illness and svmptoms can be
characterized as some combination of URS and LRS, namely: URS
with 40% probability, LRS with 40% probability, and both URS and
LRS with 20% probability. The $ value for these endpoints is the
weighted average (using the weights 0.40, 0.40, and 0.20) of the $
values derived for URS, LRS, and URS + LRS.
Uncertainty: based on variability assumed for central estimate, and
URS and LRS uncertainty distributions presented previously.
3
B-
x'
s"
a
S'
s
§
£
a
3
o
^0
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
5. Asthma - Acute
$32
Continuous uniform
distribution over the
interval [$12, $54]
Central Est: Mean of averaee WTP estimates for the four severitv
definitions of a "bad asthma day." Source: Rowe and Chestnut (1986),
a study which surveyed asthmatics to estimate WTP for avoidance of a
"bad asthma day," as defined by the subjects.
Uncertainty: based on the ranee of values estimated for each of the
four severity definitions.
6. Shortness of breath
$5.30
Continuous uniform
distribution over the
interval [$0, $10.60]
Central Est: From Ostro et al.. 1995. This is the mean of the median
estimates from two studies of WTP to avoid a day of shortness of
breath: Dickie et al., 1991 ($0.00), and Loehman et al., 1979 ($10.60).
Uncertainty: taken to be a continuous uniform distribution across the
range of values obtained from the two studies.
Restricted Activity and Work Loss Days
1. WLDs
$83
none available
Central Est: Median weeklv wage for 1990 divided bv 5 (U.S.
Department of Commerce, 1992)
Uncertainty: Insufficient information to derive an uncertainty estimate.
2. RADs
not monetized3
--
-
3. MRADs
$38
triangular distribution
centered at $38 on the
interval [$16, $61]
Central Est: Median WTP estimate to avoid 1 MRRAD ~ minor
respiratory restricted activity day — from Tolley et al. (1986)
(recommended by IEc as the mid-range estimate).
Uncertainty: ranee is based on assumption that value should exceed
WTP for a single mild symptom (the highest estimate for a single
symptom—for eye irritation—is $16.00) and be less than that for a WLD.
The triangular distribution acknowledges that the actual value is likely
to be closer to the point estimate than either extreme.
4. RRADs
not monetized3
--
-
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
Welfare Effects
Household Soiling Damage
$2.50 per
household per
Hg/m3 PM10
(annual cost)
Beta distribution with
mean=$2.50, standard
deviation=$ 1 on the interval
[$1.30, $10.00], The shape
parameters of this
distribution are a=1.2 and
P=7.3.
Central Est: Source: ESEERCO (1994). ESEERCO uses $1.26 as its
low estimate of annual cost of soiling and materials damage per
household (assuming 2.63 persons per household), taken from Manuel
et al. (1982). The Manuel study measured particulate matter as TSP
rather than PM-10. Hypothesizing that at least half of the costs of
household cleaning are for the time value of do-it-yourselfers, which
was not included in the Manuel estimate, ESEERCO multiplied the
Manuel estimate by 2 to get a point estimate of about $2.50, in 1990 $
(reported by ESEERCO as $2.70 in 1992 dollars).
Uncertainty: The Beta distribution selected is a smooth, continuous
function with its probability mass near the mean and it covers the range
of reported estimates.
Visibility
Annual household
WTP = $14 per
unit decrease in
DeciView
(decrease in
DeciView
corresponds to
increase in
visibility)
Triangular distribution
centered at $14 on the
interval [$8, $21]
Central Est: Source: IEc 1997. Calculated bv dividing the household
WTP reported in the McClelland et al. study (1991) by the
corresponding hypothesized change in DeciView.
Uncertainty: Source: IEc 1997. Calculated bv regressing reported
household WTP values on the corresponding change in DeciView
(converted from reported visual range changes) for all relevant city-
scenario combinations posed to respondents in the original studies. The
uncertainty range reflects the 25 percent adjustment for part-whole bias
applied to reported values prior to calculating the lower bound.
Worker Productivity
change in daily
wages: $1 per
worker per 10%
change in 03
none available
Central Est: Based on the elasticity of income with respect to O,
concentration derived from study of California citrus workers (Crocker
and Horst, 1981 and U.S. EPA, 1994). Elasticity applied to the average
daily income for workers engaged in strenuous outdoor labor, $73 (U.S.
1990 Census).
NOTES:
a This endpoint was not monetized because including it in the aggregation of economic benefits would result in double-counting (overlap with WLDs
and MRADs).
3
B-
x'
s"
a
S'
s
§
£
a
3
o
^0
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Results of Valuation of Health
and Welfare Effects
Table 1-3 presents the results of combining the
economic valuations described in this Appendix with
the health and welfare effects results presented in
Appendix D. As noted in Appendix D, there are alter-
native estimates for some health and welfare impacts,
which form the basis of several alternative benefit
estimates. Each of the health effects estimates also
has quantified statistical uncertainty. The range of
estimated health and welfare effects, along with the
uncertain economic unit valuations, were combined
to estimate a range of possible results. The combin-
ing of the health and economic information used the
Monte Carlo method presented in Chapter 7. Table I-
3 shows the mean estimate results, as well as the mea-
sured credible range (upper and lower five percen-
tiles of the results distribution), of economic benefits
for each of the quantified health and welfare catego-
ries.
The results for aggregate monetized benefits were
also calculated using a Monte Carlo method. The re-
sults of the Monte Carlo simulations for the economic
values for each of the major endpoint categories are
presented in Table 1-4. Note that for the upper and
lower fifth percentiles the sum of the estimated ben-
efits from the individual endpoints does not equal the
estimated total. The Monte Carlo method used in the
analysis assumes that each health and welfare end-
point is independent of the others. There is a very low
probability that the aggregate benefits will equal the
sum of the fifth percentile benefits from each of the
ten endpoints.
Table 1-5 shows the estimated total benefits ranges
for the four modeled target years of this study: 1975,
1980, 1985, and 1990. The results of the Monte Carlo
simulations of the aggregate economic benefits for
these four target years are depicted in Figure I-1.
Table 1-6 examines the impact of limiting the
scope of the analysis to locations with more certain
air quality estimates. The main analysis (as shown in
Tables 1-3 through 1-5) covers almost the entire popu-
lation of the 48 States.3 However, the air quality in-
formation is less certain for locations far from a moni-
tor. Table 1-6 presents the results of limiting the analy-
sis to people living within 50 km of an ozone, N02,
S02, or CO monitor, or in counties with a PM moni-
tor. The availability of monitors changes over time.
Hence the proportion of the population included in
this analysis changes overtime as well. Table 1-6 in-
dicates that approximately a quarter of the total ben-
efits estimated in the main analysis comes from areas
with less certain air quality information.
The results of the "all U.S. population" analysis
provides a more accurate depiction of the pattern of
economic benefits across years. The accuracy of the
scale of incidence is less certain. These results pro-
vide a better characterization of the total direct ben-
efits of the Clean Air Act in the lower 48 states than
do the "monitored area only" results because the lat-
ter completely omits historical air quality improve-
ments for about 25 percent of the population. How-
ever, the "all U.S. population" results rely on uncer-
tain extrapolations of pollution concentrations, and
subsequent exposures, from distant monitoring sites
to provide coverage for the 25 percent or so of the
population living far from air quality monitors. Thus,
the main results presented in Tables 1-3 through 1-5
include important uncertainties.
Uncertainties
The uncertainty ranges for the results on the
present value of the aggregate measured monetary
benefits reported in Table 1-3 reflect two important
sources of measured uncertainty:
uncertainty about the avoided incidence of
health and welfare effects deriving from the
concentration-response functions, including
both selection of scientific studies and statis-
tical uncertainty from the original studies; and
uncertainty about the economic value of each
quantified health and welfare effect.
These aggregate uncertainty results incorporate many
decisions about analytical procedures and specific
assumptions discussed in the Appendices to this re-
port.
In order to provide a more complete understand-
ing of the economic benefit results in Table 1-3, sen-
sitivity analyses examine several additional important
aspects of the main analysis. First, this section ex-
3 Except for lead, two to five percent (depending on pollutant) of the population who live in sparsely populated areas are
excluded from the main analysis to maximize computer efficiency. All of the population of the 48 states is included in the lead
analysis.
1-16
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Table 1-3. Criteria Pollutants Health and Welfare Benefits -- Extrapolated to Entire 48 State
Population Present Value (in 1990 using 5% discount rate) of Benefits from 1970 - 1990 (in billions of
1990 dollars).
Eimdl|p«MTrntt
IP«»II Hull torn
PircMnmtt Vallroic ((ItmlllliiifMn* lf lTOIOS))
5th "oilk
Mat
9Sttltn %i31e
Mortality
Mortality (long-term PM-10 exposure)
PM
$2,369
$16,632
$40,597
Mortality (Lead exposure)
Lead
$121
$1,339
$3,910
Chronic Bronchitis
PM
$409
$3,313
$10,401
Other Lead-induced Ailments
Lost IQ Points
Lead
$248
$377
$528
IQ< 70
Lead
$15
$22
$29
Hypertension
Lead
$77
$98
$120
Coronary Heart Disease
Lead
$0
$13
$40
Atherothrombotic brain infarction
Lead
$1
$10
$30
Initial cerebrovascular accident
Lead
$2
$16
$45
Hospital Admissions
*A11 Respiratory
PM & 03
$8
$9
$11
*COPD + Pneumonia
PM & 03
$8
$9
$10
Ischemic He art Disease
PM
$1
$4
$6
Congestive Heart Failure
PM & CO
$3
$5
$7
Other Respiratory-Related Ailments
Children
Shortness of breath, days
PM
$0
$6
$17
** Acute Bronchitis
PM
$0
$7
$18
**Upper & Lower Respiratory Symptoms
PM
$1
$2
$4
Adults
Any of 19 Acute Symptoms
PM&03
$4
$46
$117
All
Asthma Attacks
PM&03
$0
$0
$1
Increase inRespiratory Illness
NO 2
$1
$2
$4
Any Symptom
S02
$0
$0
$0
Restricted Activity and Work Loss Days
MR AD
PM&03
$50
$85
$123
Work Loss Days (WLD)
PM
$30
$34
$39
Human Wei la rc
HouseholdSoiling Damage
PM
$6
$74
$192
Visibility - EasternU.S.
particulates
$38
$54
$71
Decreased Worker Productivity
03
$3
$3
$3
Agriculture (Net Surplus)
03
$11
$23
$35
To avoid double-counting of benefits, the following endpoints weie treated as alternatives:
* Hospital admissions for COPD combined with those for pneumonia are treated as an equally-weighted alternative to hospital
admissions for allrespiratoiy illnesses.
**The definitions of acute bronchitis and upper and lower respiratory illness overlap; both studies count trouble breathing,
diy cough, and wheezing in their estimates. These two studies are treated as alternatives, which reflects the variability of
pollution-induced respiratory effects in children.
1-17
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Table 1-4. Present Value of 1970 to 1990 Monetized Benefits by Endpoint Category for 48 State
Population (billions of $1990, discounted to 1990 at 5 percent).
Present Value
Endpoint
Pollutant(s)
5th %ilc
Mean
95th %ile
Mortality
PM
$2,369
$16,632
$40,597
Mortality
Pb
$121
$1,339
$3,910
Chronic Obstructive Pulmonary Disease
PM
$409
$3,313
$10,401
IQ(LostIQ Pts. + Children w/IQ<70)
Pb
$271
$399
$551
Hypertension
Pb
$77
$98
$120
Hospital Admissions
PM, 03, Pb, & CO
$27
$57
$120
Respiratory-Related Symptoms, Restricted PM, 03, N02, & S02
$123
$182
$261
Activity, & Decreased Productivity
Soiling Damage
PM
$6
$74
$192
Visibility
particulates
$38
$54
$71
Agriculture (Net Surplus)
03
$11
$23
$35
Tabic 1-5. Monte Carlo Simulation Model Results for Target Years. Plus Present Value in 1990
Terms ofTotal Monetized Benefits for Entire 1970 to 1990 Period (in billions of 1990-valuc dollars).
Total Benefits By Year ($Billions)
1975
1980
1985
1990
Present Value (5%)
5th percentile
$87
$235
$293
$329
$5,600
Mean
$355
$930
$1,155
$1,248
$22,200
95th percentile
$799
$2,063
$2,569
$2,762
$49,400
Notes:
Present value reflects compounding of benefits from 1971 to 1990.
"Uncertainty Estimates" are results of Monte Carlo analysis combining economic and physical effects uncertainty (i.e., using both
between- and within-study variability).
Full uncertainty analysis done only for years shown. Uncertainty estimates for intermediate years computed based on ratios of 5th
to 50th percentile and 95th to 50th percentile for years shown. Ratios interpolated between years shown and applied to point
estimates for intermediate years.
1-18
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Figure 1-1. Monte Carlo Simulation Model Results for Target Years (in billions of 1990 dollars).
c
o
PQ
&
$3,000
$2,500
$2,000
$1,500
n $1,000
$500
$0
tc
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
plores the effect of selecting alternative discount rates
on the aggregate present value benefits estimation.
Second, this section examines the sources of the mea-
sured aggregate uncertainty, identifying which of the
measured uncertainty components of incidence and
valuation for individual health effects categories drive
the overall uncertainty results. Third, this section ex-
amines several issues involving the estimated eco-
nomic benefits of mortality.
The Effect of Discount Rates
The main analysis reflected in present value re-
sults shown in Table 1-3 uses a five percent discount
rate. The discount rate primarily enters the calcula-
tions when compounding the economic benefits esti-
mates from individual years between 1970 and 1990
to estimate the present value of the benefits in 1990.
The discount rate also directly enters in the calcula-
tions of the economic values of an IQ point and an
initial case of coronary heart disease.4 There is con-
siderable controversy in the economics and policy lit-
erature about the appropriate discount rate to use in
different settings. Major alternatives recommended by
various authors include a discount rate based on the
social discount rate (typical estimates are in the 2 to 3
percent range), and a discount rate based on the risk-
free rate of return on capital (typically in the 7 to 10
percent range). Table 1-7 presents the aggregate un-
certainty results using three different discount rates:
3 percent, 5 percent and 7 percent. While the aggre-
gate benefits estimates are sensitive to the discount
rate, selecting one of these alternative discount rates
affects the aggregate benefits estimates by only about
15 percent.
The Relative Importance of Different
Components of Uncertainty
The estimated uncertainty ranges in Table 1-3 re-
flect the measured uncertainty associated with both
avoided incidence and economic valuation. A better
understanding of the relative influence of individual
uncertain variables on the overall uncertainty in the
analysis can be gained by isolating the individual ef-
fects of important variables on the range of estimated
benefits. This can be accomplished by holding all the
inputs to the Monte Carlo uncertainty analysis con-
stant (at their mean values), and allowing only one
variable ~ for example, the economic valuation of
mortality ~ to vary across the range of that variable's
uncertainty. The sensitivity analysis then isolates how
this single source of variability contributes to the varia-
tion in estimated total benefits. The results are sum-
marized in Figure 1-2. The nine individual uncertainty
factors that contribute the most to the overall uncer-
tainty are shown in Figure 1-2, ordered by the relative
significance of their contribution to overall uncer-
tainty. Each of the additional sources of quantified
uncertainty in the overall analysis not shown contrib-
ute a smaller amount of uncertainty to the estimates
of monetized benefits than the sources that are shown.
Tabic 1-7. Eflcci of Alternative Discount Rates on Present Value of Total Monetized Benefits lor
1970 to 1990 (in trillions of 1990 dollars).
Present Value in 1990 of TotalBenefits
(Trillions of 1990 Dollars)
3%
5%
7%
5 th percentile
$4.9
$5.6
$6.5
Mean
$19.2
$22.2
$25.8
95th percentile
$42.7
$49.4
$57.5
No tes •
Present value reflects compounding of benefits from 1971 to 1990.
4 The estimated economic value of lost IQ points due to lead exposure is based on the present value of the impact on lifetime
earnings. A discount rate is required to calculate that present value. The impact on income primarily occurs during adulthood, which
is 20 to 70 years after the initial lead exposure. This significant lag results in the discount rate having a significant impact on the
estimated economic benefits of the IQ loss. Similarly, the cost of illness estimate for an initial case of CHD includes the present value
of the annual stream of medical costs incurred after the event, the calculation of which requires an estimate of the discount rate.
1-20
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Figure 1-2. Uncertainty Ranges Deriving From Individual Uncertainty Factors.
a
o
s
.
(/)
—
—
0
TO
¦c
o
0
O
C
0
'sz
o
c
o
c
o
'-4—'
TO
TO
O
_Q
CL
O
c
0
¦o
LO
O
c
£
—
_£=
Q_
o
J= 0
O O
c c
O 0
o
0
O o
^ 0
(/) "O
S O
T3
o
o
X
o
'-4—'
TO
TO
>
O)
o
'-4—'
TO
TO
>
o
9 £
< TO
2 «
o
CO
Because of the multiple uncertainties in the ben-
efits estimation, the total estimated present value of
the monetary benefits of the 1970 to 1990 Clean Air
Act range from a low of about $5.6 trillion to a high
of about $49.4 trillion (in 1990 dollars, discounted at
five percent). Most of the uncertainty in the total esti-
mated benefit levels comes from uncertainty in the
estimate of the economic valuation of mortality, fol-
lowed by the uncertainty in the incidence of mortality
from PM (as a surrogate for all non-lead air pollu-
tion). The incidence of lead-induced mortality also
has a significant influence on the overall uncertainty.
The importance of mortality is not surprising, because
the benefits associated with reduced mortality are such
a large share of overall monetized benefits.
The uncertainty in both the incidence and valua-
tion of chronic bronchitis are the two other signifi-
cant factors driving the overall uncertainty range. The
modeled uncertainty in the other remaining health and
welfare endpoints in the analysis contribute relatively
small amounts to the overall uncertainty in the esti-
mate of total monetary benefits of the Clean Air Act.
Most of these other endpoints account for a relatively
small proportion of the overall benefits estimates,
making it unlikely that they could contribute signifi-
cantly to the overall uncertainty. Estimates of either
the mean values or standard errors of these variables
are generally very small relative to estimated total
monetary benefits.
Economic Benefits Associated with
Reducing Premature Mortality
Because the economic benefits associated with
premature mortality are the largest source of mon-
etized benefits in the analysis, and because the uncer-
tainties in both the incidence and value of premature
mortality are the most important sources of uncertainty
in the overall analysis, it is useful to examine the
mortality benefits estimation in greater detail.
The analytical procedure used in the main analy-
sis to estimate the monetary benefits of avoided pre-
mature mortality assumes that the appropriate eco-
nomic value for each incidence is a value from the
currently accepted range of the value of a statistical
life. As discussed above, the estimated value per pre-
dicted incidence of excess premature mortality is
modeled as a Weibull distribution, with a mean value
of $4.8 million and a standard deviation of $3.2 mil-
lion. This estimate is based on 26 studies of the value
of mortal risks.
1-21
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
There is considerable uncertainty as to whether
the 26 studies on the value of a statistical life provide
adequate estimates of the value of a statistical life
saved by air pollution reduction. Although there is
considerable variation in the analytical designs and
data used in the 26 underlying studies, the majority of
the studies involve the value of risks to a middle-aged
working population. Most of the studies examine dif-
ferences in wages of risky occupations, using a wage-
hedonic approach. Certain characteristics of both the
population affected and the mortality risk facing that
population are believed to affect the average willing-
ness to pay (WTP) to reduce the risk. The appropri-
ateness of a distribution of WTP estimates from the
26 studies for valuing the mortality-related benefits
of reductions in air pollution concentrations therefore
depends not only on the quality of the studies (i.e.,
how well they measure what they are trying to mea-
sure), but also on (1) the extent to which the subjects
in the studies are similar to the population affected by
changes in pollution concentrations, and (2) the ex-
tent to which the risks being valued are similar. As
discussed below, there are possible sources of both
upward and downward bias in the estimates provided
by the 26 studies when applied to the population and
risk being considered in this analysis.
If the individuals who die prematurely from air
pollution are consistently older than the population in
the valuation studies, the mortality valuations based
on middle-aged people may provide a biased estimate
of the willingness to pay of older individuals to re-
duce mortal risk. There is some evidence to suggest
that the people who die prematurely from exposure to
ambient particulate matter tend to be older than the
populations in the valuation studies. In the general
U.S. population far more older people die than younger
people; 88 percent of the deaths are among people
over 64 years old. It is difficult to establish the pro-
portion of the pollution-related deaths that are among
the older population because it is impossible to iso-
late individual cases where one can say with even rea-
sonable certainty that a specific individual died be-
cause of air pollution.
There is considerable uncertainty whether older
people will have a greater willingness to pay to avoid
risks than younger people. There is reason to believe
that those over 65 are, in general, more risk averse
than the general population, while workers in
wage-risk studies are likely to be less risk averse than
the general population. More risk averse people would
have a greater willingness to pay to avoid risk than
less risk averse people. Although the list of recom-
mended studies excludes studies that consider only
much-higher-than- average occupational risks, there
is nevertheless likely to be some selection bias in the
remaining studies ~ that is, these studies are likely to
be based on samples of workers who are, on average,
more risk-loving than the general population. In con-
trast, older people as a group exhibit more risk averse
behavior.
In addition, it might be argued that because the
elderly have greater average wealth than those
younger, the affected population is also wealthier, on
average, than wage-risk study subjects, who tend to
be blue collar workers. It is possible, however, that
among the elderly it is largely the poor elderly who
are most vulnerable to pollution-related mortality risk
(e.g., because of generally poorer health care). If this
is the case, the average wealth of those affected by a
pollution reduction relative to that of subjects in
wage-risk studies is uncertain. In addition, the work-
ers in the wage-risk studies will have potentially more
years remaining in which to acquire streams of in-
come from future earnings.
Although there may be several ways in which job-
related mortality risks differ from air pollution-related
mortality risks, the most important difference may be
that job-related risks are incurred voluntarily whereas
air pollution-related risks are incurred involuntarily.
There is some evidence (see, for example, Violette
and Chestnut, 1983) that people will pay more to re-
duce involuntarily incurred risks than risks incurred
voluntarily. If this is the case, WTP estimates based
on wage-risk studies may be downward biased esti-
mates of WTP to reduce involuntarily incurred air
pollution-related mortality risks.
Finally, another possible difference related to the
nature of the risk may be that some workplace mor-
tality risks tend to involve sudden, catastrophic events
(e.g., workplace accidents), whereas air pollution-re-
lated risks tend to involve longer periods of disease
and suffering prior to death. Some evidence suggests
that WTP to avoid a risk of a protracted death involv-
ing prolonged suffering and loss of dignity and per-
sonal control is greater than the WTP to avoid a risk
(of identical magnitude) of sudden death. Some work-
place risks, such as risks from exposure to toxic chemi-
cals, may be more similar to pollution-related risks. It
is not clear, however, what proportion of the work-
place risks in the wage-risk studies were related to
workplace accidents and what proportion were risks
1-22
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
from exposure to toxic chemicals. To the extent that
the mortality risks addressed in this assessment are
associated with longer periods of illness or greater pain
and suffering than are the risks addressed in the valu-
ation literature, the WTP measurements employed in
the present analysis would reflect a downward bias.
The direction of bias resulting from the age dif-
ference is unclear, particularly because age is con-
founded by risk aversion (relative to the general popu-
lation). It could be argued that, because an older per-
son has fewer expected years left to lose, his WTP to
reduce mortality risk would be less than that of a
younger person. This hypothesis is supported by one
empirical study, Jones-Lee et al. (1985), that found
the value of a statistical life at age 65 to be about 90
percent of what it is at age 40. Citing the evidence
provided by Jones-Lee et al. (1985), a recent sulfate-
related health benefits study conducted for EPA (U.S.
EPA, 1995) assumes that the value of a statistical life
for those 65 and over is 75 percent of what it is for
those under 65.
There is substantial evidence that the income elas-
ticity of WTP for health risk reductions is positive
(see, for example, Alberini et al., 1994; Mitchell and
Carson, 1986; Loehman and VoHuDe, 1982;Gerking
etal., 1988; and Jones-Lee etal., 1985), although there
is uncertainty about the exact value of this elasticity.
Individuals with higher incomes (or greater wealth)
should, then, be willing to pay more to reduce risk, all
else equal, than individuals with lower incomes or
wealth. Whether the average income or level of wealth
of the population affected by PM reductions is likely
to be significantly different from that of subjects in
wage-risk studies, however, is unclear, as discussed
above.
The need to adjust wage-risk-based WTP esti-
mates downward because of the likely upward bias
introduced by the age discrepancy has received sig-
nificant attention (see Chestnut, 1995; IEc, 1992). If
the age difference were the only difference between
the population affected by pollution changes and the
subjects in the wage-risk studies, there might be some
justification for trying to adjust the point estimate of
$4.8 million downward. Even in this case, however,
the degree of the adjustment would be unclear. There
is good reason to suspect, however, that there are bi-
ases in both directions. Because in each case the ex-
tent of the bias is unknown, the overall direction of
bias in the mortality values is similarly unknown.
Adjusting the estimate upward or downward to com-
pensate for any one source of bias could therefore in-
crease the degree of bias. Therefore, the range of val-
ues from the 26 studies is used in the primary analy-
sis without adjustment.
Examining the sensitivity of the overall results to
the mortality values can help illuminate the potential
impacts of alternative mortality valuations. As men-
tioned above, a contractor study performed for EPA
used one approach to evaluate the economic value of
sulfate-related human health improvements resulting
from 1990 Clean Air Act Amendments Title IV acid
rain controls. That study assumed that 85 percent of
the people dying from sulfates (an important compo-
nent of particulate matter) were over 65, and that
people over 65 have a willingness to pay to avoid a
mortal risk that is 75 percent of the values that middle-
aged people have. Using this approach, the value of
an average statistical life (using a weighted average)
is reduced to 79 percent of the previous value.
If statistical life-years lost are used as the unit of
measure, rather than statistical lives lost, the benefit
attributed to avoiding a premature death depends di-
rectly on how premature it is. One way to estimate
the value of a statistical life-year assumes that the value
of a statistical life is directly related to remaining life
expectancy and a constant value for each life-year.
Such an approach results in smaller values of a statis-
tical life for older people, who have shorter life ex-
pectancies, and larger values for younger people. For
example, if the $4.8 million mean value of avoiding
death for people with a 35 year life expectancy is as-
sumed to be the discounted present value of 35 equal-
valued statistical life-years, the implied value of each
statistical life-year is $293,000 (using a 5% discount
rate). The average number of life-years lost by indi-
viduals dying prematurely from exposure to PM is 14
years. This average is obtained by multiplying the
predicted number of PM-related premature deaths in
each age category by the life expectancy for that age
category and dividing by the total number of PM-re-
lated premature deaths.) Using $293,000 per life-year,
the discounted present value of a statistical life for a
person with 14 years of expected life remaining (e.g.,
a 70 year old) is $2.9 million). If statistical life-years
lost are used to value fatal risks, however, other
sources of uncertainty are introduced in the valuation
process.
1-23
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
If statistical life-years lost is the unit of measure,
the value of a statistical life lost depends on (1) how
many years of expected life are lost, (2) the
individual's discount rate, and (3) whether the value
of an undiscounted statistical life-year is the same no
matter which life-year it is (e.g., the undiscounted
value of the seventy-fifth year of life is the same as
the undiscounted value of the fortieth year of life).
Each of these is uncertain. The uncertainty surround-
ing the expected years of life lost by an individual
involves the uncertainty about whether individuals
who die from exposure to air pollution are average
individuals in the demographic (e.g., age-gender-race)
classification to which they belong. The uncertainty
surrounding individuals' discount rates is well docu-
mented. Finally, even if it is assumed that all life-years
are valued the same (apart from differences due to
discounting), the value of a statistical life-year is de-
rived from the value of a statistical life (of a 40 year
old) and the discount rate, each of which is uncertain.
Using life-years lost as the unit of measure means
that, rather than estimating a single value of a statisti-
cal life lost (applicable to all ages), the analysis would
instead estimate age-specific values of statistical lives
lost. It is unclear whether the variability of estimates
of age-specific values of statistical lives lost (in par-
ticular, for ages greater than the average age of work-
ers in the wage-risk studies) would be less than or
greater than the variability of the original estimate of
the value of a statistical life lost from which they would
be derived. If there is an age-related upward bias in
the central tendency value of a statistical life that is
larger than any downward bias, then valuing life-years
rather than lives lost may decrease the bias. Even this,
however, is uncertain.
In spite of the substantial uncertainties and pau-
city of available information, this section presents an
example of a preliminary estimate of the present value
of avoided premature mortality using the life-years
lost approach. The basic approach is to (1) estimate
the number of pollution-related premature deaths in
each age category, (2) estimate the average number
of life-years lost by an individual in a given age cat-
egory dying prematurely, and (3) using the value of a
statistical life-year of $293,000, described above (as-
suming that the undiscounted value of a life-year is
the same no matter when in an individual's life it is)
and assuming a five percent discount rate, calculate
the value of a statistical life lost in each age category.
To obtain estimates of the number of air pollu-
tion-related deaths in each age cohort, it is preferable
to have age-specific relative risks. Many of the epide-
miological studies, however, do not provide any esti-
mate of such age-specific risks. In this case, the age-
specific relative risks must be assumed to be identi-
cal.
Some epidemiology studies on PM do provide
some estimates of relative risks specific to certain age
categories. The limited information that is available
suggests that relative risks of mortality associated with
exposure to PM are greater for older people. Most of
the available information comes from short-term ex-
posure studies. There is considerable uncertainty in
applying the evidence from short-term exposure stud-
ies to results from long-term (chronic exposure) stud-
ies. However, using the available information on the
relative magnitudes of the relative risks, it is possible
to form a preliminary assessment of the relative risks
by different age classes.
The analysis presented below uses two alterna-
tive assumptions about age-specific risks: (1) there is
a constant relative risk (obtained directly from the
health literature) that is applicable to all age cohorts,
and (2) the relative risks differ by age, as estimated
from the available literature. Estimates of age-spe-
cific PM coefficients (and, from these, age-specific
relative risks) were derived from the few age-specific
PM coefficients reported in the epidemiological lit-
erature. These estimates in the literature were used to
estimate the ratio of each age-specific coefficient to a
coefficient for "all ages" in such a way that consis-
tency among the age-specific coefficients is preserved
— that is, that the sum of the health effects incidences
in the separate, non-overlapping age categories equals
the health effects incidence for "all ages." These ra-
tios were then applied to the coefficient from Pope et
al. (1995). Details of this approach are provided in
Post and Deck (1996). Because Pope etal. considered
only individuals age 30 and older (instead of all ages),
the resulting age-specific PM coefficients may be
slightly different from what they would have been if
the ratios had been applied to an "all ages" coeffi-
cient. The differences, however, are likely to be mini-
mal and well within the error bounds of this exercise.
The age-specific relative risks used in the example
below assume that the relative risks for people under
65 are only 16 percent of the population-wide aver-
age relative risk, the risks for people from 65 to 74
are 83 percent of the population-wide risk, and people
1-24
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
75 and older have a relative risk 55 percent greater
than the population average. Details of this approach
are provided in Post and Deck (1996).
The life-years lost approach also requires an esti-
mate of the number of life-years lost by a person dy-
ing prematurely at each given age. The average num-
ber of life-years lost will depend not only on whether
relative risks are age-specific or uniform across all
age groups, but also on the distribution of ages in the
population in a location. As noted above, using the
same relative risk for all age categories, the average
number of life-years lost in PM-related premature
deaths in the United States was estimated to be 14
years. Using the age-specific relative risk estimates
developed for this analysis, the average number of
life-years lost becomes 9.8 years. In a location with a
population that is younger than average in the United
States, the same age-specific relative risks will pro-
duce a larger estimated average number of life-years
lost. For example, using the same age-specific rela-
tive risks, the average number of life-years lost in PM-
related premature deaths in Los Angeles County,
which has a younger population, is estimated to be
15.6 years.
The present value benefits estimates for PM-re-
lated mortality using the alternative approaches dis-
cussed above are shown in Table 1-8. Table 1-8 is based
on a single health study: Pope et al., 1995. Alterna-
tive studies, or the uncertainty approach used in the
primary analysis, would result in a similar pattern of
the relationship between valuation approaches. The
pattern of monetized mortality benefits across the dif-
ferent valuation procedures shown in Table 1-8 is es-
sentially invariant to the particular relative risk and
the particular dollar value used.
As noted above, the life-years lost approach used
here assumes that people who die from air pollution
are typical of people in their age group. The estimated
value of the quantity of life lost assumes that the people
who die from exposure to air pollution had an aver-
age life expectancy. However, it is possible that the
people who die from air pollution are already in ill
health, and that their life expectancy is less than a
typical person of their age. If this is true, then the num-
ber of life years lost per PM-related death would be
lower than calculated here, and the economic value
would be smaller.
The extent to which adverse effects of particulate
matter exposure are differentially imposed on people
of advanced age and/or poor health is one of the most
important current uncertainties in air pollution-related
health studies. There is limited information, prima-
rily from the short-term exposure studies, which sug-
gests that at least some of the estimated premature
mortality is imposed disproportionately on people who
are elderly and/or of poor health. The Criteria Docu-
ment for Particulate Matter (U.S. EPA, 1996) identi-
fies only two studies which attempt to evaluate this
disproportionality. Spix et al. (1994) suggests that a
small portion of the PM-associated mortality occurs
in individuals who would have died in a short time
anyway. Cifuentes and Lave (1996) found that 37 to
87 percent of the deaths from short-term exposure
could have been premature by only a few days, al-
though their evidence is inconclusive.
Table 1-8. Alternative Esti mates of the Present Value of Mortality Associated With PM
(based on Pope et al., 1996, in trillions of 1990 dollars).
Present Value of PM
Valuation Procedure
Mortality Benefits
Primary Analysis Method ( $4.8 million per statistical life saved)
$16.6
Life Years Lost approaches
Single relative risk, valuation using 5% discounting
$9.1
Annroximate as>c-sr>ccific relative risk, valuation usins> 5% discounting
$8 3
Notes:
Present value reflects compounding of benefits from 1971 to 1990, usinga 5 percent discount rate.
1-25
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Prematurity of death on the order of only a few
days is likely to occur largely among individuals with
pre-existing illnesses. Such individuals might be par-
ticularly susceptible to a high PM day. To the extent
that the pre-existing illness is itself caused by or ex-
acerbated by chronic exposure to elevated levels of
PM, however, it would be misleading to define the
prematurity of death as only a few days. In the ab-
sence of chronic exposure to elevated levels of PM,
the illness would either not exist (if it was caused by
the chronic exposure to elevated PM) or might be at a
less advanced stage of development (if it was not
caused by but exacerbated by elevated PM levels).
The prematurity of death should be calculated as the
difference between when the individual died in the
"elevated PM" scenario and when he would have died
in the "low PM" scenario. If the pre-existing illness
was entirely unconnected with chronic exposure to
PM in the "elevated PM" scenario, and if the indi-
vidual who dies prematurely because of a peak PM
day would have lived only a few more days, then the
prematurity of that PM-related death is only those few
days. If, however, in the absence of chronic exposure
to elevated levels of PM, the individual's illness would
have progressed more slowly, so that, in the absence
of a particular peak PM day the individual would have
lived several years longer, the prematurity of that PM-
related death would be those several years.
Long-term studies provide evidence that a por-
tion of the loss of life associated with long-term ex-
posure is independent of the death from short-term
exposures, and that the loss of life-years measured in
the long-term studies could be on the order of years.
If much of the premature mortality associated with
PM represents short term prematurity of death im-
posed on people who are elderly and/or of ill health,
the estimates of the monetary benefits of avoided
mortality may overestimate society's total willingness
to pay to avoid particulate matter-related premature
mortality. On the other hand, if the premature mortal-
ity measured in the chronic exposure studies is de-
tecting excess premature deaths which are largely in-
dependent of the deaths predicted from the short term
studies, and the disproportionate effect on the elderly
and/or sick is modest, the benefits measured in this
report could be underestimates of the total value. At
this time there is insufficient information from both
the medical and economic sciences to satisfactorily
resolve these issues from a theoretical/analytical stand-
point. Until there is evidence from the physical and
social sciences which is sufficiently compelling to
encourage broad support of age-specific values for
reducing premature mortality, EPA will continue to
use for its primary analyses a range of values for mor-
tality risk reduction which assumes society values re-
ductions in pollution-related premature mortality
equally regardless of who receives the benefit of such
protection.
1-26
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Economic Valuation References
Abbey, D.E., F. Petersen, P. K. Mills, and W. L. Beeson.
1993. "Long-Term Ambient Concentrations of
Total Suspended Particulates, Ozone, and Sul-
fur Dioxide and Respiratory Symptoms in a
Nonsmoking Population. "Archives of Environ-
mental Health 48(1): 33-46.
Abt Associates, Inc. 1992. The Medical Costs of Five
Illnesses Related to Exposure to Pollutants.
Prepared for U.S. EPA, Office of Pollution Pre-
vention and Toxics, Washington, DC.
Abt Associates, Inc. 1996. Section 812 Retrospective
Analysis: Quantifying Health and Welfare Ben-
efits. Draft. Prepared for U.S. EPA, Office of
Policy Planning and Evaluation, Washington
DC. May.
Alberini, A., A. Krupnick, M. Cropper, and W.
Harrington. 1994. "Air Quality and the Value
of Health in Taiwan." Paper presented at the
annual meeting of the Eastern Economics As-
sociation, Boston, Massachusetts, March.
Brookshire, David S., Ralph C. d'Arge, William D.
Schulze and Mark A. Thayer. 1979. Methods
Development for Assessing Air Pollution Con-
trol Benefits, Vol. IP. Experiments in Valuing
Non-Market Goods: A Case Study of Alterna-
tive Benefit Measures of Air Pollution Control
in the South Coast Air Basin of Southern Cali-
fornia. Prepared for the U.S. Environmental
Protection Agency, Office of Research and De-
velopment.
Chestnut, Lauraine G. 1995. Dollars and Cents: The
Economic and Health Benefits of Potential
Particulate Matter Reductions in the United
States. Prepared for the American Lung Asso-
ciation.
Chestnut, Lauraine G. and Robert D. Rowe. 1989. "Eco-
nomic Valuation of Changes in Visibility: A
State of the Science Assessment for NAPAP,"
as cited in National Acid Precipitation Assess-
ment Program, Methods for Valuing Acidic
Deposition and Air Pollution Effects. NAPAP
State of Science and State of Technology Re-
port No. 27, Part B. December.
Cifuentes, L. and L.B. Lave. 1996. "Association of
Daily Mortality and Air Pollution in Phila-
delphia, 1983-1988." J. Air Waste Manage.
Assoc.: in press.
Crocker T. D. and R. L. Horst, Jr. 1981. "Hours of
Work, Labor Productivity, and Environmen-
tal Conditions: a Case Study." The Review of
Economics and Statistics 63:361-368.
Cropper, M.L. and A.J. Krupnick. 1990. "The Social
Costs of Chronic Heart and Lung Disease,"
Resources for the Future Discussion Paper QE
89-16-REV.
Dickie, M. et al. 1991. Reconciling Averting Behav-
ior and Contingent Valuation Benefit Esti-
mates of Reducing Symptoms of Ozone Ex-
posure (draft), as cited in Neumann, J.E.,
Dickie, M.T., and R.E. Unsworth. 1994. In-
dustrial Economics, Incorporated. Memoran-
dum to Jim DeMocker, U.S. EPA, Office of
Air and Radiation. March 31.
Elixhauser, A., R.M. Andrews, and S. Fox. 1993.
Clinical Classifications for Health Policy
Research: Discharge Statistics by Principal
Diagnosis and Procedure. Agency for Health
Care Policy and Research (AHCPR), Center
for General Health Services Intramural Re-
search, U.S. Department of Health and Hu-
man Services.
Empire State Electric Energy Research Corporation
(ESEERCO). 1994. New York State Environ-
mental Externalities Cost Study. Report 2:
Methodology. Prepared by: RCG/Hagler,
Bailly, Inc., November.
Gerking, S., M. DeHaan, and W. Schulze. 1988. "The
Marginal Value of Job Safety: A Contingent
Valuation Study." Journal of Risk and Un-
certainty 1: 185-199.
Industrial Economics, Incorporated (IEc). 1992. Ap-
proaches to Environmental Benefits Assess-
ment to Support the Clean Air Act Section 812
Analysis. Prepared by Robert E. Unsworth,
James E. Neumann, and W. Eric Browne, for
Jim DeMocker, Office of Policy Analysis and
Review, Office of Air and Radiation, U.S.
Environmental Protection Agency. 6 Novem-
ber.
1-27
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Industrial Economics, Incorporated (IEc). 1993a.
"Analysis of Visibility Valuation Issues for
the Section 812 Study," Memorandum to Jim
DeMocker, Office of Policy Analysis and
Review, Office of Air and Radiation, U.S.
Environmental Protection Agency, prepared
by Jim Neumann, Lisa Robinson, and Bob
Unsworth. September 30.
Industrial Economics, Incorporated (IEc). 1997. "Vis-
ibility Valuation for the CAA Section 812
Retrospective Analysis," Memorandum to
Jim DeMocker, Office of Policy Analysis and
Review, Office of Air and Radiation, U.S.
Environmental Protection Agency, prepared
by Michael H. Hester and James E. Neumann.
18 February.
Irwin, Julie, William Schulze, Gary McClelland,
Donald Waldman, David Schenk, Thomas
Stewart, Leland Deck, Paul Slovic, Sarah
Lictenstein, and Mark Thayer. 1990. Valuing
Visibility: A Field Test of the Contingent Valu-
ation Method. Prepared for Office of Policy,
Planning and Evaluation, U.S. Environmen-
tal Protection Agency. March.
Jones-Lee, M.W., et al. 1985. "The Value of Safety:
Result of a National Sample Survey." Eco-
nomic Journal 95(March): 49-72.
Krupnick, A.J. and M.L. Cropper. 1992. "The Effect
of Information on Health Risk Valuations,"
Journal of Risk and Uncertainty 5(2): 29-48.
Loehman, E.T., S.V. Berg, A.A. Arroyo, R.A.
Hedinger, J.M. Schwartz, M.E. Shaw, R.W.
Fahien, V.H. De, R.P. Fishe, D.E. Rio, W.F.
Rossley, and A.E.S. Green. 1979. "Distribu-
tional Analysis of Regional Benefits and Cost
of Air Quality Control." Journal of Environ-
mental Economics and Management 6: 222-
243.
Loehman, E.T. and Vo Hu De. 1982. "Application of
Stochastic Choice Modeling to Policy Analy-
sis of Public Goods: A Case Study of Air
Quality Improvements." The Review of Eco-
nomics and Statistics 64(3): 474-480.
Manuel, E.H., R.L. Horst, K.M. Brennan, W.N. Lanen,
M.C. Duff, and J.K Tapiero. 1982. Benefits
Analysis of Alternative Secondary National
Ambient Air Quality Standards for Sulfur Di-
oxide and Total Suspended Particulates, Vol-
umes I-IV Prepared for U.S. Environmental
Protection Agency, Office of Air Quality
Planning and Standards, Research Triangle
Park, NC.
McClelland, Gary, William Schulze, Donald
Waldman, Julie Irwin, David Schenk, Tho-
mas Stewart, Leland Deck and Mark Thayer.
1991. Valuing Eastern Visibility: A Field Test
of the Contingent Valuation Method. Prepared
for Office of Policy, Planning and Evaluation,
U.S. Environmental Protection Agency. June.
Mitchell, R.C. and R.T. Carson. 1986. "Valuing Drink-
ing Water Risk Reductions Using the Con-
tingent Valuation Methods: A Methodologi-
cal Study of Risks from THM and Giardia."
Paper prepared for Resources for the Future,
Washington, DC.
Moore, M.J. and W.K. Viscusi. 1988. "The Quantity-
Adjusted Value of Life". Economic Inquiry
26(3): 369-388.
National Acid Precipitation Assessment Program
(NAPAP). 1991. Acidic Deposition: State of
Science and Technology (Summary Report).
(Washington, DC: NAPAP). September.
Neumann, J.E., M. T. Dickie, and R.E. Unsworth.
1994. Industrial Economics, Incorporated.
Memorandum to Jim DeMocker, U.S. EPA,
Office of Air and Radiation. Linkage Between
Health Effects Estimation and Morbidity
Valuation in the Section 812 Analysis —
Draft Valuation Document. March 31.
Ostro, B.D., M.J. Lipsett, J.K. Mann, H. Braxton-
Owens, and M.C. White. 1995. "Air Pollu-
tion and Asthma Exacerbations Among Afri-
can American Children in Los Angeles." In-
halation Toxicology.
Pope, C.A., III, M.J. Thun, M.M. Namboodiri, D.W.
Dockery, J.S. Evans, F.E. Speizer, and C.W.
Heath, Jr. 1995. "Particulate Air Pollution as
a Predictor of Mortality in a Prospective Study
of U.S. Adults." Am. J. Respir. Crit. Care
Med. 151: 669-674.
1-28
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Post, Ellen and L. Deck. 1996. Abt Associates Inc.
Memorandum to Tom Gillis, U.S. EPA, Of-
fice of Office of Policy Planning and Evalua-
tion. September 20.
Rowe, R.D. and L.G. Chestnut. 1986. "Oxidants and
Asthmatics in Los Angeles: A Benefits Analy-
sis—Executive Summary." Prepared by En-
ergy and Resource Consultants, Inc. Report
to the U.S. EPA, Office of Policy Analysis.
EPA-230-09-86-018. Washington, D.C.
March.
Salkever, D.S. 1995. "Updated Estimates of Earnings
Benefits from Reduced Exposure of Children
to Environmental Lead." Environmental Re-
search 70: 1-6.
Schwartz, J. 1994. "Societal Benefits of Reducing
Lead 'Exposure" Environmental Research 66:
105-124.
Spix, C., J. Heinrich, D. Dockery, J. Schwartz, G.
Volksch, K. Schwinkowski, C. Collen, and
H.E. Wichmann. 1994. Summary of the
Analysis and Reanalysis Corresponding to the
Publication Air Pollution and Daily Mortal-
ity in Erfurt, East Germany 1980-1989. Sum-
mary report for: Critical Evaluation Work-
shop on Particulate Matter—Mortality Epi-
demiology Studies; November; Raleigh, NC.
Wuppertal, Germany: Bergische Universitat-
Gesamthochschule Wuppertal.
Taylor, T.N., P H. Davis, J.C. Torner, J. Holmes, J.W.
Meyer, and M. F. Jacobson. 1996. "Lifetime
Cost of Stroke in the United States." Stroke
27(9): 1459-1466.
Tolley, G.S. et al. 1986. Valuation of Reductions in
Human Health Symptoms and Risks. Univer-
sity of Chicago. Final Report for the U.S.
Environmental Protection Agency. January.
U.S. Department of Commerce, Economics and Sta-
tistics Administration. 1992. Statistical Ab-
stract of the United States, 1992: The National
Data Book. 112th Edition, Washington, D.C.
U.S. Environmental Protection Agency (U.S. EPA).
1994. Documentation for Oz-One Computer
Model (Version 2.0). Office of Air Quality
Planning and Standards. Prepared by:
Mathtech, Inc., under EPA Contract No.
68D30030, WA 1-29. August.
U.S. Environmental Protection Agency (U.S. EPA).
1995. Human Health Benefits From Sulfate
Reductions Under Title IV of the 1990 Clean
Air Act Amendments. Prepared by Hagler
Bailly Consulting, Inc. for U.S. EPA, Office
of Air and Radiation, Office of Atmospheric
Programs. November 10.
U.S. Environmental Protection Agency (U.S. EPA).
1996. Air Quality Criteria for Particulate
Matter, Volume III of III. Office of Research
and Development, Washington DC. EPA/600/
P-95/001cF
Violette, D.M. and L.G. Chestnut. 1983. Valuing Re-
duction in Risks: A Review of the Empirical
Estimates. Report prepared for the U.S. En-
vironmental Protection Agency, Washington,
D.C. EPA-230-05-83-002.
Viscusi, W.K. 1992. Fatal Tradeoffs: Public andPri-
vate Responsibilities for Risk. (New York:
Oxford University Press).
Viscusi, W.K., W. A. Magat, and J. Huber. 1991.
"Pricing Environmental Health Risks: Survey
Assessments of Risk-Risk and Risk-dollar
Tradeoffs." Journal of Environmental Eco-
nomics and Management 201: 32-57.
Wittels, E.H., J.W. Hay, and A.M. Gotto, Jr. 1990.
"Medical Costs of Coronary Artery Disease
in the United States." The American Journal
of Cardiology 65: 432-440.
World Health Organization (WHO). 1996. Final Con-
sultation on Updating and Revision of the Air
Quality Guidelines for Europe. Bilthoven, The
Netherlands 28-31 October, 1996 ICP EHH
018VD96 2.il.
1-29
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
1-30
-------
Appendix J: Future Directions
Research Implications
In virtually any benefit analysis of environmen-
tal issues, the state of scientific information limits the
degree of coverage possible and the confidence in
benefit estimation. For most benefit categories, fur-
ther scientific research would allow for a better quanti-
fication of benefits. One of the major outcomes of the
retrospective analysis is a clear delineation of the
major limitations in the scientific and economics lit-
erature in carrying out an analysis of this scope. Of-
ten, a list of research needs is generated in studies
such as this, but there is no clear internal mechanism
to address these needs. With this study (and the ongo-
ing section 812 program), a process has been initiated
where identified research needs are to be integrated
into EPA's overall extramural research grants pro-
gram, administered by the Office of Research and De-
velopment. It is hoped that the research projects that
flow from this process will enable future analyses to
be less uncertain and more comprehensive.
Certain of the limitations in the retrospective
analysis are directly related to the historical nature of
the analysis, such as sparse information about air qual-
ity in the early 1970's in many areas in the country.
Other important limitations are related to the effects
of elevated airborne lead concentrations, which has
been virtually eliminated by the removal of lead from
gasoline. A better understanding of these relationships
would improve our understanding of the historical
impact of the Clean Air Act, but would only indirectly
contribute to developing future air pollution policy.
However, most of the research that will reduce the
major gaps and uncertainties needed to improve the
section 812 analyses will be directly relevant to EPA's
primary ongoing mission of developing and imple-
menting sound environmental policies to meet the
national goals established in the Clean Air Act and
other legislation.
There are a number of biological, physical and
economic research areas which the EPA Project Team
identified as particularly important for improving fu-
ture section 812 analyses. These research topics can
be divided into two principal categories: (1) those
which might reduce uncertainties in cost and benefit
estimates with significant potential for influencing
estimated net benefits of the Clean Air Act, and (2)
those which might improve the comprehensiveness
of section 812 assessments by facilitating quantifica-
tion and/or monetization of currently excluded cost
or benefit endpoints. The following subsections pro-
vide examples of research topics which, if pursued,
might improve the certainty and/or comprehensive-
ness of future section 812 studies.
Research Topics to Reduce Uncertainty
Scientific information about the effects of long-
term exposure to air pollutants is just beginning to
emerge, but continues to be the subject of intense sci-
entific inquiry. The relationship between chronic PM
exposure and excess premature mortality included in
the quantified results of the present analysis is one
example of such research. However, many other po-
tential chronic effects that are both biologically plau-
sible and suggested by existing research are not in-
cluded. Research to identify the relationship linking
certain known or hypothesized physical effects (e.g.,
ozone's effects on lung function or fibrosis) with the
development of serious health effects (e.g., cardiop-
ulmonary diseases or premature mortality), and the
appropriate economic valuation of the willingness to
pay to avoid the risks of such diseases, would reduce
the uncertainty caused by a major category of excluded
health effects which could have a significant impact
on the aggregate benefits estimates.
As described in Chapter 7 and Appendix I, pre-
mature mortality is both the largest source of benefits
and the major source of quantified uncertainty in the
J-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
retrospective analysis. In addition to the quantified
uncertainty, there is considerable additional
unquantified uncertainty about premature mortality
associated with air pollution. Much of the informa-
tion base about these relationships is relatively new,
more is coming out virtually daily, and there is sub-
stantial disagreement in the scientific community
about many of the key issues. EPA's Research Strat-
egy and Research Needs document for particulate
matter, currently under development, will address
many of these scientific issues as they relate to PM.
The following selection of highly uncertain issues
could have a significant impact on both the aggregate
mortality benefits estimates and the measured uncer-
tainty range:
the relationship of specific pollutants in the
overall premature mortality effect, including
the individual or interactive relationships be-
tween specific chemicals (e.g., ozone, sul-
fates, nitrates, and acid aerosols), and particle
sizes (i.e., coarse, fine and ultra-fine particles);
the degree of overlap (if any) between the
measured relationships between effects asso-
ciated with short term exposures and effects
from long term exposure;
the confounding effect of changes in historic
air pollution, including changes over time in
both pollution levels and the composition of
the pollutant mix;
the extent to which life spans are shortened
by exposure to the pollutants, and the distri-
bution of ages at the time of death;
the willingness to pay to avoid the risks of
shortened life spans; and
the extent to which total PM2 exposure in-
crementally augments the variability of out-
door PM25 and increases the dose that would
cause excess morbidity or mortality.
After premature mortality, chronic bronchitis is
the next largest health effect benefit category included
in the retrospective analysis. There is considerable
measured uncertainty about both the incidence esti-
mation and the economic valuation. Additional re-
search could reduce uncertainties about the level of
the pollutants associated with the observed effects,
the baseline incidence used to model the changes in
the number of new cases, and the correspondence be-
tween the definition of chronic bronchitis used in the
health effects studies and the economic valuation stud-
ies.
Another area of potentially useful research would
be further examination of the effects of criteria pol-
lutants on cardiovascular disease incidence and mor-
tality. Considering available epidemiological evidence
and the potential economic cost of cardiovascular dis-
ease, the value of avoiding these outcomes may sig-
nificantly influence the overall benefit estimates gen-
erated in future assessments.
Further research on the willingness to pay to avoid
the risk of hospital admissions for specific conditions
would reduce a potentially significant source of non-
measured uncertainty. The Project Team used
"avoided costs" for the value of an avoided hospital
admission, based on the avoided direct medical cost
of hospitalization (including lost wages for the em-
ployed portion of the hospitalized population).
Avoided costs are likely to be a substantial underesti-
mate of the appropriate willingness to pay, especially
for such serious health effects as hospitalization for
strokes and congestive heart failure, particularly be-
cause they omit the value of avoided pain, suffering,
and inconvenience. Furthermore, in addition to hos-
pitalization, there is evidence that some people seek
medical assistance as outpatients. It is also likely that
there are additional people adversely affected by short-
term air pollution levels who seek physician services
(but stop short of hospital admissions). Revised esti-
mates of the appropriate economic value of avoided
hospitalization and other primary care medical ser-
vices could increase the total economic benefits of
this cluster of health effects sufficiently that it could
be a much larger portion of the aggregate benefit to-
tal.
Finally, one of the challenges in preparing the
retrospective analysis was modeling the integrated
relationships between emissions of many different
chemicals, the subsequent mixture of pollutants in the
ambient air, and the resulting health and welfare ef-
fects of simultaneous exposure to multiple pollutants.
One element of the uncertainty in the analysis derives
from the limited current understanding of any inter-
active (synergistic or antagonistic) effects of multiple
pollutants. The need to better understand these com-
plex issues is not a limited scientific question to im-
prove section 812 analyses, but is the primary focus
of EPA's current activities, organized under the Fed-
J-2
-------
Appendix J: Future Directions
eral Advisory Council Act (FACA) process, to de-
velop an integrated set of attainment policies dealing
with ozone, particulate matter, sulfur and nitrogen
oxides, and visibility. Further research on multi-pol-
lutant issues may both (a) reduce a source of unmea-
sured uncertainty in the section 812 analyses and (b)
allow for effective apportionment of endpoint reduc-
tion benefits to specific pollutants or pollutant mixes.
Research Topics to Improve
Comprehensiveness
Even though research efforts falling in this cat-
egory may not result in significant changes in net mon-
etary benefit estimates, one of the goals of the section
812 studies is to provide comprehensive information
about Clean Air Act programs. For example, programs
to control hazardous air pollutants (HAPs) tend to
impose costs and yield benefits which are relatively
small compared to programs of pervasive national
applicability such as those aimed at meeting National
Ambient Air Quality Standards. Nevertheless, there
are significant social, political, financial, individual
human health, and specific ecosystem effects associ-
ated with emissions of HAPs and the programs imple-
mented to control them. Under these circumstances,
continued efforts to understand these consequences
and evaluate their significance in relation to other pro-
grammatic and research investment opportunities
might be considered reasonable, particularly in the
context of comprehensive program assessments such
as the present study.
Some cost and benefit effects could not be fully
assessed and incorporated in the net monetary benefit
estimate developed for the present study for a variety
of reasons. Various effects were excluded due to (a)
inadequate historical data (e.g., lack of data on his-
torical ambient concentrations of HAPs), (b) inad-
equate scientific knowledge (e.g., lack of concentra-
tion-response information for ecological effects of
criteria and hazardous air pollutants), or (c) resource-
intensity or limited availability of analytical tools
needed to assess specific endpoints (e.g., indirect ef-
fects resulting from deposition and subsequent expo-
sure to HAPs). Other specific examples of presently
omitted or underrepresented effect categories include
health effects of hazardous air pollutants, ecosystem
effects, any long-term impact of displaced capital on
productivity slowdown, and redirected technological
innovation.
Although the primary focus of 1970 to 1990 CAA
programs was reduction of criteria pollutants to
achieve attainment of national ambient air quality stan-
dards, emissions of air toxics were also substantially
reduced. Some air toxics were deliberately controlled
because of their known or suspected carcinogenicity,
while other toxic emissions were reduced indirectly
due to control procedures aimed at other pollutants,
particularly ozone and particulate matter. The current
analysis was able to present only limited information
on the effects of changes in air toxic emissions. These
knowledge gaps may be more serious for future sec-
tion 812 analyses, however, since the upcoming pro-
spective study will include evaluation of the effects
of an expanded air toxic program under the CAA Title
III. Existing knowledge gaps that prevented a more
complete consideration of toxics in the present study
include (a) methods to estimate changes in acute and
chronic ambient exposure conditions nationwide, (b)
concentration-response relationships linking exposure
and health or ecological outcomes, (c) economic valu-
ation methods for a broad array of potential serious
health effects such as renal damage, reproductive ef-
fects and non-fatal cancers, and (d) potential ecologi-
cal effects of air toxics.
In addition to research to improve the understand-
ing of the consequences of changes in air pollution on
human health and well-being, further research on non-
health effects could further improve the comprehen-
siveness of future assessments. Improvements in air
quality have likely resulted in improvements in the
health of aquatic and terrestrial ecosystems and the
myriad of ecological services they provide, but knowl-
edge gaps prevented them from being included in the
current analysis. Additional research in both scien-
tific understanding and appropriate modeling proce-
dures could facilitate inclusion of additional benefits
such as improvements in water quality stemming from
a reduction in acid deposition-related air pollutants.
Water quality improvements would benefit human
welfare through enhancements in certain consump-
tive services such as commercial and recreational fish-
ing, in addition to non-consumptive services such as
wildlife viewing, maintenance of biodiversity, and
nutrient cycling. Similarly, increased growth, produc-
tivity and overall health of U.S. forests could occur
from reducing ozone, resulting in benefits from in-
creased timber production, greater opportunities for
recreational services such as hunting, camping, wild-
life observation, and nonuse benefits such as nutrient
cycling, temporary C02 sequestration, and existence
J-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
value. Finally, additional research using a watershed
approach to examine the potential for ecological ser-
vice benefits which emerge only at the watershed scale
might be useful and appropriate given the broad geo-
graphic scale of the section 812 assessments.
While there are insufficient data and/or analyti-
cal resources to adequately model the short-run eco-
logical and ecosystem effects of air pollution reduc-
tion, even less is known about the long-run effects of
prolonged exposure. Permanent species displacement
or altered forest composition are examples of poten-
tial ecosystem effects that are not reflected in the cur-
rent monetized benefit analysis, and could be a source
of additional benefits. In addition to these ecological
research needs, an equally large, or larger, gap in the
benefit-cost analysis is the lack of adequate tools to
monetize the benefits of such ecosystem services.
Future Section 812 Analyses
This retrospective study of the benefits and costs
of the Clean Air Act was developed pursuant to sec-
tion 812 ofthe 1990 Clean Air Act Amendments. Sec-
tion 812 also requires EPA to generate an ongoing
series of prospective studies of the benefits and costs
of the Act, to be delivered as Reports to Congress every
two years.
Design of the first section 812 prospective study
commenced in 1993. The EPA Project Team devel-
oped a list of key analytical design issues and a
"strawman" analytical design reflecting notional de-
cisions with respect to each of these design issues.1
The analytical issues list and strawman design were
presented to the Science Advisory Board Advisory
Council on Clean Air Compliance Analysis (Coun-
cil), the same SAB review group which has provided
review of the retrospective study. Subsequently, the
EPA Project Team developed a preliminary design
for the first prospective study. Due to resource limita-
tions, however, full-scale efforts to implement the first
prospective study did not begin until 1995 when ex-
penditures for retrospective study work began to de-
cline as major components of that study were com-
pleted.
As for the retrospective, the first prospective study
is designed to contrast two alternative scenarios; how-
ever, in the prospective study the comparison will be
between a scenario which reflects full implementa-
tion of the CAAA90 and a scenario which reflects
continued implementation only of those air pollution
control programs and standards which were in place
as of passage of the CAAA90. This means that the
first prospective study will provide an estimate of the
incremental benefits and costs of the CAAA90.
The first prospective study is being implemented
in two phases. The first phase involves development
of a screening study, and the second phase will in-
volve a more detailed and refined analysis which will
culminate in the first prospective study Report to Con-
gress. The screening study compiles currently avail-
able information on the costs and benefits of the imple-
mentation of CAAA90 programs, and is intended to
assist the Project Team in the design of the more de-
tailed analysis by providing insights regarding the
quality of available data sources and analytical mod-
els, and the relative importance of specific program
areas; emitting sectors; pollutants; health, welfare, and
ecological endpoints; and other important factors and
variables.
In developing and implementing the retrospective
study, the Project Team developed a number of im-
portant modeling systems, analytical resources, and
techniques which will be directly applicable and use-
ful for the ongoing series of section 812 Prospective
Studies. Principal among these are the Criteria Air
Pollutant Modeling System (CAPMS) model devel-
oped to translate air quality profile data into quantita-
tive measures of physical outcomes; and the economic
valuation models, coefficients, and approaches devel-
oped to translate those physical outcomes to economic
terms.
The Project Team also learned valuable lessons
regarding analytical approaches or methods which
were not as productive or useful. In particular, the
Project Team plans not to perform macroeconomic
modeling as an integral part of the first prospective
analysis. In fact, there are currently no plans to con-
duct a macroeconomic analysis at all. Essentially, the
Project Team concluded, with confirmation by the
SAB Council, that the substantial investment of time
and resources necessary to perform macroeconomic
modeling would be better invested in developing high
quality data on the likely effects of the CAA on key
emitting sectors, such as utilities, on-highway vehicles,
refineries, etc. While the intended products of a mac-
1 Copies of the prospective study planning briefing materials are available from EPA.
J-4
-------
Appendix J: Future Directions
roeconomic modeling exercise - such as overall ef-
fects on productivity, aggregate employment effects,
indirect economic effects- are of theoretical interest,
the practical results of such exercises in the context
of evaluating environmental programs may be disap-
pointing for several reasons.
First, the CAA has certainly had a significant ef-
fect on several industrial sectors. However, the coarse
structure of a model geared toward simulating effects
across the entire economy requires crude and poten-
tially inaccurate matching of these polluting sectors
to macroeconomic model sectors. For example, the J/
W model used for the retrospective study has only 35
sectors, with electric utilities comprising a single sec-
tor. In reality, a well-structured analysis of the broader
economic effects of the CAA would provide for sepa-
rate and distinct treatment of coal-fired utility plants,
oil-fired plants, and so on. Furthermore, the outputs
of the macroeconomic model are too aggregated to
provide useful and accurate input information for the
sector-specific emission models used to project the
emissions consequences of CAA programs. Again, the
critical flaw is the inability to project important de-
tails about differential effects on utilities burning al-
ternative fuels.
The second critical problem with organizing a
comprehensive analysis of the CAA around a macro-
economic modeling approach is that the effect infor-
mation produced by the macroeconomic model is rela-
tively unimportant with respect to answering the fun-
damental, target variable: "How do the overall health,
welfare, ecological, and economic benefits of Clean
Air Act programs compare to the costs of these pro-
grams? " The Project Team believes that any adverse
effect, no matter how small in a global context, should
not be deemed "insignificant" if even one individual
is seriously harmed. However, the retrospective study
results themselves have shown that, when analytical
resources are limited, the resources invested in the
macroeconomic modeling would have been better
spent to provide a more complete and less uncertain
assessment of the benefit side of the equation. Even
on the cost side of the equation, it is far more impor-
tant to invest in developing accurate and reliable esti-
mates of sector-specific compliance strategies and the
direct cost implications of those strategies. This will
be even more true in the prospective study context
when the Project Team will be faced with forecasting
compliance strategies and costs rather than simply
compiling survey data on actual, historical compli-
ance expenditures.
The third and most important limitation of mac-
roeconomic modeling analysis of environmental pro-
grams is that, unlike the economic costs of protection
programs, the economic benefits are not allowed to
propagate through the economy. For example, while
productivity losses associated with reduced capital
investment due to environmental regulation are
counted, the productivity gains resulting from reduced
pollution-related illness and absenteeism of workers
are not counted. The resulting imbalance in the treat-
ment of regulatory consequences raises serious con-
cerns about the value of the macroeconomic model-
ing evaluation of environmental programs. In the fu-
ture, macroeconomic models which address this and
other concerns may be developed; however, until such
time EPA is likely to have limited confidence in the
value of macroeconomic analysis of even broad-scale
environmental protection programs.
Based on these findings and other factors, the de-
sign of the first prospective study differs in important
ways from the retrospective study design. First, rather
than relying on broad-scale, hypothetical, macroeco-
nomic model-based scenario development and analy-
sis, the first prospective study will make greater use
of existing information from EPA and other analyses
which assess compliance strategies and costs, and the
emission and air quality effects of those strategies.
After developing as comprehensive a data set as pos-
sible of regulatory requirements, compliance strate-
gies, compliance costs, and emissions consequences,
the data set will be reviewed, refined, and extended
as feasible and appropriate. In particular, a number of
in-depth sector studies will be conducted to develop
up-to-date, detailed projections of the effects of new
CAA requirements on key emitting sectors. Candi-
date sectors for in-depth review include, among oth-
ers, utilities, refineries, and on-highway vehicles.
The first prospective study will also differ from
the retrospective study in that analytical resources will
be directed toward development of a more complete
assessment of benefits. Efforts will be made to ad-
dress the deficiencies which prevailed in the retro-
spective study relating to assessment of the benefits
of air toxics control. In addition, the Project Team
will endeavor to provide a more complete and effec-
tive assessment of the ecological effects of air pollu-
tion control.
J-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
J-6
-------
November 1999
EPA-410-R-99-001
United States
Environmental Protection
Agency
Office of Air and Radiation
Office of Policy
The Benefits and Costs
of the Clean Air Act
1990 to 2010
EPA Report to Congress
November 1999
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
[This page left blank intentionally.]
-------
Executive
Summary
Section 812 of the Clean Air Act Amendments
of 1990 requires the Environmental Protection
Agency to periodically assess the effect of the Clean
Air Act on the "public health, economy, and envi-
ronment of the United States," and to report the
findings and results of its assessments to the Con-
gress. This Report to Congress, the first of a series
of prospective studies we plan to produce ever)'- two
years, presents the results and conclusions of our
analysis of the benefits and costs of the Clean Air
Act during the period from 1990 to 2010. The main
goal of this report is to provide Congress and the
public with comprehensive, up-to-date information
on the Clean Air Act's social costs and benefits, in-
cluding improvements in human health, welfare, and
ecological resources.
The first report that the EPA created under the
section 812 authority, The Benefits and Costs of the
Clean Air Act: 1970 to 1990, was published and con-
veyed to Congress in October 1997. This retrospec-
tive analysis comprehensively assessed the benefits
and costs of all requirements of the 1970 Clean Air
Act and the 1977 Amendments, up to the passage of
the Clean Air Act Amendments of 1990. The re-
sults of the retrospective analysis showed that the
nation's investment in clean air wras more than justi-
fied by the substantial benefits that were gained in
the form of increased health, environmental qual-
ity, and productivity.
The Clean Air Act Amendments of 1990 built
upon the significant progress made by the original
Clean Air Act of 1970 and its 1977 amendments in
improving the nation's air quality. The amendments
utilized the existing structure of the Clean Air Act,
but strengthened those requirements to tighten and
clarify implementation goals and timing, increase the
stringency of some requirements, revamp the haz-
ardous air pollutant regulatory program, refine and
streamline permitting requirements, and introduce
new programs for the control of acid ram precur-
sors and stratospheric ozone depleting substances.
Because the 1990 Amendments represent an incre-
mental improvement to the nation's clean air pro-
gram, the analysis summarized in this report was
designed to estimate the costs and benefits of the 1990
Amendments incremental to those assessed in the
retrospective analysis. Our intent is that this report
and its predecessor, the retrospective, together pro-
vide a comprehensive assessment of current and ex-
pected future clean air regulatory programs and their
costs and benefits.
This first prospective analysis consists of a se-
quence of six steps. These six steps, listed in order
of completion, are:
(1) estimate air pollutant emissions in 1990,
2000, and 2010;
(2) estimate the cost of emission reductions aris-
ing from the Clean Air Act Amendments;
(3) model air quality based on emissions esti-
mates;
(4) quantify air quality related health and envi-
ronmental effects;
(5) estimate the economic value of cleaner air;
and
(6) aggregate results and characterize uncertain-
ties.
The methodology and results for each step are
summarized below and described in detail in the
chapters of this report.
Air Pollutant Emissions
Estimation of reductions in pollutant emissions
afforded by the 1990 ('lean Air Act Amendments
(CAAA) serves as the starting point for this study's
subsequent benefit and cost estimates. We focused
our emissions analysis on six major pollutants: vola-
tile organic compounds (VOCs), nitrogen oxides
i
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
(NOj, sulfur dioxide (SO,), carbon monoxide (CO),
coarse particulate matter (PM10), and fine particu-
late matter (PM2S). For each of these pollutants we
forecast emissions for the years 2000 and 2010 un-
der two different scenarios: a) the Pre-CAAA sce-
nario that assumes no additional control require-
ments would be implemented beyond those that
were in place when the 1990 CAAA were passed;
and b) the Post-CAAA scenario that incorporates
the effects of controls which, when we formulated
the scenario, we expected would be likely to occur
as a result of implementing the 1990 Amendments.
Emissions estimates for both the Pre-CAAA and
Post-CAAA scenarios reflect expected growth in
population, transportation, electric power genera-
tion, and other economic activity by 2000 and 2010.
We compare the emissions estimates under each of
these scenarios to estimate the effect of the CAAA
requirements on future emissions.
The results of the emissions phase of the assess-
ment indicate that the 1990 Clean Air Act Amend-
ments significantly reduce future emissions of air
pollutants. Substantial reductions w7ill be achieved
for the two major precursors of ambient ground-
level ozone: volatile organic compounds (VOCs) and
oxides of nitrogen (NOx). Relative to the Pre-CAAA
scenario, estimated VOC emissions under the Post-
CAAA case are 35 percent lower by 2010. This
change in emissions is due largely to VOC reduc-
tions from motor vehicles and area sources (e.g., dry
cleaners, commercial bakeries, and other widely dis-
persed sources).
The NO emission reduction under the Post-
x
CAAA scenario represents the greatest proportional
emissions change estimated in our analysis. For the
vear 2010, the Post-CAAA NO emissions estimate
7 x
is 39 percent lowrer than the Pre-CAAA estimate,
representing a decrease in emissions of almost 11
million tons. Nearly half of this reduction is from
utilities, largely as a result of the particular NO
emissions cap and trading program we assumed un-
der the Post-CAAA scenario. The remaining reduc-
tions are attributable to cuts in motor vehicle and
non-utility point source emissions.
Carbon monoxide (CO) emissions contribute
directly to concentrations of carbon monoxide in
the environment. The 2010 Post-CAAA estimate
for CO emissions is 81.9 million tons, 23 percent
lower than the Pre-CAAA projection. The reduc-
tion in (X) emissions is mostly due to motor ve-
hicle emission controls.
The ("AAA also will achieve a substantial re-
duction in precursors of fine particulate matter
(PM25). Sulfur dioxide (SO,) is an important precur-
sor of PM. By 2010, SO, emissions are 31 percent
lower under the Post-CAAA scenario. Of the 8.2
million ton difference between Pre- and Post-CAAA
SO, estimates, 96 percent is attributable to additional
control of utility emissions through a national cap-
and-trade program involving marketable SO, emis-
sion allowances. Oxides of nitrogen, discussed above,
are also important fine PM precursors.
We project the 1990 Clean Air Act Amendments
to have more modest effects on emissions of par-
ticulate material which is emitted in solid form (i.e.,
"primary" or "direct" PM1() and PM,5 emissions).
Overall, emissions of primary PM and PM arc
each approximately four percent lower in 2010 un-
der the Post-CAAA scenario than under the Pre-
CAAA scenario. Although the incremental effects
of the Clean Air Act Amendments on primary PM
emissions will be relatively small, PM in the atmo-
sphere is comprised of both directly emitted primary
particles and particles that form in the atmosphere
through secondary processes as a result of emissions
of SO,, NO , and organic compounds. These PM
species, formed by the conversion of gaseous pollut-
ant emissions, are referred to collectively as "second-
ary" PM. Because, as noted above, the 1990 Amend-
ments achieve substantial reductions in these gaseous
precursor emissions, the Amendments have a much
larger effect on PM and PM levels in the atmo-
sphere than might be apparent if only the changes
in directly emitted primary particles are considered.
Compliance Costs
Our estimate of the costs of the Clean Air Act
Amendment provisions is based on an evaluation of
the increases in expenditures incurred by various
entities to meet the additional control requirements
incorporated in the Post-CAAA case. These costs
include operation and maintenance (O&M) expen-
ditures —which includes research and development
(R&D) and other similarly recurring expenditures—
plus amortized capital costs (i.e., depreciation plus
-------
Executive Summary
interest costs associated with the existing capital
stock). Relative to the Pre-CAAA case, Post-CAAA
scenario total annual compliance costs for Titles I
through V are approximately $19 billion higher by
the year 2000, rising to |27 billion by the year 2010.
Compliance with Title I, Provisions for Attain-
ment and Maintenance of National Ambient Air
Quality Standards (NAAQS), accounts for 114.5
billion, or over half, of the estimated increase in year
2010 compliance costs. Compliance with mobile
source emissions control provisions under Title II
of the Clean Air Act Amendments accounts for an
additional 30 percent of the total costs, or S9 billion
annually by 2010. Provisions to control acid depo-
sition and emissions of stratospheric ozone deplet-
ing substances account for most of the remainder of
the costs.
These direct compliance costs provide a good,
but incomplete, measure of the total effect of the
Clean Air Act Amendments on the U.S. economy.
A complete picture of the indirect impacts of these
costs would include changes in employment and
prices as well as impacts that might be experienced
among customers of the firms that must incur these
costs. While these indirect effects could be impor-
tant, we believe the direct cost estimates provide a
good initial measure of the effect of the Clean Air-
Act Amendments on the U.S. economy, as well as
an appropriate metric for comparison with the di-
rect benefits reported here.
Table ES-1
Summary Comparison of Benefits and Costs (Estimates in millions 1990$)
Titles I through V
Annual Estimates
2000
2010
Monetized Direct Costs:
Lowa
Central
$19,000
$27,000
High3
Monetized Direct Benefits:
Lowb
$16,000
$26,000
Central
$71,000
$110,000
Highb
$160,000
$270,000
Net Benefits:
Low
($3,000)
($1,000)
Central
$52,000
$83,000
High
$140,000
$240,000
Benefit/Cost Ratio:
Lowc
less than 1/1
less than 1/1
Central
4/1
4/1
High0
more than 8/1
more than 10/1
aThe cost estimates for this analysis are based on assumptions about future changes in factors such as consumption
patterns, input costs, and technological innovation. We recognize that these assumptions introduce significant
uncertainty into the cost results; however the degree of uncertainty or bias associated with many of the key factors cannot
be reliably quantified. Thus, we are unable to present specific low and high cost estimates.
b Low and high benefits estimates are based on primary results and correspond to 5th and 95th percentile results from
statistical uncertainty analysis, incorporating uncertainties in physical effects and valuation steps of benefits analysis.
Other significant sources of uncertainty not reflected include the value of unqualified or unmonetized benefits that are
not captured in the primary estimates and uncertainties in emissions and air quality modeling.
0 The low benefit/cost ratio reflects the ratio of the low benefits estimate to the central costs estimate, while the high ratio
reflects the ratio of the high benefits estimate to the central costs estimate. Because we were unable to reliably quantify
the uncertainty in cost estimates, we present the low estimate as "less than X," and the high estimate as "more than Y",
where X and Y are the low and high benefit/cost ratios, respectively.
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Human Health and
Environmental Benefits
To estimate benefits, the results of the emissions
analysis served as the principal input to a linked se-
ries of models. We used these models to estimate
changes in air quality, human health effects, ecologi-
cal effects, and, ultimately, the net economic ben-
efits of the Clean Air Act Amendments. The goals
of these steps in the analysis were to estimate the
implications of changes in emissions resulting from
compliance with the Clean Air Act Amendments
on criteria pollutant air quality throughout the lower
48 states, and the impacts on human health and the
environment that result from these changes.
We focused our air quality modeling efforts on
estimating the impact of Pre- and Post-CAAA emis-
sions on ambient concentrations of ozone, PM
PM„ SO,, NO , and CO and on acid deposition
and visibility in future years. We found that the
majority of the total monetized benefits, however,
is attributable to changes in particulate matter con-
centrations and, more specifically, to the effect of
these ambient air quality changes on avoidance of
premature mortality. We estimate that 2010 Post-
CAAA PM and PM concentrations in the east-
ern U.S. will average about 5 to 10 percent lower
than 2010 Pre-CAAA concentrations, with some
areas of the eastern U.S. experiencing much greater
reductions of up to 30 percent. The air quality mod-
eling also indicates a substantial overall reduction in
future-year PM and PM concentrations through-
out the western U.S., including most population
centers, following implementation of the Clean Air
Act Amendments.
The direct benefits of the air quality improve-
ments we estimated under the Post-CAAA scenario
include reduced incidence of a number of adverse
human health effects, improvements in visibility, and
avoided damage to agricultural crops. The estimated
annual economic value of these benefits in the year
2010 ranges from $26 to $270 billion, in 1990 dol-
lars, with a central estimate, or mean, of $110 bil-
lion. These estimates do not include a number of
other potentially important effects which could not
be readily quantified and monetized (i.e., converted
to dollar terms). These excluded effects include a
wide range of ecosystem changes, air toxics-related
human health effects, and a number of additional
health effects associated with criteria pollutants.
In addition, these results reflect the particular
choices we made with respect to interpretations of
the available scientific and economic literature and
adoption of paradigms for representing health and
environmental changes in economic terms. We re-
fer to these results, then, as our "primary" estimates;
however, in the text of this report we also present
some alternative results which reflect other available
choices for models or assumptions.
One particularly important assumption of our
primary analysis is that correlations between in-
creased air pollution exposures and adverse health
outcomes found by epidemiological studies indicate
causal relationships between the pollutant exposures
and the adverse health effects. Future research may
lead to revisions in this assumption as well as other
key assumptions, data, and models we use to esti-
mate the benefits and costs of the Clean Air Act.
Such revisions may in turn imply significant changes
in the estimates of Clean Air Act costs and benefits
presented here and in past and future assessments.
In our judgment, however, the primary results re-
flect the best currently available science and the most
up-to-date tools and data we had at our disposal —
and the most reasonable assumptions we could
adopt— as each step of the analysis was implemented.
Cleaner air also yields benefits to ecological sys-
tems. This first section 812 prospective analysis de-
votes a great deal of effort to characterizing and,
where possible, quantifying and monetizing the im-
pacts of air pollutants on natural systems. Our in-
creased effort is in part a result of the findings of the
retrospective analysis, where we identified a better
understanding of ecological effects as an important
research direction for the first prospective and sub-
sequent analyses. Quantified benefits of CAAA pro-
grams reflected in the overall monetized benefits
include: increased agricultural and timber yields; re-
duced effects of acid rain on aquatic ecosystems; and
reduced effects of nitrogen deposited to coastal estu-
aries. Many ecological benefits, however, remain
difficult or impossible to quantify, or can only be
quantified for a limited geographic area. The mag-
nitude of quantified benefits and the wide range of
unquantified benefits nonetheless suggest that as we
learn more about ecological systems and can con-
duct more comprehensive ecological benefits assess-
ments, estimates of these benefits could be substan-
tially greater.
iv
-------
Executive Summary
We developed separate estimates for the Title
VI provisions of the CAAA designed to protect
stratospheric ozone. Stratospheric ozone is the layer
of the atmosphere that protects the planet from the
harmful effects of ultraviolet radiation (UV-b). Our
primary estimate of the cumulative benefits of Title
VI is $530 billion. Using the same uncertainty esti-
mation procedure as for other parts of the analysis,
we estimate Primary Low and Primary High esti-
mates of $100 billion to $900 billion, respectively.
These estimates partially reflect potential averting
behaviors, such as remaining indoors or increasing
use of sunscreens or hats, which may mitigate the
effects of the UV-b exposure increases estimated in
the Pre-CAAA case.
Comparing Costs to Benefits
Based on the specific tools and techniques we
employed, our primary estimate of the net benefit
(benefits minus costs) over the entire 1990 to 2010
period of the additional criteria pollutant control
programs incorporated in the Post-CAAA case is
$510 billion. Our results imply that the monetizable
benefits alone exceeded the direct compliance costs
by four to one. For many of the factors contribut-
ing to this net benefit estimate (especially physical
effects and economic valuation estimates), we were
able to generate quantitative estimates of uncertainty.
By statistically combining these uncertain estimates,
we were able to develop a range of net benefit esti-
mates which provide a partial indication of the over-
all uncertainty surrounding the central estimate of
net benefits. This range, reflecting a 90 percent prob-
ability range around the mean, or central estimate,
is negative $20 billion (implying a small probability
that costs could exceed monetized benefits) to posi-
tive $1.4 trillion.
The estimates for Title VI also indicate that cu-
mulative benefits ($500 billion) well exceed cumula-
tive costs ($27 billion). The time period of our 'Title
VI analysis (175 years) suggests that these estimates
are very uncertain. Nonetheless, the conclusion that
benefits well exceed costs holds even at our Primary
Low estimate of benefits (the low end of the 90 per-
cent probability range, or $100 billion), and regard-
less of discount rate used to generate the cumulative
estimates from the perspective of the present.
The assumptions necessitated by data limitations,
by the current state of the art in each phase of the
analytical approach, by the need to predict future
conditions, and by the state of current research on
air pollution's effects imply that both the mean esti-
mate and the 90 percent probability range around
the central estimate are uncertain. While alterna-
tive choices for data, models, modeling assumptions,
and valuation paradigms may yield results outside
the range projected in our primary analysis, we be-
lieve based on the magnitude of the difference be-
tween the estimated benefits and costs that it is un-
likely that eliminating uncertainties or adopting rea-
sonable alternative assumptions would change the
fundamental conclusion of this study: the Clean Air
Act Amendments' total benefits to society exceed
its costs.
The uncertainties in the primary estimates and
the controversies which persist regarding model
choices and valuation paradigms nonetheless high-
light the need for a variety of new and continued
research efforts. Based on the findings of this study,
the highest priority research needs are:
• Improved emissions inventories and inven-
tory management systems
• A more geographically comprehensive air
quality monitoring network, particularly for
fine particles and hazardous air pollutants
• Use of integrated air quality modeling tools
based on an open, consistent model archi-
tecture
• Development of tools and data to assess the
significance of wetland, aquatic, and terres-
trial ecosystem changes associated with air
pollution
• Increased basic and targeted research on the
health effects of air pollution, especially par-
ticulate matter
• Continued development of economic valu-
ation methods and data, particularly valua-
tion of changes in risks of premature mor-
tality associated with air pollution
Properly directed and funded, such research
would improve the results of future analyses of the
benefits and costs of the Clean Air Act.
Review Process
The CAA requires EPA to consult with an out-
side panel of experts during the development and
interpretation of the 812 studies. This panel of ex-
v
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
perts was organized in 1991 under the auspices of
EPA's Science Advisory Board (SAB) as the Advi-
sory Council on Clean Air Act Compliance Analy-
sis (hereafter, the Council). Organizing the review
committee under the SAB ensured that highly quali-
fied experts would review the section 812 studies in
an objective, rigorous, and publicly open manner
consistent with the requirements and procedures of
the Federal Advisory Committee Act (FACA).
Council review of the present study began in 1993
with a review of the analytical design plan. Since
the initial June 1993 meeting, the Council has met
many times to review proposed data, proposed meth-
odologies, and interim results. While the full Coun-
cil retains overall review responsibility for the sec-
tion 812 studies, some specific issues concerning
physical effects and air quality modeling were re-
ferred to subcommittees comprised of both Council
members and members of other SAB committees.
The Council's Health and Ecological Effects Sub-
committee (HEES) met several times and provided
its own review findings to the full Council. Simi-
larly, the Council's Air Quality Modeling Subcom-
mittee (AQMS) held in-person and teleconference
meetings to review methodology proposals and
modeling results and conveyed its review recommen-
dations to the parent committee.
An interagency review was conducted, during
which a number of analytical issues were discussed.
Conducting a benefit/cost analysis of a major stat-
ute such as the Clean Air Act requires scores of meth-
odological decisions. Many of these issues are the
subject of continuing discussion within the economic
and policy analysis communities and within the
Administration. Key issues include the treatment
of uncertainty in the relationship between particu-
late matter exposure and mortality; the valuation of
premature mortality; the treatment of tax interac-
tion effects; the assessment of stratospheric ozone
recovery; and the treatment of ecological and wel-
fare effects. These issues could not be resolved within
the constraints of tins review. Thus, tins report re-
flects the findings of the EPA and not necessarily
other agencies of the Administration.
vi
-------
Table of Contents
Executive Summary /
Air Pollutant Emissions i
Compliance Costs ii
Human Health and Environmental Benefits iv
Comparing Costs to Benefits v
Review Process v
Chapter 1: Introduction[[[ 1
Background and Purpose 1
Relationship of This Report to Other Regulatory Analyses 1
Requirements of the 1990 Clean Air Act .Amendments 2
Analytical Design and Review 3
Target Variable [[[ 3
Key Assumptions 3
Analytic Sequence 4
Review Process 6
Report Organization [[[ 7
Chapter 2: Emissions [[[ 9
Overview Of Approach[[[ 9
Scenario Development 11
Emissions Estimation Results[[[ 11
Comparison of Emissions Estimates With Other Existing Data '18
Uncertainty In Emission Estimates [[[ 19
Chapter 3: Direct Costs[[[ 23
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Air Quality Model Results[[[ 39
O^one 39
Particulate Matter[[[ 41
Visibility 42
Acid Deposition[[[ 45
S02, NO, NO2, and CO 45
Uncertainty in the Air Quality Estimates 47
Chapter 5: Human Health Effects of Criteria Pollutants 51
Analytical Approach 51
Air Quality [[[ 51
Population 52
Concentration-Response functions 52
Key Analytical Assumptions [[[ 52
Exposure Analysis 54
Selection and Application
of C-R functions 55
PM-Related Mortality 57
Health Effects Modeling Results [[[ 60
Avoided Premature Mortality Estimates 60
Non-fatal Health Impacts [[[ 62
Avoided Health Effects of Other Pollutants [[[ 62
Avoided Effects of Air Toxics 62
Avoided Health Effects for Provisions to Protect Stratospheric 0%one........................................... 63
Uncertainty in the Health Effects Analysis [[[ 65
Chapter 6: Economic Valuation of Human Health Effects.............,,,,,,,,,, 69
Valuation of Benefit Estimates [[[ 69
Valuation of Premature Mortality 70
Valuation of Specific Health Effects[[[ 72
Stratospheric O^one Provisions 74
Results of Benefits Valuation 74
Valuation Uncertainties 75
-------
Table of Contents
Summary of Quantitative Results [[[ 97
Uncertainty 98
Chapter 8: Comparison of Costs and Benefits,,,,,,,......,,,,,,.......,,,,,,......,,, 99
Monetized Benefits of the CAAA[[[ 99
Overview of Benefits Analyses 99
Summary of Monetised Benefits for Human Health and Welfare Effects.................................... 100
Annual Benefits Estimates 101
Aggregate Monetised Benefits [[[ 103
Aggregate Benefits of Title VI Provisions 103
Comparison of .Monetized Benefits and Costs 104
Cost-Effectiveness Evaluation[[[ 106
Major Sources of Uncertainty 106
Quantitative Analysis of Physical Effects and Valuation Uncertainties 107
Measurement Error and Uncertainty in Direct Cost Inputs 109
PM Mortality Valuation Based on Eife-Years Eost 109
PM Mortality Incidence Using the Dockery Study[[[ 110
Uncertainties in Title VI Health Benefits Analysis 110
Uncertainties in Emissions and Air Quality Steps[[[ 111
Omission of Potentially Important Benefits Categories 113
Alternative Discount Rates[[[ 113
Appendix A: Emissions Analysis
Scenario Development[[[ A-2
Comparison of the Base Year Inventory and Emissions Projections
with Other Existing Data[[[ A-7
Post-CAAA Emissions Estimates and EPA Trends Data A-7
Prospective Analysis and GCVTC Emissions Estimates[[[ A-12
Prospective Analysis PM2.5 Emissions Estimates and Observed Data A-13
Industrial Point Sources A-14
Overview of Approach [[[ A-14
Base Year Emissions A-15
Growth Projections[[[ A-15
Control Scenarios A-16
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Growth Projections A-30
Control Scenarios A-30
Emissions Summary A-31
Area Sources A-35
O verview of Approach [[[ A-35
Base Year Emissions A-35
Growth Projections A-36
Control Scenarios A-37
Emissions Summary [[[ A-37
Reasonable Further Progress Requirements A-41
Mercury Emission Estimates [[[ A-48
Medical Waste Incinerators (MWI) A-48
Municipal Waste Combustors (MWCs) [[[ A-48
Electric Utility Generation A-49
Hazardous Waste Combustion A-49
Chlor-alkali Plants A-49
Uncertainties in the Emission Estimates A-51
Base Year Emission Estimates[[[ A-51
Growth Forecasts A-52
Future Year Control Assumptions[[[ A-53
References[[[ A-55
Appendix B: Direct Costs[[[ B-1
Introduction 13-1
Summary of Methods [[[ B-1
ERCAM Model B-1
IPM Model[[[ B-2
Additional Methods B-2
Annali^ation of Costs[[[ B-2
-------
Table of Contents
Appendix C: Air Quality Modeling[[[ C-1
Introduction[[[ C-1
Overview of Section 812 Prospective Modeling Analysis[[[C-2
Air Quality Models and Databases C-2
Methodology for the Combined Use of Observations and .Air Quality Modeling Results......... C-3
Estimating the Effects of the CAAA on Ozone Air Quality C-4
Overview of the UAM and UAM-V Photochemical Modeling Systems.............................. C-5
UAM C-5
UAM-V[[[ C-5
Regional-Scale Modeling of the Eastern US C-6
Regional-Scale Modeling of the Western U.S. [[[ C-12
Urban-Scale Modeling of the San Francisco Bay Area C-15
Urban-Scale Modeling of the Los Angeles Area C-18
Urban-Scale Modeling of the Maricopa County (Phoenix Area) C-20
Calculation of Ozone Air Quality Profiles C-23
Overview of the Methodology [[[ C-23
Description of the Observation Dataset C-23
Calculation of Percentile-Based Adjusted factors C-24
Use of Adjustment Factors to Modify Observed Concentrations C-25
Calculation of O^one Profiles[[[ C-25
Estimating the Effects of the CAAA on Particulate Matter C-38
Overview of the RADM/RPM Modeling System C-38
Application of RADM/RPM for the Eastern US C-38
Overview of the REMSAD Modeling System C-41
Application of REMSAD for the Western U.S. [[[ C-44
Calculation oj PM Air Quality Profiles C-48
Estimating the Effects of the CAAA on Visibility C-64
RADM/RPM and Visibility C-64
RADM /RPM Modeling Results C-64
REMSAD and Visibility[[[ C-66
REMSAD Modeling Results C-66
Acid Deposition C-69
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Types of Health Studies [[[D-5
Chamber Studies D-5
Epidemiological Studies[[[ D-6
Selection of C-R Functions [[[D-7
C-R F'unction General Issues D-7
C-R Function Selection Criteria [[[ D-12
Mortality D-16
Chronic Illness[[[ D-26
Hospital Administration D-30
Minor Illness D-48
C-R Functions Linking Air Pollution and Adverse Health Effects .................................... D-58
Carbon Monoxide D-58
Nitrogen Dioxide [[[ D-61
O^one D-64
Particulate Matter[[[D-70
Sulfur Dioxide D-80
Modeling Results D-82
Uncertainty[[[ D-82
Sensitivity Analyses D-83
References D-93
Appendix E: Ecological Effects of Criteria Pollutants............................ E-1
Introduction E-1
Ecological Overview of the Impacts of Air Pollutants Regulated by the CAAA E-3
Effects of Atomospheric Pollutants on Natural Systems E-3
A cidic Dep ositio n E-4
Nitrogen Deposition[[[ EA
Hazardous Air Pollutant Deposition E-5
Troposheric O^one[[[ E-7
Multiple Stresses and Patterns of Exposure E-8
Summary of Ecological Impacts from Air Pollutants Regulated by the CAAA...........................E-9
Methodological Overview E-14
Using Service Flow Endpoints for Valuation E-15
-------
Table of Contents
Timber Production Impacts From Tropospheric Ozone [[[ E-45
Ecological Effects oj O^one E-43
Modeling Timber Impacts from O^one[[[ E46
Ecological Results E-47
Economic Impacts[[[ E48
Caveats and Uncertainties E-48
Carbon Sequestration Effects E-50
Caveats and Uncertainties E-52
Aesthetic Degradation of Forests E-52
Forest Aesthetic Effects from Air Pollutants[[[ E-53
Economic Value oj Changes in Forest Aesthetics E-58
Extending Economic Estimates to a Broader Area E-59
Caveats and Uncertainties E-61
Toxification of Freshwater Fisheries E-61
Impacts of 'Toxic Air Emissions[[[ E-62
Illustration oj Economic Cost to Anglers E-63
Caveats and Uncertainties EL-65
Conclusion and Implications [[[ i . 65
Summary oj Quantitative Results E-66
Recommendations of Future Research [[[ E-68
References[[[ E-70
Appendix F: Effects of Criteria Pollutants on Agriculture F-1
Introduction F-l
Ozone Concentration Data [[[F-l
Calculation of the SUMO6 Statistic F-2
October to April O^one Concentration Data[[[ F-2
Yield Change Estimates F-3
Exposure-Response Functions F-3
Calculation of O^one Indices F-5
Calculation oj County Weights F-3
Calculation of Percent Change in Yield[[[ F-6
Economic Impact Estimates F-7
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Cost and Benefits Results with Adjusted Parameters[[[ G-25
Five Percent Discount. Rale G-26
Three Percent and Seven Percent Sensitivity Tests [[[ G-27
Two Percent Discount Rate G-30
Undiscounted Benefits[[[ G-30
Limitations and Uncertainties[[[ G-34
Long-Term Discounting G-34
Costs[[[G-3 5
Benefits G-37
References G-3 9
Appendix H: Valuation of Human Health and Welfare Effects of Criteria
Pollutants[[[ H-1
Methods Used to Value Health and Welfare Effects H-1
Valuation of Specific Health Endpoints H-3
Valuation of Premature Mortality Avoided [[[ H-3
Mortality Valuation Methodologies H-6
Valuation Strategy Chosen for this Analysis[[[ H-11
Valuation of Hospital Admissions Avoided H-14
Valuation of Chronic Bronchitis A voided[[[ H-13
Valuation oj Chronic Asthma Avoided H-16
Valuation oj Other Morbidity Endpoints Avoided H-17
-------
Tables
Table ES-1 Summary Comparison of Benefits and Costs (Estimates in millions 1990$) in
Table 2-1 .Major Emissions Source Categories 10
Table 2-2 Summary of National Annual Emissions Projections 12
Table 2-3 Summary of Source Category of National Annual Emission Projections to 2010
(thousand tons) 14
Table 2-4 Airborne Mercury Emission Estimates 15
Table 2-5 Key Uncertainties Associated with Emissions Estimation 21
Table 3-1 Summary of Direct Costs for Titles 1 to V of CAAA, By Title and Selected
Provisions 26
Table 3-2 Results of Quantitative Sensitivity Tests 32
Table 3-3 Key Uncertainty Associated with Cost Estimation 33
Table 4-1 Overview of Air Quality Models 37
Table 4-2 Comparison of Visibility in Selected Eastern Urban Areas 43
Table 4-3 Comparison of Visibility in Selected Eastern .National Parks 43
Table 4-4 Comparison of Visibility in Selected Western Urban Areas 44
Table 4-5 Comparison of Visibility in Selected Western .National Parks 44
Table 4-6 Median Values of the Distribution of ratios of 2010 Post-CAAA/Pre-CAAA
Adjustment Factors 46
Table 4-7 Key Uncertainties Associated with Air Quality Modeling 48
Table 5-1 Human Health Effects of Criteria Pollutants 53
Table 5-2 Summary of Considerations Used in Selecting C-R Functions 56
Table 5-3 Change in Incidence of Adverse Health Effects Associated
with Criteria Pollutants in 2010 (Pre-CAAA minus Post-CAAA) - 48 State U.S.
Population (avoided cases per year) 61
Table 5-4 Mortality Distribution by Age in Primary Analysis (2010 only),
Based on Pope et al. (1995) 62
Table 5-5 Major Health Benefits of provisions to Protect Stratospheric Ozone 64
Table 5-6 Key Uncertainties Associated with Human Health Effects Modeling 65
Table 6-1 Health Effects Unit Valuation (1990 Dollars) 70
Table 6-2 Summary of Mortality Valuation Estimates (millions of $1990) 72
Table 6-3 Results of Human Health Benefits Valuation, Post-CAAA 2010 75
Table 6-4 Valuation of CAAA Benefits: Potential Sources and Likely Direction of Bias 76
Table 6-5 Key Uncertainties Associated with Valuation of Health Benefits 79
Table 7-1 Classes of Pollutants and Ecological Effects 83
Table 7-2 Interactions of Mercury and. Ozone with Natural Systems at Various Levels of
Organization 84
xv
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 7-3 Interactions Between Nitrogen Deposition and Natural Systems
at Various Levels of Organization 85
Table 7-4 Interactions Between Acid Deposition and Natural Systems
at Various Levels of Organization 86
Table 7-5 Ecological Effects of Air Pollutant[[[ 88
Table 7-6 Summary of Endpoints Selected For Quantitative Analysis 89
Table 7-7 Estimated Displaced Costs For Three Estuaries[[[ 91
Table 7-8 Annual Economic Impact of Acidification in 2010 92
Table 7-9 Quantified and Unquantified Ecological and Welfare Effects
of Title VI Provisions 96
Table 7-10 Summary of Evaluated Ecological Benefits [[[ 97
Table 7-11 Summary of Other Welfare Benefits 97
Table 7-12 Key Uncertainties Associated with Ecological Effects Estimation................................. 98
Table 8-1 Criteria Pollutant Health and Welfare Benefits in 2010.................................................. 102
Table 8-2 Present Value of Monetized Benefits for 48 State Population 103
Table 8-3 Summary of Quantified primary Central Estimate Benefits and Costs ...................... 104
Table 8-4 Summary comparison of Benefits and Costs 105
Table 8-5 Summary of Impact of Alternative Methods for Calculating Costs and Benefits..... '108
Table 8-6 Effect of Alternative Discount Rates on Primary- Central Estimates 114
Table A-l Base Year Inventory- - Summary of Approach A-3
Table A-2 Analysis Approach By Major Sector[[[ A-4
Table A-3 Projection Scenario Summary- By Major Sector A-5
Table A-4 Comparison of Emissions: Prospective Analysis and GCVTC Study ..................... A-l 3
Table A-5 Industrial Point Source Control Assumptions For The Post-CAAA Scenario A-l 7
Table A-6 Industrial Point Source Emission Summaries By Pollutant
For 1990, 2000, and 2010 A-19
Table A-l Utility Emission Summary[[[ A-23
Table A-8 BEA Growth Forecasts by Major Source Category: Nonroad Engines/Vehicles ..A-26
Table A-9 Nonroad National Emission Projections By Source Category' ...................................A-28
Table A-'10 Applicability' of Mobile Source Control Programs A-32
Table A-11 National Highway7 Vehicle Emissions By7 Vehicle Type...............................................A-33
Table A-12 Area Source Emission Summary' By Pollutant For 1990, 2000, and 2010 A-39
Table A-l 3 2000 Rate of Progress Analysis [[[ A-43
Table A-l4 2010 Rate of Progress Analysis A-45
Table A-15 Discretionary Control Measures Modeled For ROP/RFP .......................................... A-47
Table A-16 Airborne Mercury7 Emission Estimates A-51
Table B-l Summary of Cost Estimates By Emissions Source B-3
-------
List of Tables and Figures
Table 13-8 Cost Estimates of Motor Vehicle Program 13-17
Table 13-9 Cost Summary of Area Source NOx and PM Controls B-19
Table 13-10 2000 Rate of Progress Analysis [[[ 13-21
Table 13-11 2010 Rate of Progress Analysis 13-22
Table 13-12 Summary of Cost Estimates By CAAA Title[[[ B 24
Table B-13 Title I National Rules, Point, and Area Source VOC Control Costs B-26
Table B-14 Summary of Costs For Title I[[[ B-26
Table B 15 Summary of Title II Motor Vehicle and Non-road
Engine/Vehicle Program Costs [[[ B-27
Table B-16 Title III, MACT Standards, Point and Area Source VOC Control Costs B-28
Table B-17 Annual Cost of Title IV [[[ B-29
Table B-18 Potential Effects of Uncertainty' on Cost Estimates B-33
Table B-19 Factors Affecting Cost of Major CAAA Provisions [[[ B-37
Table B-20 Rate-of-Progress Cost Sensitivity Summary B-42
Table B-21 Area Source PM Control Cost Sensitivity Analysis, Year 2000 ................................. B-43
Table B-22 Results of Sensitivity' Analysis of LEV Cost E-45
Table B-23 Discount Rate Sensitivity Analysis For 2010 Cost Estimates...................................... B-46
Table C-l Emission Totals by Component for each Scenario
for the OTAG Domain (tpd) C-10
Table C-2 Emissions Totals by Component for each Scenario
for the Entire U.S. (tpd) C-l3
Table C-3 Emissions Totals by Component for each Scenario
for the San Francisco Bay Area Entire U.S. (tpd) C-l7
Table C-4 Emission Totals by Component for each Scenario for Los Angeles (tpd)................ C-l 9
Table C-5 Emissions Totals by Component for each Scenario for Pheonix (tpd) C-22
Table C-6 Comparison of CASTNet and RPM Average Concentration of SCH ......................C-40
Table C-7 Comparison of CASTNet and RPM Average Concentrations
and Fractions of NO3 [[[ C-41
Table C-8 REMSAD Output File Species C-43
Table C-9 Chemical Speciation Schemes Applied for REMSAD [[[ C-45
Table C-10 Emission Totals by Component for Each Scenario for the Entire U.S. (tpd) C-46
Table C-ll Background Species Concentration used for REMSAD Initial
and Boundary Conditions C-47
Table C-l2 Geographical Regions of the U.S. [[[C-50
Table C-l3 Comparison of Visibility in Selected Eastern Cities, Metropolitan Areas,
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table I) 3 Studies and Results Selected for Adverse Effects in Fetuses, Infants,
and Young Children I) 25
Table D-4 Summary of Selected Studies for Chrome Illness [[[ D-28
Table I) 5 Studies Used to Develop Respiratory Admissions Estimates I) 32
Table D-6 Summary of Hospital Admissions Studies - Respiratory Illnesses ............................. D-33
Table I) 7 Studies Used to Develop Cardiovascular Admissions Estimates D-39
Table D-8 Summary of Hospital Admissions Studies - Cardiovascular Illness .......................... D-40
Table D 9 Studies Used to Develop Asthma Emergency Room Visits I) 42
Table D-10 Summary of Selected Studies for Emergency Room Visits
- Asthma and Acute Wheezing D-43
Table I) 1 1 Summary of Selected Studies for Emergency Room Visits
- All-Cause, All-Respiratory, COPD, and Bronchitis I) 46
Table D 12 Studies Used to Develop Minor Illness Estimates[[[ D-51
Table D-13 Summary of Selected Studies for Minor Illness D-52
Table D-14 Summary of Selected Studies for Asthmatics [[[ D-55
Table I) 15 Summary of C-R Functions for Carbon Monoxide D 59
Table D-16 Summary of C-R Functions for Nitrogen Dioxide [[[ D-62
Table D-17 Summary' of C-R Functions for Ozone D-65
Table D-18 Summary of C-R Functions for Particulate Matter[[[ D-71
Table D 19 Summary of C-R Functions for Sulfur Dioxide D-81
Table D-20 Change in Incidence of Adverse Health Effects Associated with Criteria Pollutants
(Pre-CAAA minus Post- CAAA) — 48 State U.S. Population within 50 km of a
Monitor (avoided cases per year)[[[ D-85
Table D 21 Change in Incidence of Adverse Health Effects Associated with Criteria Pollutants
(Pre-CAAA minus Post- CAAA) — 48 State U.S. Population
(avoided cases per year) D-87
Table D 22 Mortality Distribution by Age in Primary Analysis, Based
on Pope et al. (1995) D-89
Table I) 23 Illustrative Estimates of the Impact of Criteria Pollutants on Mortality
48 State U.S. Population within 50 km of a Monitor (cases per year D-90
Table I) 24 Illustrative Estimates of the Impact of Criteria Pollutants on Mortality
48 State U.S. Population (cases per year) D-90
Table D 25 Comparison of Alternative Lag Assumptions for Premature Mortality
Associated with PM Exposure D-91
Table E 1 Classes of Pollutants and Ecological Effects E-3
Table E 2 Interactions Between Acid Deposition and Natural Systems
at Various Levels of Organization E-10
Table E-3 Interactions Between Nitrogen Deposition and Natural Systems
at Various Levels of Organization E-ll
Table E-4 Interactions of Mercury and Ozone With Natural Systems
at Various Levels of Organization E-12
Table E-5 Ecological Impacts with Identifiable Human Sen-ice Flows........................................ E-16
-------
List of Tables and Figures
Table E-10 Land Use Prevalence and Pass-Through Figures[[[ E 23
Table I '. 11 Nitrogen Loading From Atmospheric Deposition 1>2.3
Table E 12 Estimated Avoided Costs for three Estuaries[[[ E-27
Table 1. 13 Avoided Cost for Atlantic Coast 1. 29
Table E-14 Summary of pH-Based Effects Threshold[[[ E-38
Table E-15 Acidification Results - 2010 E-40
Table E-16 Annual Economic Impact of Acidification in 2010[[[ E-41
Table E-17 Cumulative Economic Benefits of Acidification from 1990 to 2010 1. 42
Table E-18 Cumulative Cost of pH Stabilization from '1990 to 2010 ............................................. E-43
Table E-19 Difference in Commercial Timber Growth Rates With
and Without the CAAA E-48
Table E-20 Carbon Flux B CAAA versus No-CAAA Air Quality Scenarios E-52
Table E 21 Typical Impacts of Specific Pollutants on the Visual Quality of Forests.................. 1. 34
Table 1. 22 Forests Affected by Regional Pollution 1.55
Table E 23 Summarj' of Monetized Estimates of the Annual Value of Forest
Quality Changes E-60
Table E-24 Illustrative Value of Avoiding Forest Damage in the United States .......................... E-60
Table E 25 Summary of National Data on Toxicity Sampling for Fishing Advisories E-63
Table E-26 Estimates of the Welfare Cost of Toxification in New York State ............................ E-63
Table E-27 Summary of Evaluated Ecological Benefits E-67
Table F 1 Ozone Exposure-Response Functions for Selected Corps (SUM06) F-4
Table F 2 Relative Percent Yield Course [[[F-7
Table F-3 Ozone Exposure-Response Functions for Selected Corps (SUM06) F-8
Table G-l Six Major Sections of Title VI G-3
Table G-2 Phaseout Scenario in Clean Air Act Section 604 and Phaseout Scenario
in Amendments Added Under Clean Air Act Section 606 G-5
Table G-3 Scope of Title VI Cost Estimates[[[G-9
Table G-4 Benefits of Section 604, 606, and 609 G-l6
Table G-5 Benefits of Section 608 [[[ G-20
Table G-6 Benefits of Section 611 G-2'l
Table G-7 Sections 604 and 606: Valuation of Total Benefits from 1990 to 2165,
With a Two Percent Discount Rate G-23
Table G-8 Adjustment Strategy for Key Parameters[[[ G-25
Table G-9 Costs and Benefits of Title VI G-27
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table H-3 Unit Values Used for Economical Valuation of Health
and Welfare Endpoints 11 21
Table i I 1 Primary Estimates of Health and Welfare Benefits Due
to Criteria Pollutants - Post-CAAA 2000 11 29
Table H-5 Primary Estimates of Health and Welfare Benefits Due
to Criteria Pollutants - Post-CAAA 2010 H-30
Table H-6 Sensitivity Analysis of Alternative Discount Rates
on the Valuation of Reduced Premature Mortality H-37
Table 11 7 Elasticity Values for Conducting Sensitivity Analysis of Income Effect................. H-39
Table H-8 Illustrative Adjustment to Estimates of WPT to Avoid Morbidity H-40
Table 11 9 Illustrative Adjustment to Estimates of the Value of Statistical Life......................... H-41
Figures
Figure 1-1 Analytic Sequence For First Section 812 Prospective Analysis 4
Figure 2-1 Pre- and Post-CAAA Scenario VOC Emissions Estimates 16
Figure 2-2 Pre- and Post-CAAA Scenario NOx Emissions Estimates............................................... 16
Figure 2-3 Pre- and Post-CAAA Scenario SC32 Emissions Estimates 16
Figure 2-4 Pre- and Post-CAAA Scenario CO Emissions Estimates. ................................................ 17
Figure 2-5 Pre- and Post-CAAA Scenario PM10 Emissions Estimates 17
Figure 2-6 Pre- and Post-CAAA Scenario PM2.5 Emissions Estimates. ........................................... 17
Figure 2-7 1990 Primary PM2.5 Emissions by EPA Region 19
Figure 4-1 Schematic Diagram of the Future-Year Concentration Estimation Methodology. ... 39
Figure 4-2 Distribution of Monitor-Level Ratios for 95th Percentile
Ozone Concentration: 2010 Post-CAAA/Pre-CAAA 40
Figure 4-3 Distribution of Combined RADM/RPM- and REMSAD-Derived
Monitor Level Ratios for Annual Average PM10 Concentrations:
2010 Post-CAAA/Pre-CAAA[[[ 42
Figure 4-4 Distribution of Combined RADM/RPM- and REMSAD-Derived
Monitor-Level Ratios for Annual Average PM2.5 Concentrations:
2000 Post-CAAA / 2000Pre-CAAA 42
Figure 4-5 Distribution of Monitor-Level Ratios of Summer S02 Emissions:
2010 Post-CAAA/ 2010 Pre-CAAA 45
Figure 4-6 Distribution of Monitor-Level Ratios of NO Summer Emissions:
2010 Post-CAAA/ 2010 Pre-CAAA 45
Figure 4-7 Distribution of Monitor-Level Ratios of NO2 Summer Emissions:
2010 Post-CAAA/ 2010 Pre-CAAA 46
-------
List of Tables and Figures
Figure 7-1 Annual Economic Welfare Benefit of Mitigating Ozone Impacts
on Commercial Timber: Difference Between Pre- and Post-CAAA 93
Figure 8-1 Monte Carlo Simulation Model Primary Benefits Results
for Target Years - Titles I Through V . [[[ 101
Figure 8-2 Analysis of Contribution of Key Parameters to Quantified Uncertainty '107
Figure A-l Comparison of Pre-CAAA, Post-CAAA, and Trends
VOC Emissions Estimates [[[ A-8
Figure A-2 Comparison of Pre-CAAA, Post-CAAA, and Trends
NOx Emissions Estimates A-10
Figure A-3 Comparison of Pre-CAAA, Post-CAAA, and Trends
SO Emissions Estimates [[[ A-l 0
Figure A-4 Comparison of Pre-CAAA, Post-CAAA, and Trends
CO Emissions Estimates A-ll
Figure A-5 Comparison of Pre-CAAA, Post-CAAA, and Trends
PM10 Emissions Estimates [[[ A-ll
Figure A-6 1990 Primary PM2.5 Emissions by EPA Region A-l3
Figure C-l Schematic Diagram of the Future-year Concentration
Estimation Methodology [[[ C-4
Figure C-2 Difference in Daily Maximum Simulated Ozone Concentration (ppb)
for the 15 July 1995 OTAG Episode Day: 2010 pre-CAAA90
minus base 1990 C-26
Figure C-3 Difference in Daily Maximum Simulated Ozone Concentration (ppb)
for the 15 July 1995 OTAG Episode Day: 2010 post-CAAA90
minus base 1990 [[[C-27
Figure C-4 Difference in Daily Maximum Simulated Ozone Concentration (ppb)
for the 15 July 1995 OTAG Episode Day: 2010 post-CAAA90
minus pre-CAAA90 C-28
Figure C-5 Difference in Daily Maximum Simulated Ozone Concentration (ppb)
for the 8 July 1995 Western U.S. Simulation Day: 2010 pre-CAAA90
minus base 1990 [[[C-29
Figure C-6 Difference in Daily Maximum Simulated Ozone Concentration (ppb)
for the 8 July '1995 Western U.S. Simulation Day: 2010 post-CAAA90
minus base 1990 C-30
Figure C-7 Difference in Daily Maximum Simulated Ozone Concentration (ppb)
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Figure C-lla Distribution of Monitor - Level Ratios for 95th Percentile
Ozone Concentration 2000 Pre-CAAA90/1990 Base-Year C-35
Figure C-lib Distribution of Monitor - Level Ratios for 95th Percentile
Ozone Concentration 2000 Post-CAAA90/1990 Base-Year C-35
Figure C-12a Distribution of Monitor - Level Ratios for 95th Percentile
Ozone Concentration 2010 Pre-CAAA/1990 Base-Year C-36
Figure C-12b Distribution of Monitor - Level Ratios for 95th Percentile
Ozone Concentration 2010 Post-CAAA/1990 Base-Year C-36
Figure C-13a Distribution of Monitor - Level Ratios for 95th Percentile
Ozone Concentration 2000 Post-CAAA/2000 Pre-CAAA C-37
Figure C-13b Distribution of Monitor - Level Ratios for 95th Percentile
Ozone Concentration 2010 Post- C A AA / 2 010 Pre-CAAA C-37
Figure C-'14 80-km RADM Domain C-54
Figure C-15 Comparison of Simulated and Observed Seasonal PM'10 Concentration
(ug/m3) for REMSAD for the Western U.S.: Spring 1990...........................................C-55
Figure C-16 Comparison of Simulated and Observed Seasonal PM10 Concentration
(ug/m3) for REMSAD for the Western U.S.: Summer '1990 ....................................... C-55
Figure C-'17 Comparison of Simulated and Observed Seasonal PM10 Concentration
(ug/m3) for REMSAD for the Western U.S.: Fall 1990 ................................................C-55
Figure C-18 Comparison of Simulated and Observed Seasonal PM10 Concentration
(ug/m3) for REMSAD for the Western U.S.: Winter '1990 .........................................C-55
Figure C-19 Difference in Seasonal Average PM10 Concentration (ug/m3) for the Summer
REMSAD Simulation Period (1-10 July 1990) for 2010: post-CAAA90
minus pre-CAAA90 C-56
Figure C-20 Difference in Seasonal Average PM25 Concentration (ug/m3) for the Summer
REMSAD Simulation Period (1-10 July 1990) for 2010: post-CAAA90
minus pre-CAAA90 [[[C-57
Figure C-21a Distribution of Combined RADM/RPM- and REMSAD - derive d
Monitor-level Ratios for Annual Average PM-'IO Concentration:
2000 Pre-CAAA90 / 1990Base-Year C-58
Figure C-21b Distribution of Combined RADM/RPM- and REMSAD - derive d
Monitor-level Ratios for Annual Average PM-10 Concentration:
2000 Post-CAAA/1990 Base-Year[[[C-58
Figure C-22a Distribution of Combined RADM/RPM- and REMSAD - derive d
Monitor-level Ratios for Annual Average PM-10 Concentration:
2010 Pre-CAAA/1990 Base-Year C-59
Figure C-22b Distribution of Combined RADM/RPM- and REMSAD - derive d
Monitor-level Ratios for Annual Average PM-10 Concentration:
2010 Post-CAAA/1990 Base-Year[[[C-59
Figure C-23a Distribution of Combined RADM/RPM- and REMSAD - derive d
Monitor-level Ratios for Annual Average PM-2.5 Concentration
-------
List of Tables and Figures
Figure C-24a Distribution of Combined RADM/RPM- and REMSAD-derived
Monitor-level Ratios for Annual Average PM-2.5 Concentration:
2010 Pre-CAAA/1990 Base-Year [[[ C-61
Figure C-24b Distribution of Combined RADM/RPM- and REMSAD-derived
Monitor-level Ratios for Annual Average PM-2.5 Concentration
2010 Post-CAAA/1990 Base-Year C-61
Figure C-25a Distribution of Combined RADM/RPM- and REMSAD-derived
Monitor-level Ratios for Annual Average PM-10 Concentration:
2000 Post-CAAA/2000 Pre-CAAA[[[C-62
Figure C-25b Distribution of Combined RADM/RPM- and REMSAD-derived
Monitor-level Ratios for Annual Average PM-10 Concentration:
2010 Post-CAAA/2010 Pre-CAAA C-62
Figure C-26a Distribution of Combined RADM/RPM- and REMSAD-derived
Monitor-level Ratios for Annual Average PM-2.5 Concentration:
2000 Post-CAAA/ 2000 pre-CAAA[[[C-63
Figure C-26b Distribution of Combined RADM/RPM- and REMSAD-derived
Monitor-level Ratios for Annual Average PM-2.5 Concentration
2010 Post-CAAA/2010 Pre-CAAA C-63
Figure C-27 Seasonal Average Deciview for the Summer REMSAD Simulation
Period (1-10 July 1990): Base 1990 (Western U.S. Only) C-67
Figure C-28 Difference in Seasonal Average Deciview for the Summer REMSAD
Simulation Period (1-10 July 1990): 2010 Pre-CAAA90 Minus Base 1990
(Western United States Only)[[[C-68
Figure C-29 Annual Sulfur Deposition '1990 Base Case Scenario C-71
Figure C-30 Annual Nitrogen Deposition 1990 Base Case Scenario ................................................. C-72
Figure C-3'1 Annual Sulfur Deposition 2010 Pre CAAA Scenario C-73
Figure C-32 Annual Sulfur deposition 2010 Post CAAA Scenario [[[ C-74
Figure C-33 Annual Nitrogen Deposition 2010 Pre CAAA Scenario C-75
Figure C-34 Annual Nitrogen deposition in 2010 Post CAAA Scenario ........................................ C-76
Figure C-35 Distribution of Monitor-Level Ratios of Summer S02 Emissions:
2010 Post-CAAA / 2010 Pre-CAAA[[[C-78
Figure C-36 distribution of Monitor-Level Ratios of Summer NO Emissions:
2010 Post-CAAA / 2010 Pre-CAAA [[[C-78
Figure C-37 Distribution of Monitor-Level Ratios of Summer N02 Emissions:
2010 Post-CAAA / 2010 Pre-CAAA[[[C-79
Figure C-38 Distribution of Monitor-Level Ratios of Summer CO Emissions:
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Figure E-5 Chesapeak Bay SAY 1978-1996 [[[ E-32
Figure E-6 Percentage of Acidic Surface Waters in the National Surface
Water Survey Regions[[[ E-36
Figure E-7 Acidification of Freshwater Ecosystems E-37
Figure E-8 Annual Economic Welfare Benefit of Mitigation Ozone Impacts
on Commercial Timber E-49
Figure E-9 U.S. Major Forest Types Affected By Air Pollution-Induced Visual Injuries ......... E-57
Figure G-'l Schematic of Cost and Benefit Analysis of Title VI[[[ G-12
Figure G-2 Annual Human Health Benefits from Sections 604 and 606 (Discounted at 5%).. G-29
Figure G-3 Annual Undiscounted Human Health Benefits of Sections 604 and 606................. G-33
-------
Acronyms and Abbreviations
(iEq microequivalents
(.ig microgram
ACT average cost per ton
AG SIM AGricultural Simulation .Model
AIC Akaike information criterion
AIRS Aerometric Information Retrieval
System
ANC acid neutralizing capacity
ANOVA analysis of variance
AOL) airway obstructive disease
AP-42 EPA's Compilation of Air Pollu-
tion Emission Factors
ATDM aerosol and toxics deposition
module
AQM air quality modeling
AQMS Air Quality Modeling Subcommit-
tee
ATLAS Aggregate Timber Land Assessment
System
b light extinction coefficient
BAAQM.D Bay Area Air Quality Management
District
BACT best available control technology
BAF bioaccumulation factor
BARCT best available retrofit control
technology
BCF bioconcentration factor
BEA Bureau of Economic Analysis
BID background information document
BIES Biogenic Emissions Inventory
System
BLS Bureau of Labor Statistics
BMP best management practice
BNR biological nutrient removal
BS black smoke
OR concentration-response
CAA Clean Air Act
CAAA Clean Air Act Amendments
CAP I Clean Air Power Initiative
CAPMS Criteria Air Pollutant Modeling
System
CARB California Air Resources Board
CAS AC Clean Air Science Advisory Board
CASTNet Clean Air Act Status and Trends
Network
CB chronic bronchitis
CEM continuous emissions monitoring
CES constant elasticity of substitution
C F C chlorofluorocarbon
CFFP Clean Fuel Fleet Program
CGE computable general equilibrium
CI compression ignition
CO carbon monoxide
COH coefficient of haze
CO I cost of illness
COPD chronic obstructive pulmonary
disease
CRC capital recovery cost
CRF capital recovery factor
CTG control technique guideline
CV contingent valuation
dbh diameter at breast height
DDT dichlorodiphenyl-trichloroethane
DOE Department of Energy
dV deciview
E-GAS Economic Growth Analysis System
EC elemental carbon
EGU electrical generating unit
E.MFAC emission factors model
ER emergency room
El5A Environmental Protection Agency
EPS Emissions Processing System
ERCAM Emission Reduction and Cost
Analysis Model
ERL Environmental Research Labora-
tory
FACA Federal Advisory Committee Act
FAPRI Food and Agricultural Policy
Research Institute
FCM Fuel Consumption Model
FDA Food and Drug Administration
FEV forced expiratory volume in one
second
xxv
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
FGD flue gas desulfurization
FHWA Federal Highway Administration
FMVCP Federal .Motor Vehicle Control
Program
FORCARB forest carbon model
FORTRAN formula translation
FR Federal Register
GCVTC Grand Canyon Visibility Transport
Commission
GDP gross domestic product
GIRAS Geographic Information Retrieval
Analysis System
GIS geographic information system
GNP gross national product
GSP gross state product
H + hydrogen ions
ha hectare
HAP hazardous air pollutant
HARVCARB harvested carbon model
HB F C hydrobromofluorocarbons
IIC hydrocarbon
H C F C hydrochlorofluorocarbon
HDDV heavy-duty diesel vehicle
HDGV heavy-duty gasoline vehicle
HD V heavy-duty vehicle
HEES Health and Ecological Effects
Subcommittee
Hg mercury
IIIV-1 human immunodeficiency virus
type one
HNOj nitric acid
HPMS Highway Performance Monitoring
System
HS„0 sulfuric acid
I/M inspection and maintenance
ICI industrial/commercial/institutional
TCD International Classification of
Disease
ID identification code
IMPROVE Interagency Monitoring of
PROtected Environments
1PM Integrated Planning Model
kg kilogram
km kilometer
kWh kilowatt hour
LAER lowest achievable emission rate
lb pound
LDAR leak detection and repair
LDDT light-duty diesel truck
LD.DV light-duty diesel vehicle
LDGT light-duty gasoline truck
LDGV light-duty gasoline vehicle
LEV low emission vehicle
LRS lower respiratory symptom
LTO landing and takeoff operations
m meter
m3 cubic meter
MACT maximum achievable control
technology
MAG Maricopa Association of Govern-
ments
MAGIC Model of Acidification of Ground-
water m Catchments
MC motorcycle
MCF methyl chloroform
MDL method detection limit
MM4 mesoscale model four
MMBtu million British thermal units
MRAD minor restricted activity day
Models-3 Third Generation Air Pollution
Modeling System
MOU memorandum of understanding
.MOBILE mobile source emission factor
model
MPO metropolitan planning organization
MWC municipal waste combustor
MW.I. medical waste incinerator
N nitrogen
NAA nonattainment area
NAAQS National Ambient Air Quality
Standards
NAPAP National Acid Precipitation Assess-
ment Program
NASA National Aeronautics and Space
Administration
N CAR National ("enter for Atmospheric
Research
NCLAN National Crop Loss Assessment
Network
NE northeast
N.EMS National Energy Modeling System
NERC North American Electric Reliabil-
ity Council
NESHAP National Emission Standards for
Hazardous Air Pollutants
XXV i
-------
Acronyms and Abbreviations
NET National Emission 'Trend
NH ammonia
NIIANES National Health and Nutrition
Examination
NIH National Institutes of Health
NMOC non-methane organic compound
N O nitrogen oxide
N Oz nitrogen dioxide
NOx nitrogen oxides
NP national park
NPI National Particulates Inventory
NPP net primary productivity
NPY net present value
NSPS new source performance standard
NSR new source review
NSWS National Surface Waters Survey
NYSDEC New York Department of Environ-
mental Conservation
O ozone
O&M operation and maintenance
OBD onboard diagnostic
O C organic carbon
ODS ozone-depleting substance
OMB Office of Management and Budget
OMS Office of Mobile Sources
OPPE Office of Policy, Planning and
Evaluation
ORIS Office of the Regulatory Informa-
tion System
OSD ozone season daily
OTAG Ozone Transport Assessment
Group
OTC Ozone Transport Commission
OTR Ozone Transport Region
P-i-G plume-m-grid
PAN peroxyacetyl nitrate
Pb lead
PCB polychlorinated biphenyl
PCDD polychlorinated dibenzo-p-dioxin
PCDF polychlorinated dibenzofurans
PCE perchloroethylene
pH logarithm of the reciprocal of
hydrogen ion concentration, a
measure of acidity
PM particulate matter (both PM10 and
pm25)
PM.|()
™2,
PnET
POC
POTW
PPb
ppm
PRYL
PRZM
PSU
QALY
R&D
RACT
RAD
RADM
RELMAP
REMSAD
RE
RFG
RHC
II I A
RFP
R°2
ROP
RPM
RUM
RYP
S
SAB
SAS
SAY
SCAQMD
SCAQS
see
SCR
SEDS
SI
SIC
SIP
so.
xxvii
particulate matter less than or equal
to 10 microns in diameter
particulate matter less than or equal
to 2.5 microns in diameter
Net Photosynthesis and Evapo-
Transpiration model
parameter occurrence code
publicallv owned treatment works
parts per billion
parts per million
percentage relative yield loss
Pesticide Root Zone Model
Pennsylvania State University
quality adjusted life years
research and development
reasonable available control tech-
nology
restricted activity day
Regional Acid Deposition Model
Regional Lagrangian Model of Air
Pollution
Regulatory Modeling System for
Aerosols and Acid Deposition
rule effectiveness
reformulated gasoline
reactive hydrocarbon
regulatory impact analysis
reasonable further progress
peroxy radical
rate of progress
Regional Particulate Model
Random Utility Model
Reid vapor pressure
sulfur
Science Advisory Board
Statistical Analysis Software
submerged aquatic vegetation
South Coast Air Quality Manage-
ment District
South Coast Air Quality Study-
Source Classification Code
selective catalytic reduction
State Energy Data Systems
spark ignition
Standard Industrial Classification
State Implementation Plan
sulfur dioxide
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
SO A secondary organic aerosol
SoCAB South Coast Air Basin
SOCMI synthetic organic chemical manu-
facturing industry
SUM06 sum of hourly ozone concentra-
tions at or above 0.06 ppm
'i'AC total annualized costs
TAF temporal allocation factors
TAMM Timber Assessment Market .Model
TBRP Tampa Bay Estuary Program
TCDD tetrachlorodibenzo-p-dioxin
TEQ toxic equivalency
TLEV transitional low emission vehicle
tpd tons per day
TREGRO tree growth model
TSDF treatment, storage, and disposal
facility
TSP total suspended particulates
[JAM Urban Airshed Model
URS upper respiratory symptoms
USD A United States Department of
Agriculture
ULEV ultra-low emission vehicle
USGS United States Geological Survey
U V ultraviolet
VMT vehicle miles traveled
VNA Yoronoi neighbor averaging
VOC volatile organic compound
VR visual range
YSL value of statistical life
VSLY value of statistical life year
WE FA Wharton Economic Forecasting
Associates
WHO World Health Organization
WLD work-loss days
WT A willingness-to-accept
WT P willingne s s - to -p ay
XO, halogenated peroxy radical
yr year
ZEV zero emission vehicle
-------
Acknowledgments
This project is managed under the direction of
Robert Perciasepe, Assistant Administrator and Rob-
ert D. Brenner, Deputy Assistant Administrator for
the EPA Office of Air and Radiation. The principal
project manager is Jim DeMocker, Senior Policy
Analyst, EPA Office of Air and Radiation/Office of
Policy Analysis and Review. Brian Heninger, EPA
Office of Policy/Office of Economy and Environ-
ment, directed the ecological assessment; and Sam
Napolitano, EPA Office of Air and Radiation/Of-
fice of Atmospheric Programs directed the electric
utility emissions and cost analyses. Robin Dennis,
EPA Office of Research and Development/National
Exposure Research Laboratory directed the RADM-
RPM air quality modeling. A1 McGartland, Direc-
tor of the EPA Office of Economy and Environment
in the Office of Policy, and David Gardiner, former
Assistant Administrator for the Office of Policy pro-
vided guidance and support.
Many EPA staff contributed or reviewed portions
of this document, including Bryan Hubbcll, John
Bachmann, Ron Evans, Rosalina Rodriguez, Scott
Mathias, Ann Watkins, Rona Birnbaum, Karen Mar-
tin, Doris Price, Dm sill a Hufford, Jeff Cohen, Joe
Somers, Carl Mazza, Brett Snyder, and Tom Gillis.
A number of contractors developed key elements
of the analysis and supporting documents. Jim
Neumann of Industrial Economics, Incorporated
managed the overall integration and coordination of
the analytical work and documentation and also made
considerable substantive analytical contributions.
Other contractor members of the 812 Project Team
included Bob Unsworth, Henry Roman, Jared
Hardner, Naomi Kleckner, Nick Live say, Lauren
Fusfeld, Andre Cap, Stephen Everett, Jon Discher,
and Mike Hester of Industrial Economics, Incorpo-
rated; Leland Deck, Ellen Post, Lisa Akcson, Ken-
neth Davidson, and Don McCubbm of Abt Associ-
ates; Sharon Douglas, John Langstaff, Robert
Iwamiya, Belle Hudischewsky, and John Calcagni of
ICF Incorporated, and John Blaney of ICF Consult-
ing; and Jim Wilson, Erica Laich, and Dianne Crocker
of Pechan-Avanti Associates.
Science Advisory Board review of tins report is
supervised by Donald G. Barnes, Director of the SAB
Staff. The Designated Federal Official for the SAB
reviews is Angela Nugent. Other SAB staff who as-
sisted 111 the coordination of SAB reviews include Jack
Fowle, Robert Flaak, and Jack Kooyoomjian. Diana
Pozun provided administrative support to the SAB.
The SAB Council is chaired by Maureen Crop-
per of the World Bank. SAB Council members serv-
ing during the final review of this report include A.
Myrick Freeman of Bowdom College, Gardner
Brown, Jr. of the University of Washington, Paul
Liov of the Robert Wood Johnson School of Medi-
cine, Paulette Middleton of the RAND Center for
Environmental Sciences and Policy, Donald Fuller-
ton of the University of Texas — Austin, Lawrence
Goulder of Stanford University, Jane Hall of Cali-
fornia State University Fullerton, Charles Kolstad
of the University of California at Santa Barbara, and
Lester Lave of Carnegie-Mellon University. Alan
Krupnick of Resources for the Future served as a
Consultant to the Council. In addition, several mem-
bers of the SAB Council whose terms expired during
the development of the study provided valuable ad-
vice and ideas in the early stages of project design and
implementation. These former members include
Richard Schmalensee of MIT, William Nordhaus of
Yale University, Paul Portney of Resources for the
Future, Kip Viscusi of Harvard University, Ronald
Cummmgs of Georgia State University, Thomas
Tietenberg of Colby College, Wallace Oates of the
University of Maryland, Wayne Kachel of MHLE As-
sociates, Robert .Mendelsohn of Yale University, and
Daniel Dudek of the Environmental Defense Fund.
William Smith, a liaison to the Council from the SAB
Environmental Processes and Effects Committee also
provided valuable advice regarding the ecological as-
sessment.
xxix
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
The SAB Council is supported by two technical
subcommittees. The first of these subcommittees,
the Health and Ecological Effects Subcommittee is
chaired by Paul Liov. Members who participated in
the final review of this report included Morton
Lippmann of New York University Medical Center,
George T. Wolff of General .Motors, A. Myrick Free-
man, Timothy Larson of the University of Washing-
ton, Joseph Meyer of the University of Wyoming,
Robert Rowe of Stratus Consulting, George Taylor
of George "Mason University, jane Hall, "Michael
Kleinman of the University of California at Irvine,
and Carl Shy of the University of North Carolina at
Chapel Hill. Several former members who provided
valuable advice in the early stages of the study in-
clude Bernard Weiss of the University of Rochester
.Medical Center, David V. Bates of the University of
British Columbia, Gardner Brown, and Lester Lave.
The second technical subcommittee, the Air
Quality Modeling Subcommittee is chaired by
Paulette Middleton. Members serving during the fi-
nal review of this report include Philip Hopke of
Clarkson University, James II". Price, Jr. of the Texas
Natural Resource Conservation Commission, Harvey
Jeffries of the University of North Carolina — Chapel
Hill, Timothy Larson, and Peter Mueller of the Elec-
tric Power Research Institute. A former member who
helped guide the analysis in its early stages was George
T. Wolff.
The project managers wish to convey special ac-
knowledgment and appreciation for the valuable con-
tributions of A. Myrick Freeman. As a charter mem-
ber of the Council and as 'Vice Chair of the Health
and Ecological Effects Subcommittee, Dr. Freeman
provided wise and excellent counsel throughout the
entire course of SAB review of both this study and
the preceding retrospective study.
This report could not have been produced with-
out the support of key administrative support staff.
The project managers are grateful to Barbara Morris,
Nona Smoke, Eunice Javis, Gloria Booker, and
Wanda Farrar for their timely and tireless support
on this project.
XXX
-------
Introduction
Background and Purpose
Section 812 of the 1990 Clean Air Act Amend-
ments requires the EPA to develop periodic Reports
to Congress that estimate the benefits and costs of
the Clean Air Act (CAA). The first report EPA
created under this authority, The Benefits and Costs
of the Clean Air Act: 1970 to 1990, was published and
conveyed to Congress in October 1997. This retro-
spective analysis comprehensively assessed benefits
and costs of requirements of the 1970 Clean Air Act
and the 1977 Amendments, up to the passage of the
Clean Air Act Amendments of 1990. The results of
the retrospective analysis showed that the nation's
investment in clean air was more than justified by
the substantial benefits that were gained in the form
of increased health, environmental quality, and pro-
ductivity. The aggregate benefits of the CAA dur-
ing the 1970 to 1990 period exceeded costs by a fac-
tor of 10 to 100 times.
Before the retrospective analysis was complete,
we began the process of assessing the prospective
benefits and costs of the Clean Air Act Amendments
(CAAA), covering the period 1990 to 2010. This
report, the first of a series that we plan to produce
every two years, is the result of our prospective analy-
sis of the 1990 Amendments.
Similar to the retrospective analysis, this docu-
ment has one primary and several secondary objec-
tives. The main goal is to provide Congress and the
public with comprehensive, up-to-date information
on the CAAA's social costs and benefits, including
health, welfare, and ecological benefits. Data and
methods derived from the retrospective analysis have
already been used to assist policy-makers in refining
clean air regulations over the last two years, and we
hope the information continues to prove useful to
Congress during future Clean Air Act reauthoriza-
tions. Beyond the statutory goals of section 812,
11
EPA also intends to use the results of this study to
help support decisions on future investments in air
pollution research. In addition, lessons learned in
conducting this first prospective will help better tar-
get efforts to improve the accuracy and usefulness
of future prospective analyses.
Relationship of This Report
to Other Regulatory Analyses
The Clean Air Act Amendments of 1990 aug-
ment the significant progress made in improving the
nation's air quality through the original Clean Air
Act of 1970 and its 1977 amendments. The amend-
ments built off the existing structure of the original
Clean Air Act, but went beyond those requirements
to tighten and clarify implementation goals and tim-
ing, increase the stringency of some federal require-
ments, revamp the hazardous air pollutant regula-
tory program, refine and streamline permitting re-
quirements, and introduce new programs for the
control of acid rain and stratospheric ozone depleters.
Because the 1990 Amendments represent an addi-
tional improvement to the nation's existing clean
air program, the analysis summarized in this report
was designed to estimate the costs and benefits of
the 1990 CAAA incremental to those costs and ben-
efits assessed in the retrospective analysis. In eco-
nomic terminology, this report addresses the mar-
ginal costs and benefits of the 1990 CAAA. Our
intent is that this report and its predecessor, the ret-
rospective analysis, together provide a comprehen-
sive assessment of current and expected future clean
air regulatory programs and their costs and benefits.
Because of the time and resources necessary to
conduct this type of comprehensive prospective as-
sessment, however, and the ongoing refinements in
Clean Air Act regulatory programs, the estimates
presented in this report do not reflect some recent
1
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
major developments in EPA's clean air program.
The prospective analysis, for example, does not cap-
ture the benefits and costs of EPA's recent revision
of the particulate matter and ozone National Ambi-
ent Air Quality Standards (NAAQS), the recently
proposed Tier II tailpipe standards, or the recently
finalized regional haze standards. Neither costs nor
benefits of those actions are reflected in the estimates
presented here. In most cases, Regulatory Impact
Analyses (RIAs) for those actions did incorporate
the section 812 prospective Post-CAAA scenario as
their starting point, or baseline, from which the ac-
tions were assessed, and in most respects the RIAs
used a methodology- consistent with that used here.1
As a result, cost and benefit estimates presented in
those RIAs can be considered incremental to the
primary estimates presented in this document.
In addition to omitting these actions from the
assessment, this first prospective analysis required
locking in a set of emissions reductions to be used in
subsequent analyses at a relatively early date (late
1996), and as a result we were compelled to forecast
the implementation outcome of several pending pro-
grams. The most important of these was the then-
ongoing Ozone Transport Assessment Group
(OTAG) recommendations for achieving regional-
scale reductions of emissions of ground-level ozone
precursors. The NOx control program incorporated
in the Post-CAAA scenario may not reflect the NOx
controls that are actually implemented in a regional
ozone transport rule. We acknowledge and discuss
these types of discrepancies and their impact on the
outcome of our analysis in the document.
Finally, despite our efforts to comprehensively
evaluate the costs and benefits of all provisions of
the ("lean Air Act and its Amendments, there re-
main a few7 categories of effects that are not addressed
by either the retrospective or prospective analyses.
For example, this first prospective analysis does not
assess the effect of CAAA provisions on lead expo-
sures, primarily because the 1990 Amendments do
' There arc minor differences in the assumptions used to
construct the Post- CAAA scenario for this analysis and the
baseline used in the PM and ozone NAAQS RTA. For example,
the RIA baseline incorporates the effects of 7- and 10-year MACT
rules that are not reflected here, because of the timing of the
two analyses, and the RIA used a 95 percent rule-eliecliveness
assumption. In most respects, however, the analyses are com-
patible.
not include major new provisions for the control of
lead emissions. The vast majority of lead emissions
sources present in 1970 were addressed by programs
initiated under the original Clean Air Act and the
1977 Amendments; evaluation of the costs and health
benefits of these programs were important elements
of the retrospective analysis. In the retrospective,
however, we were unable to quantify the potentially
substantial ecological benefits of controls on lead
emissions. While this first prospective analysis re-
flects a significantly greater investment in quantify-
ing ecological effects, for the reason stated above we
did not assess the ecological effects of lead in this
analysis either. As a result, the ecological effects of
this persistent pollutant, past emissions of which may
continue to be released from soils for many years,
are not captured by either the retrospective or pro-
spective analyses. In addition, lead previously de-
posited to soils may be re-entrained in the air as road
dust, dust plumes from construction excavations, and
other particulate matter emission processes subject
to 1990 CAAA controls. Reductions in this re-en-
trainmen t of, and potential exposure to, pre-1990
emitted lead due to post-1990 control programs,
however, are not reflected in either the section 812
retrospective (1970 to 1990) or prospective (1990 to
2010) benefit analyses.
Requirements of the 1990
Clean Air Act Amendments
The first prospective analysis, despite the limi-
tations discussed above, presents a comprehensive
estimate of costs and benefits of all titles of the 1990
Clean Air Act Amendments. The 1990 Amendments
consist of the following eleven titles:
• Title I. Establishes a detailed and graduated
program for the attainment and maintenance
of the National Ambient Air Quality Stan-
dards.
• Title II. Regulates mobile sources and es-
tablishes requirements for reformulated gaso-
line and clean fuel vehicles.
• Title III. Expands and modifies regulations
of hazardous air pollutant emissions; and
establishes a list of 189 hazardous air pollut-
ants to be regulated.
2
-------
Chapter 1: Introduction
• Title IY. Establishes control programs for
reducing acid ram precursors.
• Title V. Requires a new permitting system
for primary sources of air pollution.
• Title VI. Limits emissions of chemicals that
deplete stratospheric ozone.
• Title VII. Presents new provisions for en-
forcement.
• Titles VIII through XI. Establishes miscel-
laneous provisions for issues such as disad-
vantaged business concerns, research, train-
ing, new regulation of outer continental shelf
sources, and assistance for people who lose
their jobs as a result of the Clean Air Act
Amendments.
As part of the requirements under Title VIII,
section 812 of the Clean Air Act Amendments of
1990 requires the EPA to analyze the costs and ben-
efits to human health and the environment that are
attributable to the Clean Air Act. In addition, sec-
tion 812 directs EPA to measure the effects of this
statute on economic growth, employment, produc-
tivity, cost of living, and the overall economy of the
United States.
Analytical Design and Review
Target Variable
'The prospective analysis compares the overall
health, welfare, ecological and economic benefits of
the 1990 Clean Air Act Amendment programs to
the costs of these programs. By examining the over-
all effects of the Clean Air Act, this analysis comple-
ments the Regulatory Impact Analyses (RIAs) de-
veloped by EPA over the years to evaluate individual
regulations. Resources were used more efficiently by
recognizing that these RIAs, and other EPA analy-
ses, provide complete information about the costs
and benefits of specific rules. Within this analysis,
costs can be reliably attributed to individual pro-
grams, but the broad-scale approach adopted in the
prospective study precludes reliable re-estimation of
the benefits of individual standards or programs.
Similar to the retrospective benefits analysis, this
study calculates the change in incidences of adverse
effects implied by changes in ambient concentrations
of air pollutants. However, pollutant emissions re-
ductions achieved contribute to changes in ambient
concentrations of those, or secondarily formed, pol-
lutants in ways that are highly complex, interactive,
and often nonlinear. Therefore, benefits cannot be
reliably matched to provision-specific changes in
emissions or costs.
Focusing on the broader target variables of over-
all costs and overall benefits of the Clean Air Act,
the EPA Project 'Team adopted an approach based
on construction and comparison of two distinct sce-
narios: a "Pre-CAAA" and a "Post-CAAA" scenario.
The Pre-CAAA scenario essentially freezes federal,
state, and local air pollution controls at the levels of
stringency and effectiveness which prevailed in 1990.
The Post-CAAA scenario assumes that all federal,
state, and local rules promulgated pursuant to, or in
support of, the 1990 CAAA were implemented. This
analysis then estimates the differences between the
economic and environmental outcomes associated
with these two scenarios. For more information on
the scenarios and their relationship to historical
trends, see Chapter 2 and Appendix A of this docu-
ment.
Key Assumptions
Similar to the retrospective analysis, we made
two key assumptions during the scenario design pro-
cess to avoid miring the analytical process in endless
speculation. First, as stated above, we froze air pol-
lution controls at 1990 levels throughout the Pre-
CAAA scenario. Second, we assumed that the geo-
graphic distributions of population and economic
activity remain the same between the two scenarios,
although these distributions do change over time
under both scenarios to reflect expected patterns of
high and low population and economic growth
across the country.
The first assumption is an obvious simplifica-
tion. In the absence of the 1990 CAAA, one would
expect to see some air pollution abatement activity,
either voluntary or due to state or local regulation.
It is conceivable that state and local regulation would
have required air pollution abatement equal to -or
even greater than— that required by the 1990 CAAA;
3
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
particularly since some states, most notably Califor-
nia, have in the past done so. If one were to assume
that state and local regulations would have been
equivalent to 1990 CAAA standards, then a cost-
benefit analysis of the 1990 CAAA would be a mean-
ingless exercise since both costs and benefits would
equal zero. Any attempt to predict how7 states' and
localities' regulations would have differed from the
1990 CAAA would be too speculative to support
the credibility of the ensuing analysis. Instead, the
Pre-CAAA scenario has been structured to reflect
the assumption that states and localities would not
have invested further in air pollution control pro-
grams after 1990 in the absence of the federal CAAA.
Thus, this analysis accounts for all costs and ben-
efits of air pollution control from 1990 to
2010 and does not speculate about the frac-
tion of costs and benefits attributable exclu-
sively to the federal CAAA. Nevertheless,
it is important to note that state and local
governments and private initiatives are re-
sponsible for a significant portion of these
total costs and total benefits. In the end,
the benefits of air pollution controls result
from partnerships among all levels of gov-
ernment and with the active participation
and cooperation of private entities and indi-
viduals.
The second assumption concerns chang-
ing demographic patterns in response to air
pollution. In the hypothetical Pre-CAAA
scenario, air quality is worse than the actual
1990 conditions and the projected air qual-
ity in the Post-CAAA scenario. It is pos-
sible that under the Pre-CAAA scenario
more people, relative to the Post-CAAA
case, would move away from the most
heavily polluted areas. Rather than specu-
late on the scale of population movement,
the analysis assumes no differences in demo-
graphic patterns between the two scenarios.
Similarly, the analysis assumes no differences
between the two scenarios with respect to
the spatial pattern of economic activity.
in detail later in this report. These six steps, listed in
order of completion, are:
(1) emissions modeling
(2) direct cost estimation
(3) air quality modeling
(4) health and environmental effects estimation
(5) economic valuation
(6) results aggregation and uncertainty charac-
terization
Figure 1-1 summarizes the analytical sequence
used to develop the prospective results; we describe
the analytic process in greater detail below.
Figure 1-1
Analytic Sequence for
First Section 812 Prospective Analysis
Analytic Sequence
The analysis comprises a sequence of six
basic steps, summarized below and described
Benefits
Analysis
Cost
Analysis
Analytic Design
Scenario
Development
Valuation
Physical
Effects
Direct Cost
Estimation
Emissions
Profile
Development
Air Quality
Modeling -
Criteria Pollutants
Comparison of Benefits
and Costs
4
-------
Chapter 1: Introduction
The first step of the analysis is the estimation of
the effect of the 1990 CAAA 011 emissions sources.
We generated emissions estimates through a three
step process: (1) construction of an emissions inven-
tory for the base year (1990); (2) projection of emis-
sions for the Pre-CAAA case for two target years,
2000 and 2010, assuming a freeze on emissions con-
trol regulation at 1990 levels and continued economic
progress, consistent with sector-specific Bureau of
Economic Analysis economic activity projections;
and (3) construction of Post-CAAA estimates for the
same two target years, using the same set of economic
activity projections used in the Pre-CAAA case but
with regulatory stringency, scope, and timing con-
sistent with EPA's CAAA implementation plan (as
of late 1996). The analysis reflects application of
utility and other sector-specific emissions models
developed and used in various offices of EPA's Of-
fice of Air and Radiation. These emissions models
provide estimates of emissions of six criteria air pol-
lutants2 from each of several key emitting sectors.
We provide more details in Chapter 2 and Appen-
dix A.
The emissions modeling step is a critical compo-
nent of the analysis, because it establishes consistency
between the subsequent cost and benefit estimates
that we develop. Estimates of direct compliance costs
to achieve the emissions reductions estimated in the
first step are generated as either an integral or subse-
quent output from the emissions estimation mod-
els, depending on the model used. For example, the
Integrated Planning .Model used to estimate emissions
and compliance costs for the utility sector develops
an optimal allocation of reductions of sulfur and ni-
trogen oxides taking into account the regulatory flex-
ibility inherent in the Title IY trading schemes for
emissions allocations. In a few cases, for example
the Title V permitting requirements, we estimate
public and private costs incurred to implement the
2 The six pollutants are particulate matter (separate esti-
mates for each of PM|f and PM sulfur dioxide (SO,), nitro-
gen oxides (NOJ, carbon monoxide (CO), volatile organic com-
pounds (VOCs), and ammonia (NH,). One of die CAA criteria
pollutants, ozone (O^), is formed in die atmosphere through the
interaction of sunlight and ozone precursor pollutants such as
NOv and VOCs. Ammonia is not a criteria pollutant, but is an
important input to die air quality modeling step because it af-
fects secondary particulate formation. The sixth criteria pollut-
ant, lead (Pb), is not included in this analysis since airborne
emissions of lead were virtually eliminated by pre-1990 Clean
Air Act programs.
regulatory requirements through analysis of the rel-
evant RIAs conducted to support promulgation of
the rules.
Emissions estimates also form the first step in
estimating benefits. After the emissions inventories
are developed, they are translated into estimates of
air quality conditions under each scenario. Given
the complexity, data requirements, and operating-
costs of state-of-the-art air quality models, and the
project's resource constraints, the EPA Project Team
adopts simplified, linear scaling approaches for some
gaseous pollutants. However, for particulate mat-
ter, ozone, and other air quality conditions that in-
volve substantial non-linear formation processes and/
or long-range atmospheric transport and transfor-
mation, the EPA Project Team invests the time and
resources needed to use more sophisticated model-
ing systems. For example, we exercise EPA's Re-
gional Acid Deposition Model/Regional Particulate
Model (RADM/RPM) to estimate secondarily-
formed particulate matter in the eastern U.S.
Up to this point of the analysis, modeled condi-
tions and outcomes establish the Pre-CAAA and
Post-CAAA scenarios. However, at the air quality
modeling step, the analysis returns to a foundation
based on actual historical conditions and data. Spe-
cifically, actual 1990 historical air quality monitor-
ing data are used to define the baseline conditions
from which the Pre-CAAA and Post-CAAA sce-
nario air quality projections are constructed. We
derive air quality conditions under the Pre-CAAA
scenario by scaling the historical data adopted for
the base-year (1990) by the ratio of the modeled Pre-
CAAA and base-year air quality. We use the same
approach to estimate future-year air quality for the
Post-CAAA scenario. This method takes advantage
of the richness of the monitoring data on air qual-
ity, provides a realistic grounding for the benefit
measures, and yet retains analytical consistency by
using the same modeling process for both scenarios.
The outputs of this step of the analysis are profiles
for each pollutant characterizing air quality condi-
tions at each monitoring site in the lower 48 states.
The Pre-CAAA and Post-CAAA scenario air
quality profiles serve as inputs to a modeling system
that translates air quality to physical outcomes (e.g.,
mortality, emergency room visits, or crop yield
5
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
losses) through the use of concentration-response
functions. Scientific literature 011 the health and
ecological effects of air pollutants provides the source
of these concentration-response functions. At this
point, we derive estimates of the differences between
the two scenarios in terms of incidence rates for a
broad range of human health and other effects of air
pollution by year, by pollutant, and by geographic
area.
In the next step, we use economic valuation
models or coefficients to estimate the economic value
of the reduction in incidence of those adverse effects
amenable to monetization. For example, a distribu-
tion of unit values derived from the economic litera-
ture provides estimates of the value of reductions in
mortality risk. In addition, we compile and present
benefits that cannot be expressed in economic terms.
In some cases, we calculate quantitative estimates of
scenario differences in the incidence of a
nonmonetized effect. In many cases, available data
and techniques are insufficient to support anything
more than a qualitative characterization of the change
in effects.
Next, we compare costs and monetized benefits
to provide our primary estimate of the net economic
benefits of the 1990 CAAA and associated programs,
and a range of estimates around that primary esti-
mate reflecting quantified uncertainties associated
with the physical effects and economic valuation
steps. The monetized benefits used in the net ben-
efit calculations reflect only a portion of the total
benefits due to limitations in analytical resources,
available data and models, and the state of the sci-
ence. For example, in many cases we are unable to
quantify or monetize the potentially large benefits
of air pollution controls that result from protection
of the health, structure, and function of ecosystems.
In addition, although available scientific studies dem-
onstrate clear links between air quality changes and
changes in many human health effects, the available
studies do not always provide the data needed to
quantify and/or monetize some of these effects.
Finally, we present a limited set of alternative
benefit estimates which reflect methods, models, or
assumptions that differ from those we used to de-
rive the primary net benefit estimate. We also quan-
tify some of the uncertainties surrounding these al-
ternative estimates. In addition, beyond those vari-
ables for which alternative results are estimated, we
conduct sensitivity analyses for a number of vari-
ables that may influence the primary net benefit es-
timate. The primary estimate and the range around
this estimate, however, reflect our current interpre-
tation of the available literature; our judgments re-
garding the best available data, models, and model-
ing methodologies; and the assumptions we consider
most appropriate to adopt in the face of important
uncertainties.
In addition, throughout the report at the end of
the chapter we summarize the major sources of un-
certainty for each analytic step. Although the im-
pact of many of these uncertainties cannot be quan-
tified, we qualitatively characterize the magnitude
of effect on our net benefit results by assigning one
of two classifications to each source of uncertainty:
potentially major factors could, m our estimation,
have effects of greater than five percent of the total
net benefits; and probably minor factors likely have
effects less than five percent of total net benefits.
Review Process
The CAA requires RPA to consult with an out-
side panel of experts during the development and
interpretation of the 812 studies. This panel of ex-
perts was organized in 1991 under the auspices of
EPA's Science Advisory Board (SAB) as the Advi-
sory Council on Clean Air Act Compliance Analy-
sis (hereafter, the Council). Organizing the review
committee under the SAB ensured that highly quali-
fied experts would review the section 812 studies in
an objective, rigorous, and publicly open manner
consistent with the requirements and procedures of
the Federal Advisory Committee Act (FACA).
Council review of the present study began in 1993
with a review of the analytical design plan. Since
the initial June 1993 meeting, the Council has met
many times to review proposed data, proposed meth-
odologies, and interim results. While the full Coun-
cil retains overall review responsibility for the sec-
tion 812 studies, some specific issues concerning
physical effects and air quality modeling were re-
ferred to subcommittees comprised of both Council
members and members of other SAB committees.
The Council's Health and Ecological Effects Sub-
committee (L1EES) met several times and provided
6
-------
Chapter 1: Introduction
its own review findings to the full Council. Simi-
larly, the Council's Air Quality Modeling Subcom-
mittee (AQMS) held ui-person and teleconference
meetings to review methodology proposals and
modeling results and conveyed its review recommen-
dations to the parent committee.
An interagency review was conducted, during
which a number of analytical issues were discussed.
Conducting a benefit/cost analysis of a major stat-
ute such as the Clean Air Act requires scores of meth-
odological decisions. Many of these issues are the
subject of continuing discussion within the economic
and policy analysis communities and within the
Administration. Key issues include the treatment
of uncertainty in the relationship between particu-
late matter exposure and mortality; the valuation of
premature mortality; the treatment of tax interac-
tion effects; the assessment of stratospheric ozone
recovery; and the treatment of ecological and wel-
fare effects. These issues could not be resolved within
the constraints of this review. Thus, this report re-
flects the findings of the EPA and not necessarily-
other agencies of the Administration.
Report Organization
The remainder of the main text of this report
summarizes the key methodologies and findings our
prospective study.
• Chapter 2 summarizes emissions modeling
and key elements of the regulatory scenarios.
• Chapter 3 discusses the direct cost estima-
tion.
• Chapter 4 presents the air quality modeling
methodology and sample results.
• Chapter 5 describes the approaches used and
principal results obtained through the hu-
man health effects estimation process.
• Chapter 6 describes the human health effects
economic valuation methodology and re-
sults.
• Chapter 7 summarizes the ecological and
other welfare effects analyses, including as-
sessments of commercial timber, agriculture,
visibility, and other categories of effects.
• Chapter 8 presents the aggregated results of
the cost and benefit estimates and describes
and evaluates important uncertainties in the
results.
Additional details regarding the methodologies
and results are presented in the appendices and in
the referenced supporting documents.
• Appendix A provides additional detail on the
sector-specific emissions modeling effort.
• Appendix B covers the direct costs.
• Appendix C provides details of the air qual-
ity models used and results obtained.
• Appendix 13 presents the human health ef-
fects estimation methodology and results.
• Appendix E describes the ecological benefits
estimation methods and results.
• Appendix F presents the agricultural benefits
estimation methodology and results.
• Appendix G provides details of the strato-
spheric ozone analysis.
• Appendix H describes the methods and as-
sumptions used to value quantified effects
of the CAA in economic terms.
• Appendix 1 describes areas of research which
may increase comprehensiveness and/or re-
duce uncertainties in effect estimates for fu-
ture assessments.
7
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
f This page left blank intentionally.7
8
-------
Emissions
Estimation of pollutant emissions, a key com-
ponent of this prospective analysis, serves as the start-
ing point for subsequent benefit and cost estimates.
We focused the emissions analysis on six major pol-
lutants: volatile organic compounds (VOCs), nitro-
gen oxides (NOJ, sulfur dioxide (SO,), carbon mon-
oxide (CO), particulate matter with an aerodynamic
diameter of 10 microns or less (PM10), and fine par-
ticulate matter (PM ^.1 For each of these pollut-
ants we projected 1990 emissions to the years 2000
and 2010 under two different scenarios: a) the Pre-
CAAA scenario which assumes no additional con-
trol requirements would be implemented beyond
those in place when the 1990 Amendments were
passed; and b) the ~Post-CAAA scenario which incor-
porates the effects of controls authorized by the 1990
Amendments. We compare the emissions estimates
under each of these scenarios to forecast the effect
of the CAAA requirements on future emissions.
This chapter consists of four sections. The first
section provides an overview of our approach for
developing the Pre- and Post-CAAA control sce-
narios and projecting emissions from 1990 levels to
2000 and 2010. The second section summarizes our
emissions projections for the years 2000 and 2010
and presents our estimates of changes in future emis-
sions resulting from the implementation of the 1990
Amendments. The third section compares these re-
sults with other estimates that are based upon more
1 We also estimated ammonia (NH.) emissions. NH, in-
fluences the formation of secondary PM (PM formed as a result
of atmospheric chemical processes). We used Nil, emissions
estimates as an input during the air quality modeling phase of
the prospective analysis when estimating future-year ambient
PM concentrations. However, we did not examine the human
health and environmental effects of exposure to NH . In addi-
tion to NH;, we also estimated mercury (TTg) emissions. We
qualitatively evaluated the effects of TTg emissions on ecologi-
cal systems, but we did not examine the impact of Hg on hu-
man health. We did not estimate the effect of the CAAA on
lead (Pb) emissions. By 1990 most major airborne Pb emission
sources were already controlled and the CAAA has minimal
additional impact on Pb emissions.
U
recent emissions data. Finally, we conclude this chap-
ter with a summary of the key uncertainties associ-
ated with estimating emissions.
Overview Of Approach
We projected emissions for five major source
categories: industrial point sources, utilities, nonroad
engines/vehicles, motor vehicles, and area sources
(see Table 2-1).2 The basic method involves esti-
mating emissions in the 1990 base-year, adjusting
the base-year emissions to reflect projected growth
in the level of pollution-generating activity by 2000
and 2010 in the absence of additional CAAA require-
ments, and modifying these projections to reflect
future-year control assumptions. The resulting esti-
mates depend largely upon three factors: the method
for selecting the base-year inventory, the indicators
used to forecast growth and the effectiveness of fu-
ture controls, and the specific regulatory programs
incorporated in the Pre- and Post-CAAA scenarios.
We constructed the base-year inventory using
1990 emissions levels. For all of the air pollutants
examined in this analysis except particulate matter,
we selected emissions levels from Version 3 of the
National Particulates Inventory (NPI) to serve as the
baseline. This inventory consists of emissions data
compiled primarily by the National Acid Precipita-
tion Assessment Program (NAPAP), EPA's Office
of Mobile Sources (OMS), and the Federal Highway
Administration (FHWA). For both PM,5 and PMin,
however, we updated NPI estimates to incorporate
changes in the methodology used to calculate fugi-
tive dust emissions. Adoption of this new technique,
also used to develop EPA's National Emission Trend
2 We estimated utility and industrial point source emis-
sions at die plant/facility level. We estimated nonroad engine/
vehicle, motor vehicle, and area source emissions at the county
level.
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 2-1
Major Emissions Source Categories
Source Category Examples
Industrial Point Sources boilers, cement kilns, process heaters, turbines
Utilities electricity producing utilities
Nonroad Engines/Vehicles aircraft, construction equipment, lawn and garden equipment,
locomotives, marine engines
Motor Vehicles buses, cars, trucks (sources that usually operate on roads and
highways)
Area Sources agricultural tilling, dry cleaners, open burning, wildfires
(NET) PM,. and PM.|() inventory, leads to lower
estimates of fugitive dust emissions and therefore of
overall primary PM.3
Once we established the base-year inventory, we
projected emissions to the years 2000 and 2010, ac-
counting for the influences expected to cause future
emissions to differ from 1990 levels. For all but util-
ity sources, we rely on an emissions analysis using
the Emissions Reduction and Cost Analysis Model
(ERCAM) which incorporates the effects of the level
of pollution-generating activity and the stringency
and success of regulations designed to protect air
quality. In this analysis, we view changes in eco-
nomic growth as an important indicator of future
activity levels and thus, future emissions. We used
1995 Bureau of Economic Analysis (BEA) Gross
State Product (GSP) projections to forecast the
growth of emissions from industrial point sources.
We relied on BEA GSP projections as well as data
on BEA predicted changes in population to estimate
future emissions from nonroad and area sources.4
We used BEA population growth as an indicator of
the increase in nonroad emissions from recreational
marine vessels, recreational vehicles, and lawn/gar-
den equipment as well as an indicator of the increase
in area source solvent emissions (e.g., VOC emis-
sions from dry cleaners). For motor vehicle sources,
we estimated the growth in activity based primarily
on the projected increase in vehicle miles traveled
(YM'T). We develop future YM'T estimates using
the EPA MOBILE fuel consumption model.
3 Prim my PM consists of directly emitted particles such as
wood smoke and road dust. Secondary PM forms in the atmo-
sph ere as a result of atmospheric chemical reactions.
4 The growth forecast for area source agricultural tilling is
based on projections ol acres planted, not BEA GSP and popu-
lation projections.
We estimated the impact of CAAA regulations
oil industrial point source, nonroad, motor vehicle,
and area source emissions based on expected control
efficiency and rule effectiveness. Control efficiency
represents the percentage reduction in emissions
anticipated as a result of the implementation of the
CAAA, assuming full compliance and successful
operation of all control mechanisms. The rule ef-
fectiveness factor accounts for equipment malfunc-
tion, non-compliance, and other circumstances that
influence the overall effectiveness of air pollution
regulations. We selected a rule effectiveness of 80
percent as the standard for this analysis which we
applied to stationary source NOj and VOC con-
trols.5 Rule effectiveness was not calculated for mo-
bile source controls as an adjustment factor separate
from the emissions rates estimated for the various
vehicle classes.
To estimate future utility source emissions, we
relied on the Integrated Planning Model (IPM). This
optimization model forecasts, for the 48 contiguous
states and the District of Columbia, emissions from
all existing utility power generation units, as well as
from independent power producers and other co-
generation facilities that sell wholesale power and
are included in the North American Electric Reli-
ability Council (NERQ data base for reliability plan-
ning. The model considers future capacity additions
by both utilities and independent power producers
which might cause an increase in emissions. In addi-
tion, the model is capable of producing baseline air
3 At the time we selected the general rule effectiveness for
use in this analysis, 80 percent was the standard factor applied in
air pollution modeling. More recent analyses have used higher
rule effectiveness values. If a higher rule effectiveness value had
been used in this analysis, emissions reduction estimates would
be larger and the estimated benefits associated with air quality
improvements would be greater.
10
-------
Chapter 2: Emissions
emissions forecasts and estimates of air emissions
levels under various control options at the national
and NERC regional and subregional level. We used
IPM to estimate base-year (1990) utility source emis-
sions and to project future-year (2000 and 2010)
emissions under both the Pre- and Post-CAAA sce-
narios.
Using emissions analysis or IPM, we estimated
future emissions for each of the five major source
categories under both the Pre- and Post-CAAA sce-
narios. While the selection of the base-year inven-
tory, emission growth factors, and rate of regula-
tory effectiveness all influence the emissions projec-
tions, the difference between Pre- and Post-CAAA
estimates is primarily determined by the difference
in control assumptions incorporated in the two pro-
jection scenarios.
Scenario Development
We developed two contrasting emissions con-
trol scenarios, the Pre-CAAA scenario and the Post-
CAAA scenario. The Pre-CAAA scenario maintains
the air pollution regulatory requirements which ex-
isted in 1990 through the 2000 and 2010 analytical
period and serves as a baseline against which we
measure the changes in emissions projected under
the Post-CAAA scenario.6 This latter scenario as-
sumes the implementation of the 1990 Clean Air Act
Amendments and incorporates the influences of the
following provisions:
• Title I VOC and NO reasonably available
control technology (RACT) and reasonable
further progress (RFP) requirements for
ozone nonattainment areas;
• Title II motor vehicle and nonroad engine/
vehicle provisions;
• Title III 2- and 4-year maximum achievable
control technology (MACT) standards;
• Title IV SO, and NOx emissions programs
for utilities;
6 We also attempted to incorporate in the Pre-CAAA
(baseline) scenario the non-CAAA regulations and policies we
expect will have a significant effect on emissions between 1990
and 2010. For example, the TPM, which we used to estimate
utility emissions, incorporates the effect of the deregulation of
railroad rates on SO, emissions. IPM accounts for the influence
of the future cost of low-sulfur coal prices expected to occur as
a result of lower railroad rales. The impact of prescribed burn-
ing policies lor private arid federally owned lands on PM emis-
sions is also incorporated in the Pre-CAAA scenario.
• Title V permitting system for primary
sources of air pollution; and
• Title VI emissions limits for chemicals that
deplete stratospheric ozone.''
The Post-CAAA scenario also assumes the imple-
mentation of region-wide NOx controls and a cap-
and-trade system designed to reduce emissions dur-
ing the summer months from large utility and in-
dustrial sources in the 37 easternmost states that com-
prise the Ozone Transport Assessment Group
(OTAG) domain.8 In addition, the Post-CAAA sce-
nario incorporates the effects of a similarly designed
trading program for the 11 northeast states that com-
prise the Ozone Transport Region (OTR). This trad-
ing program is consistent with Phase II of the Ozone
Transport Commission (OTC) Memorandum of
Understanding (MOU).9 We provide more detailed
discussion of both Pre- and Post-CAAA scenario
development in Appendix A.
Emissions Estimation
Results
The results of the Pre- and Post-CAAA projec-
tions indicate that the 1990 Clean Air Act Amend-
ments will likely have a significant effect on future
emissions of air pollutants. Table 2-2 displays both
base-year (1990) and future-year (2000 and 2010)
emissions estimates for the modeled scenarios along
with the percent change from Pre- to Post-CAAA
VOC, NOx, SO,, CO, PM10, and PM23 projections.
A more detailed breakout of 2010 Pre- and Post-
CAAA emissions estimates, displaying emissions for
each major source category, is contained in Table 2-
3. Figures 2-1 through 2-6 show the emissions pro-
jections for each of the pollutants examined in this
analysis.
Emissions projections for VOC, NOx, SO,, and
CO, displayed in Figures 2-1 through 2-4, follow
' For a more detailed discussion of the CAAA provisions
incorporated in the Post-CAAA scenario, see Appendix A.
8 The NO control program incorporated in the Post-
CAAA scenario may not reflect the NO controls that are actu-
ally implemented in a regional ozone transport rule.
'' The Posl-C AAA scenario does not incorporate any in-
fluences ol the recently revised PM and ozone NA.AQS regula-
tions or any impact of the recently proposed Tier II tailpipe
standards.
11
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 2-2
Summary of National Annual Emissions Projections
(thousand tons)
1990 2000 2000 2000 2010 2010 2010
Base- Pre- Post- % Pre- Post- %
Pollutant Year CAAA CAAA Change CAAA CAAA Change
VOC 22,715 24,410 17,874 -27% 27,559 17,877 -35%
NOx 22,747 25,021 18,414 -26% 28,172 17,290 -39%
S02 22,361 24,008 18,013 -25% 26,216 18,020 -31%
CO 94,385 95,572 80,919 -15% 107,034 81,943 -23%
Primary 28,289 28,768 28,082 -2% 28,993 28,035 -3%
PM-io
Primary 7,091 7,353 7,216 -2% 7,742 7,447 -4%
PM2.5
Notes: Totals reflect emissions for the 48 contiguous States, excluding Alaska and Hawaii.
Percent change between Pre-CAAA and Post-CAAA scenarios.
similar patterns. Pre-CAAA estimates indicate emis-
sions of these pollutants would increase, on average,
by almost 20 percent from 1990 to 2010. These in-
creases reflect the expectation that anticipated growth
in activity levels in the relevant emitting sectors will
more than offset reductions achieved by pre-1990
control programs. While we predict relatively steady
growth m emissions in the absence of the 1990
Amendments, projections show emissions of these
four pollutants would increase at a slightly faster rate
over the last ten years of the 20 year projection pe-
riod.
Post-CAAA estimates of VOC, NOx, SO,, and
CO emissions for the modeled regulatory scenarios
decrease significantly from 1990 to 2000 and then
plateau, remaining relatively constant from 2000 to
2010. The initial decrease is triggered by the imple-
mentation of the CAAA and the associated controls.
After cleaner means of production are adopted, bet-
ter emissions control technologies are implemented,
and other required changes and improvements are
made, emissions reduction slows and in some in-
stances stops all together; emissions may even begin
to increase. Although the Post-CAAA estimates for
each of the above mentioned pollutants show little
or no change in the level of emissions from 2000 to
2010, an overall comparison of our Pre- and Post-
CAAA projections indicates that during this tune
period the 1990 Amendments continue to have an
increasingly beneficial effect on emission levels.
Comparison of Pre- and Post- CAAA emissions
estimates reveals that by 2010, estimated VOC emis-
sions will be 35 percent lower as a result of the imple-
mentation of the CAAA than they would have been
if no new control requirements, beyond those in
place in 1990, were mandated. This sizeable change
in emissions attributable to the Amendments is due
largely to estimated VOC reductions from motor
vehicle and area sources. The 2010 Post-CAAA es-
timate for these two source categories combined is
8.2 million tons lower than the Pre-CAAA projec-
tion, a total which accounts for 84 percent of the
predicted difference in VOC emissions estimated
under the two scenarios.
Based on the regulatory programs incorporated
in the Post-CAAA scenario, we project that NO
emissions will be reduced by the greatest percent-
age. Comparison of projections for the year 2010
indicates the Post-CAAA NOx estimate is 39 per-
cent lower than the Pre-CAAA estimate, represent-
ing a decrease in emissions of 10.8 million tons. We
project nearly half of this reduction will come from
utilities, while the remaining portions will come from
cuts in motor vehicle and non-utility point source
emissions.
12
-------
Chapter 2: Emissions
Figure 2-3 shows that by 2010 we anticipate SO,
levels will be 31 percent lower than they would have
been under the Pre-CAAA scenario. We project 96
percent of the 8.2 million ton difference between
Pre- and Post-CAAA estimates will result from regu-
lation of utilities, while the remaining reduction
comes from motor vehicles.
We estimate 2010 Post-CAAA CO emissions
will be 81.9 million tons, 23 percent lower than the
Pre-CAAA projection. Much of this reduction we
project will be achieved as a result of nonattainment
(Title I) and motor vehicle provisions (Title II) of
the 1990 Amendments. The more influential pro-
grams (in order of importance) are expected to be
enhanced vehicle emission inspections, wintertime
oxygenated fuel use, and LEA7 program adoption.
Figures 2-5 and 2-6 indicate that the 1990 Clean
Air Act Amendments have more modest effects on
primary PM1() and PM5 emissions.10 For both of
these pollutants, Pre-CAAA projections increase at
a slow rate from 1990 to 2010. Post-CAAA emis-
sions estimates for primary PM10 and PM., how-
ever, follow different paths. While we estimate
implementation of the CAAA will cause primary
PM10 levels to slowly decrease from 1990 to 2010,
Post-CAAA projections indicate primary PM23 emis-
sions will actually rise despite the influence of the
CAAA. Overall, however, emissions of primary
PM and PM both will be approximately four per-
cent lower in 2010 than they would have been with-
out the CAAA.11
The significant influence of area source emissions
on primary PM emissions levels, combined with the
limited regulation of this major source category,
explains the limited effect of the CAAA on primary
particulate matter emissions. According to data used
in this analysis, area sources account for over 90 per-
cent of primary PM emissions and over 80 percent
0 EPA projected PM n and PM,5 levels holding natural
source emissions ol particulate matter constant at 1990 levels.
The estimates presented in Figures 2-5 and 2-6 have been ad-
justed; these estimates represent total PM emissions minus natu-
ral source emissions (wind erosion).
:i Directly emitted PM, such as fugitive dust, is referred to
as primary PM. Secondary PM is not directly emitted, but rather
lorms in I lie atmosphere. NOr and SO, are two examples of
secondary PM precursors.
of primary PM, _ emissions.12 As a result, even the
successful reduction of motor vehicle and nonroad
emissions have only a slight impact on overall pri-
mary PM and PM estimates developed for this
study.13 Furthermore, the CAAA's most significant
primary PM area source controls target emissions in
counties not in compliance with the National Am-
bient Air Quality Standards (NAAQS).14 Currently,
however, there are fewer than 85 counties in the
country that are not in attainment with the national
standards. Emissions changes in these areas are ca-
pable of having only a minor influence on the over-
all primary PM level in the United States. Even
minor changes in primary PM emissions leading to
minor changes in the concentrations of this pollut-
ant, however, are significant. In the subsequent
portions of this analysis, sizable benefits are estimated
to result from small reductions in PM concentra-
tions in the atmosphere.
The seemingly small impact on direct PM emis-
sions resulting from implementation of the CAAA
depicted in Figures 2-5 and 2-6 can be misleading.
While these figures illustrate the impact of the 1990
CAAA on primary PM emissions, it is important to
remember that ambient PM concentrations are in-
fluenced by the presence of both primary and sec-
ondary PM. VOCs, NOx, and SO,, all pollutants
regulated by the CAA, are secondary PM precur-
sors. The reduction in the emissions of these three
pollutants also leads to lower overall PM concentra-
tions in the atmosphere. The complete impact of
the CAAA on PM thus is not fully captured by Fig-
ures 2-5 and 2-6. Additional discussion of the influ-
ence of the CAAA on PM and ambient air quality is
provided in Chapter 4 and Appendix C.
As part of this prospective analysis we also esti-
mated future-year NH emissions The 1990 Amend-
ments, however, do not include provisions designed
*2 As discussed on pages 18 and 20 and in Table 2-5, how-
ever, some recent, data in die ale thai lire composition data used
in this analysis may underestimate the contribution from mo-
tor vehicle carbonaceous emissions.
13 The difference between 2010 Pre- and Post-CAAA esti-
mates for PM and PM motor vehicle emissions is 31 percent
and 39 percent respectively. The difference between 2010 Pre-
and Post CAAA estimates for PM 0 and PM. nonroad emis-
sions is 19 percent and 20 percent respectively.
14 The PM NAAQS referred to here is the 50 ug/m3 (an-
nual mean) 150 ug/m3 (daily mean) standard.
13
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 2-3
Summary by Source Category of National Annual Emission Projections to 2010
(thousand tons)
Source
2010
2010
Pollutant
Category
1990
Pre-CAAA
Post-CAAA
% Change
VOC
Utility
37
49
50
2%
Point
3,500
4,200
3,500
-19%
Area
10,000
13,000
8,500
-36%
Nonroad
2,100
2,600
1,900
-28%
Motor Vehicle
6,800
7,300
3,900
-46%
TOTAL
23,000
28,000
18,000
-35%
NOx
Utility
7,400
9,100
3,800
-58%
Point
2,900
3,600
2,200
-39%
Area
2,200
3,000
3,000
-1%
Nonroad
2,800
3,400
2,700
-20%
Motor Vehicle
7,400
9,100
5,600
-39%
TOTAL
23,000
28,000
17,000
-39%
CO
Utility
330
450
460
2%
Point
6,000
7,400
7,400
0%
Area
12,000
14,000
14,000
0%
Nonroad
14,000
19,000
18,000
-4%
Motor Vehicle
62,000
66,000
42,000
-37%
TOTAL
94,000
107,000
82,000
-23%
so2
Utility
16,000
18,000
9,900
-44%
Point
4,600
6,000
6,000
0%
Area
1,000
1,500
1,500
0%
Nonroad
240
240
240
0%
Motor Vehicle
570
770
410
-47%
TOTAL
22,000
26,000
18,000
-31%
Primary
Utility
280
310
280
-9%
PM-io
Point
930
1,200
1,200
0%
Area
26,000
27,000
26,000
-3%
Nonroad
340
410
340
-19%
Motor Vehicle
360
300
210
-31%
TOTAL
28,000
29,000
28,000
-3%
Primary
Utility
110
120
110
-8%
pm25
Point
590
750
750
0%
Area
5,800
6,300
6,100
-2%
Nonroad
290
360
290
-20%
Motor Vehicle
290
230
140
-39%
TOTAL
7,100
7,700
7,400
-4%
NOTES: Table may not sum due to rounding. Percentage change was calculated prior to rounding.
14
-------
Chapter 2: Emissions
to regulate Nil, As a result, the Pre- and Post-
CAAA estimates follow a similar upward trend. We
estimate NH emissions will increase roughly 55
percent from 1990 to 2010. Although we do not
estimate the costs and benefits associated with NH
controls and changes in NH3 ambient concentrations
as part of this analysis, estimation of NH, emissions
is an important part of the prospective study. NH
is a secondary PM precursor, and we relied on fu-
ture-year NH3 emissions estimates as model input
to help us estimate PM concentrations.
We also estimated the effect of CAAA provi-
sions on mercury (Hg) emissions for five separate
I Ig emissions sources: medical waste incinerators
(MWI), municipal waste combustors (MWCs), elec-
tric utility plants, hazardous waste combustors, and
chlor-alkali plants.15 Together, these sources account
for 75 to 80 percent of national anthropogenic air-
borne Hg emissions. In this analysis we qualitatively
examine the effects of mercury emissions reductions
on ecological systems (see Chapter 7 and Appendix
E). We do not, however, evaluate the impact of Hg
on human health.
Table 2-4 displays, for each emission category,
base-year (1990) and future-year (2000 and 2010) Pre-
and Post-CAAA emissions estimates. The table also
showrs the difference between Pre- and Post-CAAA
estimates for each projection year. Overall, the re-
sults of this analysis indicate that the 1990 Amend-
ments will provide a reduction in Hg emissions of
44.2 tons per year (tpy) in the year 2000 and a reduc-
tion of 56.2 tpy in 2010. These changes represent a
35 percent reduction in airborne mercury emissions
for the year 2000 and a 42 percent reduction for 2010.
We estimate that most of the reduction will be the
result of New Source Performance Standards for
MWI and MWCs.
Table 2-4
Airborne Mercury Emission Estimates
2000 Emissions (tons) 2010 Emissions (tons)
Source Category
1990
Emissions
(tons)
Pre-
CAAA
Post-
CAAA
Diff.
Pre-
CAAA
Post-
CAAA
Diff.
Medical Waste Incin.
50
17.9
1.3
16.6
22.6
1.6
21.0
Municipal Waste Comb.
54
31.2
5.5
25.7
33.8
6.0
27.8
Electric Utility Generation
51.3
63.0
61.1
1.9
68.5
65.4
3.1
Hazardous Waste Comb.
6.6
6.6
6.6
0
6.6
3.0
3.6
Chlor-Alkali Plants
9.8
6.0
6.0
0
2.0
1.3
0.7
Total CAAA Benefits (Reductions) 44.2 56.2
15 With the exception of electric utility plant Hg emissions
that were estimated using IPM, we relied on previously gener-
ated estimates (typically from recently conducted RIAs) to evalu-
ate (lie impact oi the CAAA oil Hg emissions. For a more com-
plete discussion of the methodology, see Appendix A.
15
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Figure 2-1
Pre- and Post-CAAA Scenario VOC Emissions
Estimates
1990 2000
Year
>— Pre-CAAA —Post-CAAA I
Figure 2-2
Pre- and Post-CAAA Scenario NO, Emissions
Estimates
Pre-CAAA
Fbst-CAAA
2010
Figure 2-3
Pre- and Post-CAAA Scenario S02 Emissions
Estimates
1990 2000
Year
— Pre-CAAA —¦— Post-CAAA
2010
16
-------
Chapter 2: Emissions
Figure 2-4
Pre- and Post-CAAA Scenario CO Emissions
Estimates
120
100 -
o
¦E
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Comparison of Emissions
Estimates With Other
Existing Data
Comparison of the emissions projections gener-
ated by the prospective analysis to historical emis-
sions estimates drawn from the National Air Pollut-
ant and Emissions Trends reports (Trends) provides
a check on the reasonableness of our emissions in-
ventories. In addition, comparison of emissions pro-
jections from the prospective analysis with those of
the Grand Canyon Visibility Transport Commis-
sion (GCVTC) study of western regional haze pro-
vides an initial test of the sensitivity of emissions
projections to base-year inventories and growth as-
sumptions. Analysis of PM emissions and compari-
son of estimated and observed PM data also help us
evaluate the prospective study's emissions estimation
methods.
Trends reports contain historical estimates of
annual VOC, NOx, SO,, CO, and PM10, emissions.
While the most recent report only provides emis-
sions data through the first half of the 1990s, com-
parison of these estimates from 1990 to 1996 with
emissions trends projected under the Post-CAAA
scenarios reveals that emissions figures from both
are similar. The disparity that does exist between
the two sets of estimates largely stems from the fact
that the Post-CAAA scenario trend lines running
from 1990 to 2000 consist of only two data points.
As a result, Post-CAAA trend lines cannot capture
yearly fluctuations in emissions and the exact tim-
ing of emissions cuts. Only for NOx are the Trends
and Post-CAAA estimates significantly different; this
is because the Trends report is still in the process of
incorporating the State's periodic emission inventory
into the NET database. As a result, Trends values
do not capture all the NO emission reductions that
have occurred since 1990. For example, significant
reductions attributable to reasonable available con-
trol technology' (RACT) requirements for major sta-
tionary source NO emitters areas are not reflected
X
in the Trends figures.
The Grand Canyon Visibility Transport Com-
mission conducted an air pollution analysis for West-
ern States that projected emissions for selected pol-
lutants, including NOx, SO„, and PM25, from 1990
base-year levels for the year 2000 and every tenth
subsequent year up to 2040. GCVTC estimates of
future-year emissions levels differ from Post-CAAA
projections. This disparity results from the use of
different base-year inventories in the two studies and
from specific regional reductions not incorporated
in the prospective analysis scenarios. Despite the
difference in GCVTC and Post-CAAA estimates,
the change in the level of emissions from 1990 to
2010 predicted by the two studies is similar. Com-
parison of both sets of projections illustrates the sen-
sitivity of future-year emissions estimates to the base-
year inventory.
The 1997 National Air Quality and Emissions
Trends Report provides a summary of PM con-
centration speciation data. This report shows the
relative contribution of the major PM emissions
source components (crustal material, carbonaceous
particles, nitrate, and sulfate) to ambient PM,S con-
centrations in urban and nonurban areas through-
out the U.S.16 Comparison of primary PM,5 emis-
sions estimates generated for this analysis with the
observed concentration data presented in the 1997
report indicates that the ratio in the prospective study
of crustal material to primary carbonaceous particles
is high. At least part of this apparent overestima-
tion of crustal material and underestimation of car-
bonaceous particulates, however, is due to the fact
that much of the emitted crustal material quickly
settles and does not have a quantifiable impact on
ambient air quality. In this analysis, we apply a fac-
tor of 0.2 to crustal emissions to estimate the frac-
tion of crustal PM that makes its way into the
"mixed layer" of the atmosphere and influences pol-
lutant concentrations. Figure 2-7 displays the
breakout of primary PM into its adjusted crustal
and carbonaceous (elemental carbon and organic
carbon) components. The figure divides crustal
material into two subcategories, fugitive dust or in-
dustrial sources, based on the source of the material
and also shows the fraction of primary PM„„ that is
16 Crustal material is directly emitted from fugitive dust
sources such as agricultural operations, construction, paved and
unpaved roads, and wind erosion as well as from some indus-
trial sources such as metals processing. Carbonaceous particles,
as defined in the 1997 National Air Quality and Emissions
Trends Report," are emitted directly and as condensed liquid
droplets Irom iuel comb us lion, burning of forests, rangelands,
and fields; oil highway and highway mobile sources (gas and
diesel); and certain industrial processes".
18
-------
Chapter 2: Emissions
Figure 2-7
1990 Primary PM26 Emissions by EPA Region (tons/year)
0
Region 10
©
Region 8
Region 5
Region 7
Region 9
©
Region 6
Region 4 \
Region 1
Region 2
lion 3
A
A
A
A
Crustal - Fugitive Dust Sources
Crustal - Industrial Sources
Other Primary
Elemental Carbon
Organic Carbon
o
100,000 450,000 850,000
neither crustal nor carbonaceous. The ratios of ad-
justed crustal material to primary carbonaceous par-
ticles presented in Figure 2-7 are in line with the
observed PM25 concentration data presented in the
1997 report.
Uncertainty In Emission
Estimates
Table 2-5 provides a list of sources of uncertainty
associated with estimating base-year emissions, the
expected direction of bias introduced by each un-
certainty (if known), and the relative significance of
each uncertainty in the overall 812 benefits analysis.
The emissions estimates presented in the prospec-
tive analysis are characterized by three major sources
of uncertainty: estimation of the base-year inven-
tory, prediction of the growth in pollution-generat-
ing activity, and assumptions about future-year con-
trols.
Base-year emissions were estimated using emis-
sions factors that express the relationship between a
particular human/industrial activity and the level of
emissions. The accuracy of base-year emissions esti-
mates varies from pollutant to pollutant, depending
largely on how directly the selected activity and
emissions correlate. We likely estimated 1990 S02
emissions with the greatest precision. Sulfur diox-
ide emissions are generated during combustion of
sulfur-containing fuel and are directly related to fuel
sulfur content. In addition, we were able to verify
these estimates through comparison with Continu-
ous Emission Monitoring (CEM) data. As a result,
we were able to accurately estimate S02 emissions
using emissions factors based on data on fuel usage
and fuel sulfur content. Nitrogen oxides are also a
product of fuel combustion, allowing us to estimate
emissions of this pollutant using the same general
technique used to estimate S02 emissions. However,
the processes involved in the formation of NOx
during combustion are more complicated than those
involved in the formation of SO • thus, our NO
emissions estimates are more variable and less cer-
tain than S02 estimates.
Volatile organic compounds, like SOz and NOx,
are products of fuel combustion; however, these
compounds are also a product of evaporation. To
estimate evaporative emissions of this pollutant we
19
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
used emissions factors that relate changes in emis-
sions to changes 111 temperature. Because future
meteorological conditions are difficult to predict,
the uncertainty associated with forecasting tempera-
ture influences the uncertainty in our YOG emis-
sions estimates. The likely significance of this un-
certainty, in terms of its impact on the overall mon-
etary benefit present in this analysis, is probably
minor.
In this analysis we estimated primary PM7 5 emis-
sions based on unit emissions that may not accu-
rately reflect the composition and mobility of par-
ticles. The ratio of crustal to carbonaceous particu-
late material, for example, likely is high as a result of
overestimation of the fraction of crustal material,
primarily composed of fugitive dust, and underesti-
mation of the fraction of carbonaceous material.
Because the CAAA has a greater impact 011 emis-
sions sources that generate carbonaceous particles
(mobile sources) than on sources that mainly emit
crustal material (area sources), we likely underesti-
mate the impact of the CAAA on reducing I'M.,
thereby reducing monetary benefits estimates. The
uncertainty associated with estimating the partition
of PM emissions components could conceivably
have a major impact 011 the net benefit estimate;
compared to secondary PM precursor emissions,
however, changes 111 primary PM25 emissions have a
relatively small impact on PM25 related benefits.
We estimated future-year emissions levels based
on expected growth in pollution-generating activi-
ties. Inherent uncertainties and data inadequacies/
limitations exist 111 forecasting growth for any fu-
ture period. Also, the growth indicators we used in
this analysis may not directly correlate with changes
in the factors that influence emissions. Both of these
factors contribute to the uncertainty associated with
this study's emissions results. For example, the best
indicator of pollution-generating activity is fuel use
or some other measure of input/output that most
directly relates to emissions. The key BRA indica-
tor used in this analysis, GSP, is closely correlated
with the pollution-generating activity associated with
many manufacturing industry processes (iron and
steel, petroleum refining, etc.). However, a good
portion of industrial sector emissions are from boil-
ers and furnaces, whose activity is related to produc-
tion, but not as closely as to product output. Activi-
ties such as fuel switching may produce different
emission patterns than those reflected in the results
of this study.
Our future-year control assumptions are also a
source of uncertainty. Despite our efforts to mini-
mize this uncertainty, whether each of the Post-
CAAA controls will be adopted, whether Post-
CAAA control programs will be more or less effec-
tive than estimated, and whether unanticipated tech-
nological shifts will reduce future-year emissions are
all unknown. For example, the Post-CAAA scenario
includes implementation of a region-wide NO con-
trol strategy designed to regulate the regional trans-
port of ozone. However, the control program as-
sumed under the Post-CAAA scenario may not re-
flect the NOx controls that are actually implemented
in a regional ozone transport rule.
20
-------
Chapter 2: Emissions
Table 2-5
Key Uncertainties Associated with Emissions Estimation
Potential Source of Error
Direction of Potential Bias for
Net Benefits Estimate
Likely Significance Relative to Key
Uncertainties in Net Benefit
Estimate*
PM2.5 emissions are largely
based on scaling of PM10
emissions.
Overall, unable to determine
based on current information,
but current emission factors are
likely to underestimate PM2.5
emissions from combustion
sources, implying a potential
underestimation of benefits.
Potentially major. Source-specific
scaling factors reflect the most careful
estimation currently possible, using
current emissions monitoring data.
However, health benefit estimates
related to changes in PM2.5 constitute
a large portion of overall CAAA-related
benefits.
Primary PM2.5 emissions
estimates are based on unit
emissions that may not
accurately reflect composition
and mobility of the particles.
For example, the ratio of
crustal to primary
carbonaceous particulate
material likely is high.
Underestimate. The effect of
overestimating crustal emissions
and underestimating
carbonaceous when applied in
later stages of the analysis, is to
reduce the net impact of the
CAAA on primary PM2.5
emissions by underestimating
PM2.5 emissions reductions
associated with mobile source
tailpipe controls.
Potentially major. Mobile source
primary carbonaceous particles are a
significant contributor to public
exposure to PM2.5. Overall, however,
compared to secondary PM2.5
precursor emissions, changes in
primary PM2.5 emissions have only a
small impact on PM2.5 related benefits.
The Post-CAAA scenario
includes implementation of a
region-wide NOx emissions
reduction strategy to control
regional transport of ozone
that may not reflect the NO*
controls that are actually
implemented in a regional
ozone transport rule.
Unable to determine based on
current information.
Probably minor. Overall, magnitude of
estimated emissions reductions is
comparable to that in expected future
regional transport rule. In some areas
of the 37 state region, emissions
reductions are expected to be
overestimated, bur in other areas, NO*
inhibition of ozone leads to
underestimates of ozone benefits
(e.g., some eastern urban centers).
VOC emissions are dependent
on evaporation, and future
patterns of temperature are
difficult to predict.
Unable to determine based on
current information.
Probably minor. We assume future
temperature patterns are well
characterized by historic patterns, but
an acceleration of climate change
(warming) could increase emissions.
Use of average temperatures
(i.e., daily minimum and
maximum) in estimating
motor-vehicle emissions
artificially reduces variability in
VOC emissions.
Unable to determine based on
current information.
Probably minor. Use of averages will
overestimate emissions on some days
and underestimate on other days.
Effect is mitigated in Post-CAAA
scenarios because of more stringent
evaporative controls that are in place
by 2000 and 2010.
21
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 2-5 (continued)
Key Uncertainties Associated with Emissions Estimation
Potential Source of Error
Direction of Potential Bias for
Net Benefits Estimate
Likely Significance Relative to Key
Uncertainties in Net Benefit
Estimate*
Economic growth factors used
to project emissions are an
indicator of future economic
activity. They reflect
uncertainty in economic
forecasting as well as
uncertainty in the link to
emissions.
Unable to determine based on
current information.
Probably minor. The same set of
growth factors are used to project
emissions under both the Pre-CAAA
and Post-CAAA scenarios, mitigating
to some extent the potential for
significant errors in estimating
differences in emissions.
Uncertainties in the
stringency, scope, timing, and
effectiveness of Post-CAAA
controls included in projection
scenarios.
Unable to determine based on
current information.
Probably minor. Future controls could
be more or less stringent, wide-
reaching (e.g., NOx reductions in
OTAG region - see above), or
effective (e.g., uncertainty in realizing
all Reasonable Further Progress
requirements) than projected. Timing
of emissions reductions may also be
affected (e.g., sulfur emissions
reductions from utility sources have
occurred more rapidly than projected
for this analysis).
* The classification of each potential source of error reflects the best judgement of the section 812 Project Team. The
Project Team assigns a classification of "potentially major" if a plausible alternative assumption or approach could
influence the overall monetary benefit estimate by approximately five percent or more; if an alternative assumption or
approach is likely to change the total benefit estimate by less than five percent, the Project Team assigns a classification
of "probably minor."
22
-------
Direct Costs
The costs of complying with the requirements
of the Clean Air Act Amendments (CAAA) of 1990
will affect all levels of the U.S. economy. The im-
pact, initially experienced through the direct costs
imposed by regulations promulgated under the
amendments, will also be seen in patterns of indus-
trial production, research and development, capital
investment, productivity, employment, and con-
sumption. The purpose of the analysis summarized
in this chapter is to estimate the incremental change
in annual compliance costs from 1990 to 2010 that
are directly attributable to the 1990 Clean Air Act
Amendments.
This chapter consists of four sections. The first
section summarizes our approach to estimating di-
rect compliance costs. In the second section we
present the results of the cost analysis. We first re-
port the total costs of Titles I through Y and then
present estimates for major individual provisions.
We also briefly discuss our derivation of Title VI
costs. In the third section, wre provide a qualitative
discussion of the potential magnitude of social costs
and other impacts associated with the Amendments
to characterize the potential welfare loss not cap-
tured in the direct cost approach. We conclude the
chapter with a discussion of the major analytic un-
certainties and include the results of quantitative sen-
sitivity tests of key data and assumptions.
Approach to Estimating
Direct Compliance Costs
As discussed in the previous chapter, the first
step of the prospective analysis required the devel-
opment of emission estimates for the base-year, 1990,
and for the two target years in our analytic time
period, 2000 to 2010. We developed two scenarios,
Pre-CAAA and Post-CAAA, that reflect three key
13
parameters: (i) base-year inventory selection, (li) in-
dicators of forecasted economic growth, and (iii) ef-
fects of future year controls and selected CAAA pro-
visions. The Pre-CAAA scenario applies the strin-
gency and scope of air pollution regulations as they
existed in 1990 and projects emissions and costs to
2000 and 2010. This scenario establishes a baseline
that represents projected emission levels and con-
trol costs in the absence of the 1990 Amendments.
Under the Post-CAAA scenario, costs are based on
compliance writh selected CAAA provisions. To-
gether these two scenarios form the foundation upon
which the incremental costs and benefits of comply-
ing with the 1990 Amendments are estimated. For
more information on the development of these sce-
narios, see Chapter 2.
We closely integrate the modeling of direct com-
pliance costs writh emissions projections by main-
taining consistency among control assumptions (i.e.
emissions scenarios) used as inputs in the cost esti-
mation modeling and in the analysis of emissions
projections and benefits. Wre use two models to es-
timate costs, Emission Reduction and Cost Analysis
Model (ERCAM) and Integrated Planning Model
(IPM). These models generate cost estimates for the
Post-CAAA scenarios in two projection years, 2000
and 2010. The estimates are calculated relative to
costs under the same year Pre-CAAA scenario, so
estimates represent incremental costs of compliance
with the 1990 Amendments.
We use ERCAM to estimate costs associated with
regulating particulate matter (PM), volatile organic
compounds (VOCs), and non-utility source oxides
of nitrogen (NOj.1 The model is essentially a cost-
accounting tool that provides a structure for modi-
fying and updating changes in inputs while main-
1 This model was developed by E. II. Pedum & Associ-
ates, Inc. lo lacilitate EPA's analysis oi emissions control.
23
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
taming consistency with the emission and cost analy-
ses. (lost scenarios and assumptions are developed
for each non-utilitv source category (e.g., point, area,
nonroad, and motor vehicle sources) and 111 response
to specific provisions and emission targets. The
model estimates costs based on inputs such as cost
per ton, source-specific cost equations, incremental
production, and operating cost estimates. For this
analysis, we collected data and inputs from informa-
tion presented in regulatory impact assessments
(RIAs), background information documents (BIDs),
regulatory support documents, and Federal Regis-
ter notices.
To estimate the costs of reducing utility NO
and sulfur dioxide (SO^ emissions, we use the Inte-
grated Planning Model (IPM). IPM allows us to es-
timate the control costs of several pollutants while
maintaining consistent control scenarios and eco-
nomic forecasts of the electric power industry. It
assesses the optimal mix of pollution control strate-
gies subject to a series of specified constraints. Kev
inputs and constraints in the model include targeted
emissions reductions (on a seasonal or annual basis),
costs and constraints of control technology, and eco-
nomic parameters (e.g., forecasted demand for elec-
tricity, power plant availability/capacity, costs of
fuel, etc.)
To assess the costs of reducing emission of pol-
lutants or sectors not covered by our two models,
we estimate costs using the best available cost equa-
tions or other types of analyses. For example, we
estimate non-utility S02 emission control costs for
point sources by applying source-specific cost equa-
tions for flue gas desulfurization (FGD)/scrubber
technology to affected sources in 2000 and 2010.
While we do not explicitly model (X) attainment
costs, we include in the analysis the costs of pro-
grams designed to reduce CO emissions, such as oxy-
genated fuels and a cold temperature CO motor ve-
hicle emission standard. Finally, to estimate costs
of the rate of progress/reasonable further progress
(ROP/RFP) provisions, requirements under Title I
that require ozone nonattainment areas to make
steady progress toward attainment, we first estimate
the emissions reduction shortfall that must be
achieved in each target year in each nonattainment
area, and then apply a cost per ton estimate from a
schedule of measures that could be applied locally
to meet the necessary ROP/RFP requirement. For
more detail on the specific methods used to estimate
compliance costs for each pollutant and source cat-
egory, see Appendix B.
The cost estimates in this chapter are the incre-
mental costs of the 1990 Amendments (i.e. the dif-
ference between pre- and Post-CAAA cost estimates).
We present the results as total annualized costs (TAC)
in 2000 and 2010. Annualized costs include both
capital costs, such as costs of control equipment, and
operation and maintenance (O&M) costs.2 They
do not represent actual cash flow in a given year,
but are rather an estimate of average annual burden
over the period during which firms will incur costs.
In annualizing costs, we convert total capital invest-
ment to a uniform scries of total per-year equivalent
payments over a given time period using an assumed
real cost-of-capital at five percent. We then add
O&M and other reoccurring costs to the annualized
capital cost to arrive at TAC. The discounted sum
of these annual expenditures is equal to the net
present value of total costs incurred over the time
period of this analysis.3
Direct Compliance Cost
Results
Total annual compliance costs for Titles I
through V of the 1990 Amendments in the year 2000
will be approximately $19.4 billion; the estimate in-
creases to $26.8 billion in the year 2010. These costs
reflect "annualized" operation and maintenance
(O&M) expenditures (which includes research and
development (R&D) and other similarly recurring-
expenditures) plus amortized capital costs (i.e., de-
preciation plus interest costs associated with the ex-
2 For a few VOC source categories, we estimate that capi-
tal investment will not be necessary; for these sources, compli-
ance costs reflect O&M costs only.
J We recalculate the control cost estimates from regulatory
documents that use a seven or ten percent discount rate so that
the costs will be consistent with the five percent discount rate
assumption used in this analysis. We also calculate cost using
three percent and seven percent discount rales, as sensitivity tests;
lor detail see the discussion of uncertainty later in this chapter,
in Chapter 8, and in Appendix B.
24
-------
Chapter 3: Direct Costs
isting capital stock) for the particular year.4 We
present cost estimates by title and emissions source
category (point sources, area sources, utilities,
nonroad engines and vehicles, and motor vehicles)
in Table 3-1.
In some cases, assigning costs to a single CAAA
title is complicated by the fact that there are rules
issued pursuant to more than one title.5 In addi-
tion, with the passage of the 1990 Amendments, the
States were given greater discretion in developing
CAAA compliance strategies. For example, the
States can determine how best to meet progress re-
quirements and are responsible for creating permit
programs (under Title V). As a result, a significant
portion of the costs also represent State-level strate-
gies and decisions for reducing emissions.
Title I, Provisions for Attainment and Mainte-
nance of National Ambient Air Quality Standards
(NAAQS), represents pollution controls (of VOC,
NOx, and PM emissions) implemented primarily by
point and area sources. Title I provisions also ac-
count for State programs designed to meet progress
requirements. By 2010, we project the costs of Title
I provisions will account for over half of total CAAA
direct compliance costs ($14.5 billion). An additional
34 percent of estimated total costs ($9 billion) is at-
tributed to regulating mobile source emissions un-
der Title II. Collectively, the combined direct com-
pliance costs of these two titles is 116 billion in 2000
and $23 billion by 2010.
The remaining three titles account for less than
20 percent of total CAAA direct costs. We estimate
that Title III provisions, which target hazardous air
pollutant (HAP) emissions, will cost $840 million
by the year 2010. This estimate represents total an-
nualized capital costs (PACs) for individual two- and
four-year MACT standards. While the majority of
this estimated cost reflects reducing VOC emissions
4 Capital expenditures are investments, generating a stream
of benefits and opportunity costs over an investment's lifetime.
In a cost-benefit analysis, the appropriate accounting technique
is to annualize capital expenditures. This technique involves
spreading the costs of capital equipment uniformly over the use-
ful life of the equipment, by using a discount rate to account for
the time value of money. In this analysis, all capital expendi-
tures were annualized using a real five percent interest rate.
5 In those cases, we generally assigned costs to a single title
based upon implementation dates and the year by which emis-
sion reductions are expected.
(since I LAP emissions were not included as part of
the Section 812 base- year inventor)7), Title III costs
do include some costs of final MACT rules that regu-
late non-VOC HAP emissions.
In order to estimate the costs associated with
Title IV, we considered the implications of pollu-
tion abatement controls (for S02 and NOJ on the
electric power industry's operation of generation
units and how, over tune, this would affect the de-
mand for electricity. The annual compliance esti-
mate for Title IV costs is $2.3 billion in 2000. 'This
estimate decreases to $2.0 billion by 2010. This de-
crease reflects, in part, the future compliance cost
savings resulting from the SO, allowance trading-
program.
Title V costs arc associated with new operating
permit programs. The estimate accounts for approxi-
mately one percent of total costs projected under
the Post-CAAA 2010 scenario. States are expected
to implement Title V permit programs by 2005. The
estimate reflects the costs of State-developed pro-
grams during the first five-year implementation pe-
riod. These costs include incremental administra-
tive costs incurred by the permitted sources, State
and local permitting agencies, and EPA. The esti-
mate excludes federally-implemented State programs
and state programs which were already established
in the baseline.
Our presentation of cost estimates for the strato-
spheric ozone protection provisions of Title VI is,
by necessity, different from other titles. Ideally, one
should compare the costs of actions taken in a given
year to the benefits attributable to these actions. For
Title AT, a cost-benefit comparison of any given year
requires assumptions that result in potentially mis-
leading figures. The difficulty is due to the differing
time horizons and the complexity of the process by
which ozone-depleting substances (ODSs) cause ad-
verse effects on human health and the environment.
Title VI provisions incur costs over significantly
varying tune horizons; for example, the cost analy-
sis of Sections 604 and 606 provisions spans 85 years
(from 1990 to 2075). At the same time, the analysis
of Section 611 extends from 1994 to 2015. In re-
sponse to this analytic difficulty, wre base our com-
parison of Title VI costs to Title VI benefits on net
present values.
25
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 3-1
Summary of Direct Costs for Titles I to V of CAAA, By Title and Selected Provisions
(Annual Costs in million 1990$)
Title/Provision
Primary Cost
Estimate
2000
Percentage of
Total Costs
Primary Cost
Estimate
2010
Percentage of
Total Costs
Title 1- Provisions for Attainment and Maintenance of NAAQS
Stationary NO* Controls, Utility Industry
$ 790
4%
$ 2,500
9%
Progress Requirements
1,200
6%
2,500
9%
PM NAAQS Controls
1,900
10%
2,200
8%
California LEV
320
2%
1,100
4%
National LEV
180
1%
1,100
4%
High Enhanced l/M
1,100
6%
1,400
5%
Other Title 1 Programs
3,100
16%
3,700
14%
Title 1: Total Costs
$ 8,600
44%
$ 14,500
54%
Title II- Provisions Relating to Mobile Sources
California Reformulated Gasoline
$2,000
10%
$2,400
9%
NOx Tailpipe/Extended Useful
Life Standard
1,500
8%
1,700
6%
Other Title II Programs
3,900
20%
4,900
18%
Title II: Total Costs
$ 7,400
38%
$ 9,050
34%
Title III- Hazardous Air Pollutants
Title III: Total Costs
$ 780
4%
$ 840
3%
Title IV- Acid Deposition Control
Title IV: Total Costs
$2,300
12%
$2,040
8%
Title V- Permits
Title V: Total Costs
$ 300
2%
$ 300
1%
Total Annual Cost
$ 19,400
100%
$26,800
100%
Note: Totals may not sum due to rounding. Only major provisions are listed under each title - other, less costly provisions not
listed here are nonetheless included in the totals by title and the overall total.
The net present value of Title VI program costs
reflect selected actions and their associated costs from
Sections 604, 606, 608, 609, and 611. Examples of
these actions include: replacement of ozone-deplet-
mg chemicals with alternative technologies and ma-
terials; recycling and storage of unused chlorofluo-
rocarbons; labeling; training; and administration.
Using a discount rate of five percent and a 85-year
time horizon (from 1990 to 2075), we estimate the
net present value of Title VI costs is $27 billion. For
illustrative purposes, we calculated an annualized
estimate of Title VI costs. It is, however, important
to recognize that these estimates may overestimate
actual compliance costs in those years, especially in
the year 2000, because of the phased nature of imple-
mentation— see Appendix G for more details. Our
annualized estimate of total Title VI costs is $1.4
billion. This value reflects an annualized equivalent
value of costs incurred over 85 years (from 1990 to
2075) using a five percent discount rate.
Selected Provisions
Our analysis indicates eight provisions will ac-
count for approximately 54 percent of the total di-
rect compliance costs estimate for 2010. Six are Title
I provisions that affect stationary sources and vehicle
26
-------
Chapter 3: Direct Costs
emissions. The remaining two provisions target
mobile sources under Title II. These provisions are:
• PM NAAQS controls6,
• Electric power industry compliance (station-
ary NO control),
• Progress Requirements,
• California Low Emission Vehicle
(LEV) pro gram,
• National Low Emission Vehicle (LEA7) pro-
gram,
• High Enhanced Inspection and Maintenance
(I/M) program,
• California Reformulated Gasoline, and
• NOx Tailpipe/Extended Useful Life Stan-
dard.
The 1990 CAAA regulates stationary source
emissions primarily under Title I. Among the rel-
evant provisions, PM NAAQS, utility industry com-
pliance with NO standards, and progress require-
ments are the main sources of Title I costs. From
2000 to 2010, we estimate the control costs of all
three provisions will increase by at least a factor of
two. Under the Post-CAAA scenario developed for
the emissions analysis, the utility industry's compli-
ance with NO emission standards affects all electric
x
generation units using fossil fuels. Existing oil and
gas units face Reasonable Available Control Tech-
nology (RACT) requirements and all new units must
comply with more stringent New Source Perfor-
mance Standards (NSPS) and New Source Review
(NSR) requirements. By 2010, estimated costs for
stationary NO controls more than triple ($790 mil-
lion to 52,500 million). The cost estimate indicates
that the provision will be the single largest source of
CAAA direct costs. The second largest component
of total costs in 2010 is attributed to progress re-
quirements. Annual compliance costs with progress
requirements double from 2000 to 2010 ($1.2 bil-
lion and $2.5 billion, respectively). Among the three
provisions, the annual costs associated with PM
NAAQS compliance exhibits the least amount of
growth. We estimate annual costs for PM NAAQS
compliance will grow from $1.9 billion in 2000 to
$2.2 billion in 2010.
5 We estimate the PM NAAQS provision costs based on
compliance with standards that were in effect prior to 1997 revi-
sions (62 Fed. Reg. 38,652, 1997).
Among the provisions regulating vehicle emis-
sions, only the national and California LEV pro-
grams exhibit a trend of increasing direct costs of
the same magnitude as seen with the costs of regu-
lating stationary sources under Title I. The com-
bined cost of national and California LEV programs
is $2.2 billion in 2010. For the California LEV pro-
gram, the increase in cost is largely a function of
higher per vehicle cost estimates (e.g., zero emission
vehicles (ZEV) are mandated in the year 2003). Our
cost analysis of the national LEV program assumes
that only the Northeast Ozone Transport Region
(OTR) states will incur costs in the year 2000. By
2010, however, we expect that the program will af-
fect areas outside of the OTR. As a result, 2010
national LEV costs increase with the expected ex-
pansion and increased volume of vehicle sales. Un-
like many of the other provisions, high enhanced 1/
M costs do not exhibit significant growth from 2000
to 2010. We estimate this provision accounts for
approximately six percent of total costs in 2000 and
five percent in 2010. These costs, however, are un-
certain pending State decisions regarding the design
of their programs.
Among the analyzed Title II provisions, we at-
tribute nearly 15 percent of total annual direct costs
to the California reformulated gasoline (REG) pro-
gram and NOx Tailpipe/Extended Useful Life Stan-
dard. Although the reformulated gasoline program
affects only California, the state accounts for nearly
ten percent of annual gasoline sales in the United
States. We estimate compliance costs of $1.9 billion
in the year 2000. As the program enters Phase 2,
estimated costs grow to $2.4 billion. The trend in
cost associated with NO Tailpipe/Extended Useful
Life Standard is very different. While costs increase
slightly between the years 2000 and 2010, the
provision's share of total cost slightly decreases.
Characterization of Other
Economic Impacts
In an ideal setting, a cost-benefit analysis would
not only identify, but also quantify and monetize,
an exhaustive list of social costs associated with a
regulator)' action. This would include assessing how-
regulatory actions targeting a specific industry or set
of facilities can alter the level of production and con-
sumption in the directly affected market and related
27
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
markets. For example, regulation of emissions from
the electric utility industry that results in higher elec-
tricity rates would have both supply-side and de-
mand-side responses. In secondary markets, the in-
creased electricity rates affect production costs for
various industries and initiate behavioral changes
(e.g., using alternative fuels as a substitute for elec-
tric power). With each affected market, there are
also associated externalities that should be included
in estimating social costs. Returning to the utilities
example, the externalities associated with electric
power generation versus nuclear power generation
can be very different. The mix of externalities could
change as consumers substitute nuclear power for
electric power. It is frequently difficult to accurately
characterize one or all of these dimensions of mar-
ket responses and estimate the resulting social costs.
There are three generally practiced approaches
to calculating costs associated with regulation: (i) di-
rect compliance cost, (ii) partial equilibrium model-
ing, and (ni) general equilibrium modeling. Direct
compliance cost estimates are calculated differently
than the economic welfare impact estimates that re-
sult from partial or general equilibrium modeling; a
direct cost estimate is often the most straightforward
of the three approaches. This method estimates com-
pliance expenditures or, in economic terms, how an
industry's or firm's marginal cost curve shifts due to
increased production costs associated with regula-
tory compliance. As a result, this method does not
account for firm responses and market responses,
such as adjustment of production levels and product
prices. The other two methods measure changes in
producer and consumer welfare, and incorporate
these types of adjustments.
The direct cost approach likely overstates actual
compliance expenditures, but may have an ambigu-
ous relationship to total social costs. There are two
primary reasons for the overstatement of compli-
ance expenditures. First, the direct cost approach
does not account for market responses. As a result,
total direct cost estimates reflect the incremental cost
per unit of output multiplied by the generally higher,
pre-regulation quantity produced. Second, a direct
cost approach tends to make the simplifying assump-
tion that firms rely on static pollution abatement
technology, when in fact the presence of compli-
ance costs provides an incentive to innovate. Sev-
eral ex post cost analyses suggest that the marginal
cost curve may not necessarily shift by the full
amount of the pollution abatement. For example,
firms may respond by altering production processes
to more efficiently reduce emissions.' Social cost
estimates, however, may include other costs not re-
flected in direct cost estimates (discussed below),
thereby offsetting the tendency for direct cost esti-
mates to overstate expenditures.
Measuring net welfare changes due to regulatory-
action requires either partial or general equilibrium
modeling. These more complicated approaches es-
timate social costs by accounting for a wider range
of market consequences associated with compliance
with pollution abatement requirements. The par-
tial equilibrium approach is particularly appropri-
ate when social costs are predominantly incurred in
directly affected markets. It requires modeling both
supply and demand functions in the affected eco-
nomic sector. Therefore, measures of social cost
reflect behavioral responses by both producers and
consumers in a specific market and do not necessar-
ily reflect how those changes affect related markets.
In cases where the regulatory action is known
to have an impact on many sectors of the US
economy, the general equilibrium model is a more
appropriate approach to estimating social costs. Like
the partial equilibrium model, the general equilib-
rium model estimates social costs by accounting for
direct compliance costs and producer and consumer
market behavior. The general equilibrium model
can capture first-order effects that occur in multiple
sectors of the economy, and may also provide in-
sight into unanticipated indirect effects in sectors that
might not have been included in the scope of a par-
tial equilibrium analysis.
The relationship of general equilibrium estimates
to estimates from the other two cost approaches is
not always clear. General equilibrium estimates have
a broader basis from which to estimate social costs
and can reflect the net welfare changes across the
full range of economic sectors in the U.S. Partial
equilibrium modeling tends to understate full social
costs because of its restricted scope (i.e., generally
limited to one industry). Total direct cost estimates
are likely to overstate costs in the primary market
because they do not reflect consumer and producer
responses. This is demonstrated in comparisons of
Morgcnstern et ai. (1998) estimate the ratio of incurred
abatement expenditures to estimated direct costs can be as low
as 0.8.
28
-------
Chapter 3: Direct Costs
estimates generated using a direct cost approach and
a partial equilibrium approach. The extent to which
a direct cost estimate will overstate or understate a
social cost estimate from a general equilibrium model
depends on the magnitude of the "ripple effects" in
economic sectors not targeted by a regulation.8
In the 812 retrospective analysis (FP A, 1997),
we recognized that the Clean Air Act has a perva-
sive impact on the US economy and opted for the
general equilibrium approach. The retrospective
nature of the analysis, however, provided us with
fairly well-developed historical data sets of goods and
service flows throughout the economy. These data
sets facilitated the development of detailed, year-bv-
year expenditures in all sectors of the economy, from
which we modeled producer and consumer behav-
ior and estimated net social costs. In the retrospec-
tive, our central estimate of total annualized direct
costs, from 1970 to 1990, was $523 billion. In com-
parison, we estimated the aggregate welfare effects
to be between |493 and $621 billion.9
For the prospective analysis, however, we adopt
a direct compliance cost approach. Although the
general equilibrium approach may represent a more
theoretically preferable method for measuring so-
cial costs, we use the simpler direct cost modeling
method for three reasons:
• First, we believe that the direct cost approach
provides a good first approximation of the
CAAA's economic impacts on various sec-
8 Current regulatory analyses that apply partial equilib-
rium modeling or general equilibrium modeling tend to mea-
sure costs with the assumption that markets are currently oper-
ating under optimally efficient conditions. Emerging literature
suggests that a hill accounting of the social costs and efficiency
impacts of environmental regulations could also include an as-
sessment of the incremental costs that reflect existing market
distortions, such as those imposed by the current tax code. The
distortions introduced by existing taxes, in combination with
new regulatory requirements, are collectively referred to as the
tax-interaction eflect. One oi the major conclusions of this
emerging literature is that, the social cost oi environmental policy
changes can be substantially higher when pre-existing tax distor-
tions are taken into account. Our direct cost estimates do not
reflect quantification of this effect, in part: because of the emerg-
ing nature of this literature and in part because existing esti-
mates of die magnitude of die tax-interaction effect are calcu-
lated as increments to social costs and are not necessarily appli-
cable adjustments to direct cost estimates.
9 Estimates are in 1990 dollars. The retrospective states,
"Tn general the estimated second order macroeconomic effects
were small relative to die size of die U.S. economy." The rate
of long term GNP growth between the control and no-control
scenarios amounted to roughly one-twentieth of one percent
less growth.
tors the U.S. economy. Comparison of the
direct cost approach to the partial equilib-
rium modeling suggests that the direct cost
approach likely overstates costs to the en-
tity that incurs the pollution control cost ex-
penditure. As discussed earlier, the direct
cost approach does not reflect adjustments
to prices and quantities that might mitigate
the effects of regulation. Recent research
analyzing ex ante and ex post cost estimates
of regulations suggests that ex anle analyses
are far more likely to overstate than under-
state costs.10 However, direct cost estimates
may also understate the effects of long-term
changes in productivity and the ripple effects
of regulation on other economic sectors that
are captured by a general equilibrium ap-
proach. The magnitude of those other ef-
fects, including potential magnification of
social costs by existing tax distortions, may
be substantial.
• Second, we believe that the closer approxi-
mation of social costs that might be gained
through a general equilibrium approach
could be compromised by the difficulty and
uncertainty associated with projecting future
economic and technological changes. The
general equilibrium approach could provide
many insights that the direct cost approach
cannot, but also introduces a significant level
of additional uncertainty.
• Third, the focus of the present analysis is a
comparison of direct costs and direct ben-
efits. To provide a balanced treatment of
costs and benefits in a general equilibrium
framework, the social cost model must be
designed and configured to reflect the indi-
rect economic consequences of both costly
and beneficial economic effects. None of
the general equilibrium models available in
the timeframe of this study could be config-
ured to support effective analysis of the full
range of specific direct costs and, especially,
direct benefits of the 1990 Clean Air Act
Amendments.
A° See, for example, Harrington et al (1999), referenced in
Appendix 13, for a comparative analysis of ex ante and ex post
regulatory cost estimates.
29
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
• Fourth, undertaking a general equilibrium
modeling exercise remains a very resource-
intensive task. For the purposes of compar-
ing costs to benefits we concluded that more
detailed modeling would not be the most
cost-effective use of the project resources.
Uncertainty in the Cost
Estimates
Overview
As we note at the beginning of this chapter, ex-
plicit and implicit assumptions regarding changes in
consumption patterns, input costs, and technologi-
cal innovation are crucial to framing the question of
the CAAA's cost impact. Given the nature of this
prospective study, there is no way to verify the ac-
curacy of the assumptions applied to future scenarios.
We can envision other plausible analyses with esti-
mates that differ from results in this chapter. More-
over, for many of the factors contributing to uncer-
tainty, the degree or even direction of the bias is
unknown or cannot be determined. Nevertheless,
uncertainties and/or sensitivities can be identified
and in many cases the potential measurement errors
can be quantitatively characterized. In this section
of the chapter, we first discuss several quantitative
sensitivity analyses undertaken to characterize the
impact of key assumptions on the ultimate cost analy-
sis. We conclude the chapter with a qualitative dis-
cussion of the impact of both quantified and
unquantified sources of uncertainty.
Quantitative Sensitivity Tests
In order to characterize the uncertainty in the
cost estimates, we conducted sensitivity analyses on
the key parameters and analytic assumptions of six
major provisions. The provisions are the following:
• Progress Requirements,
• California Reformulated Gasoline,
• PM NAAQS Controls,
• LEV program (the National and California
programs combined),
• Non-utility Stationary Source NO Con-
trols, and
• NOx Tailpipe/Extended Useful Life Stan-
dard.
We selected these provisions because they are
among the most significant sources of CAAA costs,
yet cost estimates for each of the provisions incor-
porate significant uncertainties. Collectively, these
provisions account for nearly 50 percent of total di-
rect compliance cost estimates for 2010. Table 3-2
summarizes the methods we used to conduct the cost
sensitivity analyses and the results.
For each test, we developed three estimates for
one or more components of costs affecting the total
cost estimate for a given provision: (1) a central esti-
mate, equal to the 2010 primary cost estimate re-
ported in this chapter11, (2) a low estimate; and (3) a
high estimate. The low and high estimates assess
the potential magnitude of the effect of the
component(s) on the provision's costs and conse-
quently, total CAAA costs, using reasonable alter-
native assumptions for each cost component. For
progress requirements, PM NAAQS controls, and
stationary source NOx controls, the cost projections
are based on models of future emissions controls.
Accurately identifying the set of adopted controls is
a key source of uncertainty. For example, cost-ef-
fective control measures for complying with progress
requirements have not yet been identified and the
sensitivity test suggests the potential for substantial
variability in progress requirement compliance costs.
In the case of motor vehicle provisions, there are
two significant sources of uncertainty, projecting
future car sales and forecasting accurate per vehicle
costs.
The results indicate that the sensitivity of our
primary cost estimates (central estimates) is not uni-
form across provisions. In addition, low and high
estimates may vary by as much as a factor of two.
These sensitivity analyses demonstrate the potential
effect of altering selected assumptions and data. We
do not assign probabilities to the likelihood of the
alternative. In other wrords, it would be inappropri-
ate simply to add up the array of low7 and high esti-
mates to arrive at an overall range of uncertainty
around the central estimates, because it is unlikely
that a plausible scenario could be constructed where
all the estimates are concurrently either at the high
11 The one exception is the central estimate of progress
requirements. Our sensitivity analysis which is based on more
recent, cosl information indicates thai our primary estimate is
more reflective oi a high estimate. See Appendix B for more
details.
30
-------
Chapter 3: Direct Costs
or low end of their individual plausible ranges. A
better interpretation of these results is that uncer-
tainty in key input parameters can have a significant
effect 011 the overall uncertainty of our estimates of
direct compliance costs and ultimately the net ben-
efits calculation.
In addition to examining specific provisions, we
conducted a sensitivity analysis of the cost of capital
used throughout the analysis. Cost estimates pre-
sented earlier in this chapter reflect application of a
cost of capital (for the purposes of annualizing total
capital costs) of five percent. We also examined the
effect on cost estimates for those provisions which
involve significant capital expenditures and where
we could recalculate annualized costs from the avail-
able information. These provisions include non-util-
ity and area source estimates for YOC, NO , and
PM control. The alternative estimates use three and
seven percent for the cost of capital. Results indi-
cate that cost estimates are only moderately sensi-
tive to the discount rate. The provisions evaluated
have a total annualized capital cost of approximately
S3 billion in 2010. Varying the cost of capital gener-
ated alternative estimates of |2.8 billion (three per-
cent) and $3.1 billion (seven percent).12
Qualitative Analysis of Key Factors
Contributing to Uncertainty
There are a wide range of other factors which
contribute to uncertainty in the overall cost esti-
mates. In most cases, the effect of these other fac-
tors cannot be quantitified, though some may have
significant influences on our overall net benefits es-
timate. We present a summary of these factors in
Table 3-3 below, and provide a characterization of
the potential effect of each uncertainty on the pri-
mary estimate of the net benefits (i.e., if costs are
overestimated, net benefits are underestimated). The
two most important factors are the potential impact
of innovation on the ultimate control costs incurred
and the conservative assumptions we made to esti-
mate RFP costs.
:2 Note that these calculations reflect the use of alternative
discount rates to estimate annual costs. Hie use of alternative
rates to calculate the total net present value of costs incurred
through the full 1990 to 2010 study period is examined sepa-
rately in Chap Let: 8, where we compare total costs to total ben-
efits.
The regulatory documents which provide cost
inputs to ERCAM and the IPM contain the most
recent data available, given existing technological
development. Between 2000 and 2010, however,
advancements in control technologies will allow
sources to comply with CAAA requirements at
lower costs. For example, we anticipate technologi-
cal improvements for complying with the multiple
tiers of proposed emission standards during the
phase-in of nonroad engine controls will likely lead
to reduced costs. In addition, the costs for certain
control equipment may decrease over tune as demand
increases and technology innovation and competi-
tion exert downward pressure on equipment prices.
For instance, selective catalytic reduction (SCR )
costs have decreased over the past three years as more
facilities begin to apply the technology. We also
believe that even in the absence of new emission stan-
dards, manufacturers will eventually upgrade engines
to improve performance or to control emissions
more cost-effectively; firms will institute technolo-
gies such as turbocharging, aftercooling, and vari-
able-valve timing, all of which improve engine per-
formance.
There is considerable uncertainty surrounding
the development of States' control plans for meet-
ing ozone NAAQS attainment requirements. We
base the RFP cost estimate on the assumption that
ozone nonattainment areas (NAAs) will take credit
for NO reductions for meeting progress require-
ments. Additional area-specific analysis would be
necessary to determine the extent to which areas find
NOx reductions beneficial in meeting attainment and
progress requirement targets. Trading of NOx for
VOC to meet RFP requirements may result in dis-
tributions of YOC and NO emission reductions
x
which differ from those used in this analysis. In
response to these uncertainties, we adopted a con-
servative strategy for estimating the costs of RFP
reductions in the primary analysis. We use a rela-
tively high cost per ton reduced estimate of $10,000
for all required reductions. Since the time we con-
ducted our primary cost analysis more information
has emerged suggesting controls could cost much less,
perhaps as little as $3,500 (see Table 3-2 and Appen-
dix B for more details). In our sensitivity analysis of
this variable, we incorporate the more recent cost
per ton estimates. The analysis suggests that the
$10,000 per ton reduced may in fact be more repre-
31
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 3-2
Results of Quantitative Sensitivity Tests
Provision
Primary Cost
Estimate in 20101
(billions 1990 $)
Strategy for Sensitivity Analysis
Range of Estimates
from Sensitivity Test
(billions 1990 $)
Progress
Requirements
$2.46
Vary unit costs for unidentified
measures
$1.07 -$2.46
(central, $1.15)
California
Reformulated
Gasoline
$2.45
Vary incremental fuel costs and
gasoline sales estimates
$1.4-$3.5
PM NAAQS Controls
$2.22
Vary model attainment plan
assumptions and cost per ton
estimates
$0.09 to $3.35
LEV costs (California
and National
Combined)
$2.16
Vary per vehicle costs and
projections of vehicle sales
$1.08 -$2.48
Non-Utility Stationary
Source NOx Costs
$2.15
Vary unit-level cost per ton
$1.1 -$3.2
NOx Tailpipe/Useful
Life Standards
$1.65
Vary per vehicle costs and vehicle
sales data
$0.83 -$2.48
Note:
1 In all cases, except progress requirements, the Post-CAAA 2010 primary cost estimates is equal to the central
estimate in the sensitivity analysis. For more details on the sensitivity analysis of progress requirements and other
provisions, see Appendix B.
scntativc of an upper bound cost estimate, rather than
a central estimate as our primary cost analysis re-
flects. The result of our conservative approach indi-
cates that we may overstate RFP costs by a factor of
two in 2010.
One other factor is also worth noting, although
its impact is likely to be less important than the pre-
vious two factors. Under the 1990 CAAA, EPA
created economic incentive provisions in several rules
to provide flexibility for affected facilities that com-
ply with the rules. These provisions include bank-
ing, trading, and emissions-averaging provisions.
Flexible compliance provisions tend to lower the cost
of compliance. For example, the emissions-averag-
ing program grants flexibility to facilities affected
by the marine vessels rule, the petroleum refinery
National Emission Standard for Hazardous Air Pol-
lutants (NRSHAP), and the gasoline distribution
NESHAP. These facilities can choose which sources
to control, as long as they achieve the required over-
all emissions reduction. In many of the cost analy-
ses, EPA does not attempt to quantify the effect that
economic incentive provisions will have on the over-
all costs of a particular rule. In these cases, to the
extent that affected sources use economic incentive
provisions to minimize compliance costs, costs may
be overstated. The major trading programs autho-
rized under the Amendments, however, governing-
sulfur and nitrogen oxide emissions reductions from
utilities and major non-utility point sources, are re-
flected in the cost estimates presented here.
32
-------
Chapter 3: Direct Costs
Table 3-3
Key Uncertainties Associated with Cost Estimation
Potential Source of Error
Direction of
Potential Bias
for Net Benefits
Likely Significance Relative to Key Uncertainties
on Net Benefits Estimate
Costs are based on today's
technologies. Innovations
in future emission control
technology and
competition among
equipment suppliers tend
to reduce costs overtime.
Underestimate Probably minor. Available evidence suggests that estimates
of pollution control costs based on current engineering can
substantially overestimate the ultimate cost incurred,
resulting in understating net benefits.2
Uncertainty of final State
strategies for meeting
Reasonable Further
Progress (RFP)
requirements.
Underestimate Probably minor. We apply a conservative estimate for costs
of RFP measures. Available evidence for identified RFP
measures suggests costs could be as much as 70 percent
lower than this value. The bias most likely results in
significantly understating net benefits.
Errors in emission
projections that form the
basis of selecting control
strategies and costs in
both the IPM and ERCAM
models.
Unable to
determine based
on current
information
Probably minor. In many cases, emissions reductions are
specified in the regulations, suggesting that errors in the
estimation of absolute levels of emissions under Pre- and
Post-CAAA scenarios may have only a small impact on cost
estimates. The effect on net benefits is unknown.
Exclusion of the impact of
economic incentive
provisions, including
banking, trading, and
emissions averaging
provisions.
Underestimate Probably minor. Economic incentive provisions can
substantially reduce costs, but the major economic programs
for trading of sulfur and nitrogen dioxide emissions are
reflected in the analysis.
Incomplete
characterization of certain
indirect costs, including
vehicle owner opportunity
costs associated with
Inspection and
Maintenance Programs
and performance
degradation issues
associated with the
incorporation of emission
control technology.
Overestimate Probably minor. Preliminary evidence suggests that the
opportunity costs of vehicle owners is most likely small
relative to other cost inputs.3 In addition, it is will vary from
State to State and is subject to a variety of influencing
factors. The potential magnitude of indirect costs associated
with performance degradation is more uncertain, because
few data currently exist to quantify this effect.
33
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 3-3 (continued)
Key Uncertainties Associated with Cost Estimation
Direction of
Potential Bias Likely Significance Relative to Key Uncertainties
Potential Source of Error for Net Benefits on Net Benefits Estimate1
Choice to model direct
costs rather than social
costs
Unable to
determine based
on current
information
Probably minor. The relationship of social cost to direct cost
estimates is influenced by multiple factors that operate in
opposite directions, suggesting the magnitude ofthe net
effect is reduced. Social cost estimates can reflect the net
welfare changes across the full range of economic sectors in
the U.S, and so may yield higher estimates of costs than a
direct cost approach. In addition, social cost estimates can
be constructed to reflect the potentially substantial cost-
magnifying effect of existing tax distortions. Direct cost
estimates, however, are likely to overstate costs in the
primary market because they do not reflect consumer and
producer responses. The extent to which a direct cost
estimate will overstate or understate a social cost estimate
depends on the magnitude ofthe "ripple effects" in economic
sectors not targeted by a regulation. In addition, assessment
ofthe effect on net benefit estimates must also account for
any economy-wide effects of direct benefits (e.g., the broader
implications of improving health status, and improving
environmental quality).
Use of costs for rules that
are currently in draft form
(i.e., not yet finalized).
Unable to
determine based
on current
information
Probably minor. Rules that are most important to the overall
cost estimate are largely finalized. For example, there is
some uncertainty as to how the cap-and-trade program
through the SIP process will lower NOx emissions in an
efficient manner. The expected effect on net benefits is
minimal.
Exclusion of costs of 7-
year and 10-year MACT
standards and the
residential risk standards
for the 2- and 4-year
MACT standards.
Unable to
determine based
on current
information
Probably minor. Costs for the 7- and 10-year MACT
standards are likely to be less than for the 2- and 4-year
standards included in the analysis and the need for, and
potential scope and stringency of, future Title III residual risk
standards remain highly uncertain. For consistency, benefits
ofthe 7- and 10-year standards and the residual risk
standards are also excluded.
Note:
1 The classification of each potential source of error reflects the best judgement ofthe section 812 Project Team. The
Project Team assigns a classification of "potentially major" if a plausible alternative assumption or approach could influence
the overall monetary benefit estimate by approximately five percent or more; if an alternative assumption or approach is
likely to change the total benefit estimate by less than five percent, the Project Team assigns a classification of "probably
minor."
2 For more detail, see Harrington et al (1999), referenced in Appendix B.
3 Preliminary evidence based on Arizona's Enhanced l/M program indicates that major components ofthe programs costs
are associated with test and repair costs rather than the costs of waiting and travel for vehicle owners. (Harrington and
McConnell, 1999.) To date, Enhanced l/M programs have been implemented in only four States.
34
-------
Air Quality
Modeling
Air quality modeling links changes in emissions
to changes in the atmospheric concentrations of pol-
lutants that may affect human health and the envi-
ronment. A crucial analytical step, air quality mod-
eling is one of the more complex and resource-in-
tensive components of the prospective analysis. This
chapter outlines how we estimated future-year pol-
lutant concentrations under both the Pre- and Post-
er AAA scenarios using air quality modeling results
and ambient monitor data. The first section of the
chapter begins with a discussion of some of the chal-
lenges faced by air quality modelers and a brief de-
scription of the models we used in this analysis. The
following section provides an overview of the gen-
eral methodology we used to estimate future-year
ambient concentrations. This methodology section
includes a description of how we used modeling re-
sults to adjust monitor concentration data and esti-
mate ambient concentrations for the years 2000 and
2010. The third section of this chapter summarizes
the results of the air quality modeling and presents
the expected effects of the CAAA on future-year
pollutant concentrations. A discussion of the key
uncertainties associated with air quality modeling
concludes the chapter.
Overview of Air Quality
Models
Air quality modelers face two key challenges in
attempting to translate emission inventories into pol-
lutant concentrations. First, they must model the
dispersion and transport of pollutants through the
atmosphere. Second, they must model pertinent at-
mospheric chemistry and other pollutant transfor-
mation processes. These challenges are particularly
acute for those pollutants that are not emitted di-
rectly, but instead form through secondary processes.
Ozone is the best example; it forms in the atmo-
sphere through a series of complex, non-linear chemi-
cal interactions of precursor pollutants, particularly
certain classes of volatile organic compounds (VOCs)
and nitrogen oxides (NOj. We faced similar chal-
lenges when estimating I'M concentrations. Atmo-
spheric transformation of gaseous sulfur dioxide and
nitrogen oxides to particulate sulfates and nitrates,
respectively, contributes significantly to ambient
concentrations of fine particulate matter. In addi-
tion to recognizing the complex atmospheric chem-
istry relevant for some pollutants, air quality mod-
elers also must deal with uncertainties associated with
variable meteorology and the spatial and temporal
distribution of emissions.
Air quality modelers and researchers have re-
sponded to the need for scientifically valid and reli-
able estimates of air quality changes by developing a
number of sophisticated atmospheric dispersion and
transformation models. Some of these models have
been employed in support of the development of
federal clean air programs, national assessment stud-
ies, State Implementation Plans (SIPs), and individual
air toxic source risk assessments. In this analysis,
we used several of these well-established models to
develop a picture of future changes in air quality re-
sulting from the implementation of the 1990 CAAA.
We focused our air quality modeling efforts on
estimating the impact of Pre- and Post-CAAA emis-
sions on future-year ambient concentrations of
ozone, PM PM SO., NO , and CO and on fu-
7 1 (r ^,.5' 2y x3
ture-year acid deposition and visibility. The ideal
model for this analysis would be a single integrated
air quality model capable of estimating ambient con-
centrations for all criteria pollutants throughout the
U.S. Although HP A is working to develop such a
model, at the time of this analysis the model was not
sufficiently developed and tested. In the absence of
a single integrated model, we employed the Urban
Airshed Model (UAM) in our analysis of ozone and
used both the Regional Acid Deposition Model/Re-
gional Particulate Model (RADM/RPM) and the
Regulatory Modeling System for Aerosols and Acid
35
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Deposition (REMSAD) model to assess PM10, PM25,
acid deposition and visibility. All three of these mod-
els are three-dimensional grid models which require
emissions and meteorological data as input. Each of
these models calculate pollutant concentrations by
simulating the physical and chemical pollution for-
mation processes that occur in the atmosphere.
We conducted separate UAM, RADM/RPM,
and REMSAD model runs for the 1990 base-year
and each future-year projection scenario. The pri-
mary model input used for each run consisted of
emissions estimates corresponding to the year and
scenario being modeled (as described in Chapter 2
and Appendix A) and historical meteorological data
corresponding to a past time period, referred to as a
simulation period. We selected previous ozone epi-
sodes, i.e., multi-day events characterized by weather
conditions conducive to ozone formation and trans-
port (and as a result, characterized by multi-day spans
with higher than average ozone concentrations), to
serve as the simulation periods for UAM model runs.
Although ozone concentrations during these simu-
lation periods exceed the seasonal average, because
the simulation periods for both the eastern and west-
ern U.S. cover roughly a two week span, ozone con-
centration peaks are largely offset by the surround-
ing lows. Overall, the selected simulation periods
reasonably represent summertime ozone forming
meteorological conditions and ozone concentrations.
RADM/RPM simulation periods used to model PM,
acid deposition, and visibility were chosen using a
random selection process, while separate simulation
periods at the beginning of each of the four seasons
were chosen for REMSAD.
Table 4-1 provides an overview of the air qual-
ity models used in this analysis. We modeled con-
centrations of all pollutants across the 48 contigu-
ous states; however due to the lack of an integrated
model, separate air quality models were used to esti-
mate ozone and PM for the eastern and western U.S.
Table 4-1 shows the domain for each model and the
simulation periods selected for use with each model
and provides an overview of the spatial resolution
of the models used as part of this analysis. The finer
the resolution (i.e., the smaller the grid cells) the
better the model can capture the effects of localized
changes in emissions and wreather conditions on
ambient air quality. Recognizing the relationship
between grid cell resolution and the certainty of re-
sults, we endeavored to estimate pollutant concen-
trations in more populated areas using higher reso-
lution models. For this reason, we used the fine grid
UAM-IY, an urban-scale model, to estimate ambi-
ent ozone levels in selected western cities. Similarly,
we used an intermediate resolution grid (12 km x 12
km) to model ozone in "inner O F AG" states where
population density is high and ozone transport is a
major problem.1
Using the three-dimensional grid cell models,
UAM, RADM/RPM, and REMSAD, we estimated
grid-cell specific, hourly ozone and daily PM and
PM, concentrations for each day of the relevant
simulation periods. We conducted separate model
runs for the 1990 base-year and 2000 and 2010 fu-
ture-year Pre- and Post-CAAA scenarios. Using
these results, we ultimately projected the impact of
the CAAA on ozone and PM ambient levels.
We relied on the same models used to predict
PM concentrations to estimate changes in future-year
acid deposition and visibility. For each model grid-
cell we predicted daily acid deposition levels and vis-
ibility. Estimates for each day of the simulation
period were generated for the base-year and both
projection years under the Pre- and Post-CAAA sce-
narios.
Wre estimated future-year Pre- and Post-CAAA
ambient S02, NO, NO,, and CO concentrations by
adjusting 1990 concentrations using future-year to
base-year emissions ratios. This technique assumes
a linear relationship between changes in emissions
in an area and changes in that area's ambient con-
centration of the emitted pollutant.2 Although this
technique does not take into account pollutant trans-
port or atmospheric chemistry, we believe linear scal-
ing generates reasonable approximations of ambient
concentrations of gaseous pollutants such as SO,,
NO , and CO.
x5
1 The Oxone Transport Assessment Group (OTAG) con-
sists of the 37 easternmost states and the District of Columbia.
Die "inner OTAG" region is comprised of the more eastern
(and more populated) states within the OTAG domain.
It is important to emphasize that she correlation expected
is between changes in emissions and changes in air quality. Di-
rect correlations between the absolute emissions estimates and
empirical air quality measurements used in the present analysis
may not be strong due Lo expected inconsistencies between the
geographically local, monitor proximate emissions densities al-
lecting air quality data.
36
-------
Chapter 4: Air Quality Modeling
Table 4-1
Overview of Air Quality Models
Air Quality
Measure
Region
Model
Spatial Resolution
Simulation Period
Ozone
Eastern
U.S.
UAM-V
a) 12 km x 12 km grid for "Inner
OTAG Region"
b) 36 km x 36 km grid for remainder
of 37-state OTAG region
July 20-30, 1993 and July 7-
18, 1995
Ozone
Western
U.S.
UAM-V
56 km x 56 km grid (regional scale)
covering the 11 westernmost states
(states west of North and South
Dakota, including western Texas)
July 1-10, 1990
Ozone
San
Francisco
Bay Area
UAM-IV
4 km x 4 km (urban scale) grid
covering the San Francisco Bay
Area, the Monterrey Bay Area,
Sacramento, and a portion of the
San Joaquin Valley
Aug. 3-6, 1990
Ozone
Los
Angeles
Area
UAM-IV
5 km x 5 km grid covering the
South Coast Air Basin from Los
Angeles to beyond Riverside and
including part of the Mojave Desert
June 23-25, 1987 and Aug.
26-28, 1987
Ozone
Maricopa
County
(Phoenix)
Area
UAM-IV
4 km x 4 km grid covering
urbanized portion of Maricopa
County
Aug. 9-10, 1992 and June 13-
14, 1993
Particulate Eastern RADM/RPM 80 km x 80 km grid (coarse 30 randomly selected 5-day
Matter U.S. resolution) covering eastern North periods spanning a four-year
America from the Rocky Mountains period
eastward to Newfoundland,
Canada and the Florida Keys (see
Fig. C-14 in Appendix C)
Particulate Western REMSAD 56 km x 56 km grid covering the 11 ten-day period for each of four
Matter U.S. westernmost states seasons:
May 1-10,
July 1-10,
Oct. 1-10, and
Dec. 1-10
Visibility
Eastern
U.S.
RADM/RPM
(same as PM)
(same as PM)
Visibility
Western
U.S.
REMSAD
(same as PM)
(same as PM)
Acid
Deposition
Eastern
U.S.
RADM
(same as RADM/RPM)
(same as RADM/RPM)
Sulfur
Dioxide
U.S.
linear scaling
56 km x 56 km REMSAD grid
covering 48 contiguous states
not applicable
Oxides of
Nitrogen
U.S.
linear scaling
56 km x 56 km REMSAD grid
covering 48 contiguous states
not applicable
Carbon
Monoxide
U.S.
linear scaling
56 km x 56 km REMSAD grid
covering 48 contiguous states
not applicable
37
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
General Methodology
The air quality modeling component of the 812
prospective analysis involved the application of a
variety of complex, sophisticated air quality model-
ing tools and techniques. Overall, however, the
method we used to estimate the impact of changes
in emissions on air quality was relatively straight
forward. We began by gathering 1990 air quality
monitor data for the six criteria pollutants analyzed
in this study. These observational data served as the
air quality baseline for both the Pre- and Post-CAAA
scenarios. We then estimated 2000 and 2010 con-
centrations of each pollutant under each emissions
scenario by applying adjustment factors to the 1990
monitor data. The adjustment factors for each fu-
ture-year projection scenario were based on the rela-
tive change in pollutant concentration between 1990
and the desired future-year, as predicted by air qual-
ity simulation modeling. This section presents an
overview of the methodology we used to estimate
future-year ambient concentrations. For a more
detailed description, please refer to Appendix C.
The diagram in Figure 4-1 illustrates the meth-
odology used to estimate ozone and PM concentra-
tions. First, we compiled distributions of observed
pollutant concentrations recorded at each air qual-
ity monitor in 1990. We obtained these data from
EPA's Aerometric Information Retrieval System
(AIRS), a publicly accessible database of air quality
information. Separately, we then developed distri-
butions of estimated concentrations for each pollut-
ant in 1990 using 1990 emissions data and the appro-
priate air quality model. Unlike the 1990 observed
concentrations that were measured at monitoring
sites, the 1990 estimated concentrations were calcu-
lated at the center of each cell of a grid covering the
domain of the applicable air quality model. Using
future-year emission inventory estimates for the Pre-
CAAA and Post-CAAA scenarios (developed as de-
scribed in Chapter 2 and Appendix A) and the ap-
propriate air quality models, we next developed dis-
tributions of model-estimated concentrations at each
grid cell for each of four future-year projection sce-
narios: 2000 Pre-CAAA, 2010 Pre-CAAA, 2000 Post-
CAAA, and 2010 Post-CAAA. These results were
used to derive adjustment factors for each air qual-
ity monitor, based on the simulation results for the
grid cell in which the monitor is located. The fu-
ture-year/scenario adjustment factor for each pol-
lutant represents the ratio of the simulated future-
year/scenario concentration to the 1990 model-esti-
mated concentration. These factors were calculated
by matching future-year and 1990 concentrations at
regular intervals in each distribution. Finally, four
sets of model-derived adjustment factors were applied
to the distribution of observed 1990 concentrations
at each monitor to forecast distributions of concen-
trations for each of the four future-year projection
scenarios. It is these concentrations that serve as
inputs into the CAAA benefits modeling.
An illustrative example follows. Assume the
median observed concentration of Pollutant A at
Monitor X in 1990 was 0.24 ppm. Air quality mod-
eling for the grid cell in which Monitor X is located
predicts a median Pollutant A concentration of 0.30
ppm in 1990 and 0.15 ppm in 2010 under the post-
CAAA scenario. The 2010 Post-CAAA adjustment
factor for the median Pollutant A concentration
would be 0.5, and the predicted 2010 Post-CAAA
median concentration at Monitor X would be 0.5
(=0.15/0.30) times the 1990 monitor value of 0.24
ppm, or 0.12 ppm.
Our approach for forecasting concentrations of
SO,, NO , and CO involved the same basic approach
described above. However, instead of applying
model-derived adjustment factors to the 1990 ob-
served distribution of concentrations, we adjusted
the 1990 distribution using the ratio of future-year
emissions to 1990 emissions in the vicinity of the
monitor for each of the four future-year projection
scenarios. For more information about this ap-
proach, please refer to Appendix C.
38
-------
Chapter 4: Air Quality Modeling
Figure 4-1
Schematic diagram of the future-year concentration estimation methodology
1990 2000 2010
(Base Case)
AQ Model
Predictions
Cone
crtyl
Cone
c/'fj
countyl
Ratio to Base Case
Adjustments
Factors
cityl
AQ
Observations
cone
cone
county1
cone
AQ
Results
pre-CAAA|
post-CAAA
cone
cone
county1
Ratio to Base Case
www
post-CAAA
cone
countyl
Concentration
distributions
Concentration
distributions
Concentration
distributions
NOTE: Figure illustrates how model results and observations are used to produce the air quality profiles (concentration distributions) for the
benefits analysis. The figure shows model runs at the top; four sets of "ratios" of model results in space in the middle; and frequency
distributions of pollutant monitor concentrations and the space-dependent scaling ofthese by the ratios ofthe model predictions on the bottom.
Air Quality Model Results
This section presents a summary representation
of the air quality modeling results. We discuss the
model-simulated concentration estimates and the
adjusted future-year concentration predictions with
a focus on the change in air quality resulting from
the implementation of the 1990 CAAA.
Ozone
We modeled ozone concentrations separately for
the eastern U.S., western U.S., San Francisco Bay-
area, Los Angeles area, and Maricopa County (Phoe-
nix, AZ) area. Examination of base-year and future-
year model concentration estimates shows expected
increases in Pre-CAAA ozone concentrations and
expected decreases in Post-CAAA ozone concentra-
tions in the eastern U.S. In this part of the country,
UAM-V predicts Pre-CAAA ozone concentration
increases will occur primarily over the states of Vir-
ginia, North Carolina, Kentucky, Tennessee, Geor-
gia, and Alabama; while Post-CAAA decreases will
be more widespread. Comparison of Pre- and Post-
CAAA model estimates shows that, with the excep-
tion of a few isolated areas, ambient ozone levels
throughout the East will be reduced in the year 2010
as a result of the CAAA. These lowrer levels are
largely due to significant reductions in area source
and motor vehicle VOC emissions and utility, point
source, and motor vehicle NO emissions.
5 X
Regional-scale model results for the western U.S.
indicate that ozone concentrations in this portion
of the country, just as in the eastern U.S., will gen-
erally increase from the 1990 base-year under the
Pre-CAAA scenario and decrease from 1990 levels
under the Post-CAAA scenario. In the West, we
anticipate widespread changes under both scenarios;
however, wc project that the increases in Pre-CAAA
ozone concentrations and decreases in Post-CAAA
model concentrations will be smaller than the pre-
39
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
dieted changes in ambient ozone lev-
els in the eastern U.S Furthermore,
comparison of 2010 Pre- and Post-
CAAA model estimates shows that
future-year western ozone concentra-
tions will be lower as a result of the
1990 Amendments, but UAM-V re-
sults indicate that the reductions in the
West will likely be about half the size
of the reductions in the eastern por-
tion of the country. The difference
between the change in western ozone
concentrations and the change in east-
ern ozone concentrations is largely
due to the more aggressive NOx con-
trols expected in the East. Specifi-
cally, the Post-CAAA scenario incorporates the ef-
fects of a NO cap-and-trade system for the eastern
U.S. (OTAG region). Another reason for the dif-
ference between the modeled change in eastern and
western ozone concentrations is that we estimated
ozone levels in the East and West using different
model grid resolutions. The coarser the resolution,
the less responsive the model concentration estimates
are to localized changes in emissions. Thus, the
smaller estimated change in western ozone concen-
trations than in eastern ozone concentrations may,
in part, be attributable to the fact that UAM-V grid-
cells covering the western U.S. are larger than those
covering the eastern U.S.
Western urban-area modeling results differ from
the regional scale results described above. Examina-
tion of Pre- and Post-CAAA modeling estimates
shows that, in some portions of the urban centers of
San Francisco and Los Angeles, future-year Post-
CAAA ozone concentrations are expected to be
higher than Pre-CAAA estimates. This ozone
"disbenefit" is the result of inhibiting a complex
chemical reaction termed "NO scavenging," during
which a reduction in NOx, an ozone precursor, leads
to an increase in ozone production instead of the
typical decrease.3 In the area immediately surround-
ing the two cities, however, and in Maricopa County,
3 Scavenging occurs in areas, typically cities, with limited
VOC and abundant NO . In VOC-limited areas where I he re is
a relatively high NO concentration (regions where the concen-
tration of VOC, not NO.., dictates the amount of ozone that
can he formed), these two ozone precursors (VOC and NOj
compete to react with a particular gaseous compound. To pro-
duce ozone, this compound must combine with VOC. As a
result, if the compound joins with NO , ozone production is
impeded; thus, a decrease in NO., leads to an increase in ozone
concentrations.
Figure 4-2
Distribution of Monitor Level Ratios for 95th Percentile Ozone
Concentrations: 2010 Post-CAAA/Pre-CAAA
median: 0.883
ll. 20
1 10
0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.30
Ratio
model results show that scavenging is not expected
to be influential, if it occurs at all, and future-year
Post-CAAA ozone concentration estimates are pre-
dicted to be lower than Pre-CAAA estimates.
As described above, we used the UAM-V model
results to calculate adjustment factors for each of the
four future-year projection scenarios. Wre estimated
future-year monitor-level ozone concentrations by
applying these factors to 1990 observed concentra-
tions. Examination of the distribution of adjusted
monitor concentration ratios for 95th percentile
ozone concentrations is one means of analyzing the
impact of the CAAA on air pollution. The distri-
bution of ratios of 2010 Pre-CAAA to 1990 base-
year ozone concentrations reveals that the majority
of future year Pre-CAAA ozone concentration esti-
mates are between zero and 10 percent greater than
1990 levels, with most concentrations falling in the
middle of this range. The distribution of ratios of
2010 Post-CAAA to 1990 base-year shows that in
nearly all areas of the U.S. ozone concentrations will
be lower in 2010 than in the base-year; in the major-
ity of the country, future-year concentrations will
be five to 20 percent lower than in the base-year.4
The histogram in Figure 4-2 depicts the distribution
of ratios of 2010 Post-CAAA ozone estimates to 2010
Pre-CAAA ozone estimates. Most of the ratios in
the distribution are less than one, with a median of
0.883. This indicates that the 95th percentile level
Post-CAAA concentrations, with few exceptions, are
lower than the corresponding Pre-CAAA values.
The smaller the ratio, the greater the difference be-
tween future-year scenarios.
4 See Appendix C for histograms illustrating the change ill
ozone concentrations Irom the base-year.
40
-------
Chapter 4: Air Quality Modeling
Particulate Matter
To model Pre- and Post-CAAA particulate mat-
ter (PM1() and PM25) concentrations, we used
RADM/RPM for the eastern U.S. and REMSAD
for the western U.S. Results from both models show
PM concentrations are expected to be lower under
the Post-CAAA scenario than under the Pre-CAAA
scenario, 'l'his projected improvement in air quality
is widespread throughout the eastern U.S., with 2010
Post-CAAA PM estimates in some parts of the East
up to 15 to 30 percent lower than 2010 Pre-CAAA
estimates. In the West, projected reductions in fu-
ture-year PM concentrations (Pre-CAAA minus
Post-CAAA) are largely restricted to urban areas.5
The broad scale improvement in eastern PM con-
centrations is driven largely by reductions in utility
source sulfur dioxide emissions throughout this por-
tion of the country.6 In the West, however, sulfur
dioxide emissions have a much smaller impact on
overall PM concentrations. Western PM concen-
trations are more significantly influenced by area,
motor vehicle, and nonroad source emissions of ni-
trogen oxides and directly emitted PM. These
sources are more concentrated in urban areas. As a
result, the impact of the CA A A on PM concentra-
tions in the West is primarily restricted to urban
areas.
Examination of the distribution of adjusted
monitor-level concentration ratios for annual aver-
age PM concentrations reveals that 2010 Pre-CAAA
PM10 and PM25 estimates are both higher than 1990
base-year estimates in almost all areas of the coun-
try. Pre-CAAA 2010 PM]0 and PM, 3 estimates are
generally zero to 10 percent greater than 1990 base-
year estimates. The average estimated increase m
PM, concentrations, however, is slightly larger than
the average estimated increase in PM1Q.7 The esti-
mated change in PM concentrations from the base-
year to 2010 under the Post-CAAA scenarios is less
uniform. While the majority of areas experience a
reduction in annual average PM and PM concen-
trations, in a number of areas ambient PM levels,
more frequently PM, increase from the base-year
under the Post-CAAA scenario. On average, how-
ever, 2010 Post-CAAA PM10 and PM2 5 concentra-
tions are between zero and five percent and zero
and 10 percent, respectively, lower than 1990 base-
year concentrations.8
As shown in Figures 4-3 and 4-4, the percentage
reduction 111 PM, 5 concentrations across the U.S.
between the Pre- and Post-CAAA scenarios vary
more widely than the percentage reduction in PM .
In the emissions analysis we focus on the impact of
the CAAA on anthropogenic emissions and, so, hold
natural source PM emissions constant at 1990 levels.
Natural source emissions make up a much larger
portion of PM concentrations than PM concen-
trations and dampen the influence of changes in an-
thropogenic emissions on ambient PM1() concentra-
tions.
Comparison of the twro distributions in Figures
4-3 and 4-4 shows that, despite the greater variation
of PM reductions, the percentage reduction in PM
concentrations are larger on average than the per-
centage reduction in PM10 concentrations. The rea-
son for this difference is two fold. First, as described
above, PM, 5 concentrations are more susceptible to
the influence of changes in anthropogenic emissions,
which are regulated by the CAAA. Second, the
CAAA provisions that influence PM emissions (regu-
lations that focus on secondary PM precursors such
as NOx, and SO,, and primary PM sources such as
diesel engine exhaust standards) affect the fine par-
ticulate (PM, j) subset of PM to a much greater ex-
tent than the coarser fraction that makes up the rest
of PM . As a result of these two factors, the pro-
jected difference in ambient concentrations between
the Pre-CAAA and Post-CAAA scenarios reflect a
larger percentage reduction in PM than PM .
r> Outside the larger urban areas in the West, REMSAD
results show little or no change in PM concentrations between
Pre- and Post-CAAA estimates,
0 Sulfur dioxide is a secondary PM precursor.
In some of the figures in this chapter the Pre-CAAA and
Post-CAAA scenarios are referred to as Pre-CAAA9G and Post-
C A A A90, re s p ec if ully.
8 See Appendix C lor histograms illustrating the change in
PM concentrations from die 1990 base-year to each of the Pre-
CAAA and Post-CAAA luture year scenarios.
41
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Figure 4-3
Distribution of Combined RADM/RPM and REMSAD Derived
Monitor Level Ratios for Annual Average PM10 Concentrations:
2010 Post-CAAA/Pre-CAAA
median: 0.946
0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.30
Ratio
Figure 4-4
Distribution of Combined RADM/RPM and REMSAD Derived
Monitor Level Ratios for Annual Average PM Concentrations:
2010 Post-CAAA/Pre-CAAA
median: 0.919
0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00
Ratio
Visibility
We also relied on RADM/RPM and REMSAD
to estimate the impact of the CAAA on future-year
visibility. Tables 4-2 and 4-3 compare the mean an-
nual visibility (expressed in deciviews)9 in selected
eastern urban areas and National Parks, respectively,
as estimated bv RADM/RPM under the 1990 base-
year and 2010 Pre- and Post-CAAA
scenarios. Comparison of these val-
ues reveals that, in the eastern U.S.,
we anticipate that future-year visibil-
ity in both urban and rural areas is
projected to improve under the Post-
CAAA scenario. RADM/RPM pre-
dicts that Post-CAAA visibility in
2010 will not only be better than Pre-
CAAA visibility, but also, in many
areas, it will be better than the vis-
ibility in the 1990 base-year. This
improvement in visibility is attrib-
utable to reductions in the concen-
tration of gaseous and suspended par-
ticles, such as PM, that scatter and
absorb light, and thus influence vis-
ibility.
Visibility in the West is also sig-
nificantly better under the Post-
CAAA scenario than under the Pre-
CAAA scenario (see Tables 4-4 and
4-5). Base-year model runs show that
visibility in the western U.S. is the
poorest in larger metropolitan areas
such as Los Angeles, CA; San Fran-
cisco, CA; Denver, CO; and Phoe-
nix, AZ. Under the 2010 Pre-CAAA
scenario, REMSAD estimates that,
throughout much of the West, vis-
ibility will remain relatively un-
changed from the base-year, and in
some cases will even improve. In the
metropolitan areas, however, the
model predicts visibility degradation.
Under the Post-CAAA scenario, however,
REMSAD estimates widespread improvement in
future-year visibility in the West. In both metro-
politan and non-urban areas, deciview levels esti-
mated for 2010 are lower under the Post-CAAA sce-
nario than under the Pre-CAAA scenario. The
model suggests Los Angeles and Las Vegas will ex-
perience the greatest improvement.
1.05 1.10 1.15 1.20 1.25 1.30
9 The deciview is a measure of visibility which captures (lie
relationship between air pollution and human perception of vis-
ibility. When air is free of the particles that cause visibility deg-
radation, the DeciView Haze Index is zero. The higher the
deciview level, the poorer (lie visibility; a one to two deciview
change translates to a just, noticeable change in visibility for most
individuals.
42
-------
Chapter 4: Air Quality Modeling
Table 4-2
Comparison of Visibility in Selected Eastern Urban Areas
Mean Annual Deciview*
Area Name
State
1990
Base-Year
2010
Pre-CAAA
2010
Post-CAAA
Atlanta Metro Area
GA
20.9
22.8
20.0
Boston Metro Area
MA
13.2
14.0
11.9
Chicago Metro Area
IL
17.5
19.1
17.0
Columbus
OH
16.5
17.7
15.1
Detroit Metro Area
Ml
16.0
18.5
15.3
Indianapolis
IN
20.1
21.1
19.0
Little Rock
AR
15.0
17.2
15.1
Milwaukee Metro Area
Wl
15.6
18.4
15.3
Minn.-St. Paul Metro Area
MN
10.1
12.4
10.3
Nashville
TN
20.4
21.5
19.0
New York City Metro Area
NY/NJ
15.2
18.0
13.9
Pittsburgh Metro Area
PA
15.8
16.9
14.2
St. Louis Metro Area
MO
16.5
17.8
16.0
Syracuse
NY
12.4
13.2
11.5
Washington, DC Metro Area
DC/VA/MD
17.5
19.2
16.3
*For cities or metropolitan areas not contained by a single RADM/RPM grid cell, the visibility measure
presented in this table is a weighted average of the mean annual deciview level from each of the grid
cells that together completely contain the selected area. Weighting is based upon the spatial
distribution of an area over the various grid cells.
Table 4-3
Comparison of Visibility in Selected Eastern National Parks
Mean Annual Deciview*
Area Name
State
1990
Base-Year
2010
Pre-CAAA
2010
Post-CAAA
Acadia NP
ME
11.1
12.0
10.4
Everglades NP
FL
7.6
9.2
6.9
Great Smoky Mtns. NP
TN
20.4
22.3
19.6
Shenandoah NP
VA
16.5
17.8
15.2
*For national parks not contained by a single RADM/RPM grid cell, the visibility measure presented in
this table is a weighted average of the mean annual deciview level from each of the grid cells that
together completely contain the selected area. Weighting is based upon the spatial distribution of an
area over the various grid cells.
43
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 4-4
Comparison of Visibility in Selected Western Urban Areas
Mean Annual Deciview*
Area Name
State
1990
Base-Year
2010
Pre-CAAA
2010
Post-CAAA
Denver
CO
19.4
22.6
21.0
Las Vegas
NV
14.6
17.9
15.2
Los Angeles
CA
22.7
24.6
22.0
Phoenix
AZ
15.4
17.1
15.3
Salt Lake City
UT
12.5
14.8
13.4
San Francisco
CA
24.4
26.1
24.6
Seattle
WA
20.5
22.2
21.0
*For cities not contained by a single REMSAD grid cell, the visibility measure presented in this table is a
weighted average of the mean annual deciview level from each of the grid cells that together completely
contain the selected area. Weighting is based upon the spatial distribution of an area over the various grid
cells.
Table 4-5
Comparison of Visibility in Selected Western National Parks
Mean Annual Deciview*
Area Name
State
1990
Base-Year
2010
Pre-CAAA
2010
Post-CAAA
Glacier NP
MT
11.2
11.9
11.5
Grand Canyon NP
AZ
8.3
8.8
8.3
Olympic NP
WA
11.1
11.8
11.7
Yellowstone NP
WY
9.0
9.7
9.5
Yosemite NP
CA
11.5
13.2
12.2
Zion NP
UT
8.0
9.0
8.4
*For national parks not contained by a single REMSAD grid cell, the visibility measure presented in this
table is a weighted average of the mean annual deciview level from each of the grid cells that together
completely contain the selected area. Weighting is based upon the spatial distribution of an area over the
various grid cells.
44
-------
Chapter 4: Air Quality Modeling
Acid Deposition
Figure 4-5
Distribution of Monitor
60 ¦
Level Ratios of SO, Emissions
S 50 i
2
0)
•S 40 H
>
0
1 30 H
o-
0)
^ 20 H
0)
>
1 10 i
Note:
Figure 4-6
Distribution of Monitor
We estimated nitrogen and sulfur
deposition for the 1990 base-year and
each of the future-year emissions sce-
narios. Using RADM, we focused on
acid deposition in the eastern U.S.
where the acidification problem is the
most acute. Under the Pre-CAAA
scenario, model results show an in-
crease m both nitrogen and sulfur
deposition between 1990 and 2010.
However, under the Post-CAAA sce-
nario, 2010 deposition projections are
not only lower than 2010 Pre-CAAA
projections, but also below 1990 base-
year levels as well. Average annual
acid deposition is expected to decrease
as a result of the CAAA. Motor ve-
hicle tailpipe emissions standards and
Title IY Acid Rain provisions are ex-
pected to significantly reduce both
NO and SO, emissions thus contrib-
X 1
uting to significant reductions in
downwind deposition of acidic nitro-
gen and sulfur compounds. The dif-
ferences between the Pre-CAAA and
Post-CAAA projections, however,
imply that the 1990 Amendments will
have a larger impact on the percent-
age reduction in nitrogen deposition
than on the percentage reduction in
sulfur deposition. One reason for the greater change
in nitrogen deposition is the region-wTide NO emis-
sions cap-and-trade program that is part of the Post-
CAAA scenario.
S02, NO, N02, and CO
To estimate future-year SC)2, NO, NC)2, and CO
concentrations we relied on linear emissions scaling,
adjusting 1990 base-year concentrations using ra-
tios of future-year to base-year emissions. Ratios
greater than one indicate an increase in ambient con-
centrations relative to the base-year, while ratios less
than one indicate a decrease.10
:J The values in this section represent ratios for actual moni-
toring site locations. Interpolated data are not included in these
figures. We believe, however, that liie values presented in this
section accurately reflect the impact ol the 1990 Amendments
on SO , NO, NO., and CO ambient concentrations.
median: 0.892
n i r-rn~T-
toJ
i i i i
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20
Ratio
2.4 percent of the distribution of ratios is less than 0.40.
Level Ratios of NO Emissions
| 50 -
0)
40-
>
o
§ 30 -
o-
0)
median: 0.666
n-nHrr
Note:
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20
Ratio
3.3 percent of the distribution of ratios is less than 0.40.
Our results indicate that compared to the base-
year, future-year concentrations of SO,, NO, N()2,
and CO tend to increase under the Pre-CAAA sce-
nario, while Post-CAAA concentrations for all four
pollutants except SO, tend to decrease. For example,
the median 2010 Pre-CAAA emission-based ratio for
SO, is roughly 1.35, indicating an increase in me-
dian 2010 Pre-CAAA SO, concentration of approxi-
mately 35 percent from the 1990 base-year. The
median ratios for NO, NO,, and CO are roughly
1.13, 1.17, and 1.05 respectively. Under the Post-
CAAA scenario we estimate that in 2010 NO, NOz,
and CO concentrations will tend to be approxi-
mately 25 and 30 percent below base-year levels. The
median 2010 Post-CAAA emission-based ratios for
these three pollutants are roughly 0.74, 0.70, and 0.76
respectively.
45
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Figure 4-7
Distribution of Monitor
Level Ratios of NO, Emissions
60
g 5°
o
c
<1)
median: 0.575
30
U"
o
20
o
>
ro
£
10
0
- m irn-HrrHn rrhn n n
i—i—i—i—i—i—i—i
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20
Ratio
Note: 2.7 percent of the distribution of ratios is less than 0.40.
Figure 4-8
Distribution of Monitor
60
8 50 H
v
— 40 i
>
o
§ 30 -
20 -
Level Ratios of CO Emissions
0
>
JH
-------
Chapter 4: Air Quality Modeling
tion in NO and NO concentrations. Title I
iionattainment area controls and Title II motor ve-
hicle provisions are responsible for much of the
change in CO concentrations, while regulation of
utility and motor vehicle emissions account for ma-
jority of the decrease in S02 concentrations.
Uncertainty in the Air Quality
Estimates
.Many sources of uncertainty affect the precision
and accuracy of the projected changes in air quality
presented in this study. These uncertainties arise
largely from potential inaccuracies in the emissions
inventories used as air quality modeling inputs and
potential errors m the structure and parameteriza-
tion of the air quality models themselves. For ex-
ample, we estimated changes in PM concentrations
in the eastern U.S. based exclusively on changes in
the concentrations of sulfate and nitrate particles.
By not accounting for changes in organic and pri-
mary particulate fractions, we likely underestimate
the impact of the CAAA 011 PM concentrations.
Also, by using separate air quality models for indi-
vidual pollutants and different geographic regions,
as opposed to a single integrated model, we were
unable to fully capture the interaction among air
pollutants or reflect transport of pollutants or pre-
cursors across the boundaries of the models cover-
ing the western and eastern states. The direction
and magnitude of bias these limitations impose on
net benefits estimate presented in this analysis can
not be determined based on current information.
Some model-related uncertainties, however, may
be mitigated because this analysis uses the air qual-
ity modeling results in a relative, not absolute, sense.
We focus on the change in air quality between the
Pre- and Post-CAAA scenarios and not on the am-
bient concentrations projected by the individual
models themselves. Therefore, uncertainties that
affect a model's ability7 to accurately predict the rela-
tive change in concentration of a pollutant from one
scenario to another are more important in the con-
text of this study than those that affect only the ab-
solute model results.
The relatively coarse grid cells used to model
ozone in most areas of the U.S. represents a poten-
tial source of uncertainty affecting a model's sensi-
tivity to changes in emissions. Grid size affects chem-
istry, transport, and diffusion processes that in turn
determine the response of pollutant concentrations
to changes in emissions. The less accurately a model
can predict the impact of changes 111 emissions 011
ambient levels, the greater the uncertainty associ-
ated with predicted differences between Pre- and
Post-CAAA concentration estimates.
Table 4-7 presents the most important specific
sources of uncertainty and Appendix C further de-
scribes the uncertainties associated with air quality
modeling. While the list of potential errors presented
in Table 4-7 is not exhaustive, it includes discussion
of those factors with the greatest likelihood of con-
tributing to any potential bias in the primary net
benefit estimates.
47
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 4-7
Key Uncertainties Associated with Air Quality Modeling
Potential Source of Error
Direction of
Potential Bias for
Net Benefits
Estimate
Likely Significance Relative to Key
Uncertainties in
Net Benefit Estimate*
PM-io and PM2.5 concentrations in
the East (RADM domain) are
based exclusively on changes in
the concentrations of sulfate and
nitrate particles, omitting the
effect of anticipated reductions in
organic or primary particulate
fractions.
Underestimate. Potentially major. Nitrates and sulfates
constitute major components of PM, especially
PM2.5, in most of the RADM domain and
changes in nitrates and sulfates may serve as a
reasonable approximation to changes in total
PM10 and total PM2.5. Of the other components,
primary crustal particulate emissions are not
expected to change between scenarios;
primary organic carbon particulate emissions
are expected to change, but an important
unknown fraction of the organic PM is from
biogenic emissions, and biogenic emissions are
not expected to change between scenarios. If
the underestimation is major, it is likely the
result of not capturing reductions in motor
vehicle primary elemental carbon and organic
carbon particulate emissions.
The number of PM2.5 ambient
concentration monitors
throughout the U.S. is limited. As
a result, cross estimation of PM25
concentrations from PM10 (or
TSP) data was necessary in
order to complete the "monitor-
level" observational dataset
used in the calculation of air
quality profiles.
Unable to
determine based on
the current
information.
Potentially major. PM2.5 exposure is linked to
mortality, and avoided mortality constitutes a
large portion of overall CAAA benefits. Cross
estimation of PM25, however, is based on
studies that account for seasonal and
geographic variability in size and species
composition of particulate matter. Also, results
are aggregated to the annual level, improving
the accuracy of cross estimation.
Use of separate air quality
models for individual pollutants
and for different geographic
regions does not allow for a fully
integrated analysis of pollutants
and their interactions.
Unable to
determine based on
current information.
Potentially major. There are uncertainties
introduced by different air quality models
operating at different scales for different
pollutants. Interaction is expected to be most
significant for PM estimates. However,
important oxidant interactions are represented
in all PM models and the models are being
used as designed. The greatest likelihood of
error in this case is for the summer period in
areas with NOx inhibition of ambient ozone
(e.g., Los Angeles).
Future-year adjustment factors
for seasonal or annual monitoring
data are based on model results
for a limited number of simulation
days.
Overall, unable to
determine based on
current information.
Probably minor. RADM/RPM and REMSAD
PM modeling simulation periods represent all
four seasons and characterize the full seasonal
distribution. Potential overestimation of ozone,
due to reliance on summertime episodes
characterized by high ozone levels and applied
to the May-September ozone season, is
mitigated by longer simulation periods, which
contain both high and low ozone days. Also,
underestimation of UAM-V western and UAM-
IV Los Angeles ozone concentrations (see
below) may help offset the potential bias
associated with this uncertainty.
48
-------
Chapter 4: Air Quality Modeling
Table 4-7
Key Uncertainties Associated with Air Quality Modeling (continued)
Potential Source of Error
Direction of
Potential Bias for
Net Benefits
Estimate
Likely Significance Relative to Key
Uncertainties in
Net Benefit Estimate*
Comparison of modeled and
observed concentrations
indicates that ozone
concentrations in the western
states were somewhat under-
predicted by the UAM-V model,
and ozone concentrations in
the Los Angeles area were
underestimated by the UAM-IV
model.
Unable to
determine based
on current
information.
Probably minor. Because model results are
used in a relative sense (i.e., to develop
adjustment factors for monitor data) the
tendency for UAM-V or UAM to underestimate
absolute ozone concentrations would be unlikely
to affect overall results. To the extent that the
model is not accurately estimating the relative
changes in ozone concentrations across
regulatory scenarios, the effect could be greater.
Ozone modeling in the eastern
U.S. relies on a relatively
coarse 12 km grid, suggesting
NO* inhibition of ambient ozone
levels may be under
represented in some eastern
urban areas. Coarse grid may
affect both model performance
and response to emissions
changes.
Unable to
determine based
on current
information.
Probably minor. Though potentially major for
eastern ozone results in those cities with known
NO* inhibition, ozone benefits contribute only
minimally to net benefit projections in this study.
Grid size affects chemistry, transport, and
diffusion processes which in turn determine the
response to changes in emissions, and may also
affect the relative benefits of low-elevation
versus high-stack controls. However, the
approach is consistent with current state-of-the-
art for regional-scale ozone modeling.
UAM-V modeling of ozone in
the western U.S. uses a
coarser grid than the eastern
UAM-V (OTAG) or UAM-IV
models, limiting the resolution
of ozone predictions in the
West.
Unable to
determine based
on current
information.
Probably minor. Also, probably minor for ozone
results. Grid cell-specific adjustment factors for
monitors are less precise for the west and may
not capture local fluctuations. However,
exposure tends to be lower in the predominantly
non-urban west, and models with finer grids
have been applied to three key population
centers with significant ozone concentrations.
May result in underestimation of benefits in the
large urban areas not specifically modeled (e.g.,
Denver, Seattle) with finer grid.
Emissions estimated at the
county level (e.g., area source
and motor vehicle NOx and
VOC emissions) are spatially
and temporally allocated based
on land use, population, and
other surrogate indicators of
emissions activity. Uncertainty
and error are introduced to the
extent that area source
emissions are not perfectly
spatially or temporally
correlated with these indicators.
Unable to
determine based
on current
information.
Probably minor. Potentially major for estimation
of ozone, which depends largely on VOC and
NO* emissions; however, ozone benefits
contribute only minimally to net benefit
projections in this study.
The REMSAD model under-
predicted western PM
concentrations during fall and
winter simulation periods.
Unable to
determine based
on current
information.
Probably minor. Because model results are
used in a relative sense (i.e., to develop
adjustment factors for monitor data) REMSAD's
underestimation of absolute PM concentrations
would be unlikely to significantly affect overall
results. To the extent that the model is not
accurately estimating the relative changes in PM
concentrations across regulatory scenarios, or
the individual PM components (e.g., sulfates,
primary emissions) do not vary uniformly across
seasons, the effect could be greater.
49
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 4-7
Key Uncertainties Associated with Air Quality Modeling (continued)
Direction of
Potential Bias for Likely Significance Relative to Key
Net Benefits Uncertainties in
Potential Source of Error Estimate Net Benefit Estimate*
Lack of model coverage for acid Underestimate Probably minor. Because acid deposition tends
deposition in Western states. to be a more significant problem in the eastern
U.S. and acid deposition reduction contributes
only minimally to net monetized benefits, the
monetized benefits of reduced acid deposition
in the western states would be unlikely to
significantly alter the total estimate of
monetized benefits.
Probably minor. Potentially major impacts for
ozone outputs, but ozone benefits contribute
only minimally to net benefit projections in this
study. Uncertainties in biogenics may be as
large as a factor of 2 to 3. These biogenic
inputs affect the emissions-based VOC/NOx
ratio and, therefore, potentially affect the
response of the modeling system to emissions
changes.
* The classification of each potential source of error reflects the best judgement of the section 812 Project Team. The
Project Team assigns a classification of "potentially major" if a plausible alternative assumption or approach could
influence the overall monetary benefit estimate by approximately five percent or more; if an alternative assumption or
approach is likely to change the total benefit estimate by less than five percent, the Project Team assigns a classification
of "probably minor."
Uncertainties in biogenic Unable to
emissions inputs increase determine based on
uncertainty in the AQM current information,
estimates.
50
-------
Human Health
Effects of
Criteria Pollutants
Health benefits resulting from improved air qual-
ity constitute a significant portion of the overall ben-
efits of the Clean Air Act Amendments of 1990. As
part of the prospective analysis of these amendments,
we have identified and, where possible, estimated
the magnitude of the health benefits that Americans
are likely to enjoy in future years as a result of the
CAAA. These health benefits are expressed as
avoided cases of air-pollution related health effects
such as premature mortality, heart disease, and res-
piratory illness. This chapter presents an overview
of our approach to modeling these changes in ad-
verse health effects, discusses key assumptions asso-
ciated with this approach, and summarizes model-
ing results for major health effect categories. Al-
though this chapter focuses predominantly on the
human health effects associated with exposure to
criteria pollutants, the final section of this chapter
presents a discussion of the effects associated with
air toxics and stratospheric ozone.
In general, this analysis finds that the CAAA
will result 111 significant reductions in mortality, res-
piratory illness, heart disease, and other adverse
health effects, with much of these reductions result-
ing from decreases in ambient particulate matter con-
centrations.
Analytical Approach
We estimate the impact of the CAAA on hu-
man health by analyzing the difference in the ex-
pected incidence of adverse health effects between
the Pre-CAAA and Post-CAAA regulatory sce-
narios. As described in Chapter 2, the Pre-CAAA
scenario assumes no further controls on criteria pol-
lutant emissions besides those already in place in
1990, while the Post-CAAA scenario assumes full
implementation of the 1990 CAAA. For each regu-
latory scenario, we use the Criteria Air Pollutant
Modeling System (CAPMS) to estimate the incidence
of health effects for 1990 (base-year), 2000, and 2010.
Modeling the incidence of adverse health effects re-
sulting from exposure to criteria air pollutants re-
quires three types of inputs: (1) estimates of the
changes in air quality for the Pre- and Post-CAAA
scenarios in 2000 and 2010; (2) estimates of the num-
ber of people exposed to air pollutants at a given
location; and (3) concentration-response (C-R) func-
tions that link changes in air pollutant concentra-
tions with changes in adverse health effects. We dis-
cuss each of these inputs in greater detail below.
Air Quality
The development of criteria pollutant concen-
tration estimates for use in the CAPMS model con-
sists of two steps. First, air quality modeling and
1990 base-year monitoring data arc used to project
ambient pollution levels at monitors throughout the
48 contiguous states. Second, because air quality
monitors are neither uniformly nor pervasively dis-
tributed across the country, concentration data at
monitors are extrapolated to non-monitored areas
in order to generate a more comprehensive air qual-
ity data set covering the 48 contiguous states and the
District of Columbia.
The projections of criteria pollutant concentra-
tions at air pollution monitors are developed as sum-
marized in Chapter 4 and described in detail in Ap-
pendix C. Briefly, baseline 1990 concentrations at
each monitor are adjusted using monitor- and pol-
lutant-specific adjustment factors to produce esti-
mates of concentrations in 2000 and 2010 for each
regulatory scenario. Each adjustment factor reflects
the relative change in the concentration of a pollut-
ant in a specific geographic area between 1990 and
the target year, as predicted by air quality modeling.
51
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
To develop pollutant concentration estimates for
the entire continental U.S. we extrapolate the 1990
monitor data and the future-year estimates to the
eight kilometer by eight kilometer CAPMS grid cells
that cover the 48 contiguous states. Within each of
these cells, we calculate an estimated pollutant con-
centration using data from nearby monitors accord-
ing to a distance-weighted averaging method de-
scribed in Appendix D. We then use these grid cell
pollutant concentration estimates to predict changes
in health effects among the population residing
within each cell.
Population
Health benefits resulting from the CAAA are
related to the change in air pollutant exposure expe-
rienced by individuals. Because the expected changes
in pollutant concentrations vary from location to
location, individuals in different parts of the coun-
try may not experience the same level of health ben-
efits. This analysis apportions benefits among indi-
viduals by matching the change in air pollutant con-
centration in a CAPMS grid cell with the size of the
population that experiences that change.
As a result, we require an estimate of the distri-
bution of the U.S. population among CAPMS grid
cells. The grid-cell-specific population counts for
1990 are derived from U.S. Census Bureau block level
population data. Grid cell population estimates for
future years are extrapolated from 1990 levels using
the ratio of future-year and 1990 state-level popula-
tion estimates provided by the U.S. Bureau of "Eco-
nomic Analysis.1
Concentration-Response Functions
We calculate the benefits attributable to the
CAAA as the avoided incidence of adverse health
effects. Such benefits can be measured using C-R
functions specific to each health effect. C-R func-
tions are equations that relate the change in the num-
ber of individuals in a population exhibiting a "re-
sponse" (in this case an adverse health effect such as
respirator)'- disease) to a change in pollutant concen-
tration experienced by that population. The C-R
1 U.S. Bureau of Economic Analysis. 1995. BEA Regional
Projections to 2045: Volume 1, States. U.S. Department of Com-
merce. Washington, DC. July.
functions used in CAPMS generate changes in the
incidence of an adverse health effect using three val-
ues: the grid-cell-specific change in pollutant concen-
tration, the grid-cell-specific population, and an esti-
mate of the change in the number of individuals that
suffer an adverse health effect per unit change in air
quality.2 As described in Appendix D, we derive
this last factor, as well as the specific form of the C-
R equation, from the published scientific literature
for each pollutant/health effect relationship of in-
terest.
Using the appropriate C-R functions, CAPMS
generates estimates for each grid-cell of the change
in incidence of a set of adverse health effects result-
ing from the incremental change in exposure between
the Pre- and Post-CAAA scenarios in 2000 and 2010.
For each health effect, CAPMS then generates na-
tional health benefits estimates by summing the an-
nual incidence change across all grid cells.
Each criteria pollutant evaluated in the 812 pro-
spective analysis has been associated with multiple
adverse health effects. The published scientific lit-
erature contains information that supports the esti-
mation of some, but not all, of these effects. Thus,
it is not possible currently to estimate all of the hu-
man health benefits attributable to the CAAA. In
addition, for some of the health effects we do quan-
tify, the current economic literature does not sup-
port the estimation of the economic value of these
effects. For each of the criteria pollutants we evalu-
ate in this analysis, Table 5-1 presents the health ef-
fects that are quantitatively estimated and those that
can not currently be quantified. The sixth criteria
pollutant, lead (Pb), is not included in this analysis
since airborne emissions of lead were virtually elimi-
nated by pre-1990 ('lean Air Act programs.
Key Analytical Assumptions
The modeling of health benefits attributable to
the CAAA involves numerous judgments and as-
sumptions to address data limitations and other con-
straints. Each of these analytical assumptions affects
both the accuracy and precision with which we can
estimate health benefits of the CAAA, but some as-
An estimate of the baseline incidence of the adverse health
eilect may also he required for certain C-R functions.
52
-------
Chapter 5: Human Health Effects of Criteria Pollutants
Table 5-1
Human Health Effects of Criteria Pollutants
Pollutant
Quantified Health Effects
Unquantified Health Effectst
Ozone
Respiratory symptoms
Minor restricted activity days
Respiratory restricted activity days
Hospital admissions-
All Respiratory and
All Cardiovascular
Emergency room visits for asthma
Asthma attacks
Mortality*
Increased airway responsiveness to stimuli
Inflammation in the lung
Chronic respiratory damage / Premature aging of the lungs
Acute inflammation and respiratory cell damage
Increased susceptibility to respiratory infection
Non-asthma respiratory emergency room visits
Particulate
Matter
(PM10,
PM2.5)
Mortality*
Bronchitis - Chronic and Acute
New asthma cases
Hospital admissions -
All Respiratory and
All Cardiovascular
Emergency room visits for asthma
Lower respiratory illness
Upper respiratory illness
Shortness of breath
Respiratory symptoms
Minor restricted activity days
All restricted activity days
Days of work loss
Moderate or worse asthma status
(asthmatics)
Neonatal mortality*
Changes in pulmonary function
Chronic respiratory diseases
other than chronic bronchitis
Morphological changes
Altered host defense mechanisms
Cancer
Non-asthma respiratory emergency room visits
Carbon
Monoxide
Hospital Admissions -
All Respiratory and
All Cardiovascular
Behavioral effects
Other hospital admissions
Other cardiovascular effects
Developmental effects
Decreased time to onset of angina
Non-asthma respiratory emergency room visits
Nitrogen
Oxides
Respiratory illness
Hospital Admissions -
All Respiratory and
All Cardiovascular
Increased airway responsiveness to stimuli
Chronic respiratory damage / Premature aging of the lungs
Inflammation of the lung
Increased susceptibility to respiratory infection
Acute inflammation and respiratory cell damage
Non-asthma respiratory emergency room visits
Sulfur Hospital Admissions - Changes in pulmonary function
Dioxide All Respiratory and Respiratory symptoms in non-asthmatics
All Cardiovascular Non-asthma respiratory emergency room visits
In exercising asthmatics:
Chest tightness,
Shortness of breath, or
Wheezing
f Some of the unqualified adverse health effects of air pollution may be associated with adverse health endpoints that we
have quantitatively evaluated (e.g., chronic respiratory damage and premature mortality). However, it is likely that the value
assigned to the quantified endpoint may not fully capture the value of the associated health effect (e.g., chronic respiratory
damage may result in significant pain and suffering prior to mortality). As a result, we include such effects separately in the
unqualified health effects column.
^Appendix D includes detailed discussion of the scientific evidence for these potential health effects and includes illustrative
benefit calculations for them. Current uncertainties in our understanding of these effects do not support including these
quantitative estimates in the overall CAAA benefits estimate. However, ozone-related mortality may be implicitly quantified in
the overall analysis as part of the PM mortality estimate because of the significant correlation between ozone and PM
concentrations.
* This analysis estimates avoided mortality using PM as an indicator of the criteria air pollutant mix to which individuals were
exposed.
53
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
sumptions introduce greater uncertainty into the
results than others. This section characterizes these
key assumptions and the associated uncertainties to
allow the reader to gain a better understanding of
the potential for misestimation of avoided health
effects. In addition, health benefits are presented as
ranges to reflect the aggregate effect of the uncer-
tainty in key variables (see Results section below).
This section discusses the most important analytical
assumptions of this modeling effort, grouped into
the following categories: (1) exposure analysis, (2)
selection and application of C-R functions, and (3)
estimation of changes in PM-related mortality.
Exposure Analysis
The key analytical assumptions involved in esti-
mating exposure to criteria air pollutants relate to
two steps: the extrapolation of air quality data from
monitors and the mapping of population data to air
quality data.
As discussed above, actual ambient air pollution
data arc available only for a limited number of moni-
tor sites that arc not uniformly distributed across
the U.S. Thus, to estimate the impact of air pollu-
tion changes on the health of the U.S. population,
data from monitors are extrapolated to the cells of a
grid that covers the 48 contiguous states and are
matched with population data for each grid-cell.
Essentially, the extrapolation method uses data from
the closest set of monitors surrounding a grid-cell to
compute a weighted average concentration for that
cell. Monitors closer to the grid cell are assumed to
yield a more accurate estimate of air quality in the
cell; thus data from these monitors receive more
weight than data from more distant monitors when
calculating an air quality estimate for the cell.3 The
resulting estimates are uncertain because the geogra-
phy, weather, land use, and other factors influenc-
ing air pollution may differ significantly between a
grid cell and the monitor or monitors used to gener-
ate estimates of air quality, especially as the moni-
tor-to-grid-cell distance grows.4 As a result, they may
3 Specifically, monitor data are weighted based, on the
inverse of the distance between the monitor and the grid-cell
center. Additional information on (lie extrapolation method is
provided in Appendix D.
4 In order to address tliis issue for long-distance extrapola-
tion (i.e., grid cells greater than 50 kilometers from a monitor),
the method is modified to also incorporate air quality modeling
predictions for she source and target locations. See Appendix D
for details.
not sufficiently capture local variation in air pollu-
tion levels (e.g., hot spots).
However, since the uncertainty in these extrapo-
lated values is inversely proportional to the density
of monitors in a given area, and since air quality
monitors are more prevalent in high pollution areas
than in low pollution areas, this extrapolation
method estimates the air quality in high pollution
areas (where the potential benefits of the CAAA are
greatest) with greater certainty than in low pollu-
tion areas. Thus, grid-cell ozone estimates in the
eastern U.S., where ozone levels and ozone moni-
tor density are higher, are likely to be more accurate
than those in the west, where monitor coverage is
more sparse. Also, estimates of concentrations of
criteria pollutants, which are measured by a greater
number of monitors nationwide (PM, ozone, SO^,
are expected to be less uncertain than estimates for
CO and NO , which are measured bv considerably
fewrer monitors.
Air pollutant concentration changes are mapped
to grid-cell population data derived from U.S. Cen-
sus bureau data, and extrapolated to future years
using population growth estimates from the U.S.
Bureau of Economic Analysis. There arc two key-
assumptions associated with this population map-
ping. First, we assume the population in each grid
cell grows at the same rate as the state population as
a whole. As a result, exposures (and potential ben-
efits) in individual grid cells may be either under- or
over-estimated if population growth varies from the
state average during the 1990 to 2010 period. This
uncertainty is likely to be more significant in larger
states such as California and Texas, which may have
more geographic variability in growth patterns.
Also, the effect of this assumption may be less sig-
nificant for large population centers because their
growth rate better approximates the growth rate of
the state as a whole. Second, we assume 111 the ex-
posure analysis that the population in the grid cell is
similar in terms of its activity patterns and demo-
graphic characteristics to the populations in the stud-
ies from which the C-R functions are derived. This
is a potentially significant uncertainty which is dis-
cussed further in the next section and in Appendix D.
54
-------
Chapter 5: Human Health Effects of Criteria Pollutants
Selection and Application
of C-R Functions
We rely on the most recent available, published
scientific literature to ascertain the relationship be-
tween air pollution and adverse human health ef-
fects. The uncertainties underlying those published
studies and our method for selecting studies that
could be used to derive C-R functions likely con-
tributes to the uncertainty of the health effects re-
sults. For example, the uncertainty associated with
the current state of the published scientific litera-
ture could potentially have two contradictory influ-
ences on the results of this analysis. First, to the
extent that the published literature may collectively
overstate the effects of pollution, our analysis will
overstate the benefits of CAAA-related pollution
reduction. This overestimation is possible because
scientific journals tend to publish research report-
ing significant associations between pollution and
disease more often than research that fails to find
such associations. On the other hand, our analysis
may underestimate overall health benefits of the
CAAA because, as the state of the science evolves,
current pollutant/health effect associations may be
found to be stronger than previously thought, and
new associations may be identified. For example, in
recent years, studies have showrn the potential health
benefits from reductions in ambient PM to be much
greater than previously believed. To the extent that
the present analysis does not include health effects
whose link to air pollution has not been subject to
adequate scientific inquiry, this analysis may under-
state CAAA-related health benefits.
Our method of identifying appropriate C-R func-
tions for use in the benefits analysis may also intro-
duce uncertainty. We evaluate studies using the nine
selection criteria summarized in Table 5-2 and de-
scribed in detail in Appendix D. These criteria in-
clude consideration of whether the study wras peer-
reviewed, the study design and location, and charac-
teristics of the study population, among others. The
selection of C-R functions for the benefits analysis
is guided by the goal of achieving a balance between
comprehensiveness and scientific defensibility. How-
ever, to the extent that this selection process may
lead to the exclusion of valid studies, the process in-
troduces uncertainty into the analysis. The overall
effect of this uncertainty is expected to be minor,
given the emphasis of the selection process 011 scien-
tific validity. Appendix D lists the studies selected
for each category of health effects, and presents the
associated C-R functions for each criteria pollutant.
Once the C-R functions have been selected, un-
certainty may also enter the analysis due to both
within-study and across-study variation in C-R func-
tions for individual health effects. Within-study
variation refers to the uncertainty and error that may
surround a given study's estimate of a C-R function.
Health effects studies provide both "best estimates"
of the relationship between air quality changes and
health effects and a measure of the statistical uncer-
tainty of the relationship. We use statistical simula-
tion modeling techniques to evaluate the overall
uncertainty of the results given the uncertainties as-
sociated with individual studies. Across-study varia-
tion refers to the fact that different published stud-
ies of the same pollutant/health effect relationship
typically do not report identical findings; in some
instances the differences are substantial. These dif-
ferences can exist even between equally reputable
studies and may result in health effect estimates that
vary considerably.
Across-study variation can result from two pos-
sible causes. One possibility is that studies report
different estimates of the single true relationship
between a given pollutant and a health effect due to
differences in study design, random chance, or other
factors. For example, a hypothetical study conducted
in New York and one conducted in Seattle may re-
port different C-R functions for the relationship
between PM and mortality in part because of differ-
ences between these two study populations (e.g.,
demographics, activity patterns). Alternatively,
study results may differ because they are in fact esti-
mating different relationships; that is, the same re-
duction in PM in New York and Seattle may result
in different reductions in premature mortality. This
may result from a number of factors, such as differ-
ences in the relative sensitivity of these two popula-
tions to PM pollution and differences in the compo-
sition of PM in these two locations.5 In either case,
where we identify multiple studies that are appro-
5 PM is a mix of particles of varying size and chemical
properties. The composition ol PM can vary considerably Irorn
one region lo another depending on the sources of particulate
emissions in each region.
55
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 5-2
Summary of Considerations Used in Selecting C-R Functions
Consideration Comments
Peer reviewed Peer reviewed research is preferred to research that has not undergone the peer review
research process.
Study type Among studies that consider chronic exposure (e.g., over a year or longer) prospective
cohort studies are preferred over cross-sectional studies (a.k.a. "ecological studies")
because they control for important confounding variables that cannot be controlled for in
cross-sectional studies. If the chronic effects of a pollutant are considered more important
than its acute effects, prospective cohort studies may also be preferable to longitudinal time
series studies because the latter type of study is typically designed to detect the effects of
short-term (e.g. daily) exposures, rather than chronic exposures.
Study period Studies examining a relatively longer period of time (and therefore having more data) are
preferred, because they have greater statistical power to detect effects. More recent
studies are also preferred because of possible changes in pollution mixes, medical care,
and life style overtime.
Study population Studies examining a relatively large sample are preferred. Studies of narrow population
groups are generally disfavored, although this does not exclude the possibility of studying
populations that are potentially more sensitive to pollutants (e.g., asthmatics, children,
elderly). However, there are tradeoffs to comprehensiveness of study population.
Selecting a C-R function from a study that considered all ages will avoid omitting the
benefits associated with any population age category. However, if the age distribution of a
study population from an "all population" study is different from the age distribution in the
assessment population, and if pollutant effects vary by age, then bias can be introduced
into the benefits analysis.
Study location U.S. studies are more desirable than non-U.S. studies because of potential differences in
pollution characteristics, exposure patterns, medical care system, and life style.
Pollutants Models with more pollutants are generally preferred to models with fewer pollutants, though
included in careful attention must be paid to potential collinearity between pollutants. Because PM has
model been acknowledged to be an important and pervasive pollutant, models that include some
measure of PM are highly preferred to those that do not.
Measure of PM PM2.5 and PM10 are preferred to other measures of particulate matter, such as total
suspended particulate matter (TSP), coefficient of haze (COH), or black smoke (BS) based
on evidence that PM2.5 and PM10 are more directly correlated with adverse health effects
than are these other measures of PM.
Economically Some health effects, such as forced expiratory volume and other technical measurements
valuable health of lung function, are difficult to value in monetary terms. These health effects are not
effects quantified in this analysis.
Non-overlapping Although the benefits associated with each individual health endpoint may be analyzed
endpoints separately, care must be exercised in selecting health endpoints to include in the overall
benefits analysis because of the possibility of double counting of benefits. Including
emergency room visits in a benefits analysis that already considers hospital admissions, for
example, will result in double counting of some benefits if the category "hospital
admissions" includes emergency room visits.
priate for estimating a given health effect, we use
the multiple C-R estimates, applied to the entire U.S.,
to derive a range of possible results for that health
effect.
Whether this analysis estimates the C-R relation-
ship between a pollutant and a given health endpoint
using a single function from a single study or using
multiple C-R functions from several studies, each
C-R relationship is applied throughout the U.S. to
generate health benefit estimates. However, to the
extent that pollutant/health effect relationships are
region-specific, applying a location-specific C-R func-
tion at all locations in the U.S. may result in overes-
timates of health effect changes in some locations
and underestimates of health effect changes in other
locations. It is not possible, however, to know the
extent or direction of the overall effect 011 health
benefit estimates introduced by application of a single
OR function to the entire U.S. This may be a sig-
56
-------
Chapter 5: Human Health Effects of Criteria Pollutants
nificant uncertainty in the analysis, but the current
state of the scientific literature does not allow for a
region-specific estimation of health benefits.
PM-Related Mortality
'l'his section discusses the estimation of one of
the most serious health impacts of air pollution: pre-
mature mortality associated with PM exposure. This
section consists of three parts. It begins with a dis-
cussion of the uncertainties surrounding the PM/
mortality relationship. Then, it presents specific
factors to consider when selecting a PM mortality
C-R function. It ends with a brief discussion of the
advantages and disadvantages of the study we selected
for the PM mortality analysis: Pope et al., 1995.
Uncertainties In the PM Mortality
Relationship
Health researchers have consistently linked air
pollution, especially PM, with excess mortality. A
substantial body of published scientific literature
recognizes a correlation between elevated PM con-
centrations and increased mortality rates. However,
there is much about this relationship that is still un-
certain.6 These uncertainties include:
• Causality. For this analysis, we assume a
causal relationship between exposure to el-
evated PM and premature mortality, based
on the evidence of a correlation between PM
and mortality reported in the scientific lit-
erature. This assumption is necessary be-
cause the epidemiological studies on which
this analysis relies, by design, can not defini-
tively prove causation.
• Other Pollutants. PM concentrations are
correlated with the concentrations of other
criteria pollutants, such as ozone and CO,
and it is unclear how much each pollutant
may influence elevated mortality rates. Re-
cent studies have explored whether ozone
and CO may have mortality effects indepen-
dent of PM, but we do not view the evidence
as sufficient to include such effects in the
overall CAAA-related health benefits esti-
5 The morbidity studies used in this analysis may also be
subject to many of tlic uncertainties listed in this section.
mate.7 As a result, we use the reported PM/
mortality relationship as a proxy for the
mortality effects of the air pollutant mixture.
• Shape of the C-R Function. The shape of
the true PM mortality C-R function is un-
certain, but this analysis assumes the C-R
function to have a log-linear form (as derived
from the literature) throughout the relevant
range of exposures.8 If this is not the cor-
rect form of the C-R function, or if certain
scenarios (e.g., 2010 Pre-CAAA) predict con-
centrations well above the range of values
for which the C-R function was fitted,
avoided mortality may be mis-estimated.
• Regional Differences. As discussed earlier,
significant variability exists in the results of
different PM studies. This variability may-
re fleet regionally-specific C-R functions re-
sulting from regional differences in factors
such as the physical and chemical composi-
tion of I'M. If true regional differences ex-
ist, applying these C-R functions to regions
other than the study location would result
in mis-estimation of effects in these regions.
• Exposure/Mortality Lags. It is currently
unknown whether there is a time lag — a
delay between changes in PM exposures and
changes in mortality rates — in the chronic
PM/mortality relationship. The existence
of such a lag could be important for the valu-
ation of benefits, if one were to assume that
lagged incidences of premature mortality
should be discounted over the period be-
tween when the fatal increment of exposure
is experienced and premature mortality ac-
tually occurs. Although there is no specific
scientific evidence of the existence or struc-
ture of a PM effects lag, current scientific
literature on adverse health effects such as
those associated with PM (e.g., smoking-re-
lated disease) leaves us skeptical that all inci-
Appendix D discusses the evidence linking both ozone
and CO with mortality. It also describes and presents the re-
sults of an illustrative analysis estimating CAAA-related reduc-
tions in ozone-related mortality using currently available stud-
ies.
8 C-R functions for other health effects may be assumed to
be linear or log-linear. See Appendix D for more details.
57
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
dences of premature mortality associated
with a given incremental change in PM ex-
posure would occur in the same year as the
exposure reduction. This same literature
implies that lags of up to a few years are plau-
sible, and we chose to assume a five-year lag
structure, with 25 percent of deaths occur-
ring in the first year, another 25 percent in
the second year, and 16.7 percent in each of
the remaining three years.
• Cumulative Effects. We attribute the PM/
mortality relationship used in this study
(Pope et al., 1995) primarily to PM-associ-
ated cumulative damage to the cardiopulmo-
nary system, since the short-term mortality
estimates reported in time-series studies ac-
count for only a minor fraction of total ex-
cess mortality. However, the relative roles
of exposure duration and exposure level re-
main unknown at this time.
Selection of a PM Mortality C-R
Function
In addition to the study selection criteria listed
in Table 5-2, we consider three additional factors
when selecting a PM mortality function. The first
focuses on the PM indicator (i.e., PM or PM25), the
second focuses on whether the stud}' measured short-
term or long-term PM exposure, and the third fo-
cuses on whether the study used a cohort or eco-
logic design.
Current research suggests that particle size mat-
ters when estimating the health impacts of PM. Par-
ticulate matter is a heterogeneous mixture that in-
cludes particles of varying sizes. Fine PM is gener-
ally viewed as having a more harmful impact than
coarse PM, especially for coarse particles larger than
10|im in aerodynamic diameter, although it is not
clear to what extent this may differ by the type of
health effect or the exposed population. While one
cannot necessarily assume that coarse PM has no
adverse impact on health, we prefer the use of PM,
as the best currently available measure of the impact
of PM on mortality.9
s Due to the relative abundance of studies using PMir, how-
ever, and the reasonably good correlation between PM. 5 and
PM , the 812 prospective analysis also uses PM.. studies to esti-
mate the impact ol PM on non-mortality health effects.
Two types of exposure studies (short-term and
long-term) have been used to estimate a PM/mortal-
ity relationship. Short-term exposure studies attempt
to relate short-term (often day-to-day) changes in PM
concentrations and changes in daily mortality rates
up to several days after a period of elevated PM con-
centrations. Long-term exposure studies examine
the potential relationship between longer-term (e.g.,
annual) changes in exposure to PM and annual mor-
tality rates. Researchers have found significant cor-
relations using both types of studies; however, for
this analysis, we rely exclusively on long-term stud-
ies to quantify PM mortality effects, though the
short-term studies provide additional scientific evi-
dence supporting the PM/mortality relationship.
Because short-term studies focus only on the
acute effects associated with daily peak exposures,
they are unable to evaluate the degree to which ob-
served excess mortality is premature,10 and they may
underestimate the C-R coefficient because they do
not account for the cumulative mortality effects of
long-term exposures (i.e., exposures over many years
rather than a few days). Long-term studies, on the
other hand, are able to discern changes in mortality
rates due to long-term exposure to elevated air pol-
lution concentrations, and are not limited to mea-
suring mortalities that occur within a few days of a
high-pollution event (though they may not predict
cases of premature mortality that were only has-
tened by a few days). Consequently, the use of C-R
functions derived from long-term studies is likely to
result in a more complete assessment of the effect of
air pollution on mortality risk. However, to the
extent that long-term studies fail to capture acute
mortality effects related to peak exposures, the use
of long-term mortality studies may underestimate
CAAA-related avoided mortality benefits.
Among long-term PM' studies, we prefer studies
using a prospective cohort design to those using an
ecologic or population-level design. Prospective
10 This can be important in cost-benefit analysis if benefits
are estimated ill terms of life-years lost. In short-term studies
evaluating peak pollution events, it is likely that many of the
"excess mortality" cases represented individuals who were al-
ready suffering impaired health, and lor whom the high-pollu-
tion event represented an exacerbation ol an already serious
condition. Based on the episodic studies only, however, it is
unknown how many of the victims would have otherwise lived
only a few more days or weeks, or how many would have re-
covered to enjoy many years of a healthy life in the absence of
the high-pollution event.
58
-------
Chapter 5: Human Health Effects of Criteria Pollutants
cohort studies follow individuals forward in time for
a specified period, periodically evaluating each
individual's exposure and health status. Population-
level ecological studies assess the relationship between
population-wide health information (such as counts
for daily mortality) and ambient levels of air pollu-
tion. Prospective cohort studies are preferred be-
cause they are better at controlling a source of un-
certainty known as "confounding." Con founding-
is the mis-estimation of an association that results if
a study does not control for factors that are corre-
lated with both the outcome of interest (e.g., mor-
tality) and the exposure of interest (e.g., PM expo-
sure). For example, smoking is associated with mor-
tality-. If populations in high PM areas tend to smoke
more than populations in low PM areas, and a PM
exposure study does not include smoking as a factor
in its model, then the mortality effects of smoking
may be erroneously attributed to PM, leading to an
overestimate of the risk from PM. Prospective co-
hort studies are better at controlling for confound-
ing than ecologic studies because the former follow
a group of individuals forward in time and can gather
individual-specific information on important risk
factors such as smoking. However, it is always pos-
sible, even in well-designed studies, that a relevant
risk factor (e.g., climate, the presence of other pol-
lutants) may not have been adequately considered
or controlled for. As a result, it is possible that dif-
ferences in mortality rates ascribed to differences in
average PM levels may be due, in part, to some other
factor or factors (e.g., differences among communi-
ties in diet, exercise, ethnicity, climate, industrial
effluents, etc.) that have not been adequately ad-
dressed in the exposure models.
The Pope Study
Three recent studies have examined the relation-
ship between mortality and long-term exposure to
PM: Pope et al. (1995), Dockery et al. (1993), and
Abbey et al. (1991). Of these three studies, we pre-
fer using the Pope et al. study as the basis for devel-
oping the primary PM mortality estimates in this
analysis. Pope et al. studied the largest cohort, had
the broadest geographic scope, and effectively con-
trolled for potentially significant sources of con-
founding.
Pope et al. examined a much larger population
(over 295,000) and many more locations (50 metro-
politan areas) than either the Dockery study or the
Abbey study. The Dockery study covered a cohort
of over 8,000 individuals in six U.S. cities, and the
Abbey study covered a cohort of 6,000 people in
California. In particular, the cohort in the Abbey
study was considered substantially too small and too
young to enable the detection of small increases in
mortality risk. The study was therefore omitted
from consideration in this analysis. Even though
Pope et al. (1995) reports a smaller premature mor-
tality response to elevated PM than Dockery et al.
(1993), the results of the Pope study are nevertheless
consistent with those of the Dockery study.
Pope et al., (1995) is unique in that it followed a
largely white and middle class population. The use
of this study population reduces the potential for
confounding because it decreases the likelihood that
differences in premature mortality across locations
were attributable to differences in socioeconomic
status or related factors rather than PM. However,
the demographics of the study population may also
produce a downward bias in the PM mortality coef-
ficient, because short-term studies indicate that the
effects of PM tend to be significantly greater among
groups of lower socioeconomic status.
Although it is the strongest of the PM cohort
studies, Pope et al. does have some limitations. For
example, Pope et al. did not consider the migration
of cohort members across study cities, which would
cause exposures to be more similar across individu-
als than those indicated by assigning city-specific
annual average pollution levels to each member of
the cohort. As intercity migration increases among
cohort members, the exposure experienced by mi-
grating individuals will tend toward an intercity
mean. If this migration is significant and is ignored,
approximating true differences in exposure levels
by differences in city-specific annual average PM lev-
els will exaggerate changes in exposure, resulting in
a downward bias of the PM coefficient. 'This occurs
because a given difference in mortality rates is being
associated with a larger difference in PM levels than
that actually experienced by individuals in the study
cohort. When the relationship between elevated PM
exposure and premature mortality derived from the
59
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Pope et al. study is applied in the present analysis,
the effect of the potential mis-specification of expo-
sure due to migration in the underlying study is to
underestimate PM-related mortality reduction ben-
efits attributable to the CAAA.
Also, Pope et al. only included PM when esti-
mating a C-R function. Because PM concentrations
are correlated with the concentrations of other cri-
teria air pollutants (e.g., ozone), and because these
other pollutants may be correlated with premature
mortality (see Appendix D), the PM risk estimate
may be overestimated because it includes the mor-
tality impacts of these confounders. However, in
an effort to avoid overstating benefits, and because
the evidence associating mortality with PM expo-
sure is stronger than for other pollutants, the 812
Prospective analysis uses PM as a surrogate for PM
and related criteria pollutants.
Although we use the Pope study exclusively to
derive our primary estimates of avoided mortality,
die C-R function based on Dockery et al. (1993) may
provide a reasonable alternative estimate. While the
Dockery et al. study used a smaller sample of indi-
viduals from fewer cities than the study by Pope et
al., it features improved exposure estimates, a slightly
broader study population (adults aged 25 and older),
and a follow-up period nearly twice as long as that
of Pope et al. We present an alternative estimate of
the premature adult mortality associated with long-
term PM exposure based on Dockery et al. (1993) in
Chapter 8 and in Appendix D. We emphasize, how-
ever, that the estimate based on Pope et al. (1995) is
our primary estimate of the effect of the 1990 Amend-
ments on this important health effect.
Health Effects Modeling
Results
This section presents a summary of the differ-
ences in health effects resulting from improvements
in air quality between the Pre-CAAA and Post-
CAAA scenarios. Table 5-3 summarizes the CAAA-
related avoided health effects in 2010 for each study
included in the analysis. The mean estimate is pre-
sented as the Primary Central estimate, the 5th per-
centile observation from the statistical uncertainty
modeling is presented as the Primary Low estimate,
and the 95th percentile observation is presented as
the Primary High estimate of the number of avoided
cases of each endpoint.11 To provide context for these
results, Table 5-3 also expresses the mean reduction
in incidence for each adverse health effect as a per-
centage of the baseline incidence of that effect (ex-
trapolated to the appropriate future year) for the
population considered (e.g., adults over 30 years of
age). In general, because the differences in air qual-
ity between the Pre- and Post-CAAA scenarios are
expected to increase from 1990 to 2010 and because
population is also expected to increase during that
time, the health benefits attributable to the CAAA
are expected to increase consistently from 1990 to
2010. .More detailed results are presented in Appen-
dix D.
Avoided Premature Mortality Estimates
Table 5-3 summarizes the avoided mortality due
to reductions in PM exposure in 2010 between the
Pre- and Post-CAAA scenarios. As this table shows,
our Primary Central estimate implies that PM re-
ductions due to the CAAA in 2010 will result in
23,000 avoided deaths, with a Primary Low and Pri-
mary High bound on this estimate of 14,000 and
32,000 avoided deaths, respectively. The Primary
Central estimate of 23,000 avoided deaths represents
roughly one percent of the projected annual non-
accidental mortality of adults aged 30 and older in
the year 2010. Additionally, Table 5-4 summarizes
the distribution of avoided mortality for 2010 by
age cohort, along with the expected remaining life-
span (i.e., the life years lost) for the average person
in each age cohort. The majority of the estimated
deaths occur in people over the age of 65 (due to
their higher baseline mortality rates), and this group
has a shorter life expectancy relative to other age
groups. The life years lost estimates might be higher
if data were available for PM-related mortality in
the under 30 age group.
1' The Primary Low, Primary Central and Primary High
health benefit estimates represent points on a distribution of
estimated incidence changes for each health effect. This distri-
bution reflects the uncertainty associated with the coefficient of
die C-R function for each health endpoint. More information
about C-R function uncertainty and the uncertainty modeling
that generates the results distributions is presented in Appendix
D.
60
-------
Chapter 5: Human Health Effects of Criteria Pollutants
Table 5-3
Change in Incidence of Adverse Health Effects Associated with Criteria Pollutants in 2010
(Pre-CAAA minus Post-CAAA) - 48 State U.S. Population (avoided cases per year)
2010
% of Baseline
Incidences for
the mean
estimates a
Endpoint
Pollutant
5th %
mean
95th %
2010
Mortality
ages 30 and older
PM
14,000
23,000
32,000
1.00%
Chronic Illness
chronic bronchitis
PM
5,000
20,000
34,000
3.14%
chronic asthma
o3
1,800
7,200
12,000
3.83%
Hospitalization
respiratory
admissions
PM, CO, N02,
SO2, O3
13,000
22,000
34,000
0.62%
cardiovascular
admissions
PM, CO, N02,
SO2, O3
10,000
42,000
100,000
0.86%
emergency room
visits for asthma
PM, O3
430
4,800
14,000
0.55%
Minor Illness
acute bronchitis
PM
0
47,000
94,000
5.06%
upper respiratory
symptoms
PM
280,000
950,000
1,600,000
0.86%
lower respiratory
symptoms
PM
240,000
520,000
770,000
3.57%
respiratory illness
N02
76,000
330,000
550,000
10.44%
moderate or worse
asthma0
PM
80,000
400,000
720,000
0.24%
asthma attacks0
03, PM
920,000
1,700,000
2,500,000
1.04%
chest tightness,
shortness of breath,
or wheeze
S02
290
110,000
520,000
0.003%
shortness of breath
PM
26,000
91,000
150,000
1.69%
work loss days
PM
3,600,000
4,100,000
4,600,000
0.94%
minor restricted
activity days / any of
19 respiratory
symptomsd
03, PM
25,000,000
31,000,000
37,000,000
2.15%
restricted activity
days0
PM
10,000,000
12,000,000
13,000,000
1.00%
a The baseline incidence generally is the same as that used in the C-R function for a particular health effect. However, there are a few
exceptions. To calculate the baseline incidence rate for respiratory-related hospital admissions, we used admissions for persons of all
ages for International Classification of Disease (ICD) codes 460-519; for cardiovascular admissions, we used admissions for persons of
all ages for ICD codes 390-429; for emergency room visits for asthma, we used the estimated ER visit rate for persons of all ages; for
chronic bronchitis we used the incidence rate for individuals 27 and older; for the pooled estimate of minor restricted activity days and
any-of-19 respiratory symptoms, we used the incidence rate for minor restricted activity days.
b Percentage is calculated as the ratio of avoided mortality to the projected baseline annual non-accidental mortality for adults aged 30
and over. Non-accidental mortality was approximately 95% of total mortality for this subpopulation in 2010.
0 These health endpoints overlap with the "any-of-19 respiratory symptoms" category. As a result, although we present estimates for
each endpoint individually, these results are not aggregated into the total benefits estimates.
d Minor restricted activity days and any-of-19 respiratory symptoms have overlapping definitions and are pooled.
61
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Non-Fatal Health Impacts
We report non-fatal health effects estimates 111 a
similar manner to estimates of premature mortali-
ties: as a range of estimates for each quantified health
endpoint, with the range dependent 011 the quanti-
fied uncertainties in the underlying concentration-
response functions. The range of results for 2010
only is characterized in Table 5-3 with 5th percen-
tile, mean, and 95th percentile estimates which cor-
respond to the Primary Low, Primary Central, and
Primary High estimates, respectively. All estimates
are expressed as new cases avoided in 2010, with the
following exceptions. Hospital admissions reflect
admissions for a range of respiratory and cardiovas-
cular diseases, and these results, along with emer-
gency room visits for asthma, do not necessarily rep-
resent the avoidance of new cases of disease (i.e., air
pollution may simply exacerbate an existing condi-
tion, resulting in an emergency room visit or hospi-
tal admission). Further, each admission is only
counted once, regardless of the length of stay in the
hospital. "Shortness of breath" is expressed in terms
of symptom days: that is, one "case" represents one
child experiencing shortness of breath for one day.
Likewise, "Restricted Activity Days" and "Work
Loss Days" are expressed in person-days.
Avoided Health Effects of
Other Pollutants
This section discusses the health effects associ-
ated with non-criteria air pollutants regulated by the
Clean Air Act Amendments of 1990. It first dis-
cusses the effects of pollutants known as "air tox-
ics", and then summarizes the effects associated with
stratospheric ozone depleting substances.
Avoided Effects of Air Toxics
In addition to addressing the control of criteria
pollutants, the Clean Air Act Amendments re-
vamped regulations for air toxics — defined as non-
criteria pollutants which can cause adverse effects to
human health and to ecological resources — under
section 112 of the Act. Among other changes, the
1990 Amendments establish a list of air toxics to be
regulated, require EPA to establish air toxic emis-
sions standards based on maximum achievable con-
trol technology (MACT standards), and include a
provision that requires EPA to establish more strin-
gent air toxic standards if MACT controls do not
sufficiently protect the public health against residual
risks. Control of air toxics is expected to result both
from these changes and from incidental control due
to changes in criteria pollutant programs.
Table 5-4
Mortality Distribution by Age in Primary Analysis (2010 only), Based on Pope et al. (1995)a
Age Group
Proportion of Premature Mortality by Age b
Life Expectancy (years)
Infants
not estimated
-
1-29
not estimated
-
30-34
1%
48
35-44
4%
38
45-54
6%
29
55-64
12%
21
65-74
24%
14
75-84
30%
9
85+
24%
6
a Results based on PM-related mortality incidence estimates for the 48 state U.S. population.
b Percentages may not sum to 100 percent due to rounding.
62
-------
Chapter 5: Human Health Effects of Criteria Pollutants
For several decades, the primary focus of risk
assessments and control programs designed to reduce
air toxics has been cancer. According to present EPA
criteria, over 100 air toxics are known or suspected
carcinogens. EPA's 1990 Cancer Risk study indi-
cated that as many as 1,000 to 3,000 cancers annu-
ally may be attributable to the air toxics for which
assessments were available (virtually all of this esti-
mate came from assessments of about a dozen well-
studied pollutants).12 We note, however, that the
results of this analysis are based, in part, on conser-
vative, upper-bound estimates of chemical specific
risk factors.
In addition to cancer, inhalation of air toxics
compounds can cause a wide variety of health ef-
fects, including neurotoxicity, respiratory problems,
and adverse reproductive and developmental effects.
However, there has been considerably less work
done to assess the magnitude of non-cancer effects
from air toxics.
Air toxics can also cause adverse health effects
via non-inhalation exposure routes. Persistent
bioaccumulating pollutants, such as mercury and
dioxins, can be deposited into wrater or soil and sub-
sequently taken up by living organisms. The pol-
lutants can biomagnify through the food chain and
exist in high concentrations when consumed by
humans in foods such as fish or beef. The resulting
exposures can cause adverse effects in humans.
Finally, there are a host of other potential eco-
logical and welfare effects associated with air toxics,
for which very little exists in the way of quantita-
tive analysis. Toxic effects of these pollutants have
the potential to disrupt both terrestrial and aquatic
ecosystems and contribute to adverse welfare effects
such as fish consumption advisories in the Great
Lakes.13
2 These pollutants included PIC (products of incomplete
combustion), 1,3-butadiene, hexavalent chromium, benzene,
formaldehyde, chloroform, asbestos, arsenic, ethylene
dibromide, dioxin, gasoline vapors, and ethylene dichloride. See
U.S. EPA, Cancer Risk from Outdoor Exposure to Air Toxics,
EPA-450/ 1-90-G041 Prepared by EPA/OAR/OAQPS.
13 U.S. EPA, Office of Air Quality Planning and Standards.
"Deposition of Air Pollutants to the Great Waters, First Report
to Congress/'' May 1994. EP A -453/ R-93-055.
Unfortunately, the effects of air toxics emissions
reductions could not be quantified for the present
study. Unlike criteria pollutants, monitoring data
for air toxics are relatively scarce, and the data that
do exist cover only a handful of pollutants. Emis-
sions inventories are very limited and inconsistent,
and air quality modeling has only been performed
for a few source categories. In addition, the scien-
tific literature on the effects of air toxics is generally
much weaker than that available for criteria pollut-
ants. Appendix I presents a list of research needs
identified by the Project Team which, if met, would
enable at least a partial assessment of air toxics ben-
efits in future section 812 prospective studies.
Avoided Health Effects for Provisions to
Protect Stratospheric Ozone
We estimate benefits of stratospheric ozone pro-
tection programs by relying on analyses conducted
to support a scries of regulatory support documents
for these provisions. The series of basic steps to ar-
rive at physical effects estimates — from emissions
estimation, atmospheric modeling, exposure assess-
ment, and dose-response characterization — is simi-
lar to that used to estimate effects of criteria pollut-
ants, but the details of each modeling step are vastly
different. The emissions and atmospheric modeling
yields estimates of changes in ultraviolet-b (UV-b)
radiation, and the exposure and dose-response analy-
ses then yield estimates of the effects of changes in
UV-b radiation, including human health, welfare,
and ecological effects. Appendix G provides a de-
tailed description of the methodology and sources
used to generate these estimates. Several of the ben-
efits can be identified but cannot yet be reliably quan-
tified, and so are described qualitatively.
The quantified physical effects estimates of sec-
tions 604 and 606 of Title VI, the provisions that
provide the primary controls on production and re-
lease of CFCs and HCFCs generate about 98 per-
cent of the monetized quantified benefits estimate.
The quantified health benefits include the follow-
ing: reduced incidences of mortality and morbidity
associated with skin cancer (melanoma and
nonmelanoma); and reduced incidences of cataracts
63
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
and their associated pain and suffering.14 Using the
change in UV radiation dose, we estimate the num-
ber of additional cases of skin cancer (melanoma and
nonmelanoma) and cataracts. Because the baseline
levels of all of these UV-related health effects tend
to be higher for older people and for those with
lighter skins, EPA's method for projecting future
incremental skin cancers and cataracts incorporates
these factors in its benefits estimates.15 We present
a brief summary of these benefits in Table 5-5, and
the analysis is described in detail in Appendix G.
To calculate the number of deaths from mela-
noma, the model uses a dose response function simi-
lar to the C-R functions for criteria pollutants. For
nonmelanoma, the model estimates the number of
deaths by assuming that a fixed percentage of the
total nonmelanoma cases will result in death.16 We
estimate that from 1990 to 2165 sections 604 and
606 will result in 6.3 million avoided deaths from
skin cancer, 27.5 million avoided cataract cases, and
299.0 million cases of non-fatal skin cancers (mela-
noma and nonmelanoma). The unquantified effects
of sections 604 and 606 include avoided pain and
suffering from skin cancer and human health and
environmental benefits outside the United States.
Table 5-5
Major Health Benefits of Provisions to Protect Stratospheric Ozone
(CAAA Sections 604, 606, And 609)
Health Effects- Quantified
Estimate
Basis for Estimate
Melanoma and nonmelanoma
skin cancer
(fatal)
6.3 million lives saved from skin
cancer in the U.S. between 1990
and 2165
Dose-response function based on UV
exposure and demographics of
exposed populations.1
Melanoma and nonmelanoma
skin cancer
(non-fatal)
299 million avoided cases of non-
fatal skin cancers in the U.S.
between 1990 and 2165
Dose-response function based on UV
exposure and demographics of
exposed populations.1
Cataracts
27.5 million avoided cases in the
U.S. between 1990 and 2165
Dose-response function uses a
multivariate logistic risk function based
on demographic characteristics and
medical history. 1
Health Effects- Unquantified
Skin cancer: reduced pain and suffering
Reduced morbidity effects of increased UV. For example,
reduced actinic keratosis (pre-cancerous lesions resulting from excessive sun exposure)
reduced immune system suppression.
Notes:
1 For more detail see EPA's Regulatory Impact Analysis: Protection of Stratospheric Ozone (1988).
2 Note that the ecological effects, unlike the health effects, do not reflect the accelerated reduction and
phaseout schedule of section 606.
Benefits due to the section 606 methyl bromide phaseout are not included in the benefits total because
annual incidence estimates are not currently available.
4 Quantitative estimates presented in Appendix G also
include reduced crop damage associated with UV-b radiation
and tropospheric ozone; reduced damage to fish harvests associ-
ated with UV-b radiation; and reduced polymer degradation
from UV-b radiation. The derivation of these effects is described
ill more detail in Chapter 7.
13 The dose-response equation is (fractional change in inci-
dence) = (fractional change ill UV-b dose + 1)D -1, where b (the
biological amplification factor) equals the percent change in in-
cidence associated with a one percent change in dose. More
information about the origins of the models can be found in
Appendix G.
16 Scotto, Fears, Mid Fraumeiii, U.S. Department, of Health
and Human Sendees, N1H, "Incidence of Nonmelanoma Skill
Cancer in the United States," 1981, pages 2, 7, and 13.
64
-------
Chapter 5: Human Health Effects of Criteria Pollutants
Uncertainty in the Health
Effects Analysis
As discussed above, and in greater detail in Ap-
pendix D, a number of important assumptions and
uncertainties in the physical effects analysis may in-
fluence the estimate of monetary benefits presented
in this study. Several of these key uncertainties, their
potential directional bias (i.e., overestunatioii or
underestimation), and the potential significance of
each of these uncertainties for the overall net ben-
efit results of the analysis are summarized in Table
5-6. As shown in this table, the decisions made to
overcome the problems of limited data, the inad-
equacy of the currently available scientific literature,
and other constraints do not clearly bias the overall
results of this analysis 111 one particular direction.
Table 5-6
Key Uncertainties Associated with Human Health Effects Modeling (continued)
Potential Source of Error
Direction of Potential
Bias for Net Benefits
Estimate
Likely Significance Relative to
Key Uncertainties in Net Benefit Estimate*
Application of C-R
relationships only to those
subpopulations matching the
original study population.
Underestimate.
Potentially major. The C-R functions for several
health endpoints (including PM-related premature
mortality) were applied only to subgroups of the U.S.
population (e.g., adults over 30) and thus may
underestimate the whole population benefits of
reductions in pollutant exposures. In addition, the
demographics of the study population in the Pope et
al. study (largely white and middle class) may result
in an underestimate of PM-related mortality, because
the effects of PM tend to be significantly greater
among groups of lower socioeconomic status.
No quantification of health
effects associated with
exposure to air toxics.
Underestimate
Potentially major. According to EPA criteria, over
100 air toxics are known or suspected carcinogens,
and many air toxics are also associated with adverse
health effects such as neurotoxicity, reproductive
toxicity, and developmental toxicity. Unfortunately,
current data and methods are insufficient to develop
(and value) quantitative estimates of the health
effects of these pollutants.
Use of long-term global
warming estimates in Title VI
analysis that show more
severe warming than is now
generally anticipated.
Overestimate (for Title
VI estimate only)
Potentially major. Global warming can accelerate
the pace of stratospheric ozone recovery; if warming
is less severe than anticipated at the time the Title VI
analyses were conducted, the modeled pace of
ozone recovery may be overestimated, suggesting
benefits of the program could be delayed, perhaps
by many years. The magnitude of estimated Title VI
benefits suggests that the impact of delaying
benefits could be major.
The quantitative analysis of
Title VI (see next section)
does not account for
potential increases in
averting behavior (i.e.,
people's efforts to protect
themselves from UV-b
radiation).
Unable to determine
based on current
information.
Potentially major. Murdoch and Thayer (1990)
estimate that the cost-of-illness estimates for
nonmelanoma skin cancer cases between 2000 and
2050 may be almost twice the estimated cost of
averting behavior (application of sunscreen). Our
Title VI analysis relies on epidemiological studies,
which incorporate averting behavior as currently
practiced. Omission of future increases in averting
behavior, however, may overstate the benefits of
reduced emissions of ozone-depleting chemicals.
Benefits could be understated if individuals alter their
behaviors in ways that could increase exposure or
risk (e.g., sunbathing more frequently). A recent
European study by Autier et al. (1999) found that the
use of high sun protection factor (SPF) sun screen is
associated with increased frequency and duration of
sun exposure.
65
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 5-6
Key Uncertainties Associated with Human Health Effects Modeling (continued)
Potential Source of Error
Direction of Potential
Bias for Net Benefits
Estimate
Likely Significance Relative to
Key Uncertainties in Net Benefit Estimate*
Analysis assumes a causal
relationship between PM
exposure and premature
mortality based on strong
epidemiological evidence of
a PM/mortality association.
However, epidemiological
evidence alone cannot
establish this causal link.
Unable to determine
based on current
information.
Potentially major. A basic underpinning of this
analysis, this assumption is critical to the estimation
of health benefits. However, the assumption of
causality is suggested by the epidemiologic evidence
and is consistent with current practice in the
development of a best estimate of air pollution-
related health benefits. At this time, we can identify
no basis to support a conclusion that such an
assumption results in a known or suspected
overestimation bias.
Across-study variance /
application of regionally
derived C-R estimates to
entire U.S.
Unable to determine
based on current
information.
Potentially major. The differences in the expected
changes in health effects calculated using different
underlying studies can be large. If differences reflect
real regional variation in the PM/mortality
relationship, applying individual C-R functions
throughout the U.S. could result in considerable
uncertainty in health effect estimates.
Estimate of non-melanoma
skin cancer mortality
resulting from reductions in
stratospheric ozone is
calculated indirectly, by
assuming the mortality rate
is a fixed percentage of non-
melanoma incidence.
Unable to determine
based on current
information.
Potentially major. New data on the death rate for
non-melanoma skin cancer may significantly
influence the Title VI mortality estimate. Some
preliminary estimates suggest that this estimate may
need to be adjusted downward.
The baseline incidence
estimate of chronic bronchitis
based on Abbey et al. (1995)
excluded 47 percent of the
cases reported in that study
because those reported
"cases" experienced a
reversal of symptoms during
the study period. These
"reversals" may constitute
acute bronchitis cases that
are not included in the acute
bronchitis analysis (based on
Dockery et al., 1996).
Underestimate.
Probably minor. The relative contribution of acute
bronchitis cases to the overall benefits estimate is
small compared to other health benefits such as
avoided mortality and avoided chronic bronchitis.
CAAA fugitive dust controls
implemented in PM non-
attainment areas would
reduce lead exposures by
reducing the re-entrainment
of lead particles emitted prior
to 1990. This analysis does
not estimate these benefits.
Underestimate
Probably minor. While the health and economic
benefits of reducing lead exposure can be
substantial (e.g., see section 812 Retrospective
Study Report to Congress), most additional fugitive
dust controls implemented under the Post-CAAA
scenario (e.g., unpaved road dust suppression,
agricultural tilling controls, etc) tend to be applied in
relatively low population areas.
Exclusion of C-R functions
from short-term exposure
studies in PM mortality
calculations.
Underestimate
Probably minor. Long-term PM exposure studies
may be able to capture some of the impact of short-
term peak exposure on mortality; however the extent
of overlap between the two study types is unclear.
66
-------
Chapter 5: Human Health Effects of Criteria Pollutants
Table 5-6
Key Uncertainties Associated with Human Health Effects Modeling (continued)
Direction of Potential
Bias for Net Benefits
Potential Source of Error Estimate
Age-specific C-R functions Unable to determine
for PM related premature based on current
mortality not reported by information.
Pope etal. (1995).
Estimation of the degree of
life-shortening associated
with PM-related mortality
used a single C-R function
for all applicable age groups.
Assumption that PM-related
mortality occurs over a
period of five-years following
the critical PM exposure.
Analysis assumes that 25
percent of deaths occur in
year one, 25 percent in year
two, and 16.7 percent in
each of the remaining three
years.
Extrapolation of criteria Unable to determine
pollutant concentrations to based on current
populations distant from information,
monitors.
Exposure analysis in areas Unable to determine
beyond 50 km is based on a based on current
new technique that relies on information.
the direct use of air quality
modeling results in
combination with adjusted
monitor data.
Pope et al. (1995) study did Unable to determine
not include pollutants other based on current
than PM. information.
Likely Significance Relative to
Key Uncertainties in Net Benefit Estimate*
Unknown, possibly major when using a value of life
years approach. Varying the estimate of degree of
prematurity has no effect on the aggregate benefit
estimate when a value of statistical life approach is
used, since all incidences of premature mortality are
valued equally. Under the alternative approach
based on valuing individual life-years, the influence
of alternative values for numbers of average life-
years lost may be significant.
Probably minor. If the analysis underestimates the
lag period, benefits will be overestimated, and vice-
versa. However, available epidemiological studies
do not provide evidence of the existence or potential
magnitude of a lag between exposure and incidence.
Thus, an underestimate of the lag seems unlikely. If
the assumed lag structure is an overestimate, even if
benefits are fully discounted from the future year of
death, application of reasonable discount rates over
this period would not significantly alter the monetized
benefit estimate.
Probably minor. Extrapolation method is most
accurate in areas where monitor density is high.
Monitor density tends to be highest in areas with
high criteria pollutant exposures; thus most of this
uncertainty affects low exposure areas where
benefits are likely to be low. In addition, an
enhanced extrapolation method incorporating
modeling results is used for areas far (> 50 km) from
a monitor.
Probably minor. The new technique is used for less
than 10 percent of the country for PM exposure, and
less than 15 percent for ozone. The approach we
use should be more accurate than the alternative
approach of linear interpolation over long distances.
The new method nonetheless requires further testing
against monitor data to access its accuracy.
Probably minor. If ozone and other criteria pollutants
correlated with PM contribute to mortality, that effect
may be captured in the PM estimate. Thus, PM is
essentially used as a surrogate for a mix of
pollutants. This uncertainty does make it difficult to
disaggregate avoided mortality benefits by pollutant,
however other studies (besides Pope) suggest that
PM is the dominant factor in premature mortality.
Unable to determine
based on current
information.
* The classification of each potential source of error reflects the best judgement of the section 812 Project Team. The
Project Team assigns a classification of "potentially major" if a plausible alternative assumption or approach could influence
the overall monetary benefit estimate by approximately five percent or more; if an alternative assumption or approach is
likely to change the total benefit estimate by less than five percent, the Project Team assigns a classification of "probably
minor."
67
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
f This page left blank intentionally.7
68
-------
Economic Valuation
of Human Health
Effects
The reduced incidence of physical effects is a
valuable measure of health benefits for individual
endpoints; however, to compare or aggregate ben-
efits across endpoints, the benefits must be mon-
etized. Assigning a dollar value to avoided incidences
of each effect permits us to sum monetized benefits
realized as a result of the CAAA, and compare them
with the associated costs.
In the 812 prospective analysis, we have two
broad categories of benefits, health and welfare ben-
efits. Human health effects include mortality and
morbidity endpoints, which are presented in this
chapter. Welfare effects include agricultural and eco-
logical benefits, visibility, and worker productivity,
which are covered in the following chapter. We
obtain valuation estimates from the economic lit-
erature, and report them in "dollars per case reduced
for health effects" and "dollars per unit of avoided
damage for welfare effects".1 Similar to estimates of
physical effects provided by health studies, we re-
port each of the monetary values of benefits applied
111 this analysis in terms of a central estimate and a
probability distribution around that value. The sta-
tistical form of the probability distribution varies
by endpoint. For example, we use a Weibull distri-
bution to describe the estimated dollar value of an
avoided premature mortality, while we assume the
estimate for the value of a reduced case of acute bron-
chitis is uniformly distributed between a minimum
and maximum value.
Although human health benefits of the 1990
Amendments are attributed to reduced emissions of
criteria pollutants (Titles 1 through V) and reduced
emission of ozone depleting substances (Title VI),
this chapter focuses only on the valuation of human
health effects attributed to the reduction of criteria
1 The literature reviews arid process lor developing valua-
tion estimates are described ill detail ill Appendix I and in refer-
enced supporting reports.
pollutants. The chapter begins with an brief review
of the economic concepts behind valuing human
health effects in a cost-benefit context and a sum-
maw of the unit values applied to health endpoints.
We follow with a discussion of how we derive valu-
ation estimates for specific health effects. We then
present the results of this analysis. W'e conclude the
chapter with a review of the uncertainties associated
with benefits valuation.
Our analysis indicates that the benefit of avoided
premature mortality risk reduction dominates the
overall net benefit estimate. This is, in part, due to
the high monetary value assigned to the avoidance
of premature mortality relative to the unit value of
other health endpoints. Because of the critical im-
portance of this endpoint in the study's results, this
chapter pays particular attention to the major chal-
lenges to valuing mortality risk reductions and the
limitations of the estimate we apply in this analysis.
There are also significant reductions in short term
and chronic health effects and a substantial number
of health (and welfare) benefits that we could not
quantify or monetize.
Valuation of Benefit
Estimates
In an environmental benefit-cost analysis, the
dollar value of an environmental benefit (e.g., a
health-related improvement due to environmental
quality) enjoyed by an individual is the dollar amount
such that the person would be indifferent between
experiencing the benefit and possessing the money.
In general, the dollar amount required to compen-
sate a person for exposure to an adverse effect is
roughly the same as the dollar amount a person is
willing to pay to avoid the effect. Thus, economists
speak of "willingness-to-pay" (WTP) as the appro-
priate measure of the value of avoiding an adverse
69
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
effect. For example, the value of an avoided respira-
tory symptom would be a person's WTP to avoid
that symptom.
For most goods, WTP can be observed by ex-
amining actual market transactions. For example, if
a gallon of bottled drinking water sells for one dol-
lar, it can be observed that at least some persons are
willing to pay one dollar for such water. For goods
that are not exchanged in the market, such as most
environmental "goods," valuation is not so straight-
forward. Nevertheless, a value may be inferred from
observed behavior, such as through estimation of
the WTP for mortality risk reductions based 011
observed sales and prices of products that result in
similar effects or risk reductions, (e.g., non-toxic
cleaners or bike helmets). Alternatively, surveys may
be used in an attempt to directly elicit WTP for an
environmental improvement.
Wherever possible 111 this analysis, we use esti-
mates of mean WTP. In cases where WIT estimates
are not available, wre use the cost of treating or miti-
gating the effect as an alternative estimate. For ex-
ample, for the valuation of hospital admissions we
use the avoided medical costs as an estimate of the
value of avoiding the health effects causing the ad-
mission. These costs of illness (COI) estimates gen-
erally understate the true value of avoiding a health
effect. They tend to reflect the direct expenditures
related to treatment and not the utility an individual
derives from improved health status or avoided
health effect. As noted above, we use a range of
values for most environmental effects, to support
the primary central estimate of net benefits. Table
6-1 summaries the mean unit value estimates that
we use in this analysis. We present the full range of
values in Appendix H, including those used to de-
rive the primary low and primary high estimates, as
well as values used to generate an alternative value
for avoiding premature mortality.
Valuation of Premature Mortality
Some forms of air pollution increase the prob-
ability that individuals will die prematurely. We use
concentration-response functions for mortality that
express the increase in mortality risk as cases of "ex-
Table 6-1
Health Effects Unit Valuation (1990 dollars)
Endpoint Pollutant Valuation (mean est.)
Mortality
PM10
$4,800,000
per case
Chronic Bronchitis
PM10
$260,000
per case
Chronic Asthma
03
$25,000
per case
Hospital Admissions
All Respiratory
SO2, NO2, PM10 & O3
$6,900
per case
All Cardiovasular
SO2, NO2, & CO PM10 &
O3
$9,500
per case
Emergency Room Visits for Asthma
PM10 & O3
$194
per case
Respiratory Illness and Symptoms
Acute Bronchitis
PM10
$45
per case
Asthma Attack or Moderate or
PM10 & O3
$32
per case
Worse Asthma Day
Acute Respiratory Symptoms
SO2, NO2, PM-i, & O3
$18
per case
Upper Respiratory Symptoms
PM1
$19
per case
Lower Respiratory Symptoms
PM10
$12
per case
Shortness of Breath, Chest
PM10 & SO2
$5.30
per day
Tightness, or Wheeze
Work Loss Days
PM10
$83
per day
Mild Restricted Activity Days
PM10 & O3
$38
per day
70
-------
Chapter 6: Economic Valuation of Human Health Effects
cess premature mortality" per time period (e.g., per
year). The benefit, however, is the avoidance of small
increases 111 the risk of mortality. By summing indi-
viduals' WTP to avoid small increases in risk over
enough individuals, we can infer the value of a sta-
tistical premature death avoided.2 For expository
purposes, we express this valuation as "dollars per
mortality avoided," or "value of a statistical life"
(VSL), even though the actual valuation is of small
changes in mortality risk experience by a large num-
ber of people. The economic benefits associated with
avoiding premature mortality were the largest cat-
egory of monetized benefits in the section 812 CAA
retrospective analysis (U.S. "EPA 1997) and continue
to be the largest source of monetized benefits for
this prospective analysis. .Mortality benefits, how-
ever, are also the largest contributor to the range of
uncertainty in monetized benefits. For a more de-
tailed discussion of the factors affecting the valua-
tion of premature mortality see Appendix H.
The health science literature on air pollution
indicates that several human characteristics affect the
degree to which mortality risk affects an individual.
For example, some age groups appear to be more
susceptible to air pollution than others (e.g., the eld-
erly and children). Health status prior to exposure
also affects susceptibility. At risk individuals include
those who have suffered strokes or are suffering from
cardiovascular disease and angina (Rowlatt, et al.
1998). An ideal economic benefits estimate of mor-
tality risk reduction would reflect these human char-
acteristics, in addition to an individual's willingness
to pay (WTP) to improve one's own chances of sur-
vival plus WTP to improve other individuals' sur-
vival rates.3 The ideal measure would also take into
account the specific nature of the risk reduction com-
modity that is provided to individuals, as well as the
context in which risk is reduced. To measure this
value, it is important to assess how reductions in air
pollution reduce the risk of dying from the time that
reductions take effect onward, and how individuals
2 Because people are valuing small decreases in the risk of
premature mortality, it is expected deaths that are inferred. For
example, suppose that a given reduction in pollution confers on
each exposed individual a decrease in mortal risk of 1/100,000.
Then among 100,000 such individuals, one fewer individual can
be expected to die prematurely , If each individual's WTP for
that risk reduction is $50, then the implied value of a statistical
premature death avoided is $50 x 100,000 = $5 million.
- For a more detailed discussion of altruistic values related
to the value of l.iie, see Jones-Lee (1992).
value these changes. Each individual's survival curve,
or the probability of surviving beyond a given age,
should shift as a result of an environmental quality
improvement. For example, changing the current
probability of survival for an individual also shifts
future probabilities of that individual's survival. This
probability shift will differ across individuals because
survival curves are dependent on such characteris-
tics as age, health state, and the current age to which
the individual is likely to survive
Although a survival curve approach provides a
theoretically preferred method for valuing the eco-
nomic benefits of reduced risk of premature mortal-
ity associated writh reducing air pollution, the ap-
proach requires a great deal of data to implement.
The economic valuation literature does not yet in-
clude good estimates of the value of this risk reduc-
tion commodity. As a result, in this study we value
avoided premature mortality risk using the value of
statistical life approach, supplemented by an alter-
native valuation based on a value of statistical life
years lost approach. We provide a review7 of the
relevant literature and a more detailed discussion of
our selected approach in Appendix 11.
As in the retrospective, we use a mortality risk
valuation estimate which is based on an analysis of
26 policy-relevant value-of-life studies (see Table 6-
2). Five of the 26 studies are contingent valuation
(CV) studies, which directly solicit WTP informa-
tion from subjects; the rest are wage-risk studies,
which base WTP estimates on estimates of the addi-
tional compensation demanded in the labor market
for riskier jobs. We used the best estimate from each
of the 26 studies to construct a distribution of mor-
tality risk valuation estimates for the section 812
study. A Weibull distribution, with a mean of $4.8
million and standard deviation of 13.24 million, pro-
vided the best fit to the 26 estimates. There is con-
siderable uncertainty associated with this approach.
We discuss this issue in detail later in this chapter
and in Appendix H.
In addition, we developed alternative calculations
based on a life-years lost approach. To employ the
value of statistical life-year (VSLY) approach, we first
estimated the age distribution of those lives projected
to be saved by reducing air pollution. Based on life
expectancy tables, we calculate the life-years saved
71
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 6-2
Summary of Mortality Valuation Estimates (millions of $1990)
Study
Type of
Estimate
Valuation
(millions 1990$)
Kneisner and Leeth (1991) (US)
Labor Market
0.6
Smith and Gilbert (1984)
Labor Market
0.7
Dillingham (1985)
Labor Market
0.9
Butler (1983)
Labor Market
1.1
Miller and Guria (1991)
Cont. Value
1.2
Moore and Viscusi (1988a)
Labor Market
2.5
Viscusi, Magat, and Huber (1991b)
Cont. Value
2.7
Gegax et al. (1985)
Cont. Value
3.3
Marin and Psacharopoulos (1982)
Labor Market
2.8
Kneisner and Leeth (1991)
(Australia)
Labor Market
3.3
Gerking, de Haan, and Schulze
(1988)
Cont. Value
3.4
Cousineau, Lacroix, and Girard
(1988)
Labor Market
3.6
Jones-Lee (1989)
Cont. Value
3.8
Dillingham (1985)
Labor Market
3.9
Viscusi (1978, 1979)
Labor Market
4.1
R.S. Smith (1976)
Labor Market
4.6
V.K. Smith (1976)
Labor Market
4.7
Olson (1981)
Labor Market
5.2
Viscusi (1981)
Labor Market
6.5
R.S. Smith (1974)
Labor Market
7.2
Moore and Viscusi (1988a)
Labor Market
7.3
Kneisner and Leeth (1991) (Japan)
Labor Market
7.6
Herzog and Schlottman (1987)
Labor Market
9.1
Leigh and Folson (1984)
Labor Market
9.7
Leigh (1987)
Labor Market
10.4
Garen (1988)
Labor Market
13.5
Source: Viscusi, 1992 and EPA analysis.
from each statistical life saved within each age and
gender cohort. To value these statistical life-years,
we hypothesized a conceptual model which depicted
the relationship between the value of life and the
value of life-years. As noted in Chapter 5, the aver-
age number of life-years saved across all age groups
for which data were available is 14 for PM-related
mortality. The average for PM, in particular, differs
from the 35-year expected remaining lifespan derived
from existing wage-risk studies.4 Using the same
distribution of value of life estimates used above (i.e.
the Wei bull distribution with a
mean estimate of $4.8 million),
we estimated a distribution for the
value of a life-year and combined
it with the total number of esti-
mated life-years lost. The details
of these calculations are presented
in Appendix H.
Valuation of Specific
Health Effects
Chronic Bronchitis
The best available estimate of
WTP to avoid a case of chronic
bronchitis (CB) comes from
Viscusi et al. (1991). The Viscusi
study, however, describes to the
respondents a severe case of CB.
We employ an estimate of WTP
to avoid a pollution-related case
of CB that is based on adjusting
the WTP to avoid a severe case,
as estimated by Viscusi et al.
(1991), to account for the likeli-
hood that an average case of pol-
lution-related CB is not as severe.
(1992).
See, lor example, Moore and Viscusi (1988) or Viscusi
We use the mean of a distri-
bution of WTP estimates as the
central tendency estimate of WTP
to avoid a pollution-related case
of chronic bronchitis in this
analysis. The distribution incor-
porates uncertainty from three
sources: (1) the WTP to avoid a
case of severe CB, as described by
Viscusi et al., 1991; (2) the seventy level of an aver-
age pollution-related case of CB (relative to that of
the case described by Viscusi et al., 1991); and (3) the
elasticity of WTP with respect to severity of the ill-
ness. Based on assumptions about the distributions
of each of these three uncertain components, we
derive a distribution of WTP to avoid a pollution-
related case of CB by statistical uncertainty analysis
techniques.5 The expected value of this distribution,
3 The statistical uncertainty analysis technique we used,
which is also known as simulation modeling, is a probabilistic
approach to characterizing the uncertainty or the distribution
ol potential values around a central estimate.
72
-------
Chapter 6: Economic Valuation of Human Health Effects
which is about §260,000, is taken as the central ten-
dency estimate of WTP to avoid a pollution-related
case of CB. We describe the three underlying distri-
butions, and the generation of the resulting distri-
bution of WTP, in Appendix H.
Chronic Asthma
'l'he valuation of this health endpoint requires
an estimate which reflects an individual's desire to
avoid the effects of chronic asthma throughout his
or her lifetime. We derive this valuation estimate
from two studies that solicit values from individuals
diagnosed as asthmatics. Blumenschem and
Johannesson (1998) generate an estimate of monthly
WTP, while O'Conor and Blomquist (1997) gener-
ate an annual WTP estimate. In order to extend
monthly and annual WTP estimates over an
individual's lifetime, we adjusted the reported esti-
mates to reflect the average life-years remaining and
age distribution of the adult U.S. population, given
that chronic asthma is not expected to affect the av-
erage life expectancy. The mean estimate of WTP
to avoid a case of chronic asthma resulting from this
method is approximately $25,000.
Respiratory-Related Ailments
In general, the values we assign to the respira-
tory-related ailments in Table 6-1 are a combination
of WTP estimates for individual symptoms compris-
ing each ailment. For example, a willingness to pay
estimate to avoid the combination of specific upper
respiratory symptoms defined in the concentration-
response relationship measured by Pope et al. (1991)
is not available. While that study defines upper res-
piratory symptoms as one suite of ailments (runny
or stuffy nose; wet cough; and burning, aching, or
red eves), the valuation literature reports individual
WTP estimates for three closely matching symptoms
(head/sinus congestion, cough, and eye irritation).
We therefore use these available WTP estimates and
a benefits transfer procedure to estimate the value
of avoiding those symptoms defined in the concen-
tration-response study.
To capture the uncertainty associated with the
valuation of respiratory-related ailments, we incor-
porated a range of values reflecting the fact that an
ailment, as defined in the concentration-response
relationship, could be comprised of just one symp-
tom or several. At the high end of the range, the
valuation represents an aggregate of WTP estimates
for several individual symptoms. The low end rep-
resents the value of avoiding a single mild symptom.
Minor Restricted Activity Days
An individual suffering from a single severe pol-
lution-related symptom or combination of symp-
toms may experience a Minor Restricted Activity
Day (MRAD). Krupnick and Kopp (1988) argue that
mild symptoms will not be sufficient to result in a
MRAD, so that WTP to avoid a MRAD should ex-
ceed WTP to avoid any single mild symptom. On
the other hand, WTP to avoid a MRAD should not
exceed the WTP to avoid a work loss day (which
results when the individual experiences more severe
symptoms). No studies report an estimate for WTP
to avoid a day of minor restricted activity. There-
fore, we derive for this analysis a value from WTP
estimates for avoiding combinations of symptoms
which may result in a day of minor restricted activ-
ity ($38 per day). The uncertainty range associated
with this value extends from the highest value for a
single symptom to the value for a work loss day.
Furthermore, a distributional form is used which
reflects our expectations that the actual value is likely
to be closer to the central estimate than either ex-
treme.
Hospital Admissions. Cardiovascular
and Respiratory
The valuation of this benefits category reflects
the value of reduced incidences of hospital admis-
sions due to respiratory or cardiovascular conditions.
We use avoided hospital admissions as a measure as
opposed to the number of avoided cases of respira-
tory or cardiovascular conditions, because of the
availability of C-R relationships for the hospital ad-
missions endpoint. Hospital admissions reflect a class
of health effects linked to air pollution which are
acute in nature but more severe than the symptom-
day measures discussed above.
As described in Chapter 5, our approach to esti-
mating the number of incidences for this category-
involves reliance on several concentration-response
(C-R) functions. "Each concentration response func-
73
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
tion provides an alternative definition of either res-
piratory effects or cardiovascular effects, and may
be based on different pollutants. For valuation of
the incidences, the current literature provides well-
developed and detailed cost estimates of hospitaliza-
tion by health effect or illness. Using illness-specific
estimates of avoided medical costs and avoided costs
of lost work-time, developed by Elixhauser (1993),
we construct cost of illness (COI) estimates that are
specific to the suite of health effects defined by each
C-R function. For example, we use twelve distinct
C-R functions to quantify the expected change in
respiratory admissions.6 Consequently in this analy-
sis, we develop twelve separate COI estimates, each
reflecting the unique composition of health effects
considered in the individual studies.
The use of COI estimates suggests we likely un-
derstand the WTP to avoid these effects. The valu-
ation of any given health effect, such as hospitaliza-
tion, should reflect the value of avoiding associated
pain and suffering and lost leisure time, in addition
to medical costs and lost work time. While the prob-
ability distributions in this analysis characterize a
range of potential costs associated with hospitaliza-
tion, they do not account for the omission of fac-
tors from the COI estimates such as pain and suffer-
ing. Consequently, the valuations for these end-
points most likely understate the true social values
for avoiding hospital admissions due to respiratory
or cardiovascular conditions.
Stratospheric Ozone Provisions
We develop monetary estimates of the health
benefits due to stratospheric ozone provisions based
on estimated incidences presented in a series of ex-
isting regulatory support analyses. To ensure con-
sistency with the valuation strategy of this analysis,
however, we adjust certain parameters used in the
existing regulatory analyses of Title VI provisions.
Specifically, we re-evaluate the physical effects change
projected in the RIAs using the discount rate and
the value of statistical life adopted throughout the
rest of our present study. The net effect of these
changes is to reduce the estimates of benefits from
those found in the regulatory source support docu-
c For more detailed discussion of liie various health effects
considered by each C-R function and methodology for estimat-
ing the number of avoided hospital admissions, see Appendix
D.
ments. The most important change is the discount
rate. Because the benefits of stratospheric ozone
protection accrue over several hundred years, the
discount rate chosen can have an especially large in-
fluence on the benefits estimate. The central esti-
mate employed in this analysis is five percent; the
rate used in the source documents is two percent.
The value of statistical life (VSL) estimate is also
an important factor in the calculations, because the
vast majority of benefits of stratospheric ozone pro-
tection result from avoided fatal skin cancer cases.
To reflect the uncertainty of the YSL estimates, we
employ the same statistical uncertainty aggregation
approach used in the criteria pollutant analysis, us-
ing a Weibull distribution of VSL estimates as an
input. Appendix G describes the details of these and
other changes made to ensure consistency between
our stratospheric ozone provision benefits analysis
and our criteria pollutant analyses.
Results of Benefits Valuation
We combine the number of reduced incidences
of our health endpoints with our estimated values
of avoiding the health effect to generate total annual
monetized human health benefits in 2000 and 2010.
We attribute to Titles I through V of the CAAA
total annual human health benefits of 168 billion in
2000 and $110 billion in 2010. We summarize the
Post-CAAA 2010 monetized benefit in Table 6-3.
The table provides our central estimate, in addition
to the 5th and 95th percentile estimates for each ben-
efit category.
There are two aspects of our results that war-
rant discussion. The first is the valuation of prema-
ture mortality due to PM exposure. The second is
our strategy to avoid double-counting when aggre-
gating health benefits. As discussed in Chapter 5,
premature mortality is attributed to PM exposure
and our primary estimate reflects a lag between PM
exposure and premature mortality. While this lag
does not alter the number of estimated incidences, it
does alter the monetization of benefits. Because we
value the "event" rather than the present risk, in this
analysis we assume that the value of avoided future
premature mortality should be discounted. There-
fore, the type of lag structure employed plays a di-
rect role in the valuation of this endpoint.
74
-------
Chapter 6: Economic Valuation of Human Health Effects
Table 6-3
Results of Human Health Benefits Valuation, 2010
Monetary Benefits
(in millions 1990$)
5th %ile
Mean
95th %ile
Mortality
Ages 30+
$ 14,000
$ 100,000
$ 250,000
Chronic Illness
Chronic Bronchitis
$ 360
$ 5,600
$ 18,000
Chronic Asthma
40
180
300
Hospitalization
All Respiratory
$ 75
$ 130
$ 200
Total Cardiovascular
93
390
960
Asthma-Related ER Visits
0.1
1
3
Minor Illness
Acute Bronchitis
$ 0
$2
$ 5
URS
4
19
39
LRS
2
6
12
Respiratory Illness
1
6
15
Mod/Worse Asthma1
2
13
29
Asthma Attacks1
20
55
100
Chest tightness, Shortness of
Breath, or Wheeze
0
0.6
3
Shortness of Breath
0
0.5
1.2
Work Loss Days
300
340
380
MRAD/Any-of-19
680
1,200
1,800
Total Benefits in 20102
-
$ 110,000
-
Note:
1 Moderate to worse asthma and asthma attacks are endpoints included in the
definition of MRAD/Any-of-19 respiratory effects. Although valuation estimates are
presented for these categories, the values are not included in total benefits to avoid
the potential for double-counting.
2 Summing 5th and 95th percentile values would yield a misleading estimate of the
5th and 95th percentile estimate of total health benefits. For example, the likelihood
that the 5th percentile estimates for each endpoint would simultaneously be drawn
during the statistical uncertainty analysis is much less than 5 percent. As a result,
we present only the total mean.
The primary analysis reflects a five-year lag struc-
ture. Under this scenario, 50 percent of the esti-
mated cases of avoided mortality occur within the
first two years. The remaining 50 percent are then
distributed across the next three years. Our valua-
tion of avoided premature mortality applies a five
percent discount rate to the lagged estimates over
the periods 2000 to 2005 and 2010 to 2015. We dis-
count over the period between the initial PM expo-
sure change (either 2000 or 2010) and timing of the
incidence.
Many of the monetized health
benefit categories include overlapping
health endpoints, creating the poten-
tial for double-counting. In an effort
to avoid overstating the benefits, we
do not aggregate all of the quantified
health effects. For example, asthma
attacks and moderate to worse asthma
are considered components of the
endpoint, "Any of 19 Respiratory-
Symptoms". Consequently, we
present the results but do not include
them in our reported total benefits
figures. In other cases, there are end-
points included in our aggregation of
benefit that appear to have overlap-
ping health effects. For those ben-
efit categories that describe similar
health effects, it is important to keep
in mind that estimated incidences are
based on unique portions of the popu-
lation.
Valuation
Uncertainties
We addressed many valuation
uncertainties explicitly and quantita-
tively by expressing values as distri-
butions (see Appendix II for a com-
plete description of distributions
employed), using a computerized sta-
tistical technique to apply the valua-
tions to physical effects (see Chapters
5 and 8) with the mean of each valu-
ation distribution providing the foun-
dation for the primary central esti-
mate of total net benefits. This approach does not,
of course, guarantee that all uncertainties have been
adequately characterized, nor that the valuation es-
timates are unbiased. It is possible that the actual
WTP to avoid an air pollution-related impact is out-
side of the range of estimates used in this analysis.
Nevertheless, we assume that the distributions em-
ployed are reasonable approximations of the ranges
of uncertainty, and that there is no compelling rea-
son to believe that the mean values employed arc
systematically biased (except for the cost of illness
values, which probably underestimate WTP). There
are, however, a limited number of health endpoints
75
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
for which a different valuation approach may yield
results significantly different from out primary cen-
tral benefit estimate. For example, using a value of
statistical life year approach in lieu of the value of
statistical life method for valuing avoided premature
mortality yields a mean estimate for this benefit
which is approximately 45 percent lower than our
primary central estimate. For those few endpoints
where reasonable alternative valuation paradigms
yield significantly different results from our preferred
approach, see our discussion in Chapter 8.
The potential for biases as introduced by ben-
efits transfer methodology is applicable to all ben-
efits categories and, as noted in Table 6-4, the direc-
tion of its bias is unknown. Because changes in
mortality risk are the single most important compo-
nent of aggregate benefits, mortality risk valuation
is also the dominant component of the quantified
uncertainty. This category accounts for over 90
percent of total annual estimates under the Post-
CAAA scenario. The second largest benefits cat-
egory, reduced risk of chronic bronchitis, valued at
approximately $5.6 billion per year in 2010, accounts
for roughly five percent of the total estimated ben-
efits. Consequently, any uncertainty concerning
mortality risk valuation beyond that addressed by
the quantitative uncertainty assessment (i.e., that
related to the Weibull distribution with a mean value
of $4.8 million) deserves note.
Mortality Risk Benefits Transfer
One issue that merits special attention is the
uncertainties and possible biases related to the "ben-
efits transfer" from the 26 valuation source studies
to valuation of reductions in PM-related mortality
rates. Given the limitations of the current litera-
ture, we address this source of uncertainty qualita-
tively in this section. Although each of the mortal-
ity risk valuation source studies (see Table 6-2) esti-
mate the average WTP for a given reduction in mor-
tality risk, the degree of reduction in risk being val-
ued varies across studies and is not necessarily the
same as the degree of mortality risk reduction esti-
mated in this analysis. The transferability of esti-
mates of the value of a statistical life from the 26
studies to the section 812 benefit analysis rests on
the assumption that, within a reasonable range, WTP
for reductions in mortality risk is linear in risk re-
duction. For example, suppose a study estimates that
the average WTP for a reduction in mortality risk
of 1/100,000 is $50, but that the actual mortality
risk reduction resulting from a given pollutant re-
duction is 1/10,000. If WTP for reductions in mor-
tality risk is linear in risk reduction, then a WTP of
150 for a reduction of 1/100,000 implies a WTP of
1500 for a risk reduction of 1/10,000 (which is ten
times the risk reduction valued in the study). Un-
der the assumption of linearity, the estimate of the
value of a statistical life does not depend on the par-
ticular amount of risk reduction being valued. This
assumption has been shown to be reasonable pro-
vided the change in the risk being valued is within
the range of risks evaluated in the underlying stud-
ies (Rowlatt et al. 1998).
Although the particular amount of mortality risk
reduction being valued in a study may not affect the
transferability of the WTP estimate from the study
Table 6-4
Valuation of CAAA Benefits: Potential Sources and Likely Direction of Bias
Benefits Category
Factor
Likely Direction of Bias in WTP
Estimates Used in this Study
Premature Mortality
Age
Uncertain, perhaps overestimate
Degree of Risk Aversion
Underestimate
Income
Uncertain
Voluntary vs. Involuntary
Underestimate
Catastrophic vs. Protracted Death
Uncertain, perhaps underestimate
Discounting over a latency period
Uncertain, perhaps underestimate
Chronic Bronchitis
Severity-level
Uncertain
Elasticity of WTP with respect to
severity
Uncertain
All other benefit endpoints
Benefits Transfer
Uncertain
76
-------
Chapter 6: Economic Valuation of Human Health Effects
to the benefit analysis, the characteristics of the study
subjects and the nature of the mortality risk being
valued in the study could be important. Certain char-
acteristics of both the population affected and the
mortality risk facing that population are believed to
affect the average WTP to reduce risk. The appro-
priateness of the mean of the WTP estimates from
the 26 studies for valuing the mortality-related ben-
efits of reductions in pollutant concentrations there-
fore depends not only on the quality of the studies
(i.e., how well they measure what they are trying to
measure), but also on (1) the extent to which the
subjects in the studies are similar to the population
affected by changes in air pollution and (2) the ex-
tent to which the risks being valued are similar.
The substantia] majority of the 26 studies relied
upon are wage-risk (or labor market) studies. Com-
pared with the subjects in these wage-risk studies,
the population most affected by air pollution-related
mortality risk changes is likely to be, on average,
older and probably more risk averse. Some evidence
suggests that approximately 85 percent of those iden-
tified in short-term ("episodic") studies who die pre-
maturely from PM-related causes are over 65.7 The
average age of subjects in wage-risk studies, in con-
trast, would be well under 65, and probably closer
to 40 years of age.
The direction of bias resulting from the age dif-
ference is unclear. We could argue that, because an
older person has fewer expected years left to lose,
his or her WTP to reduce mortality risk would be
less than that of a younger person. This hypothesis
is supported by one empirical study, Jones-Lee et al.
(1985), which found WTP to avoid mortality risk at
age 65 to be about 90 percent of what it is at age 40.
On the other hand, there is reason to believe that
those over 65 are, in general, more risk averse than
the general population. This would imply that older
populations are likely to select occupations that are
relatively less risky than workers represented in
wage-risk studies or the general population. Al-
though the list of 26 studies used here excludes stud-
ies that consider only much-higher-than-average oc-
cupational risks, there is nevertheless likely to be
some selection bias in the remaining studies, because
these studies are likely to be based on samples of
See SchwarL2 and Dockery (1992), Oslro et al. (1995),
an d Chestnut (1995).
workers who are, 011 average, more risk-loving than
the general population. In contrast, older people as
a group exhibit more risk-averse behavior.
There is substantial evidence that the income
elasticity of WTP for health risk reductions is posi-
tive (although there is uncertainty about the exact
value of this elasticity). This implies that individu-
als with higher incomes and/or greater wealth should
be willing to pay more to reduce risk, all else equal,
than individuals with lower incomes or wealth. 'The
comparison between the income, both actual and
potential, or wealth of the workers 111 the wage-risk
studies versus that of the population of individuals
most likely to be affected by changes in pollution
concentrations, howrever, is unclear. One could ar-
gue that because the elderly are relatively wealthy,
the affected population is also wealthier, on aver-
age, than are the wage-risk study subjects, who tend
to be middle-aged (on average) blue-collar workers.
On the other hand, the workers m the wage-risk
studies will have potentially more years remaining
in which to acquire streams of income from future
earnings. On net, the potential income comparison
is unclear.
Although there may be several ways in which
job-related mortality risks differ from air pollution-
related mortality risks, the most important differ-
ence may be that job-related risks are incurred vol-
untarily, or generally assumed to be, whereas air
pollution-related risks are incurred involuntarily.
There is some evidence8 that people will pay more
to reduce involuntarily incurred risks than risks in-
curred voluntarily. If this is the case, WTP estimates
based on wage-risk studies may understate WTP to
reduce involuntarily incurred air pollution-related
mortality risks.
Another important difference related to the na-
ture of the risk may be that some workplace mortal-
ity risks tend to involve sudden, catastrophic events,
whereas air pollution-related risks tend to involve
longer periods of disease and suffering prior to death.
Some evidence suggests that WTP to avoid a risk of
a protracted death involving prolonged suffering and
loss of dignity and personal control is greater than
the WTP to avoid a risk (of identical magnitude) of
sudden death. To the extent that the mortality risks
8 See, for example, VioletLe and Chestnut, 1983.
77
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
addressed in this assessment are associated with longer
periods of illness or greater pain and suffering than
are the risks addressed in the valuation literature,
the WTP measurements employed in the present
analysis would reflect a downward bias.
Economic assessment of WTP for lagged mor-
tality effects also introduces uncertainty. For lack
of a more refined technique, our analysis relies on
the simplifying assumption that lagged mortality
risks can be valued at the tune of the occurrence of
death, rather than at the tune of exposure. In subse-
quent development of the annual and present value
estimates, we therefore discount the dollar benefits
estimate as if the full benefit accrues only in the year
of death. There are several reasons to believe that
this approach underestimates willingness to pay.
Most importantly, while death may occur after a lag
period, morbidity effects may appear at any time
prior to death, including immediately upon expo-
sure. It is not clear that other dose-response assess-
ments capture the full range of morbidity effects,
direct and indirect, that might be associated with a
latent fatal exposure. Other potentially important
factors include the use of a financial discount rate,
which may or may not accurately represent the rate
at which individuals might discount delayed health
benefits and the effect of knowledge of a fatal expo-
sure 011 valuation of a delayed effect, in other words
whether the valuation is affected by a prior diagno-
sis of a fatal condition.
We summarize the potential sources of bias in-
troduced by relying on wage-risk studies to derive
an estimate of the WTP to reduce air pollution-re-
lated mortality risk 111 Table 6-4; the overall effect of
these multiple biases is addressed in Table 6-5.
Among these potential biases, it is disparities in age
and income between the subjects of the wage-risk
studies and those affected by air pollution which have
thus far motivated specific suggestions for quantita-
tive adjustment;9 however, the appropriateness and
the proper magnitude of such potential adjustments
remain unclear given presently available information.
These uncertainties are particularly acute given the
possibility that age and income biases might offset
each other in the case of pollution-related mortality
risk aversion. Furthermore, the other potential bi-
9 Chestnut, 1995; lEc, 1992.
ases discussed above, and summarized in Table 6-4,
add additional uncertainty regarding the transferabil-
ity of WTP estimates from wage-risk studies to en-
vironmental policy and program assessments.
78
-------
Chapter 6: Economic Valuation of Human Health Effects
Table 6-5
Key Uncertainties Associated with Valuation of Health Benefits
Direction of Potential
Potential Source of Error Bias for Net Benefits
Likely Significance Relative to Key
Uncertainties on Net Benefits Estimate1
Benefits transfer for mortality
risk valuation, including
differences in age, income,
degree of risk aversion, the
nature of the risk, and
treatment of latency between
mortality risks presented by PM
and the risks evaluated in the
available economic studies.
Unable to determine based
on currently available
information
Potentially major. The mortality valuation
step is clearly a critical element in the net
benefits estimate, so any uncertainties can
have a large effect. As discussed in the text,
however, information on the combined effect
of these known biases is relatively sparse,
and it is therefore difficult to assess the
overall effect of multiple biases that work in
opposite directions.
Benefits transfer for chronic
bronchitis, including
adjustments made to better
match the severity of the risks
modeled in the available
economic studies.
Unable to determine based
on currently available
information
Probably minor. Benefits of avoided chronic
bronchitis account for about five percent of
total benefits, limiting the effect on net
benefits to a maximum of about seven
percent. Steps taken in the study to adjust
for severity using the best available empirical
information likely limit the effect to much less
than this maximum value.
Inability to value some
quantifiable morbidity
endpoints, such as impaired
lung function.
Underestimate
Probably minor. Reductions in lung function
are a well-established effect, based on
clinical evaluations of the impact of air
pollutants on human health, and the effect
would be pervasive, affecting virtually every
exposed individual. There is therefore a
potential for a major impact on benefits
estimates. The lack of a clear symptomatic
presentation of the effect, however, could
limit individual WTP to avoid lung function
decrements.
Note: 1 The classification of each potential source of error reflects the best judgement of the section 812 Project Team. The
Project Team assigns a classification of "potentially major" if a plausible alternative assumption or approach could influence
the overall monetary benefit estimate by approximately five percent or more; if an alternative assumption or approach is likely
to change the total benefit estimate by less than five percent, the Project Team assigns a classification of "probably minor."
79
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
f This page left blank intentionally.7
80
-------
Ecological and
Other Welfare
Effects
EPA's traditional focus in environmental ben-
efits assessment has been on quantifying beneficial
impacts of environmental regulation on human
health. As we have learned more about the effects
of anthropogenic stressors on ecological systems,
however, pursuit of environmental programs tar-
geted on reductions of damage to the environment
have become more common. The CAAA Title IV
provisions, collectively referred to as the Acid Ram
Program, are a good example. These provisions are
in place largely as the result of a major research ef-
fort to better understand and quantify the effects of
sulfur and nitrogen oxides on natural systems sus-
ceptible to acid rain. Although the benefits of this
program include improvements in human health, the
initial impetus was protection of ecological resources.
We have designed this first section 812 prospec-
tive analysis to be responsive to the increased focus
on the importance of ecological resources by devot-
ing a great deal of effort to characterizing and, where
possible, quantifying and monetizing the impacts of
air pollutants on natural systems. This increased fo-
cus is also partly a result of the outcome of EPA's
retrospective analysis, in which we identified an in-
creased understanding of and focus on ecological ef-
fects as one of the important research directions for
the first prospective and subsequent analyses. This
chapter presents the results of these efforts.
This chapter consists of four sections. First, we
provide an overview of our approach to estimating
the effects of air pollution on ecological systems.
Second, we provide a characterization of these ef-
fects in qualitative terms. The second section con-
cludes with a summary of the process for selecting
specific impacts which can be quantified and mon-
etized using currently available methods. Third, we
present the results of our quantitative and economic
analyses. Finally, we discuss major uncertainties of
the ecological and other welfare effects analyses.
Overview of Approach
Our analysis of ecological effects involves three
major steps:
• First, we identify and characterize ecologi-
cal effects from air pollution.
• Second, we develop and implement selection
criteria for more in-depth assessment of eco-
logical impacts.
• 'Third, we perform quantitative and qualita-
tive analyses to characterize a portion of the
benefits of the 1990 CAAA provisions.
The first step involves taking a broad view of
pollutants controlled under the CAAA and their
documented effects 011 ecological systems, both as
individual pollutants and, to the extent possible, as
one component 111 multiplc-strcssor effects on eco-
systems and their components. We organize our
analysis in terms of major pollutant classes and by
the level of biological organization at which impacts
are measured (e.g., regional ecosystem, local ecosys-
tem, community, population, individual, etc.).
After completing the first step on a broad level,
the second step involves narrowing the scope of sub-
sequent analyses. While it is desirable to focus ef-
fort on those impacts that are of greatest importance,
in practice the state of the science in ecological as-
sessment largely dictates the subsequent focus of the
analysis. There exist only a handful of comprehen-
sive ecological assessments from which to draw con-
clusions about those effects that are most important
either ecologically or in economic terms, and those
studies are potentially controversial in their meth-
ods and conclusions, in part because of the incom-
plete understanding of many of these effects. As a
result, the categories of effects ultimately chosen for
assessment here are necessarily limited by available
81
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
methods and data. As scientific understanding and
impact assessment methods grow more comprehen-
sive, however, we expect that the focus of subsequent
analyses will be on those effects whose avoidance
would have the greatest potential ecological and/or
economic value.
The third step involves implementing a wide
range of analyses to more exhaustively characterize
specific effects of air pollution on ecological systems.
We provide quantitative estimates of the benefits of
the 1990 CAAA for the following effects:
• eutrophication of estuaries associated with
airborne nitrogen deposition;
• acidification of freshwater bodies associated
with airborne nitrogen and sulfur deposition;
and
• reduced forest growth associated with ozone
exposure.
In addition, in this chapter wre present the meth-
ods and results for quantitative analysis of other
welfare effects, including reduced agricultural yields
associated with ozone exposure, the impact of am-
bient particulate matter on visibility, the effects of
ozone on farm worker productivity, and the effects
of stratospheric ozone on crop and fisheries yields.
These effects have been identified as important cat-
egories of benefits in many previous analyses, includ-
ing the section 812 retrospective analysis. As a re-
sult, these effects were not considered in the same
three step process used for other sendee flows.
We attempted to conduct quantitative analyses
of two other benefits categories: the accumulation
of toxics in freshwater fisheries associated with air-
borne toxics deposition; and aesthetic degradation
of forests associated with ozone and airborne toxics
exposure. However, we found that, while some
quantitative methods exist to evaluate these benefits,
key links are missing in the analytic process. This in
turn prevents development of defensible benefits es-
timates which can be reasonably associated with the
air quality and air pollutant deposition patterns de-
veloped from our Post-CAAA and Pre-CAAA sce-
narios. See Appendix E for more detailed discus-
sion of these service flows. In addition, in assessing
nitrogen deposition impacts to estuarine systems, we
relied on a displaced cost approach with results that
we chose to omit from the primary benefits estimate
because of uncertainties in the methodology. These
results are nonetheless reported in this chapter, but
are used for the purposes of sensitivity testing only.
Because the breadth and complexity of air pol-
lutant-ecosystem interactions do not allow for com-
prehensive quantitative analysis of all the ecological
benefits of the CAAA, we stress the importance of
continued consideration of those impacts not val-
ued in this report in policy decision-making and in
further technical research. Judging from the geo-
graphic breadth and magnitude of the relatively
modest subset of impacts that we find sufficiently
well-understood to quantify and monetize, it is ap-
parent that the economic benefits of the CAAA's
reduction of air pollution impacts on ecosystems are
substantial.
Characterization of Impacts
of Air Pollution on Ecological
Systems
The purpose of this section is to provide an over-
view of potential interactions between air pollutants
and the natural environment. We identify major
single pollutant-environment interactions, as well as
the synergistic impacts of ecosystem exposure to
multiple air pollutants. Although a wide variety of
complex air pollution-environment interactions are
described or hypothesized in the literature, for the
purposes of this analysis we focus on major aspects
of ecosystem-pollutant interactions. We do this by
limiting our review according to the following crite-
ria:
• Pollutants regulated by the CAAA.
• Known interactions between pollutants and
natural systems as documented in
peer-reviewed literature.
• Pollutants present in the atmosphere in suf-
ficient amounts after 1990 to cause signifi-
cant damages to natural systems.
Our understanding of air pollution effects on
ecosystems has progressed considerably during re-
cent decades. Previously, air pollution was regarded
primarily as a local phenomenon and concern was
associated with the vicinity of industrial facilities,
82
-------
Chapter 7: Ecological arid Other Welfare Effects
power plants or urban areas. The pollutants of con-
cern were gaseous (e.g., sulfur dioxide and ozone) or
heavy metals (e.g., lead) and the observed effects were
visible stress- specific symptoms of injury (e.g., fo-
liar chlorosis). The most typical approach to docu-
menting the effects of specific pollutants was a
dose-response experiment, where the objective was
to develop a regression equation describing the rela-
tionship between exposure and some easily measured
effect (e.g., growth, yield or mortality). As analytic
methods unproved and ecology progressed, a broader
range of effects of air pollutants was identified and
understanding of the mechanisms of effect improved.
Observations made on various temporal scales (e.g.,
long-term studies) and spatial scales (e.g., watershed
studies) led to the recognition that air pollution can
affect all organizational levels of biological systems.
Our current understanding of ecosystem impacts
can be organized by the pollutants of concern and
by the level of biological organization at which im-
pacts are directly measured. We attempt to address
both dimensions of categorization in this overview.
In Table 7-1 we summarize the major pollutants of
concern, and the documented acute and long-term
ecological impacts associated with them.
The summary 111 Table 7-1 is a highly condensed
version of the results of our characterization of eco-
logical impacts. In addition to the pollutant-specific
effects outlined in the table, it is important to iden-
tify the level of biological organization and types of
ecosystems that are susceptible to these types of ef-
fects. Tables 7-2 through 7-4 provide more detail on
pollutant-specific impacts at a range of levels of bio-
logical organization. It is important to note that the
interactions listed are intended to illustrate the range
of possible adverse effects. For a more complete re-
view of air-pollutant-induced effects on ecosystems,
see Appendix E.
Effects of Mercury and Ozone
Table 7-2 summarizes the effects of mercury and
ozone on ecological systems. To illustrate the na-
ture of our review of effects, consider the second
row in Table 7-2. This row summarizes the effects
of the air pollutants mercury and ozone at the "indi-
vidual" level of biological organization. As indicated
in the table, in a general sense air pollutants can in-
duce a direct physiological response in individuals
(analogous to that experienced by humans exposed
to pollutants), or an indirect effect either through
impacts on the individual's surroundings or by weak-
ening the individual and making it more susceptible
to other stressors. "Mercury has several direct effects
to fauna, including effects to the central nervous
system and the liver, while the documented direct
effects of ozone tend to be to a variety of plant func-
tions. Indirect effects of mercury are not well un-
derstood, but the indirect effects of ozone may serve
to compound the direct effects to plants by also
making the plants more susceptible to drought or
heat stress, for example. This type of cataloging of
Table 7-1
Classes of Pollutants and Ecological Effects
Pollutant
Class
Major Pollutants and
Precursor Emissions
Acute Effects
Long-term Effects
Acidic
Deposition
Sulfuric acid, nitric acid
Precursor emissions: Sulfur
dioxide, nitrogen oxides
Direct toxic effects to
plant leaves and
aquatic organisms.
Progressive deterioration of soil
quality. Chronic acidification of
surface waters.
Nitrogen
Deposition
Nitrogen compounds (e.g.,
nitrogen oxides)
Saturation of terrestrial ecosystems
with nitrogen. Progressive nitrogen
enrichment of coastal estuaries.
Hazardous Air
Pollutants
(HAPs)
Mercury, dioxins
Direct toxic effects to
animals.
Conservation of mercury and dioxins
in biogeochemical cycles and
accumulation in the food chain.
Ozone
Tropospheric ozone
Precursor emissions: Nitrogen
Oxides and Volatile Organic
Compounds (VOCs)
Direct toxic effects to
plant leaves.
Alterations of ecosystem wide
patterns of energy flow and nutrient
cycling.
83
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 7-2
Interactions of Mercury and Ozone with Natural Systems At Various Levels of Organization
Examples of Interactions
Spatial Scale
Type of Interaction
Mercury in
streams and lakes
Ozone
Molecular and
cellular
Chemical and
biochemical processes
Mercury enters the body of
vertebrates and binds to
sulfhydril groups (i.e.
proteins).
Oxidation of enzymes of plants.
Disruption of the membrane
potential.
Individual Direct physiological Neurological effects in Direct injuries include visible
response. vertebrates. Behavioral foliar damage, premature needle
abnormalities. Damages to senescence, reduced
the liver. photosynthesis, altered carbon
allocation, and reduction of
growth rates and reproductive
success.
Indirect effects: Few interactions known. Increased sensitivity to biotic
Response to altered Damages through and abiotic stress factors like
environmental factors or increased sensitivity to pathogens and frost. Disruption
alterations of the other environmental stress of plant-symbiont relationship
individual's ability to cope factors could occur, for (mychorrhiza), and symbionts.
with other kinds of stress, example, through
impairment of immune
response.
Population
Change of population
characteristics like
productivity or mortality
rates.
Reduced reproductive
success offish and bird
species. Increased
mortality rates, especially in
earlier life stages.
Reduced biological productivity.
Selection for less sensitive
individuals. Possibly
microevolution for ozone
resistance.
Community
Changes of community
structure and competitive
patterns
Loss of species diversity of
benthic invertebrates.
Alteration of competitive
patterns. Selective advantage
for ozone-resistant species.
Loss of ozone sensitive species
and individuals. Reduction in
productivity.
Local Ecosystem
(e.g..landscape
element)
Changes in nutrient
cycle, hydrological cycle,
and energy flow of lakes,
wetlands, forests,
grasslands, etc.
Not well understood.
Alterations of ecosystem-wide
patterns of energy flow and
nutrient cycling.
Regional Ecosystem
(e.g., watershed)
Biogeochemical cycles
within a watershed.
Region-wide alterations
of biodiversity.
Not well understood.
Region-wide loss of sensitive
species.
effects, while limited in its direct usefulness in a cost-
benefit framework, nonetheless does convey the wide
range of documented effects of air pollutants on eco-
logical resources. These tables and the accompany-
ing text, found in Appendix E, also provide a frame-
work for determining the extent to which impor-
tant factors may not be well characterized by quan-
titative analysis, setting the stage for prioritization
of research needs.
Effects of Nitrogen Deposition
Table 7-3 provides a summary of the effects of
nitrogen deposition on natural systems. These im-
pacts are manifest in both terrestial and coastal es-
tuarine systems. In both types of systems, nitrogen
can be a growth-enhancing nutrient. As shown in
the rows characterizing individual and population
level impacts, the effects on many varieties of plants
are beneficial. This growth can have other harmful
effects, however. For example, excessive growth of
84
-------
Chapter 7: Ecological arid Other Welfare Effects
Table 7-3
Interactions Between Nitrogen Deposition and Natural Systems
At Various Levels of Organization
Examples of Interactions
Spatial Scale
Type of Interaction
Eutrophication and
Nitrogen Saturation of
Terrestrial Landscapes
Eutrophication of Coastal
Estuaries
Molecular and
cellular
Chemical and
biochemical processes
Assimilation of nitrogen by
plants and microorganisms
Assimilation of nitrogen by
plants and microorganisms.
Individual
Direct physiological
response.
Increases in leaf- size of
terrestrial plants.
Increase in growth of marine
plants.
Indirect effects:
Response to altered
environmental factors or
alterations of the
individual's ability to
cope with other kinds of
stress.
Decreased resistance to
biotic and abiotic stress
factors like pathogens and
frost. Disruption of plant-
symbiont relationships with
mycorrhiza fungi.
Injuries to marine fauna through
oxygen depletion of the
environment. Loss of physical
habitat due to loss of sea-grass
beds. Injury through increased
shading. Toxic blooms of
plankton.
Population
Change of population
characteristics like
productivity or mortality
rates.
Increase in biological
productivity and growth
rates of some species.
Increase in biological
productivity. Increase of growth
rates (esp. of algae and marine
plants).
Community
Changes of community
structure and
competitive patterns
Alteration of competitive
patterns. Selective
advantage for fast growing
species and individuals
that efficiently use
additional nitrogen. Loss
of species adapted to
nitrogen-poor
environments.
Excessive algal growth.
Changes in species
composition. Decrease in sea-
grass beds.
Local Ecosystem
(e.g., landscape
element)
Changes in nutrient
cycle, hydrological cycle,
and energy flow of lakes,
wetlands, forests,
grasslands, etc.
Magnification of the
biogeochemical nitrogen
cycle. Progressive
saturation of
microorganisms, soils, and
plants with nitrogen.
Magnification of the nitrogen
cycle. Depletion of oxygen,
increased shading through
algal growth.
Regional Ecosystem
(e.g., watershed)
Biogeochemical cycles
within a watershed.
Region-wide alterations
of biodiversity.
Leaching of nitrogen from
terrestrial sites to streams
and lakes. Acidification of
aquatic bodies.
Eutrophication of estuaries.
Additional input of nitrogen
from nitrogen-saturated
terrestrial sites within the
watershed.
marine organisms can lead to eutrophy, a state where
the enhanced surface growth of plants shields bot-
tom growing plants from sunlight, causing them to
die and, in extreme cases, lead to low dissolved oxy-
gen, or anoxic, conditions that impair a wide range
of species and ecological functions. These effects
are described in the table in the rowrs characterizing
effects at the community and ecosystem levels. For
this reason, isolated analysis of the effects of nitro-
gen on individuals or populations may provide mis-
leading results; by the same token, analyses which
ignore the beneficial effects of nitrogen in certain
types of systems may lead to similarly misleading-
results. These complex linkages across biological
levels of organization suggest that, when feasible, a
systems level approach to ecological assessments is
preferable to isolated analyses of effects at lower or-
ders of organization.
Effects of Acid Deposition
Table 7-4 provides a summary of the effects of
acid deposition on forest and freshwater systems.
The direct effects of acid deposition in lakes and
streams include effects on fish species, as charaterized
in the row describing individual-level effects. These
85
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 7-4
Interactions Between Acid Deposition and Natural Systems At Various Levels of Organization
Examples of Interactions
Acidification of Streams
Spatial Scale
Type of Interaction
Acidification of Forests
and Lakes
Molecular and
cellular
Chemical and
biochemical processes
Damages to epidermal
layers and cells of plants
through deposition of
acids.
Impairment of ion interactions
of fish at the cellular level.
Individual
Direct physiological
response
Increased loss of nutrients
via foliar leaching.
Decreases in pH and increase
in aluminum ions causes
pathological changes in gill
structure of fish.
Indirect effects:
Response to altered
environmental factors or
alterations of the
individual's ability to
cope with other kinds of
stress.
Cation depletion in the soil
causes nutrient
deficiencies in plants.
Concentrations of
aluminum ions in soils can
reach phytotoxic levels.
Increased sensitivity to
other stress factors like
pathogens and frost.
Aluminum ions in the water
column can be toxic to many
aquatic organisms through
impairment of gill regulation.
Acidification can indirectly
affect submerged plant species,
because it reduces the
availability of dissolved carbon
dioxide (CO2).
Population
Change of population
characteristics like
productivity or mortality
rates.
Decrease of biological
productivity of sensitive
organisms. Selection for
less sensitive individuals.
Microevolution of
resistance.
Decrease of biological
productivity of sensitive
organisms. Selection for less
sensitive individuals.
Microevolution of resistance.
Community
Changes of community
structure and
competitive patterns
Alteration of competitive
patterns. Selective
advantage for acid-
resistant species. Loss of
acid sensitive species and
individuals. Decrease in
productivity. Decrease of
species richness and
diversity.
Alteration of competitive
patterns. Selective advantage
for acid-resistant species. Loss
of acid sensitive species and
individuals. Decrease in
productivity. Decrease of
species richness and diversity.
Local Ecosystem
(e.g., landscape
element)
Changes in nutrient
cycle, hydrological cycle,
and energy flow of lakes,
wetlands, forests,
grasslands, etc.
Progressive depletion of
nutrient cations in the soil.
Increase in the
concentration of mobile
aluminum ions in the soil.
Measurable declines of
decomposition of some forms
of organic matter, potentially
resulting in decreased rates of
nutrient cycling.
Regional Ecosystem
(e.g., watershed)
Biogeochemical cycles
within a watershed.
Region-wide alterations
of biodiversity.
Leaching of sulfate, nitrate,
aluminum, and calcium to
streams and lakes.
Acidification of aquatic
bodies.
Additional acidification of
aquatic systems through
processes in terrestrial sites
within the watershed.
effects are not as straightforward as they might ap-
pear, however, because it is not only the acidity (pH)
of the water itself that causes the effect but the in-
creased leaching of metals, particularly aluminum,
which takes place in acidic (low pH) environments
that contributes substantially to the effects on fish.
These effects will vary widely from place to place
according to the mineral content of the soil near the
lake and the lakebed sediment, as well as the natural
resistance of the lake in absorbing acid deposition
(i.e., its buffering capacity). Other important effects
characterized in the table include the ability of acid
deposition to deplete cation concentrations in ter-
restrial ecosystems; increase the concentration of
aluminum in soils; and leach nutrients, sulfates, and
metals to surrounding streams and lakes. Effects of
note at the individual level include foliar damage to
trees.
86
-------
Chapter 7: Ecological arid Other Welfare Effects
A few general points emerge from our review of
ecological effects:
• Air pollutants have indirect effects that are
at least as important as direct toxic effects
on living organisms. Indirect effects include
those in which the pollutant alters the physi-
cal or chemical environment (e.g., soil prop-
erties), the plant's ability to compete for lim-
ited resources (e.g., water, light), or the
plant's ability to withstand pests or patho-
gens. Examples are excessive availability of
nitrogen, depletion of nutrient cations in the
soil by acid deposition, mobilization of toxic
elements such as aluminum, and changes in
winter hardiness. As is true for other com-
plex interactions, indirect effects are more
difficult to observe than direct toxic relation-
ships between air pollutants and biota, and
there may be a variety of interactions that
have not yet been detected.
• There is a group of pollutants that tend to
be conserved in the landscape after they have
been deposited to ecosystems. These con-
served pollutants are transformed through
biotic and abiotic processes within ecosys-
tems, and accumulate in biogeochemical
cycles. These pollutants include, but are not
limited to, hydrogen ions (H+), sulfur (S)
and nitrogen (N) containing substances, and
mercury (I Ig). Chronic deposition of these
pollutants can result in progressive increases
in concentrations and cause injuries due to
cumulative effects. Indirect, cumulative
damages caused by chronic exposure (i.e.,
long-term, moderate concentrations) to these
pollutants may increase in magnitude over
time frames of decades or centuries with very
subtle annual increments of change. Ex-
amples are N-saturation of terrestrial ecosys-
tems, cation depletion of terrestrial ecosys-
tems, acidification of streams and lakes, and
accumulation of mercury in aquatic food
webs.
• Damages to ecosystems are most likely
caused by a combination of environmental
stress factors. These include anthropogenic
factors such as air pollution and other envi-
ronmental stress factors such as low tempera-
ture, excess or limited water, and limited
availability of nutrients. The specific com-
binations of factors differ among regions and
ecosystems where declines have been ob-
served. Accurately predicting the impacts
of multiple stress factors is an extremely dif-
ficult task, but this is an area of very active
research among ecologists.
• Pollutant-environment interactions are com-
plicated by the fact that biotic and abiotic
factors in ecosystems change dramatically
over time. Besides oscillations on a daily
basis, and changes in a seasonal rhythm, there
are long-range successional developments
over time periods of years, decades, or even
centuries. These temporal variations occur
111 polluted and pristine ecosystems, and 110
single point in time or space can be defined
as representative of the entire system.
Selection of Service Flows
Potentially Amenable to
Economic Analysis
Based on this broad overview of effects, we iden-
tify a set of pollutant-environment interactions
which are amenable to more detailed quantification
and monetization. We evaluate the long list of ef-
fects and seek categories where a defensible link ex-
ists between changes in air pollution emissions and
the quality or quantity of the ecological service flow,
and where economic models are available to mon-
etize these changes. The use of these criteria greatly
constrains the range of impacts that can be treated
quantitatively. While the previous section identi-
fies many pollutant-ecosystem interactions, only a
handful are understood and have been modeled to
an extent sufficient to reliably quantify their impact.
The theoretical basis of economic benefits as-
sessment is that ecosystems provide services to man-
kind, and that those services have economic value.
The application of this theory requires the isolation
of service flows that have market values or are oth-
erwise amenable to available methods for determin-
ing value in the absence of formal markets. Avail-
able methods do not exist to comprehensively value
all service flows for any particular ecosystem or ag-
gregation of ecosystems. Generally, we are limited
87
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
to those service flows that are either sources of ma-
terial inputs or associated with natural amenities that
involve active recreation. Impacts to these service
flows that can be valued tend to manifest themselves
immediately and can be readily measured and assessed
in terms of the established cause and effect relation-
ships.
Based on the constraints of economic valuation
methods and data, we select from the host of ecosys-
tem impacts identified in the previous section a set
of service flows as candidate endpouits for analysis.
The list of service flows establishes the potential
scope of economic analysis for ecological effects fea-
sible in the context of the present study. Table 7-5
presents the service flow impacts that we quantita-
tively estimate in this analysis plus those effects that
currently cannot be quantified for each of the four
ecological pollutant categories discussed in Table 7-1.
From the list of effects in Table 7-5, we further
limited the quantitative and qualitative analyses con-
ducted to reflect the available model coverage. The
results are summarized in Table 7-6. The relatively
short list of effects in Tables 7-5 and 7-6 demonstrates
that, of the great number of known impacts of air
pollution, only a subset can be assessed quantita-
tively. Note that for one category of effects, nitro-
gen deposition impacts to estuarine systems, we re-
lied on a displaced cost approach (described below)
Table 7-5
Ecological Effects of Air Pollutants
Pollutant
Quantified Effects
Unquantified Effects
Acidic Deposition
Impacts to recreational
freshwater fishing
Impacts to commercial forests
(e.g., timber, non-timber forest products)
Impacts to commercial freshwater fishing
Watershed damages (water filtration
flood control)
Impacts to recreation in terrestrial
ecosystems (e.g. forest aesthetics,
nature study)
Reduced existence value and option
values for nonacidified ecosystems (e.g.
biodiversity values)
Nitrogen
Deposition
Additional costs of alternative or
displaced nitrogen input controls
for eastern estuaries
Impacts to commercial fishing,
agriculture, and forests
Watershed damages (water filteration,
flood control)
Impacts to recreation in estuarine
ecosystems (e.g. Recreational fishing,
aesthetics, nature study)
Reduced existence value and option
values for non-eutrophied ecosystems
(e.g. biodiversity values)
Tropospheric
Ozone Exposure
Reduced commercial timber
yields and reduced tons of carbon
sequestered
Impacts to recreation in terrestrial
ecosystems (e.g. forest aesthetics,
nature study)
Reduced existence value and option
values for ozone-impacted ecosystems
Hazardous Air
Pollutant (HAPS)
Deposition
No service flows quantified
Impacts to commercial and recreational
fishing from toxification of fisheries
Reduced existence value and option
values for non-toxified ecosystems (e.g.
biodiversity values)
88
-------
Chapter 7: Ecological arid Other Welfare Effects
Table 7-6
Summary of Endpoints Selected for Quantitative Analysis
Endpoint Analysis Geographic Scope
Lake acidification impacts on Quantification of improved fishing Case study of New York State
recreational fishing with monetization of recreational
value
Estuarine eutrophication Quantification of improved fishing Case studies of Chesapeake Bay,
impacts on recreational and with monetization of displaced Long Island Sound, and Tampa
commercial fishing costs of alternative Bay (with illustrative extensions to
eutrophication control methods East Coast estuaries)
Ozone impacts on commercial Quantification of improved timber National assessment
timber sales growth with monetization of
commercial timber revenues
Ozone impacts on carbon Quantification of improved National assessment
sequestration in commercial carbon sequestration
timber
that we chose to omit from the primary benefits es-
timate because of uncertainties in the methodology.
These results are nonetheless reported in this chap-
ter, but are used for the purposes of sensitivity test-
ing only. In the next section we discuss the meth-
ods, results, and caveats of the analyses of these se-
lected endpoints.
Results
In this section we summarize the methods used
for, and results obtained from, our quantitative and
economic analyses of selected service flows. We first
review the methods for each analysis, and then
present a summary of key quantitative results. For
a more detailed description of methods and results,
see Appendix E.
Estuarine Eutrophication Associated
with Airborne Nitrogen Deposition
Atmospherically derived nitrogen makes up a
sizable fraction of total nitrogen inputs in estuaries
in the eastern United States. Airborne nitrogen depo-
sition accounts for a significant fraction of the total
nitrogen loads to coastal estuaries, particularly on
the East and Gulf coasts. For example, the most
recent estimates for the Chesapeake Bay indicate air-
borne deposition accounts for over 40 percent of the
total nitrogen load to the estuary; in Galveston Bay,
the share is almost 50 percent. When nitrogen en-
ters estuaries it can cause eutrophication, or an in-
creased nutrient load that, in excess, changes the
ecosystem's structure and function and affects eco-
logical service flows. Many state governments and
multi-state regional authorities have expressed in-
creasing concern about the control of airborne ni-
trogen deposition as an important source of nitro-
gen loading.
Our analysis of the effects of nitrogen deposi-
tion followed two tracks. We first attempted to quan-
tify the service flows affected by and the damages
associated with eutrophication, and derive dose-re-
sponse relationships and valuation strategies for each
of the key service flow categories (for example, rec-
reational fishing). The derivation of dose-response
relationships between atmospheric nitrogen loading
and ecological effects, however, is complicated by
the dynamic nature of ecological systems. In addi-
tion to being characterized by non-linear, "thresh-
old" type responses, estuarine ecosystems are simul-
taneously influenced by a variety of stressors (both
anthropogenic and natural). This makes it difficult
to quantify the nature and magnitude of ecological
changes expected to result from a change in a single
stressor such as nutrient loading. Further, if the state
of the ecosystem has changed (as from oligotrophic1
to eutrophic) the removal of the initial stressor does
not necessarily mean a rapid return to the prior state.
This complicates the quantitative benefits assessment
of controlling nitrogen deposition through the
CAAA.
1 Oligotrophy refers to a state of relatively low nutrient
enrichment and low productivity of aquatic ecosystems. In con-
trast, eutrophy refers to a state of relatively high nutrient load-
ing and higher productivity, sometimes leading to
overenrichment and reduction in ecological service flows due
l.o water quality decline.
89
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Out: second track relies on a displaced cost ap-
proach to benefit estimation. To reduce excess nu-
trient loads (including nitrogen) to local estuaries,
many coastal communities are pursuing a range of
abatement options. These options include waste-
water and stormwater discharge point source con-
trols as well as urban non-point and agricultural
non-point source controls for runoff from the land.
If atmospheric nitrogen depostion is reduced, the
need for these types of expenditure to control other
sources of nitrogen loading is also lessened, and the
displaced control expenditures represent a benefit
to society.
Displaced or avoided cost approaches are not
always justified. In order to establish that the costs
would truly be avoided, and to ensure that the avoid-
ance of that cost represents a real benefit to society,
we need to show that realistic and enforceable nitro-
gen reduction goals exist for each evaluated estuary.
Without specific targets or reduction goals, it is not
possible to suggest that there are specific control
expenditures to be displaced. Therefore, we choose
case study estuaries that most closely meet this crite-
rion: Chesapeake Bay, Long Island Sound, and
Tampa Bay. These areas have established nitrogen
reduction programs that rely primarily on reductions
of effluent from point sources as well as reductions
in non-point source discharges. Information on the
reduction goal and potential abatement options for
meeting those goals allows us to estimate the por-
tion of the goal that can be met by the CAAA, as
well as the associated cost savings.2
The benefits valuation derived using the dis-
placed-costs approach should be interpreted cau-
tiously for two reasons. First, it is an estimation of
capital costs that serve more purposes than mitigat-
ing nitrogen inputs into the estuaries of concern.
Water treatment works are intended to provide waste
water treatment for a variety of pollutants and may
be required even in the absence of deposition of air-
borne nitrogen. Second, the nitrogen loading tar-
gets for the estuaries are not concrete, strictly en-
forced limits, based on certain knowledge of the ca-
pacity of the estuaries to accept nitrogen inputs.
z With increasing populations, controls of alternative
sources (e.g., automobile and utility emissions) mav be needed
simply to meet the original target or goal. If the CAA amend-
ments are necessary just to achieve the target reductions, then
we are actually measuring alternative costs and not avoided costs.
Instead, the targets may change over time as knowl-
edge of the effects of nitrogen to these estuaries
change. For these reasons, and because of the un-
certainty about the ability of local and regional enti-
ties to enforce the nitrogen reduction targets, we
calculate estimates of displaced costs for these three
estuaries but do not include them in the primary
benefits estimate for the CAAA.
Our approach involves three basic steps. First,
we estimate the total loading of nitrogen to each of
the three target estuaries. We use nitrogen deposi-
tion estimates from the RADM model, generated
for each 80 km x 80 km grid cell 111 the eastern U.S.
We then estimate the ultimate fate of deposited ni-
trogen through a GIS-based model of nitrogen "pass-
through." The pass-through is the share of nitrogen
deposited that is ultimately transported to the cstua-
rine waters rather than retained by the land. Pass-
through factors vary by land use, from about 20 per-
cent (for forests and wetlands) to 100 percent (for
open wrater). We estimate the nitrogen loading for
each scenario, and the within-year, cross-scenario
differences are the reduced nitrogen deposition at-
tributed to the CAAA. We present these estimates
in the second column of Table 7-7.
Second, we estimate the marginal costs of alter-
native abatement actions which could be imple-
mented in the three case study estuaries. We develop
our displaced-cost estimate by assuming that deci-
sion makers will choose to forego the most costly
nitrogen abatement projects first. That is, wre as-
sume that reduced deposition and the resulting load-
ings reduction will eliminate the need for additional
point or non-point source controls at the high end
of the marginal cost curve. We summarize those re-
sults in the third and fourth columns of Table 7-7.
Third, we multiply the reduced nitrogen load-
ing attributed to the CAAA by the marginal cost
estimate to arrive at a range of estimates of displaced
cost, ensuring that the reduction in airborne nitro-
gen is less than or equal to the potential tonnage
reduction achieved by the displaced, high marginal
cost abatement strategies. We present our results in
the last column of Table 7-7. Our estimates suggest
that the displaced cost is substantial for the large
Chesapeake Bay and Long Island Sound estuaries,
and more modest for Tampa Bay. The Chesapeake
90
-------
Chapter 7: Ecological arid Other Welfare Effects
Table 7-7
Estimated Displaced Costs for Three Estuaries
Estuary
Reduced N Deposition in
2010(millions of pounds)
Low Marginal
Cost($/lb/yr.)
High Marginal Cost
($/lb/yr.)
Estimated Annual
Displaced Costs in
2010 ($millions)
Long Island
12.8
$2
$8
$26-$ 100
Sound
Central Estimate: $63
Chesapeake
58.1
$6
$22
$350-$1,300
Bay
Central Estimate: $820
Tampa Bay
1.8
$6
$38
$11 -$68
Central Estimate: $40
Bay and Long Island Sound watersheds together ac-
count for about 40 percent of the total estuarine
watershed area on the East (Atlantic) coast that is
sensitive to nitrogen deposition, while Tampa Bay
accounts for about two percent of the sensitive wa-
tershed area for the Gulf coast.
Acidification of Freshwater Fisheries
During the 1970s and 1980s, "acid rain" came to
be known to the public as a phenomenon that in-
jures trees, forests, and water bodies throughout
Europe and in some areas of the United States and
Canada. One of the goals of the CAAA was to ad-
dress the problem of acidification of terrestrial and
aquatic ecosystems caused by acidic deposition. To
assess this effect we conducted a quantitative analy-
sis of benefits derived from a reduction in acidifica-
tion of aquatic bodies as they relate to recreational
fishing in die Adirondacks region of New York State.
As discussed earlier in this chapter, acidification
of water bodies is a complex process. Airborne ac-
ids, in the form of sulfur and nitrogen compounds,
are deposited to water bodies and surrounding drain-
age areas, with the potential to change the pi I of the
water body. Many water bodies are relatively resis-
tant and can absorb a great deal of deposition before
pH changes substantially. This buffering capacity is
referred to as acid neutralizing capacity (ANC).
Once pH begins to be affected, a scries of interac-
tions occur, the most important of which is the leach-
ing of aluminum from sediments and surrounding-
soil and the suspension of this metal in the water
column. While acidic pH presents a direct stress to
aquatic organisms, it is the combined effect of pH
and aluminum exposure that presents the greatest
risk. Lakes in the Adirondacks region of New York
State are particularly susceptible to acidification be-
cause they have low baseline ANC, relative to wa-
ter bodies in other areas of the country.
Because of these complex physical and chemical
interactions, acidification stress is typically evaluated
by application of a model that simulates these pro-
cesses, and requires data 011 individual lake chemis-
try and sediment composition. We relied on the
scenario-specific atmospheric deposition data (both
sulfur and nitrogen) from the RADM air quality
model (see Chapter 4 and Appendix C) as an input
to EPA's Model of Acidification of Groundwater in
Catchments (MAGIC). MAGIC generates several
measures of the impact of sulfur and nitrogen depo-
sition on lake acidity, including ANC and pIT.3 We
used the pi I outputs to classify lakes where recre-
ational fishing might be impaired, and those estimates
were used 111 an economic model of recreational fish-
ing behavior in New York State to develop economic
estimates of the impact of acid ram on recreational
fishing resources in that state.
We summarize the results of our analysis of eco-
nomic benefits of avoided Adirondacks acidification
attributable to the CAAA in 2010 in Table 7-8. The
range of annual benefits from the CAAA are $12
million to |49 million using the low-end assump-
tions on the threshold of effect (pH 5.0), and |82 to
$88 million for the high-end assumptions on the ef-
fects threshold (pH 5.4). Higher pH (or, less acidic)
threshold assumptions lead to greater damage esti-
mates, because more lakes cross the less acidic thresh-
old. We calculate our benefits results by comparing
2 For more ini.onnalio.ti on EPA's MAGIC model see Cosby
el al. (1985a), as referenced in Appendix E.
91
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 7-8
Annual Economic Impact of Acidification in 2010 (Millions of 1990 Dollars)
Range of Economic Impact
Year Scenario Low Estimate Central Estimate High Estimate
1990 Base Year $61 $320
2010 Post-CAAA $24 to $61 $261 to $281
Pre-CAAA $73 $349 to $363
Range of CAAA Benefits in 2010 $12 to $49 $50 $82 to $88
the suite of Post-CAAA 2010 estimates of total dam-
ages to the corresponding suite of estimates using
Pre-CAAA deposition. The impact of nitrogen satu-
ration in the surrounding terrestrial environment is
reflected in the range of estimates presented in Table
7-8. If surrounding soils are saturated, less deposited
nitrogen will remain oil the land and more nitrogen
will enter the water bodies, increasing the stress on
the aquatic ecosystem. This phenomenon is reflected
by the higher damage estimates for saturated versus
non-saturated scenarios, other factors equal, although
our model shows no effect of saturation in the 2010
Pre-CAAA low estimate. The results we present
are in line with those generated from previous analy-
ses that find annual benefits to the Adirondacks of
halving utility emissions to be approximately in the
millions to tens of millions of dollars.4
Reduced Timber Growth Associated
with Ozone Exposure
The third category of effects we quantify is im-
proved commercial timber growth through the re-
duction of tropospheric ozone concentrations attrib-
utable to the CAAA. There is substantial scientific
evidence to suggest that elevated ozone concentra-
tions in the troposphere disrupt ecosystems by dam-
aging and slowing the growth of vegetation. In this
analysis, we examine one aspect of these impacts,
reduced commercial timber growth. Much of the lit-
erature on the effects of ozone on tree growth is
based on laboratory exposures of seedlings or leaf-
scale experiments in the field. Estimates from those
studies have been used in previous analyses, making-
use of professional judgment as an interpretive tool,
but always with strong caveats about the potential
applicability of the seedling and leaf-scale results to
4 For alternative estimates see, lor example, Engliii el al.
(1991), Mullen and Menz (1985), and Morey and Shaw (1990), as
referenced ill Appendix E.
tree growth and, in particular, the rate of accumula-
tion of wood mass that is important for commercial
timber production.5 In an attempt to overcome these
issues, we sought to find a concentration-response
relationship that would provide a more defensible
and broadly applicable basis for estimating effects
on tree growth.
Our analysis makes use of the Net Photosyn-
thesis and Evapo-Transpiration model II (PnET II),
a biological model of timber stand productivity to
estimate the impacts of ozone on timber yields. The
PnET II model was designed to estimate the com-
bined effects of several stressors on the rate of net
primary productivity (Nl'P), a measure of the rate
of photosynthesis. Nl'P in a tree does not necessar-
ily all go towards accumulation of wood mass; some
may be allocated to root growth, leaf growth, or
other tree functions. The PnET II model provides a
means to measure both NPP and wood mass growth,
as well as the effect on trees of several stressors com-
bined. One important stressor to acknowledge in
an analysis of the effects of ozone on trees is drought
stress. Ozone has the effect of reducing water loss in
trees by stimulating the closing of stomata through
which water is transpired. As a result, in drought
stress conditions, ozone can have beneficial effects
on tree growth. The PnET II model reflects the
impact of this factor in combination with other di-
rect effects of ozone on tree function.
We used the PnET II model to provide estimates
of timber stand responses to ozone exposure under
each of the scenarios examined in this analysis. We
aggregated tree growth results by region, with sepa-
rate estimates for hardwoods and softwoods, and used
them as inputs to the Timber Assessment Market
See de Steiger el al. (1990) for an example ol the genera-
tion of tree growth dose-response estimates based on professional
judgement.
92
-------
Chapter 7: Ecological and Other Welfare Effects
o
Q
Model (TAMM), an economic model of the for-
est sector maintained by the United States For-
est Service. There are three stages to the eco-
nomic estimation. First, forest growth rate in-
formation generated by PnET II is provided to
the Aggregate Timber Land Assessment System
(ATLAS), the forest inventory tracking com-
ponent of TAMM. Growth rate information is
provided for each of the forest production re-
gions defined by TAMM.6 Second, ATLAS
generates an estimate of forest inventories in
each major region, which in turn serves as in-
put to the market component of TAMM.
Third, TAMM estimates the future harvests and
market responses in each region.
Our analysis suggests that there is a signifi-
cant and measurable difference in timber har-
vests attributable to ozone exposure under the Post-
CAAA and Pre-CAAA scenarios. At the outset of
our modeling period, the early 1990s, virtually no
change is measured in forest harvest volumes. This
result occurs because increases in growth rates do
not substantively affect timber volume over a short
period of time. By the end of our modeling period,
nearing 2010, increased growth rates over the previ-
ous decade(s) begin to affect overall forest yields of
harvestable timber. This is observed in Figure 7-1 as
an increasing annual benefit estimate over the mod-
eling period. The shape of the benefits time-series
reveals a production spike in the 2007 to 2008 pe-
riod. This spike is due to a large anticipated harvest
of Southeast U.S. timber due to forest maturity dur-
ing this period. The spike would occur even in the
absence of the CAAA, but is elevated by the CAAA
due to increased growth rates projected under the
Post-CAAA scenario. Although this change is small
in percentage terms relative to total economic sur-
plus generated by the timber sector, it contributes
to a large portion of the commercial timber benefits
estimate over the 1990-2010 period.
We calculate the cumulative value of annual ben-
efits based on the discounted stream of the annual
differences in consumer and producer surplus from
6 TAMM includes Canadian as well as U.S. timber produc-
tion regions because of the important influence of Canadian tim-
ber supply on the U.S. market. This analysis reflects modeling
of Canadian timber regions and their impact on U.S. produc-
tion, but we did not simulate changes in ozone in Canadian re-
gions.
Figure 7-1
Annual Economic Welfare Benefit of Mitigating Ozone
Impacts on Commercial Timber: Difference Between
the Pre-CAAA and Post-CAAA Scenarios
$1,150t
$950'
$750'
$550'
$350'
$150'
$(50)
commercial timber harvests under the Post-CAAA
and Pre-CAAA ozone exposure scenarios from 1990
to 2010. Discounting annual benefits to 1990 using
a five percent discount rate, the total cumulative
benefits estimate is approximately $1.9 billion. These
estimates are incorporated into the primary central
estimate by developing a range of annual estimates
for the year 2000, based on model results for the
period 1998 to 2002, and the year 2010, based on
model results for the period 2005 to 2010. The aver-
aging of results across several years to generate our
target year results avoids the potential problem of a
particular year's results (such as for 2010)
mischaracterizing the full time series of estimates
when we later calculate the net present value of ef-
fects.
Reduced Carbon Sequestration
Associated with Reduced Timber
Growth
Forest ecosystems help mitigate increasing atmo-
spheric concentrations of carbon dioxide by seques-
tering carbon from the atmosphere. These ecosys-
tems convert atmospheric carbon into biological
structures (e.g., wood) or substances needed in the
tree's physiological processes. As described above,
however, ozone reduces the growth of forests,
thereby limiting the amount of carbon that is se-
questered. Sequestered carbon can help mitigate glo-
bal climate change that has been linked to anthro-
pogenic emissions of carbon and other greenhouse
gases.
93
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
We used the timber inventory output of the
TAMM/ATLAS modeling system (described above),
in combination with a forest carbon model
(FORCARB), to estimate changes in carbon storage
in each of four ecosystem components: trees, forest
understory, forest floor, and soil. The estimates from
FORCARB, however, do not account for 'leakages"
of carbon back to the atmosphere as wood or wood
products decay and decompose over time. To esti-
mate the amount of carbon that is sequestered over
the long-term, we used a second model,
HARYCARB, to estimate the life-cycle of harvested
forest timber and thereby adjust the forest carbon
sequestration estimates of FORCARB.
The results of these calculations yield estimates
of long-term increases in carbon storage as a result
of the CAAA provisions of 8 million metric tons of
carbon per year by the year 2000, and 29 million
metric tons of carbon per year by the year 2010.
Because of the great uncertainties in assessing the
mitigating effect of carbon sequestration on global
climate change, and the economic value of avoiding
climate change, we do not attempt to monetize this
category of benefit.
Other Categories of Ecological Benefits
There were two additional categories of ecologi-
cal effects for which we considered developing eco-
nomic estimates; however, we abandoned the exer-
cise when key portions of the analysis proved to be
excessively problematic. Aesthetic degradation of
forests, the first of these additional categories, was
supported by a benefits transfer of contingent valu-
ation studies of individual willingness to pay to avoid
foliar damage. This category of effects, however,
proved too difficult to link to the specific air quality-
scenarios we evaluated. In other words, available
scientific methods and data on the visual appearance
of forest stands and their impact on perceived forest
aesthetics make it difficult to precisely describe
changes in forest aesthetics. Evaluation of the sec-
ond additional effect category, toxification of fresh-
water fisheries, was limited by the lack of toxic depo-
sition and exposure data as well as by the limitations
of available economic estimates of the impacts of
toxics on recreational and commercial fishery re-
sources. (See Appendix E for a more detailed dis-
cussion of these service flows). These and many other
ecological benefit categories could not be quantified
given current data and methods and are thus not re-
flected in our overall benefits estimates.
Valuation of Other Effects
Agricultural Benefits
As discussed earlier in this chapter, tropospheric
ozone affects the growth of a wide range of plant
species, including agricultural crops. Our agricul-
tural benefits analysis relies on crop-yield loss C-R
functions derived from the National Crop Loss As-
sessment Network (NCLAN) research and a national
economic model of the agricultural sector (AGSIM).
The NCLAN-derived. relationships use a sum of
hourly ozone concentration at or above 0.06 ppm
(SUM06) as a measure of ozone exposure for the .May
to September ozone season; these exposure estimates
arc derived from the ozone air quality modeling re-
sults discussed in Chapter 4. Where the C-R func-
tions require a longer time period of ozone concen-
trations, for example, for winter crops or when the
growing or harvest season for summer crops extends
beyond the end of September, we rely on 1990 moni-
tor data to estimate ozone exposure, conservatively
using the same estimates for both Pre-CAAA and
Post-CAAA scenarios. The NCLAN functions
cover the following crops: corn, cotton, peanuts,
sorghum, soybeans, and winter wheat.
The AGSIM agricultural sector model takes the
yield loss information, incorporates agricultural
price, farm policy, and other data for each year, and
then estimates production levels for each crop and
the economic benefits to consumers and producers
associated with these production levels. The crop
coverage in the AGSIM model includes a wider range
of crops than the NCLAN data inputs, adding bar-
ley, oats, hay, rice, and cottonseed. The broader-
crop coverage ensures that the model addresses price
and production quantity effects on potential substi-
tute crops that might be related to the effects in the
six NCLAN crops. We estimate economic effects
using a range of C-R outcomes for several crops, to
reflect the variation in ozone sensitivity among the
various crop cultivars. Our central estimate is the
expected value of the range of results that emerge
from the economic model.
94
-------
Chapter 7: Ecological arid Other Welfare Effects
Our results indicate significant beneficial effects
of ozone reductions in the agricultural sector. Our
Primary Central estimate of the benefit in 2000 is
$450 million; the annual benefit rises to $550 mil-
lion in 2010. Our estimated uncertainty around the
Primary Central estimates, however, is very broad.
For example, in 2010, the Primary Low estimate is
$7.1 million, and the Primary High is $1,100 mil-
lion. The uncertainty7 range reflects variation in the
ozone response of crop cultivars and uncertainty
about the suitability of alternative crop cultivars for
the soil types and climate conditions in various agri-
cultural regions. See Appendix F for more details
on the methods and results of the C-R functions and
economic modeling for agricultural effects.
Visibility
As outlined in Chapter 4, air pollution impairs
visibility in both residential and recreational settings.
An individual's willingness to pay to avoid reduc-
tions in visibility differs in these two settings. Im-
pairments in residential visibility are experienced
throughout an individual's daily life and activities.
Visibility in recreational settings, on the other hand,
is experienced by visitors to areas with notable vis-
tas. For the purposes of this report, we interpret
recreational settings applicable for this category of
effects to include National Parks throughout the
nation. Other recreational settings may also be ap-
plicable, for example National Forests, state parks,
or even hiking trails or roadside areas, but a lack of
suitable economic valuation literature to identify
these other areas, as well as a lack of visitation data
in some cases, prevents us from generating estimates
for those recreational vista areas.
We derive a residential visibility valuation func-
tion from the Chestnut and Dennis (1997) published
estimates for the Eastern U.S. These estimates are
based on original research conducted by McClelland
et al. (1990) in two Eastern cities (Atlanta and Chi-
cago). Because of technical concerns about the
study's methodology, however, we calculate a ben-
efits estimate but omit the results from the primary-
benefits estimates.1' For recreational visibility, we
' The two technical concerns involve the method of adjust-
ing the contingent valuation survey results for non-response, and
the failure to include adjustments for the "warm glow" effect,
or the tendency of respondents to indicate higher willingness to
pay for an environmental good because of a strong desire to
improve the environment in general.
derive values from the the Chestnut and Rowe (1989)
study of WTP for visibility in three park regions in
the Western, Southwestern, and Eastern U.S.8 In
both cases, the valuation function takes the follow-
ing form:
HHWTP = E * lti(VR1/VR2)
where:
HHWTP = annual WTP per household for
visibility changes
YR1 = the starting annual average visual
range
VR2 = the annual average visual range after
the change in air quality
B = the estimated visibility coefficient.
The form of this valuation function is designed
to reflect the way individuals perceive and express
value for changes in visibility. In general terms, ex-
pressed WTP for visibility changes varies with the
percentage change in visual range, a measure that is
closely related to, though not exactly analogous to,
the Deciview index used in Chapter 4. We use a
central B coefficient for residential visibility of 141,
as reported in Chestnut and Dennis (1997). For rec-
reational visibility, the coefficients vary based on the
region of study and whether the household is within
or outside of the National Park region studied. In-
region coefficients are higher than those for out-of-
region households. The in-region estimates for Cali-
fornia, the Southwest, and Southeast are $105, $137,
and $65, respectively; the corresponding out-of-re-
gion estimates are $73, $110, and $40, respectively.
The derivation and application of these valuation
functions are described in more detail in Appendix
H. The results of this procedure suggest visibility is
an important category of CAAA benefits; the Pri-
mary Central estimate for 2010, for example, indi-
cates annual recreational visibility benefits of $2.9
billion.
Worker Productivity
We base the valuation of worker productivity
on a study that measures the decline in worker pro-
3 The visibility valuation function, and the sources of esti-
mates for the coefficients for the functions, were origin ally de-
veloped as part of die National Acid Precipitation Assessment
Program (NAPAP), and were subjected to peer-review as part
of that program.
95
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
ductivity among outdoor farm workers exposed to
ozone (Crocker and Horsf, 1981). In our analysis,
we estimate the value of reduced productivity at f 1
per 10 percent increase in ozone concentration. This
estimate reflects valuing reduced productivity in
terms of the reduction in percentage of daily income
incurred by the average worker engaged in strenu-
ous outdoor labor.
Stratospheric Ozone Provisions
The quantified benefits of stratospheric ozone
protection provisions are dominated by the reduced
health effects expected from reductions in UV-b ra-
diation; the derivation of health benefits of these
provisions is discussed in Chapter 5. We summarize
other categories of benefits associated with reduced
UV-b radiation in Table 7-9. The quantified ben-
efits include: reduced crop damage; and reduced poly-
mer degradation. To estimate crop damage, we ap-
ply the results of existing studies on the relationship
between crops and UV-b radiation to the changes in
UV-b radiation predicted by the emissions and at-
mospheric models.9 The polymer damage function
is based on a study by Horst (1986). The estimated
total cumulative benefits associated with these eco-
logical and other welfare effects are about 2 percent
of the total cumulative benefits of the 'Title VI pro-
visions.
Sources ol dose-response relationship for crops and UV-
b: Teramura and Murali (1986) and Rowe and Adams (1987).
Source of dose-response relationship for crops and tropospheric
ozone: Rowe and Adams (1987).
Table 7-9
Quantified and Unquantified Ecological and Welfare Effects of Title VI Provisions
Ecological Effects- Quantified Estimate Basis for Estimate
American crop harvests Avoided 7.5 percent decrease Dose-response sources: Teramura and Murali
from UV-b radiation by 2075 (1986), Rowe and Adams (1987)
American crop harvests Avoided decrease from Estimate of increase in tropospheric ozone:
tropospheric ozone Whitten and Gery (1986). Dose-response
source: Rowe and Adams (1987)
Polymers Avoided damage to materials Source of UV-b/stabilizer relationship: Horst
from UV-b radiation (1986)
Ecological Effects- Unquantified
Ecological effects of UV. For example, benefits relating to the following:
• recreational fishing
• forests
• marine ecosystem and fish harvests
• avoided sea level rise, including avoided beach erosion, loss of coastal wetlands, salinity of estuaries
and aquifers
• other crops
• other plant species
• fish harvests
Ecological benefits of reduced tropospheric ozone relating to the overall marine ecosystem, forests, man-made
materials, crops, other plant species, and fish harvests
Benefits to people and the environment outside the U.S.
Notes:
1) For more detail see EPA's Regulatory Impact Analysis: Protection of Stratospheric Ozone (1988).
2) Note that the ecological effects, unlike the health effects, do not reflect the accelerated reduction and phaseout schedule
of section 606.
3) Benefits due to the section 606 methyl bromide phaseout are not included in the benefits total because the EPA provides
neither annual incidence estimates nor a monetary value.
96
-------
Chapter 7: Ecological arid Other Welfare Effects
Summary of Quantitative
Results
Although the effects of air pollutants on eco-
logical systems are likely to be widespread, many
effects may be poorly understood and lack quantita-
tive effects characterization methods and support-
ing data. In addition, many of our quantitative re-
sults reflect an incomplete geographic scope of analy-
sis; for example, we generated monetized acidifica-
tion results only for the Adirondacks region of New-
York State. As a result, the quantitative results we
generate for the purposes of estimating the benefits
of the CAAA reflect only a small portion of the over-
all impacts of air pollution on ecological systems or
ecological service flows.
Despite these limitations, it is important to rec-
ognize the magnitude of the monetized ecological
benefits that we could estimate and reflect those re-
sults in the overall estimates of benefits generated
in the larger analysis. Table 7-10 provides a tabular
summary of the results documented earlier in this
chapter. It is not possible to indicate the degree to
which ecological benefits are underestimated, but
considering the magnitude of benefits estimated for
the select endpoints considered in our analysis, it is
reasonable to conclude that a comprehensive ben-
efits assessment would yield substantially greater total
benefits estimates.
In Table 7-11 we provide a summary of benefits
estimates for other welfare effects, including reduced
agricultural yields, impaired visibility, and decreased
Table 7-10
Summary of Evaluated Ecological Benefits (millions 1990$)
Description
of Effect
Air
Pollutant
Geographic
Scale of
Economic
Estimate
Range of
Annual Impact
Estimates in
2010
Primary
Central
Estimate
for 2010
Primary Central
Cumulative
Impact Estimate
1990-2010
Key Limitations
Freshwater Sulfur and Regional
acidification nitrogen (Adirondacks)
oxides
$12 to $88 $50 $260 - Captures only
recreational fishing
impact
- Incomplete
geographic coverage
leads to underestimate
of benefits
Reduced tree
growth - Lost
commercial
timber
Ozone National $190to$1000 $600 $1,900 - Uncertainties in
stand-level response to
ozone exposure
- Uncertainty in future
timber markets
TOTAL MONETIZED
ECONOMIC BENEFIT
$200 to $1,100 $650
$2,200 - Partial estimate that
omits major
unquantifiable benefits
categories; see text
Note: Estimates reflect only those benefits categories for which quantitative economic analysis was supported. A
comprehensive total economic benefit estimate would likely greatly exceed the estimates in the table. Range of
estimates for timber assessment is based on variation in annual point estimates for 2005 through 2010.
Table 7-11
Summary of Other Welfare Benefits (millions 1990$)
Description
of Effect
Air
Pollutant
Geographic
Scale of
Economic
Estimate
Primary Central
Annual Estimate
2000 2010
Primary Central
Cumulative
Estimate
1990-2010
Key Limitations
Reduced
Agricultural
Yields
Ozone
National
$450
$550
$3,900
- Covers only major grain crops
- Omits effects on fruits and
vegetables
Impaired
Recreational
Visibility
Particulate
Matter
National
$2,000
$2,900
$19,000
- National Parks only
- Omits residential visibility
benefits
Reduced
Worker
Productivity
Ozone
National
$460
$710
$4,400
- Reflects effects on workers
engaged in strenuous outdoor
employment
Note: Estimates reflect only those benefits categories for which quantitative economic analysis was supported.
97
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
worker productivity. These estimates add substan-
tially to the total 11011-health benefits of the CAAA.
In particular, our estimates for the annual value of
avoiding visibility impairments is $2,900 million by
2010, even through this estimate does not reflect the
value of residential visibility improvements.
Uncertainty
Because of the limitations in the available meth-
ods and data, the benefits assessment in this report
does not represent a comprehensive estimate of the
economic benefits of the CAAA. Moreover, the
potential magnitude of long-term economic impacts
of ecological damages mitigated by the CAAA sug-
gests that great care must be taken to consider those
ecosystem impacts that are not quantified here. Sig-
nificant future analytical work and basic ecological
and economic research must be performed to build
a sufficient base of knowledge and data to support
an adequate assessment of ecological benefits. For
the current analysis, this incomplete coverage of
effects represents the greatest source of uncertainty
in the ecological assessment. This and other key
uncertainties are summarized in Table 7-12.
Because the chronic ecological effects of air pol-
lutants may be poorly understood, difficult to ob-
serve, or difficult to discern from other influences
on dynamic ecosystems, our analysis focuses on acute
or readily observable impacts. Disruptions that may
seem inconsequential in the short-term, however, can
have hidden, long-term effects through a series of
interrelationships that can be difficult or impossible
to observe, quantify, and model. This factor sug-
gests that many of our qualitative and quantitative
results may underestimate the overall, long-term ef-
fects of pollutants on ecological systems and re-
sources.
Table 7-12
Key Uncertainties Associated with Ecological Effects Estimation
Direction of
Potential
Bias for Net
Benefits
Estimate
Potential Source of Error
Likely Significance Relative to Key Uncertainties in Net
Benefit Estimate*
Incomplete coverage of
ecological effects
identified in existing
literature, including the
inability to adequately
discern the role of air
pollution in multiple
stressor effects on
ecosystems.
Underestimate Potentially major. The extent of unquantified and unmonetized
benefits is largely unknown, but the available evidence suggests
the impact of air pollutants on ecological systems may be
widespread and significant. At the same time, it is possible that a
complete quantification of effects might yield economic valuation
results that remain small in comparison to the total magnitude of
health benefits.
Omission of the effects of
nitrogen deposition as a
nutrient with beneficial
effects.
Incomplete assessment of
long-term bioaccumulative
and persistent effects of
air pollutants.
The PnET II modeling of
the effects of ozone on
timber yields relies on a
simplified mechanism of
response (i.e., changes in
net primary productivity).
Overestimate Probably minor. Although nitrogen does have beneficial effects
as a nutrient in a wide range of ecological systems, nitrogen in
excess also has significant and in some cases persistent
detrimental effects that are also not adequately reflected in the
analysis.
Underestimate Potentially major. Little is currently known about the longer-term
effects associated with the accumulation of toxins in ecosystems,
but what is known suggests the potential for major impacts.
Future research into the potential for threshold effects is
necessary to establish the ultimate significance of this factor.
Overestimate Probably minor. Existing evidence suggests that the growth
changes PnET II projects are relatively large, however none of
the currently available points of conparison fully address such
issues as the impact of stand-level competition, and the net
primary productivity results are within the range of results of other
studies of environmental and anthropogenic stressors.
The classification of each potential source of error reflects the best judgement of the section 812 Project Team. The Project
Team assigns a classification of "potentially major" if a plausible alternative assumption or approach could influence the
overall monetary benefit estimate by approximately five percent or more; if an alternative assumption or approach is likely to
change the total benefit estimate by less than five percent, the Project Team assigns a classification of "probably minor."
98
-------
Comparison of
Costs and Benefits
In this chapter we present our summary of the
primary estimates of monetized benefits of the
CAAA from 1990 to 2010, compare the benefits es-
timates with the corresponding costs, and explore
some of the major sources of uncertainty in the ben-
efits estimates. We also present the results of our
calculations using alternative assumptions for sev-
eral key input variables.
Monetized Benefits of the
CAAA
In this section we provide an overview of the
three types of analyses conducted to estimate ben-
efits, present the annual estimates of monetized ben-
efits for the human health, ecological, and welfare
analyses, and then present an aggregate measure of
benefits from all titles of the CAAA for the fall study
period.
Overview of Benefits Analyses
Our primary estimates of the monetized eco-
nomic benefits for the 1990 to 2010 period derive
from three distinct analyses: (1) the analysis of
changes in human health effects associated with re-
duced exposures to criteria pollutants and the subse-
quent valuation of these changes, summarized and
described in Chapters 5 and 6; (2) the analysis of
monetized ecological and other welfare benefits (e.g.,
visibility), described in Chapter 7; and (3) the analy-
sis of the benefits of stratospheric ozone protection
provisions, summarized briefly in Chapters 5, 6, and
7 and described in detail in Appendix G.
We measure the benefits and present the results
from each of these analyses in slightly different ways.
For the first two analyses, we generate annual esti-
mates of benefits that result from changes in expo-
sures in two target years of the study, 2000 and 2010.
These estimates can be directly compared to the es-
timates of costs incurred in the target years, because
for the most part the annual benefits accrue in the
same year as the costs are incurred. There is one
exception, however: we model the effect of particu-
late matter on premature mortality to occur over a
period of five years from the time of exposure. In
this case, we have accounted for the incidence of
premature mortality over the assumed lag period,
and discounted the valuation of this effect back to
the target year.
The annual estimates provide an indication of
the trend in benefits accrued over the 20-year study-
period. To generate a cumulative measure of ben-
efits over the full 20-year period, we must make an
assumption about the level of benefits that would
be realized in the years between the target years. We
interpolate these values, assuming a linear trend in
both costs and benefits over the 1990 to 2000 and
2000 to 2010 periods (assuming benefits and costs in
the starting year, 1990, are zero). In one portion of
the ecological benefits analysis, acidification, we gen-
erate only a single annual estimate for the target year
2010. In that case, we assume a linear trend in an-
nual benefits over the full 20-year study period.
The third analysis, assessing changes in strato-
spheric ozone and the resulting health effects, is dif-
ferent from the criteria pollutant analyses. The long-
term nature of the program, and the significant lag
effects associated with the processes of ozone deple-
tion over decades-long time scales, make it difficult
to generate a meaningful estimate for any single tar-
get year. As a result, we could not generate an an-
nual benefit estimate that could be reliably linked to
emissions reductions in a single year and, by exten-
sion, compared to the costs incurred to achieve that
year's allocation of reductions in stratospheric ozone
depleting substances. Instead, we generate an annu-
alized equivalent of the cumulative present value of
99
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
benefits and costs of the Title VI program. These
annualized equivalents cannot be ascribed to any
particular target year.
These fundamental differences in the measure-
ment of benefits affect our presentation of benefits
estimates in this chapter. Although we generate and
report an annual estimate of costs and benefits of
Title VI provisions, we encourage the reader to in-
terpret aggregations of these annual estimates with
those from other titles of the CAAA with caution.
In particular, we discourage the use of these CAAA
Title-specific benefit-cost ratios as the sole, or even
primary, basis for comparing the relative economic
value of Title VI versus other CAAA titles. The
comparative benefit-cost ratios are too sensitive to
important, highly uncertain analytical assumptions
such as the discount rate.
Summary of Monetized Benefits for
Human Health and Welfare Effects
As discussed above, we generate annual estimates
for the human health and welfare effects based on
exposure analysis conducted for each of the two tar-
get years of the analysis, 2000 and 2010. The range
of estimates we generate for the monetized benefits
of human health effects incorporates both the quan-
tified uncertainty associated with each of the health
effect estimates and the quantified uncertainty asso-
ciated with the corresponding economic valuation
strategy. Quantitative estimates of uncertainties in
earlier steps of the analysis (i.e., emissions and air
quality changes) could not be developed adequately
and are therefore not applied in the present study.
As a result, the range of estimates for monetized ben-
efits presented in this chapter is more narrow than
would be expected with a complete accounting of
the uncertainties in all analytical components. The
characterization of the uncertainty surrounding eco-
nomic valuation is discussed in detail in Appendix
IT. The characterization of the uncertainty surround-
ing specific health effect estimates is discussed in
Appendix D. Below, we discuss the combined ef-
fect of these two categories of uncertainty and our
techniques for aggregating uncertainty across end-
points and analyses.
We assume that for each endpomt-pollutant com-
bination there are distributions for both the con-
centration-response function and the valuation co-
efficients. We combine these distributions by using
a computerized, statistical aggregation technique to
estimate the mean of the monetized benefit estimate
for each endpomt-pollutant combination and to char-
acterize the uncertainty surrounding each estimate.1
In the first step of our procedure, we employ
statistical analysis to generate mean estimates and
quantified uncertainty measures for the C-R func-
tion for each endpomt-pollutant combination. For
many health and welfare effects, only a single study
is available to use as the basis for the C-R function.
In this case, the best estimate of the mean of the
distribution of C-R coefficients is the reported esti-
mate in the study. The uncertainty surrounding the
estimate of the mean C-R coefficient is character-
ized by the standard error of the reported estimate.
This yields a normal distribution, centered at the
reported estimate of the mean. If multiple studies
are considered for a given C-R function, a normal
distribution is derived for each study, centered at
the mean estimate reported in the study. On each
iteration of the aggregation procedure, a C-R coeffi-
cient is selected from an aggregate distribution of C-
R estimates for that endpoint. The aggregate distri-
bution of C-R coefficients is determined by a vari-
ance-weighted aggregate distribution of values.
In the second step, we estimate incidence for each
exposure analysis unit (i.e., 8 km by 8 km cell in a
grid pattern) in the 48 contiguous states, and aggre-
gate the results into an estimate of the change in
national incidence of the health or welfare effects.
Through repeated iterations from the distribution
of mean OR coefficients, we generate a distribution
of the estimated change in incidence for each health
and welfare effect due to the change in air quality
between the Post-CAAA and Pre-CAAA scenarios.
Finally, in the third step we use computerized
statistical aggregation methods once again to charac-
* The statistical aggregation technique applied is commonly
referred to as simulation modeling. The technique involves many
re-calculations of results, using different combinations of input
parameters each time. For each calculation, values from each
input parameter's statistical distribution are selected at random
to ensure that the calculation does not always result in extreme
values, or rely solely on low end or solely on high end input
parameters. The aggregate distribution more accurately reflects
a reasonable likelihood of the joint, occurrence ol multiple in-
put parameters.
100
-------
Chapter 8: Comparison of Costs and Benefits
terize the overall uncertainty surrounding monetized
benefits. For each distinct health and welfare effect,
the aggregation procedure selects an estimated inci-
dence change from the distribution of changes for
that endpoint, selects a unit value from the corre-
sponding distribution of economic valuation unit
values, and multiplies the two to generate a mon-
etized benefit estimate. We then repeat the process
many times to generate a distribution of estimated
monetized benefits for each endpoint-pollutant com-
bination. Combining the results for the individual
endpoints using the aggregation procedure yields a
distribution of total estimated monetized benefits for
each target year (2000 and 2010).2 We present the
results of this analysis of health effects in Table 6-3
in Chapter 6.
The ecological and welfare results are not cur-
rently amenable to the same type of uncertainty
analysis. The modeling procedures for estimating
the effects of sulfur and nitrogen deposition in acidi-
fying lakes, the effects of ozone in reducing timber
and agricultural production, and the effects of par-
ticular matter on visibility are all subject to uncer-
tainty and require substantial resources simply to
develop single estimates. We describe key uncer-
tainties in Chapter 7 and they are reflected in the
ranges of values we present at the end of that chap-
ter. The sources of uncertainty in these estimates,
however, cannot as easily be dis-
aggregated among physical ef-
fects modeling and valuation
components. The endpoints of
the ranges we present reflect rea-
sonable alternative choices in
key input variables, but the
ranges cannot currently be inter-
preted as points on a statistical
distribution of results. For these
ecological effects, the central es-
timate is the midpoint of the
ranges of values. We then inter-
pret the endpoints of the range
of estimates as the upper and
lower bounds of a uniform dis-
tribution of values. The uni-
form distribution is used when we aggregate the eco-
logical and other welfare effects analyses with the
analyses of human health.
Annual Benefits Estimates
We present the results of our aggregation of pri-
mary annual benefits estimates for Titles I through
V in Figure 8-1 below. The figure provides a charac-
terization of both the primary central estimate and
the range of values generated by the aggregation
procedure described above, for each of the two tar-
get years of the analysis (2000 and 2010). The Pri-
mary High estimate corresponds to the 95th percen-
tile value from the aggregation, and the Primary Low
estimate corresponds to the 5th percentile value. The
total benefits estimates are substantial; the Primary
Central estimate in 2010 is $110 billion.
Table 8-1 shows the detailed breakdown of ben-
efits estimates for one of the two target years, 2010.
As shown in the table, $100 billion of the $110 bil-
lion total benefit estimate in 2010, or roughly 90
percent, is attributable to reductions in premature
mortality associated with reductions in ambient par-
ticulate matter and associated criteria pollutants. The
remaining benefits are divided among two broad
categories of benefits: avoided morbidity, the larg-
est component of which is avoided chronic bron-
Figure 8-1
Central, Low, and High Primary Benefits Results for
Target Years (in billions of 1990 dollars) - Titles I through V
300-
Benefits
¦
£
O
— 200-
m
(/)
d>
£
d>
CO
Benefits
150-
100-
jS 50-
< High 160
^ Central 71
^Low 16
^Central 110
^ Low 26
2000
2010
2 This procedure implicitly assumes independence between
the specific aggregation simulation draws from the distribution
of health and economic valuation estimates.
101
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 8-1
Criteria Pollutant Health and Welfare Benefits in 2010
Monetary Benefits (in millions 1990$)*
Primary
Benefits Category Primary Low Central Primary High
Mortality
Ages 30+
14,000
100,000
250,000
Chronic Illness
Chronic Bronchitis
360
5,600
18,000
Chronic Asthma
40
180
300
Hospitalization
All Respiratory
76
130
200
Total Cardiovascular
93
390
960
Asthma-Related ER Visits
0.1
1.0
2.8
Minor Illness
Acute Bronchitis
0.0
2.1
5.2
URS
4.2
19
39
LRS
2.2
6.2
12
Respiratory Illness
0.9
6.3
15
Mod/Worse Asthma1
1.9
13
29
Asthma Attacks1
20
55
100
Chest Tightness, Shortness of
Breath, or Wheeze
0.0
0.6
3.1
Shortness of Breath
0.0
0.5
1.2
Work Loss Days
300
340
380
MRAD/Any-of-19
680
1,200
1,800
Welfare
Decreased Worker
Productivity
710
710
710
Visibility - Recreational
2,500
2,900
3,300
Agriculture (Net Surplus)
7.1
550
1,100
Acidification
12
50
76
Commercial Timber
180
600
1,000
Aggregate Range of Benefits2
26,000
110,000
270,000
Note:
* The estimates reflect air quality results for the entire population in the US.
1 Moderate to worse asthma, asthma attacks, and shortness of breath are endpoints
included in the definition of MRAD/Any of 19 respiratory effects. Although valuation
estimates are presented for these categories, the values are not included in total benefits to
avoid the potential for double-counting.
2 The Aggregate Range reflects the 5th, mean, and 95th percentile of the estimated credible
range of monetary benefits based on quantified uncertainty, as discussed in the text.
102
-------
Chapter 8: Comparison of Costs and Benefits
Table 8-2
Present Value of Monetized Benefits for 48 State Population
Present Value (millions 1990$, discounted to 1990 at 5 percent)
Primary Low Primary Central Primary High
Titles I through V (1990 through 2010) $160,000 $690,000 $1,600,000
Title VI (1990 through 2165) $100,000 $530,000 $900,000
chitis, comprises about 60 percent of the non-mor-
tality benefits; and avoided ecological and other wel-
fare effects, the largest component of which is im-
proved recreational visibility, comprises about 40
percent. Note that, because of the aggregation pro-
cedure used, and because we round all intermediate
results to two significant digits for presentation pur-
poses, the columns of Table 8-1 may not sum to the
total estimate presented in the last row.3
Aggregate Monetized Benefits
As discussed earlier in this chapter, we linearly
interpolate benefit estimates between 1990 and 2000
and between 2000 and 2010 and then aggregate the
resulting annual estimates across the entire 1990 to
2010 period of the study to yield a present discounted
value of total aggregate benefits for the period. In
this section we discuss issues involved in each stage
of aggregation, as well as the results of the aggrega-
tion.
As noted earlier, air quality modeling was car-
ried out only for the two target years (2000 and 2010).
The resulting annual benefit estimates provide a tem-
poral trend of monetized benefits across the period
resulting from the annual changes in air quality.
They do not, however, characterize the uncertainty
associated with the yearly estimates for intervening
years. In an attempt to capture uncertainty associ-
ated with these estimates, we relied on the ratios of
the 5th percentile to the mean and the 95th percen-
tile to the mean in the two target years. In general,
these ratios were fairly constant across the target
¦' The sum of benefits across eiidpoints at a given percentile
level does not result in the total monetized benefits estimate at
the same percentile level in fable 8-1. For example, if the fifth
percentile benefits of the endpoints shown in Table 8-1 were
added, the resulting total would be substantially less than $30
billion, the fifth percentile value of the distribution of aggregate
monetized benefits reported in Table 8-1. This is because the
various health and welfare effects are treated as stochastically
independent, so that the probability that the aggregate monetized
benefit is less than or equal to tie sum of the separate live per-
centile values is substantially less than five percent.
years, for a given endpoint. The ratios were inter-
polated between the target years, yielding ratios for
the intervening years. Multiplying the ratios for each
intervening year by the central estimate generated
for that year provided estimates of the 5th and 95th
percentiles, which we use to characterize uncertainty
about the Primary Central estimate.
In Table 8-2 we present the cumulative mon-
etized benefits aggregated from 1990 to 2010. We
present the mean estimate from the aggregation pro-
cedure, along with the Primary Low7 (i.e., 5th per-
centile of the distribution) and Primary High (i.e.,
95th percentile of the distribution) estimates, for all
provisions of Titles I through V and, then, separately
for Title YI. Aggregating the stream of monetized
benefits across years involved discounting the stream
of monetized benefits estimated for each year to the
1990 present value (using a five percent discount rate).
Aggregate Benefits of Title ¥1 Provisions
As described in summary form in Chapters 5, 6,
and 7 and in detail in Appendix G, expected human
health benefits from Title VI provisions are substan-
tial. The analysis we conducted is based largely on
existing results from EPA Regulatory Impact Analy-
ses for individual rules promulgated under Title VI.
To the extent possible, we adjusted existing estimates
to reflect both the central estimates and uncertainty
characterizations used in the criteria pollutant analy-
sis. We made major adjustments for both the value
of statistical life (VSL) and the discount rate. We
adjusted the VSL estimate to reflect the Weibull dis-
tribution of VSL used in our analysis for other pro-
visions. As discussed in the appendix, the choice of
the discount rate for estimated benefits which ac-
crue over decades to century-long time spans pre-
sents special problems. Although we argue that a
two percent discount rate is more appropriate for
such long-term discounting, for consistency in this
chapter we present estimates using the five percent
discount rate used throughout the rest of this study.
103
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
The results of the benefits calculations in Ap-
pendix G indicate a cumulative central benefit esti-
mate of $530 billion for Title VI (see Appendix G
for details). Using the same aggregation techniques
for the valuation analysis described above, but only
for the mortality valuation step, we generate a 90
percent confidence interval around this central esti-
mate to derive Primary Low and Primary High esti-
mates of $100 billion to 1900 billion, respectively.
We present these estimates in Table 8-2 above. The
annual human health benefits from Title YI provi-
sions steadily increase until about 2045, then decrease
until 2165, the last year in the analysis. About 93
percent of the benefits accrue from 2015 to 2165.
These benefit estimates only partially reflect poten-
tial averting behaviors, such as remaining indoors
or increasing use of sun screens or hats, which may
mitigate the effects of the UV-b exposure increases
estimated under the Prc-CAAA scenario.
Comparison of Monetized
Benefits and Costs
Table 8-3 presents summary quantitative results
for the prospective assessment, with costs disaggre-
gated by Title and benefits disaggregated by major
category. We present annual,
primary estimate results for
each of the two target years of
the analysis, with all dollar fig-
ures expressed as inflation-ad-
justed 1990 dollars. The final
columns provide net present
value estimates for costs and
benefits from 1990 to 2010 or,
in the case of stratospheric
ozone protection provisions,
1990 to 2165, discounted to
1990 at five percent. The re-
sults indicate that the Primary
Central estimate of benefits
clearly exceeds the costs of the
CAAA, for each of the two
target years and for the cumu-
lative estimates of present
value over the 1990 to 2010 pe-
riod.
gating benefits by CAAA Title or even by pollut-
ant. As the table indicates, a very high percentage
of the benefits is attributable to reduced premature
mortality associated with reductions in ambient par-
ticulate matter and associated criteria pollutants. The
CAAA achieves ambient PM reductions through a
wide range of provisions controlling emissions of
both gaseous precursors of PM that form particles
in the atmosphere (sulfur and nitrogen oxides as well
as, to a lesser extent, organic constituents) and di-
rectly emitted PM (i.e., dust particles). Because the
effects of these constituents on ambient PM are non-
linear, and because some precursor pollutants inter-
act with each other in ways which influence the to-
tal concentration of particulates in the atmosphere,
separating the effects of individual pollutants on the
change in ambient PM would require many itera-
tions of our air quality modeling system. These dif-
ficulties in separating the effects of individual emis-
sions reductions on the benefits estimates also high-
light the need for an integrated air quality modeling
system that can more readily analyze multiple sce-
narios within reasonable tune and resource con-
straints. A tool of this nature could allow us to more
reliably and cost-effectively estimate incremental
contributions to ambient PM and ozone concentra-
tion reductions.
Table 8-3
Summary of Quantified Primary Central Estimate Benefits and Costs
(Estimates in million 1990$)
Cost or Benefit
Annual Estimates
Category
2000
2010
Present Value
Costs:
Title I
$8,600
$14,500
$85,000
Title II
$7,400
$9,000
$65,000
Title III
$780
$840
$6,600
Title IV
$2,300
$2,000
$18,000
Title V
$300
$300
$2,500
Total Costs, Title l-V
$19,000
$27,000
$180,000
Title VI
$1,400'
k
$27,000*
Monetized Benefits:
Avoided Mortality
$63,000
$100,000
$610,000
Avoided Morbidity
$5,100
$7,900
$49,000
Ecological and
Welfare Effects
$3,000
$4,800
$29,000
Total Benefits, Title l-V
$71,000
$110,000
$690,000
Stratospheric Ozone
$25,00C
I*
$530,000*
The estimates in Table 8-3
reflect the difficulty we en-
countered in reliably disaggre-
* Annual estimates for Title VI stratospheric ozone protection provisions are annualized
equivalents of the net present value of costs over 1990 to 2075 (for costs) or 1990 to 2165
(for benefits). The difference in time scales for costs and benefits reflects the persistence of
ozone depleting substances in the atmosphere, the slow processes of ozone formation and
depletion, and the accumulation of physical effects in response to elevated UV-b radiation
levels.
104
-------
Chapter 8: Comparison of Costs and Benefits
Table 8-4 provides the results of our compari-
son of primary benefits estimates to primary cost
estimates. In the top half of the table we show both
annual and present value estimates for Titles I
through V, present value estimates for Title VI, and
a total present value for all titles. The "monetized
benefits" indicate both the Primary Central estimate
(the mean) from our statistical aggregation model-
ing analysis and the Primary Low and Primary High
estimates (5th and 95th percentile values, respec-
tively). In the bottom half of the table we present
two alternative methods for comparing benefits to
costs. "Net benefits" are the Primary Central esti-
mates of monetized benefits less the Primary Cen-
tral estimates of costs. The table also notes the ben-
efit/cost ratios implied by the benefit ranges.
The conclusion we draw from Table 8-4 is that,
given the particular data, models and assumptions
we believe are most appropriate at this time, our
analysis indicates that the benefits of the CAAA sub-
stantially exceed its costs. Furthermore, the results
of the uncertainty analysis imply that it is extremely
unlikely that the monetized benefits of the CAAA
over the 1990 to 2010 period could be less than its
costs. Looking at Titles I through V, the central
benefits estimate exceeds costs by a factor of four to
one, whether we are looking at annual or present
value measures, and the high estimate exceeds costs
by more than twice that factor (a ratio of nine or ten
to one). Using the Primary Low estimate of ben-
efits, the annual estimates of benefits in 2000 and
2010 are slightly less than the annual costs for that
year. The data also suggest that costs for criteria
Table 8-4
Summary Comparison of Benefits and Costs (Estimates in millions 1990$)
Titles I through V
Title VI
All Titles
Annual Estimates
2000 2010
Present Value
Estimate
1990-2010
Present Value
Estimate
1990-2165
Total Present
Value
Monetized Direct Costs:
Lowa
Central
High3
$19,000
$27,000
Not Estimated
$180,000
Not Estimated
$27,000
$210,000
Monetized Direct Benefits:
Lowb
Central
Highb
$16,000
$71,000
$160,000
$26,000
$110,000
$270,000
$160,000
$690,000
$1,600,000
$100,000
$530,000
$900,000
$260,000
$1,200,000
$2,500,000
Net Benefits:
Low
Central
High
($3,000)
$52,000
$140,000
($1,000)
$93,000
$240,000
($20,000)
$510,000
$1,400,000
$73,000
$500,000
$870,000
$50,000
$1,000,000
$2,300,000
Benefit/Cost Ratio:
Lowc
less than 1/1
less than 1/1
less than 1/1
less than 4/1
1/1
Central
4/1
4/1
4/1
20/1
6/1
High0
more than 8/1
more than 10/1
more than 9/1
more than 33/1
12/1
aThe cost estimates for this analysis are based on assumptions about future changes in factors such as
consumption patterns, input costs, and technological innovation. We recognize that these assumptions introduce
significant uncertainty into the cost results; however the degree of uncertainty or bias associated with many of the
key factors cannot be reliably quantified. Thus, we are unable to present specific low and high cost estimates.
b Low and high benefits estimates are based on primary results and correspond to 5th and 95th percentile results
from statistical uncertainty analysis, incorporating uncertainties in physical effects and valuation steps of benefits
analysis. Other significant sources of uncertainty not reflected include the value of unqualified or unmonetized
benefits that are not captured in the primary estimates and uncertainties in emissions and air quality modeling.
0 The low benefit/cost ratio reflects the ratio of the low benefits estimate to the central costs estimate, while the high
ratio reflects the ratio of the high benefits estimate to the central costs estimate. Because we were unable to reliably
quantify the uncertainty in cost estimates, we present the low estimate as "less than X," and the high estimate as
"more than Y", where X and Y are the low and high benefit/cost ratios, respectively.
105
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
pollutant programs grow somewhat more rapidly
than benefits from 1990 to 2000, but that benefits
grow more rapidly from 2000 to 2010.
The estimates for Title VI indicate that benefits
well exceed costs, even at the low benefits estimate.
This conclusion holds despite the relatively high dis-
count rate used for the estimates in Table 8-4 (5 per-
cent) a value that most analysts would consider too
high for the long time period over which benefits of
this program are discounted (175 years).4 The total
estimates for all titles of the CAAA also indicate ben-
efits in excess of costs for the full range of primary
benefits.
Cost-Effectiveness Evaluation
The approach to premature mortality valuation
used in our primary estimates is a method that al-
lows us to aggregate the benefits of reducing mortal-
ity risks with other monetized benefits of the CAAA.
One of the great advantages of the benefit-cost para-
digm is that a wide range of quantifiable benefits can
be compared to costs to evaluate the economic effi-
ciency of particular actions. Some analysts suggest,
however, that presentation of the results of a cost-
benefit analysis may mask the key assumptions that
are made to quantify all benefits in monetary terms.
Another evaluative paradigm, cost-effectiveness
analysis, is sometimes suggested as further evidence
of whether the benefits of a regulatory program jus-
tify its costs. Cost-effectiveness analysis involves es-
timation of the costs per unit of benefit (e.g., lives
saved). This type of analysis is most useful for com-
paring programs that have similar goals, for example,
alternative medical interventions or treatments that
can save a life or cure a disease. They arc less readily
applicable to programs with multiple categories of
benefits, such as the CAAA, because the cost-effec-
tiveness calculation is based on quantity of a single
benefit category. In other words, we cannot readily
convert reductions in new cases of chronic bronchi-
tis, reduced hospital admissions, improvements in
visibility, and increased commercial timber and crop
yields to a single metric such as "lives saved." For
4 The primat)' central benefit-cost ratio for Tide VT using
a 3 percent discount rate is 44 to 1, higher than any of those
presented in Table 8-4 (see Table 8-6 below), in addition, the
ratio using a 2 percent discount rale, liie rate used in liie under-
lying RIAs, is 75 to 1. See Appendix G for more detail on the
sensitivity ol Tide VI benefits to the choice oi discount rale.
these reasons, we prefer to present our results in
terms of monetary benefits.
Despite the risks of oversimplification of ben-
efits, cautiously interpreted cost-effectiveness calcu-
lations may provide farther evidence of whether the
costs incurred to implement the CAAA are a rea-
sonable investment for the nation. The most com-
mon cost-effectiveness metric, costs per life saved,
can be readily calculated from the information pre-
sented in this report. For example, we estimate the
total annual direct costs of implementation of Titles
I through Y in 2010 to be approximately 127 bil-
lion. In exchange for this expenditure, in the year
2010 we avoid 23,000 cases of premature mortality
and gain estimated non-mortality benefits of about
$20 billion. We can generate a net cost per life saved
by subtracting from costs the total non-mortalitv
benefits, and then dividing by lives saved. For Titles
I through Y, we estimate a net cost per life saved of
approximately $300,000 ($27 billion minus $20 bil-
lion divided by 23,000).3 Although we are also con-
cerned about many of the uncertain assumptions
required to generate cost per life-year saved estimates,
we include an estimate for illustrative purposes. For
the year 2010, the net cost per life-year saved esti-
mate implied by the primary central case results is
$23,000 per life-year ($7 billion divided by 310,000
life-years saved).6
Major Sources of Uncertainty
We can obtain additional insights into key as-
sumptions and findings of the present study through
further analysis of potentially important variables
and inputs. The estimated uncertainty ranges for
each endpomt category summarized in Table 8-1
reflect the measured uncertainty associated with two
aspects of the analysis: avoided physical effects (both
health and welfare benefits) and economic valuation
of benefits. In addition, in Chapter 3 we conduct
quantitative sensitivity analyses of key components
of the direct cost estimates. For many other aspects
of our analysis, however, including emissions esti-
= Hie illustrative calculations presented here do not. reflect
discounting of the physical incidence of mortality.
6 Because of Agency concerns regarding discounting of
physical effects, the ratio presented here reflects undiscounted
life-years saved. II iulure years were discounted, the implicit
cost per lile-year saved would be significantly higher.
106
-------
Chapter 8: Comparison of Costs and Benefits
mates, air quality modeling, and unquantified cat-
egories of benefits, we are unable to conduct quanti-
tative analysis of uncertainty. Instead, we have at-
tempted throughout this report to identify and char-
acterize major sources of uncertainty — we present
the results of these efforts at the end of Chapters 2
through 7. In this section, we provide a summary
evaluation of the relative importance of key sources
of uncertainty.
Table 8-5 below provides a summary of both
quantified and unquantified sources of uncertainty
and our estimates of the impact of these sources of
uncertainty on the primary central estimates of ben-
efits and costs. The table covers seven major catego-
ries of uncertainties: measurement uncertainties in
physical effects and valuation components of the
benefits analysis; measurement uncertainties in esti-
mation of direct costs; alternative assumptions for
PM-rclatcd mortality valuation; alternative assump-
tions for PM-relatcd mortality risk; unquantified
sources of error in emissions and air quality model-
ing; and omissions of key benefits categories. The
table entries cover quantitative analyses of uncer-
tainty, characterization of unquantified uncertainty,
and the potential effect of alternative modeling para-
digms for costs and benefits. Additional treatment
of alternative paradigms is necessary because reason-
able people may disagree with our methodological
choices regarding these issues, and these choices
might be considered to significantly influence the
results of the study.
Quantitative Analysis of Physical
Effects and Valuation Uncertainties
As discussed previously in this chapter, we have
conducted quantitative uncertainty analysis of our
benefits estimates to reflect measurement error in
two key steps of the analysis: estimation of physical
effects and economic valuation. We present the re-
sults of our analysis in Figure 8-1 and Table 8-1 above.
The procedure used to generate these estimates is
well-suited to analysis of uncertainties where the
probability of alternative outcomes can be quantita-
tively characterized in an objective manner. For
example, most studies that estimate concentration-
response relationships report an estimate of the sta-
tistical uncertainty around the central estimate. Be-
cause many estimates are available for the value of
statistical life, wre can use the discrete distribution of
the best available estimates as a basis for quantita-
tively characterizing the probability of alternative
values. It is important to recognize, however, that
this procedure reflects only a portion of the range of
possible sources of uncertainty in our benefits esti-
mates. Other, nonqualified sources of uncertainty
must also be factored into conclusions about the ra-
tio of benefits to costs.
As part of our analysis of key contributors to
uncertainty in benefits estimates, we also conducted
a sensitivity analysis to determine the physical ef-
fects estimation and economic valuation variables
with the greatest contribution to the quantified mea-
surement uncertainty range. We present the results
of this sensitivity analysis in Figure 8-2. In this sen-
Figure 8-2
Analysis of Contribution of Key Parameters to Quantified Uncertainty
o
Ov
Ov
$300
$250
S $200
¦B $150
a>
S
M $100
| $50
1 $0
H
P 95th %ile
Mean
<
) <
) <
>£«»»»
5th %ile
i
i i i i
e
0
1
o
U
•I? .2
I 1
tsi ^
0 o
3 2 |
1 ^ £
0 S|
03 >
9 O
a a a
P y
O
m
.1? O
.S3 ^
£ £
5: os a
9 £ %
3 ° S
107
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 8-5
Summary of Key Sources of Uncertainty and Their Impact on Costs and Benefits
Source of
Description of Alternative Parameter
Inputs
Impact on Annual Estimates in 2010
Uncertainty
Costs
Benefits
Measurement
error and
uncertainty in the
physical effects
and economic
valuation steps
Use a range of input assumptions to
reflect statistical measurement
uncertainty in concentration-response
functions, modeling of physical effects,
and estimation of economic values.
Most important input parameters are
value of statistical life and estimated
relationship between particulate matter
and premature mortality (see Chapters
5, 6, and 7).
None
For Titles I through V,
effect of the use of
alternative input
assumptions ranges
from a $84 billion
decrease (5th
percentile) to a $160
billion increase (95th
percentile).
Measurement
error and
uncertainty in
direct cost inputs
Use alternative assumptions for key
input parameters for six of the highest
cost provisions. Conduct sensitivity
tests for each provision separately (see
Chapter 3, pages 30 to 32). As
discussed in Chapter 3 and in this
chapter, aggregation of provision-
specific results would be inappropriate.
High estimates for
some provisions are
$1 billion higher
than primary
estimate. Low
estimates are as
much as $2 billion
below primary
estimate
None
Value of
statistical life-
based estimates
do not reflect
age at death
Use estimates of the incremental
number of life-years lost from exposure
to ambient PM and a value of statistical
life-year as opposed to measuring
number of lives lost and a value of
statistical life (see Chapters 5 and 6).
None
Decrease by $47
billion
Basis of estimate
of avoided
mortality from
PM exposure
The Dockery et al. study provides an
alternative estimate of the long-term
relationship between chronic PM
exposure and mortality (see Chapter 5).
None
Increase by $100 to
$150 billion
Uncertainties in
Title VI health
benefits analysis
Major uncertainties include: estimating
fatal cancer cases resulting from UV-b
exposure; not accounting for future
averting behavior; and not accounting
for future improvements in the early
detection and treatment of melanoma
(see Table 5-6).
None
Not quantified, but net
effect is probably that
benefits estimates are
too high.
Uncertainties in
emissions and
air quality steps
Major uncertainties include:
underestimation of PM2.5 emissions;
omission of changes in primary and
organic PM in eastern U.S.; emissions
estimation uncertainties in the western
U.S.; scarcity of PM25 monitors; and
lack of a fully integrated air quality and
emissions modeling system (see
Tables 2-5 and 4-7).
Uncertainties in
emissions estimates
affects some costs,
but net effect is
minor.
Not quantified, but net
effect is probably that
benefits estimates are
too low.
Omission of
potentially
important
benefits
categories from
primary estimate
Non-quantified categories of impacts
summarized in Chapters 5 and 7.
Quantified but omitted categories
include household soiling, nitrogen
deposition, and residential visibility (see
Chapter 7).
None
Increase by at least $8
billion, (does not
reflect unquantified
categories)
108
-------
Chapter 8: Comparison of Costs and Benefits
sitivity analysis, we hold constant all inputs to the
probabilistic uncertainty analysis except one — for
example, the economic valuation of mortality. We
allow that one variable to vary across the estimated
range of that variable's uncertainty. The sensitivity
analysis isolates the effect of this single source of un-
certainty on the total measured uncertainty in esti-
mated aggregate benefits. The first uncertainty bar
represents the range associated with the total mon-
etized benefits of the Clean Air Act, based on analy-
sis of quantifiable components of uncertainty, as
reported above. This range captures the multiple
measurement uncertainties in the quantified benefits
estimation. The rest of the uncertainty bars repre-
sent the quantified measurement uncertainty ranges
generated by single variables. As shown in Figure 8-
2, the most important contributors to aggregate quan-
tified measurement uncertainty are mortality valua-
tion and incidence, followed by chronic bronchitis
valuation and incidence.
Measurement Error and Uncertainty in
Direct Cost inputs
As noted in Chapter 3, explicit and implicit as-
sumptions about changes in consumption patterns,
input costs, and technological innovation are cru-
cial to estimating the direct compliance costs of the
CAAA. For many of the factors contributing to
uncertainty, the degree and, in some cases, the di-
rection of the bias are unknown or cannot be deter-
mined. Uncertainties and sensitivities can be identi-
fied, however, and in many cases the potential mea-
surement errors can be quantitatively characterized.
We designed our sensitivity analyses of key input
parameters to provide a sense of the relative impor-
tance of various input parameters and assumptions
necessary to generate estimates of direct costs. The
sensitivity tests use ranges of input parameters that
include all reasonable alternative estimates that we
could identify.
The results indicate that the sensitivity of our
primary central cost estimates is not uniform across
provisions. Tow and high estimates may vary by as
much as a factor of two. Unlike our quantitative
analysis of benefits, we do not assign probabilities
to the likelihood of alternative input parameters. In
our judgement, assignment of probabilities to these
alternative outcomes would be a largely subjective
task; we know of no objective means to develop these
probabilities. As a result, it would be inappropriate
simply to add up the array of low and the array of
high estimates to arrive at an overall range of uncer-
tainty around the central estimates, because it is un-
likely that a plausible scenario could be constructed
where all the estimates are concurrently either at the
high or low end of their individual plausible ranges.
A better interpretation of these results is that uncer-
tainty in key input parameters can have a significant
effect on the overall uncertainty of our estimates of
direct compliance costs and ultimately the net ben-
efits calculation.7
PM Mortality Valuation Based
on Life-Years Lost
The primary analytical results we present ear-
lier in this chapter assign the same economic value
to incidences of premature mortality regardless of
the age and health status of those affected. Although
this has been the traditional practice for benefit-cost
studies conducted within EPA, some argue this may
not be the most appropriate method for valuation
of premature mortality caused by PM exposure.
Some short-term PM exposure studies suggest that a
significantly disproportionate share of PM-related
premature mortality occurs among persons 65 years
of age or older. Combining standard life expectancy
tables with the limited available data on age-specific
incidence allows rough approximations of the num-
ber of life-years lost by those who die prematurely
as a result of exposure to PM or, alternatively, the
changes in life expectancy of those who are exposed
to PM.
The ability to estimate, however roughly,
changes in age-specific life expectancy raises the is-
sue of whether available measures of the economic
value of mortality risk reduction can, and should,
be adapted to measure the value of specific numbers
Although the analysis conducted here is a direct cost analy-
sis, other sources of uncertainty would also need to be consid-
ered for a social cost analysis. For example, forecasts of key
economic variables (e.g., interest rates), specification of produc-
tion functions, and the reliability of key supply and demand
elasticities are all important factors in social cost modeling that
contribute to measurement uncertainty, in addition, most cur-
rent social cost analyses assume that markets are currently oper-
ating under optimally efficient conditions. Emerging literature
suggests that a full accounting of the social costs and efficiency
impacts of environmental regulations could also include an as-
sessment of the incremental costs that reflect existing market
distortions, such as those imposed by the current tax code. Our
assessment of uncertainties in direct cost estimates do not re-
flect these considerations.
109
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
of life-years saved.8 As stated 111 our retrospective
analysis, we have on occasion performed sensitivity
calculations that adjust mortality values for those
over age 65. Nonetheless, as discussed in Appendix
H, the current state of knowledge and available ana-
lytical tools do not conclusively support using a life-
years lost approach or any other approach which
assigns different risk reduction values to people of
different ages or circumstances. While we prefer an
approach which makes no valuation distinctions
based on age or other characteristics of the affected
population, we present alternative results based on
a VSLY approach below. The method used to de-
velop life years lost estimates is described briefly in
Chapter 5 and Appendix D. The method used to
develop VSLY estimates is described in Appendix
H.
The fourth row of Table 8-5 summarizes the ef-
fect of using a VSLY approach on results for 2010.
The results indicate that the choice of valuation meth-
odology significantly affects the estimate of the mon-
etized value of reductions in air pollution-related pre-
mature mortality. However, the downward adjust-
ment which would result from applying a VSLY ap-
proach in lieu of a YSL approach does not change
the basic conclusion of this study7, since the central
estimate of monetized benefits of the CAAA still
substantially exceeds the costs of compliance.
We emphasize that the results of the VSLY ap-
proach to valuing avoided mortality benefits repre-
sent a crude estimate of the value of changes in age-
specific life expectancy. These results should be in-
terpreted cautiously, due to the several significant
assumptions required to generate a monetized esti-
mate of life years lost from the relative risks reported
in the Pope et al., 1995 study and the available eco-
nomic literature. These assumptions include, but
are not limited to: extrapolation of the age distribu-
tion of the U.S. population in future years; assump-
tions about the age-specificity of the relative risk
reported by Pope et al., 1995; assumptions about the
life expectancy of different age groups, adjustment
8 This issue was extensively discussed during the Science
Advisory Board Council review of drafts of the retrospective
study. The Council suggested it would be reasonable and ap-
propriate to show PM mortality benefit estimates based on value
of statistical life-years (VSLY) saved as well as the value of statis-
tical life (YSL) approach traditionally applied by the Agency to
all incidences of pretrial lire morlalily. Consistent with SAB
Council review advice for the present study, we apply the same
approach in lids analysis.
of the life years lost estimates by an appropriate lag
period (if any); assumptions about the age-specific -
ity of the lag period (if any); derivation of VSLY
estimates from VSL estimates; assumptions about the
variation in VSLY with age; and selection of an ap-
propriate rate at which to discount the lagged esti-
mates of life years lost. Changes in any of these
assumptions could significantly affect the VSLY ben-
efit estimate. For example, if we were to assume no
lag period for PM-related mortality effects instead
of the five-year lag structure described in Chapter 5,
VSLY benefit estimates would increase from $53
billion to |61 billion. The specific assumptions we
used in generating these results are discussed in Ap-
pendix H.
PM Mortality Incidence Using
the Dockery Study
As described in Chapter 5, we chose to use the
results of the Pope et al. (1995) study to estimate the
magnitude of the effect of ambient PM exposure on
the incidence of premature mortality. Alternative
estimates do exist in the literature, however. Al-
though we chose the Pope study because of its cov-
erage of the largest number of cities and other tech-
nical advantages, the Dockery et al. (1993) study
provides a credible and reasonable alternative to the
Pope study. The Dockery study used a smaller
sample of individuals in fewer U.S. cities than the
Pope study, but it features improved exposure esti-
mates, a slightly broader study population (includ-
ing adults aged 25 to 30), and a follow-up period
nearly twice as long as that used in the Pope study.
Use of the Dockery study in place of the Pope
study would substantially increase the benefits esti-
mate. As shown in the fifth row of Table 8-5, we
estimate that using the Dockery study estimates
would increase the annual central benefits estimate
by $100 to $150 billion, more than doubling the to-
tal annual benefits for Titles 1 through V and, in
turn, doubling the estimated benefit-cost ratio.
Uncertainties in Title VI
Health Benefits Analysis
As discussed in Chapter 5 and Appendix G,
health benefits such as avoided mortality from mela-
noma and non-melanoma skin cancers constitute the
majority of monetized benefits resulting from Title
110
-------
Chapter 8: Comparison of Costs and Benefits
VI regulations 011 stratospheric ozone-depleting
chemicals. Estimates of avoided mortality from skin
cancer due to reduced UV-b exposure between 1990
and 2165 represent over 90 percent of the total health
benefits of Title VI. As a result, uncertainties re-
lated to avoided mortality estimation under Title VI
represent key uncertainties for our overall CAAA
benefits estimate. Three main areas of uncertainty
are important for our avoided mortality estimates
for Title VI: dose-response relationships; predicting
averting behavior; and predicting future medical
advancements.
Because the literature on the relationship be-
tween exposure to ultraviolet rays and melanoma
and non-melanoma mortality is not as well devel-
oped as that for other health effects, the dose-response
functions for both of these endpoints are character-
ized by significant uncertainty. The association of
UV-b exposure with melanoma is controversial, al-
though studies suggest that sunlight exposure is a
major environmental risk factor for melanoma. If
one assumes that a causal relationship exists between
UV-b rays and melanoma, uncertainty still remains
about three aspects of the nature of the dose-response
relationship. Specifically, the relative contribution
of different wavelengths of light to melanoma de-
velopment, the critical exposure period (e.g., acute,
intermittent, or chronic), and the existence (and
length) of a latency period between U V exposure
and disease are all unclear. The effect of the first
two uncertainties on our results cannot be deter-
mined from available information. If a significant
latency period exists, then the third uncertainty may-
indicate that our analysis, which does not include a
latency period, overestimates avoided melanoma
mortality benefits. Because limited data on non-
melanoma mortality precluded the development of
a dose-response function for this endpoint 111 the
current analysis, our estimate of non-melanoma skin
cancer mortality resulting from UV-b exposure is
calculated indirectly, by assuming the mortality rate
is a fixed percentage of non-melanoma incidence.
New data on the death rate for non-melanoma skin
cancer may significantly influence this mortality es-
timate.
Our analysis of avoided mortality also does not
incorporate adjustments for future increases 111 avert-
ing behavior (i.e., efforts by individuals to protect
themselves from UV-b radiation ). Our estimates
rely on epidemiological studies that incorporate
averting behavior as currently practiced. However,
if people would react to increased skin cancer risk 111
the future by applying sun screen more frequently,
spending more time indoors or otherwise reducing
their UV-b exposure, then our estimate of avoided
mortality would significantly overestimate Title VI
benefits. It is not certain, though, that individuals
will pursue such behavior, and studies show that
those engaging 111 averting behavior may also alter
their behavior 111 ways that may increase exposure
or risk, counteracting the benefits of averting be-
havior. For example, a recent study of young Euro-
peans by Autier et al. (1999) found that the use of
high sun protection factor (SPF) sun screen is associ-
ated with increased frequency and duration of sun
exposure.
Finally, our analysis does not adjust estimates of
future mortality for possible advances in medical
technology that could lead to earlier detection and
more effective treatment of melanomas. Such ad-
vancements could significantly reduce the expected
future melanoma mortality, and by not adjusting for
such developments, we may be overestimating
avoided melanoma mortality. However, future re-
search may also identify additional adverse human
health outcomes associated with UV exposure that
we have not considered in this analysis, resulting in
an underestimate of Title VI benefits.
Uncertainties in Emissions
and Air Quality Steps
The emissions estimates presented in this analy-
sis are a critical component of the overall analysis.
As the starting point for both costs and benefits, they
provide a consistent basis for evaluating the economic
efficiency of the CAAA. Characterizing emissions
can be very difficult, however, particularly for those
source categories where emissions monitoring data
are sparse or nonexistent. In general, all our emis-
sions estimates are affected by three major sources
of uncertainty: estimation of the base-year inven-
tory, prediction of the growth 111 pollution-generat-
ing activity, and assumptions about future-year con-
trols.
Base-year emissions were estimated using emis-
sions factors that express the relationship between a
particular human/industrial activity and the level of
111
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
emissions. The accuracy of base-year emissions esti-
mates varies from pollutant to pollutant, depending
largely on how directly the selected activity and
emissions correlate. We likely estimated 1990 SO„
emissions with the greatest precision. Sulfur diox-
ide emissions are generated during combustion of
sulfur-containing fuel and are directly related to fuel
sulfur content. In addition, we were able to verify
these estimates through comparison with Continu-
ous Emission Monitoring (CEM) data. As a result,
we were able to accurately estimate SO, emissions
using emissions factors based on data on fuel usage
and fuel sulfur content. Nitrogen oxides are also a
product of fuel combustion, allowing us to estimate
emissions of this pollutant using the same general
technique used to estimate SO, emissions. However,
the processes involved in the formation of NO
during combustion are more complicated than those
involved in the formation of SO,; thus, our NOx
emissions estimates arc more variable and less cer-
tain than SO, estimates.
Volatile organic compounds, like SO, and NOx,
are products of fuel combustion; however, these
compounds are also a product of evaporation, 'l'o
estimate evaporative emissions of this pollutant we
used emissions factors that relate changes in emis-
sions to changes in temperature. Because future
meteorological conditions are difficult to predict,
the uncertainty associated with forecasting tempera-
ture influences the uncertainty in our VOC emis-
sions estimates. The likely significance of this un-
certainty, in terms of its impact on the overall mon-
etary benefit present in this analysis, is probably
minor.
Of particular importance, however, are uncer-
tainties that affect the estimation of future year emis-
sions of particulate matter and secondarily formed
PM precursors. In this analysis we estimated primary
PM 5 emissions based on unit emissions that may
not accurately reflect the composition and mobility
of particles. The ratio of crustal to carbonaceous
particulate material, for example, likely is high as a
result of overestimation of the fraction of crustal
material, primarily composed of fugitive dust, and
underestimation of the fraction of carbonaceous
material. Because the CAAA have a greater impact
on emissions sources that generate carbonaceous par-
ticles (mobile sources) than on sources that mainly
emit crustal material (area sources), we likely under-
estimate the impact of the CAAA on reducing PM,
thereby reducing monetary benefits estimates. The
uncertainty associated with estimating the partition
of PM emissions components could conceivably
have a major impact on the net benefit estimate.
Compared to secondary PM precursor emissions,
however, changes in primary PM emissions have a
relatively small impact on PM, related benefits.
Our future-year control assumptions are also a
source of uncertainty. Despite our efforts to mini-
mize this uncertainty, whether each of the Post-
CAAA controls will be adopted, whether Post-
CAAA control programs will be more or less effec-
tive than estimated, and whether unanticipated tech-
nological shifts will reduce future-year emissions are
all unknown. For example, the Post-CAAA scenario
includes implementation of a region-wide NO con-
trol strategy designed to regulate the regional trans-
port of ozone. However, the control program as-
sumed under the Post-CAAA scenario may not re-
flect the NO controls that are actually implemented
in a regional ozone transport rule.
In addition to potential inaccuracies in the emis-
sions inventories used as air quality modeling inputs,
there are at least three sources of air quality model-
ing uncertainty that may have a major effect on the
precision and accuracy of our projected changes in
air quality. First, we estimate changes in PM con-
centrations in the eastern U.S. based exclusively on
changes in the concentrations of sulfate and nitrate
particles. By not accounting for changes in organic
and primary particulate fractions, we likely under-
estimate the impact of the CAAA on PM concen-
trations. Second, by using separate air quality mod-
els for individual pollutants and different geographic
regions, as opposed to a single integrated model, we
were unable to fully capture the interaction among
air pollutants or reflect transport of pollutants or
precursors across the boundaries of the models cov-
ering the western and eastern states. Third, the lack
of a well-developed modeling network for PM, 5
means we must estimate monitored concentrations
of this pollutant based on PM monitor estimates.
The direction and magnitude of bias these limita-
tions impose on net benefits estimate presented in
this analysis can not be determined based on current
information.
Some model-related uncertainties, however, may
be mitigated because this analysis uses the air qual-
112
-------
Chapter 8: Comparison of Costs and Benefits
ity modeling results in a relative, not absolute, sense.
We focus on the change in air quality between the
Pre- and Post-CAAA scenarios and not on the am-
bient concentrations projected by the individual
models themselves. Therefore, uncertainties that
affect a model's ability to accurately predict the rela-
tive change in concentration of a pollutant from one
scenario to another are more important in the con-
text of this study than those that affect only the ab-
solute model results. In addition, as summarized in
the previous chapters, most of the uncertainties in
emissions estimation and air quality modeling con-
tribute to a conservative bras 111 our benefits results.
When faced with alternative approaches to emissions
and air quality modeling, we made explicit attempts
to choose parameters, assumptions and modeling-
strategies that would tend to understate benefits.
Omission of Potentially Important
Benefits Categories
As described in Chapters 5 through 7 above, and
in more detail in Appendix H, the primary estimate
reflects application of a strict set of criteria for inclu-
sion of monetized benefits categories. For example,
estimates of the value of improved visibility in U.S.
residential areas indicate a positive value for this ser-
vice flow, but the best available residential visibility
estimates rely on an unpublished study of values in
the eastern U.S. Although our physical effects analy-
sis indicates significant visibility improvements in
all regions of the U.S., our application of the results
of the economic valuation literature reflect a con-
servative approach to valuation of improved visibil-
ity in the U.S. While we believe our conservative
inclusion criteria for the primary benefits reflects
the greater uncertainty in measuring some economic
values, we also believe that the statutory language of
section 812 clearly warns against the practice of as-
suming a default value of zero for demonstrated cat-
egories of benefits. Therefore, the last row of Table
8-5 presents the effect of using a somewhat more
inclusive set of criteria for accepting benefits trans-
fer-based economic values. In this alternative case,
we included estimates for improved residential vis-
ibility, displaced costs from reduced airborne nitro-
gen loadings to estuaries, and reduced expenditures
for household soiling (which are not included in any
form in the primary estimate).
In addition to these quantified but omitted cat-
egories of benefits, there is a wide range of benefits
of the CAAA that we can identify but cannot quan-
tify. We present summaries of unquantified health
effects in Chapter 5 (Tables 5-1 and 5-5) and
unquantified ecological and welfare effects in Chap-
ter 7 (Tables 7-5 and 7-9). Two of the most impor-
tant omissions, in our judgement, are the lack of any
quantified estimates for the health benefits of air
toxics control and the omission of the systemic and
long-term ecological effects of mercury and other
persistent air pollutants. The importance of these
two categories of effects are discussed in Chapters 5
and 7, respectively.
Alternative Discount Rates
In some instances, the choice of discount rate
can have an important effect on the results of a ben-
efit-cost analysis; for example, when the distribution
of costs and benefits throughout the tune period are
very different from one another. In this assessment,
the discount rate affects annualized costs (i.e., amor-
tized capital expenditures), and the discounting of
all costs and benefits to 1990. Table 8-6 summarizes
the effect of alternative discount rates on the Pri-
mary Central estimate results of this analysis. The
estimates we present show that altering the discount
rate has only a small effect on annual cost and ben-
efit estimates. In part, this is due to limitations in
our ability to conclusively identify costs as annual-
ized capital expenditures or annual operating costs
in the underlying estimates. As described in Chap-
ter 3, about $3 billion (or roughly 10 percent) of the
2010 estimate is annualized capital costs. Varying
the discount rate, which we also use to represent the
cost of capital, affects only this component of costs.
The benefits estimates that employ a discount rate
include the mortality estimate, where it is used as
part of our valuation of the lag effect of PM mortal-
ity, and the chronic asthma value, where we use a
discount rate to develop a lump-sum value for avoid-
ance of incidence from an annual payment value in
the underlying literature.
Not surprisingly, the effect of discount rates on
the net present value benefit calculations is greater.
Nonetheless, the estimates we present in Table 8-6
show that varying the discount rate assumption also
does not change our overall conclusion that the ben-
efits of the CAAA exceed its costs.
113
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table 8-6
Effect of Alternative Discount Rates on Primary Central Estimates
(Estimates in million 1990$)
Discount Rate Assumption
3%
5%
7%
Annual Costs in 2010:
Titles I through V
$26,600
$26,800
$26,900
Annua! Benefits:
Titles I through V
$110,000
$110,000
$107,000
Present Value of Costs:
Titles I through V
$230,000
$180,000
$140,000
Title VI
$43,000
$27,000
$20,000
Present Value of Benefits:
Titles I through V
$890,000
$690,000
$520,000
Title VI
$1,900,000
$530,000
$240,000
Cumulative Net Benefits:
Titles I through V
$650,000
$510,000
$380,000
Title VI
$1,860,000
$500,000
$220,000
Benefit/Cost Ratio:
Titles I through V
4/1
4/1
4/1
Title VI
44/1
20/1
12/1
114
-------
Valuation of Human Health
and Welfare Effects of
Criteria Pollutants
This appendix describes the derivations of the
economic valuations for health and welfare endpoints
considered in the benefits analysis. It includes three
primary sections. First, we introduce the method for
monetizing improvements in health and welfare.
Second, we summarize dollar estimates used to value
benefits and outline the derivation of each estimate.
Valuation estimates were obtained from the literature
and reported in dollars per case avoided for health
effects, and dollars per unit of avoided damage for
welfare effects. Economic valuations are characterized
in terms of a central (point) estimate as well as a
probability distribution which reflects the uncertainty
around the central estimate. Third, we present the
results of the economic benefits analysis. All dollar
values are in 1990 dollars. This third section
concludes with an exploration of the uncertainties in
valuing the benefits attributable to the CAAA.
Methods Used to Value Health
And Welfare Effects
The general approach to benefits analysis involves
a three-step process— (i) identification of potential
physical effects (i.e., individual health and welfare
endpoints); (ii) quantification of significant endpoints;
and (iii) monetization of benefits. The first two steps,
identification and quantification of physical effects, are
described in Appendix D, Human Health and Welfare
Effects of Criteria Pollutants. The third step is
detailed in this appendix. Monetization of benefits
attributed to the CAAA involves applying dollar
estimates obtained from economic literature to
individual health and welfare endpoints relevant for
the 812 prospective analysis. As context to
understanding the methodology for transferring
estimated values of physical effects, this section
provides a brief discussion of the theoretical economic
foundation of, and general approach to, valuing the
benefits of improved air quality.
Economists equate the dollar value of a benefit to
the level of well-being an individual enjoys from the
provision or consumption of a particular good or
composite good (i.e., bundle or mix of goods). A
fundamental assumption in economic theory is that
individuals can trade between different consumption
levels of these goods, services, or money, and
maintain the same level of welfare. Typically, this
willingness to trade-off between goods is measured as
willingness to pay (WTP) or willingness to accept
compensation (WTA). These measures are essentially
dollar equivalents to changes in the level of
consumption of a good or service so that the
individual maintains the same level of well-being. In
other words, the individual is indifferent between his
or her current bundle of goods and the alternative
bundle of goods.
While WTP and WTA represent an individual's
own assessment of the dollar value of better health,
they are not necessarily equivalent measures.1 WTP,
in the case of health, is the largest amount of money
a person would pay to obtain an improvement (or
avoid a decline) in health. When faced with two
1 The measures differ for several reasons. For example the
measures have different points of reference from which to
evaluate changes in welfare. WTP's reference point is the level of
utility without the improvement. WTA's reference point is the
level of utility with the improvement. Moreover, the measures have
different upper bound constraints. WTP measures what a person
would pay to obtain better health and is bound by the person's
wealth and income. WTA, on the other hand, measures what a
person must be paid to forego better health. WTA does not have
an upper bound, but it must be at least as large as WTP.
Economists, however, do not expect significant differences
between WTP and WTA when the dollar amounts are small
relative to the individual's wealth and income.
H-1
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
options, to either (1) pay a certain dollar amount to
enjoy the health improvement or (2) abstain from
paying the dollar amount and not experience the
health improvement, the individual feels either choice
provides the same degree of well-being. Alternatively,
willingness to accept compensation (WTA) is the
smallest amount of money a person would voluntarily
accept as compensation to forego an improvement, or
endure a decline, in health. The individual feels that to
accept the payment and not experience the health
improvement or refuse the compensation and
experience improved health will provide the same
degree of well-being. In practice, WTP is generally
used to value benefits because it is often easier to
measure and quantify.2 In this report, we refer to all
valuation estimates as WTP values, even though the
underlying economic valuation literature is based on
studies which elicited expressions of WTP and/or
WTA.3
In the context of cost-benefit analysis, WTP is
useful for estimating the monetary value of non-
market, public goods. A major characteristic of public
goods is that they are nonrival (i.e., one person's
consumption of the good does not reduce the amount
available to others). In the case of health-related
improvements due to environmental quality, the
benefits are also nonexclusive. Benefits are not (and
to some extent, cannot be) regulated. As a result, the
benefits are actually reductions in the probabilities or
risk of enduring certain health
2It is worth noting that the appropriateness of either WTP or
WTA also depends on property rights. In the case of a policy
aimed at reducing existing pollution levels, a WTP measure
implicidy assumes that the property rights rest with the polluting
firm. Alternatively, WTA measures implicitly assume that the
property rights rest with the public. (Carson and Mitchell, 1993.)
3In some cases (e.g., hospital admissions), neither WTA nor
WTP estimates are available. In those cases, cost of illness (COI)
estimates are applied in lieu of WTP values. COI estimates
understate the true welfare change since important value
components (e.g., pain and suffering associated with the health
effect) are not reflected in the out-of-pocket costs for the hospital
stay.
problems. In theory, the total social value associated
with the decrease in risk is
N
V (number of units of risk reduction ) * (WTP per unit risk reduction )
where (number of units of risk reduction); is the
number of units of risk reduction conferred on the ith
exposed individual as a result of the pollution
reduction, (WTP per unit risk reduction); is the ith
individual's willingness to pay for a unit risk reduction,
and N is the number of exposed individuals. The
units are in terms of cases reduced per unit of time
(usually one year).
Using mortality risk as an example, suppose that
a given reduction in PM concentrations results in
lowering the risk of death by 1/10,000 per year. Then
for every 10,000 individuals, one less death would be
expected if ambient PM concentrations are reduced.
If an individual's WTP for this 1/10,000 decrease in
mortality risk is $500 (assuming, for now, that all
individuals' WTPs are the same), then the value of a
statistical life is 10,000 x $500, or $5 million.
While the estimation of WTP for a market good
(i.e., the estimation of a demand schedule) is not a
simple matter, the estimation of WTP for a
nonmarket good, such as a decrease in the risk of
having a particular health problem, is substantially
more difficult. Estimation of WTP for decreases in
very specific health risks (e.g., WTP to decrease the
risk of a day of coughing or WTP to decrease the risk
of admission to the hospital for a respiratory illness)
is further complicated by several factors, such as
wealth, income, age, pre-existing health impairments,
or other personal characteristics. There are many
policy contexts where distinguishing among WTP
estimates based on categorical differences (e.g.,
distinguishing between WTP of a low-income group
and a high-income group) is controversial. Given the
consideration of these influencing factors and the
limitations on information available for developing
WTP estimates, EPA sought to develop the most
appropriate and accurate estimates possible.
Derivations of the dollar value estimates for this study
are discussed below.
H-2
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Valuation of Specific Health
Endpoints
Since the Section 812 CAA retrospective analysis
(U.S. EPA 1997), there have been significant
advances made in economic valuation methodologies
for both mortality and morbidity effects. Much of the
literature presents emerging new approaches for
characterizing the effects of potentially important
determinants of WTP, such as age, income, risk
perception, and current health status. Despite this
progress, many of the more recent studies test
techniques that are in the development stage and use
data from work reviewed and incorporated in the
Section 812 retrospective analysis. This section
reviews the sources and methodology used to derive
WTP estimates for premature mortality and a variety
of morbidity effects valued in the present study. In
addition, there are brief discussions of more recent
advances relevant to particular endpoints.
Valuation of Premature Mortality
Avoided
The economic benefits associated with premature
mortality were the largest category of monetized
benefits in the Section 812 CAA retrospective analysis
(U.S. EPA 1997).4 In addition, EPA identified
valuation of mortality benefits as the largest
contributor to the range of uncertainty in monetized
benefits. Because of the uncertainty in estimates of
the value of premature mortality avoidance, it is
important to adequately characterize and understand
the various types of economic approaches available
for mortality valuation. Such an assessment also
requires an understanding of how alternative valuation
4As noted in the methods section, it is actually reductions in
mortality risk that are valued in a monetized benefit analysis.
Individual WTPs for small reductions in mortality risk are summed
over enough individuals to infer the value of a statistical life saved.
This is different from the value of a particular, identified life saved.
The "value of a premature death avoided," then, should be
understood as shorthand for "the value of a statistical premature
death avoided."
approaches reflect that some individuals may be more
susceptible to air pollution-induced mortality.
The health science literature on air pollution
indicates that several human characteristics affect the
degree to which mortality risk affects an individual.
For example, some age groups are more susceptible to
air pollution than others (e.g., the elderly and
children). Health status prior to exposure also affects
susceptibility — at risk individuals include those who
have suffered strokes or are suffering from
cardiovascular disease and angina (Rowlatt, et al.
1998).
To reflect the full range of knowledge of air
pollution-induced mortality, an ideal estimate of
mortality risk reduction benefits would be an ex ante
willingness to pay (WTP) to improve one's own
chances of survival plus WTP to improve other
individuals' survival rates.5 The measure would take
into account the specific nature of the risk reduction
commodity that is provided to individuals, as well as
the context in which risk is reduced. To measure this
value, it is important to assess how reductions in air
pollution reduce the risk of dying from the time that
reductions take effect onward, and how individuals
value these changes. Each individual's survival curve,
or the probability of surviving beyond a given age,
should shift as a result of an environmental quality
improvement. That is, changing the current
probability of survival for an individual also shifts
future probabilities of that individual's survival. This
probability shift will differ across individuals because
survival curves are dependent on such characteristics
as age, health state, and the current age to which the
individual is likely to survive. For example, Figure H-
1 illustrates how a risk reduction may change a
survival curve for a given population. In this figure,
the solid line shows a survival curve for white males,
from California 1980 life tables (adapted from Selvin,
1996), up to age 80. The dashed line shows that the
probability of survival beyond a given age increases
with a reduction in mortality risk.
5 For a more detailed discussion of altruistic values related to
the value of life, see Jones-Lee (1992).
H-3
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
While the change in a survival curve represents a
cumulative effect of a change in risk over time, the
annual change in risk of death represents a static effect
of a risk reduction. As discussed in Appendix D in
greater detail, the instantaneous risk of death at a
specific age is often used to illustrate the effects of
changes in risk. The annual risk of death is related to
die probability of survival in that it represents the rate
at which die survival probability changes at any given
age, divided by die probability of surviving beyond
that age. Figure H-2 shows how a constant risk
reduction reduces annual risk of deadi across various
age cohorts. The baseline risk of deadi increases with
each cohort (solid line). As a result, die reduction in
risk (in this hypothetical example a constant 25
percent reduction) lowers each cohorts' risk level at a
different rate. The elderly experience a greater
reduction in risk than younger cohorts as can be seen
by the increasing difference between the solid and
dashed line. It is important to note that diis example
shows die effect of a uniform risk reduction, and air
pollution controls may have risk reduction effects that
vary across age cohorts.
An alternative way to view the age-dependent
effect of risk reduction is to consider changes in the
cumulative effect of risk as measured by changes in
remaining life expectancy. Remaining life expectancy
is measured as the average number of additional years
expected to be lived by those individuals alive at a
given age, and derives from die area under die survival
curve at any given age. The age-dependent effects of
a hypodietical change in risk are portrayed in Figure
H-3. Consider the effect of risk reduction on two
cohorts, aged 10 years apart. When each cohort was
at age 40 both had the same life expectancy shown in
Figure H-3 as point A'. Given a risk reduction in the
future that occurs when one cohort is at age 60 and
the other at age 70, the life expectancy of the 60 year
old increases by the amount A'B', and the life
Figure H-1
Hypothetical Survival Curve Shift
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Baseline survival curve
0.2
Survival curve post-risk reduction
0.1
0.0
0
20
40
60
80
H-4
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Figure H-2
Change in 1990 Annual Risk of Death by 25 Percent
3
2.5
2
1.5
1
0.5
0
40
Age
Figure H-3
Increase in 1990 Remaining Life Expectancy
40
35
30
l 25
c
CD
x 20
LU
B"
_i
ra
£ 15
c
1
A"
£C
10
5
0
40
Age
H-5
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
expectancy of the 70 year old increases by the amount
A"B". The change in life expectancy is greater for
the younger cohort than the older cohort because
these measures represent a cumulative accrual of
increased life expectancy (i.e., the younger cohort will
benefit from the lower risk environment for more
years).
Because the risk reduction results in various
changes in risk levels, individual values for risk
reduction are likely to vary as well. Some individuals
having a greater change in risk, and hence life
expectancy, may have different values for the change
than those individuals experiencing a smaller change
in risk. Note that future generations may hold values
for health as well. Cropper and Sussman (1990)
develop theoretical models formalizing these concepts
when investigating how an individual's values for
reduction of a future risk to oneself and to future
generations should be discounted to the present.
While these theoretical models reflect the types of
values necessary to estimate the impact of the CAAA,
they are difficult to implement. First, they require an
estimate of individuals' survival curves. In order to
develop these survival probabilities, it is necessary to
characterize the dose/response relationship for the
regulated pollutants and know how this information
varies with age and health states over time. Second, it
is necessary to estimate values for risk reductions,
considering the key dimensions in which risk and
valuation of risk reduction may vary (e.g., with age and
health state).
Mortality Valuation Methodologies
This section summarizes alternative approaches to
mortality risk valuation, and outlines the approach
used to measure the economic value of these types of
benefits for air pollution reductions associated with
the CAAA. The first part provides background on the
methods that individuals have developed to estimate
the value of risk reduction benefits, including
commonly-applied approaches to valuation as well as
approaches that are beginning to be established in the
risk valuation literature. The second part discusses the
appropriateness of using these methodologies for
assessing the economic value of mortality benefits
associated with air pollution reduction. The Agency
has concluded that recent advances in the literature
show promise in incorporating several of the factors
that are likely to influence value, but problems with
the methodological approaches and lack of data
needed to reliably to appropriately estimate values
with the newer models leads us to adopt a value of
statistical life approach for the primary estimate of air
pollution-related mortality benefits.
Commonly Applied Approaches
The preferred approach researchers have taken to
estimate values for avoiding premature mortality is
based on individual WTP for risk reduction.
Although some cost-benefit analyses have based
values on avoided lost earnings (i.e., the human capital
approach), the WTP approach is preferred because it
more closely conforms to economic theory.6 The
common WTP measures of the value of life-saving
programs include the value of statistical life (VSL) and
the value of a statistical life year (VSLY). Newer
approaches to estimate values incorporate changes in
life expectancy, risk of dying, life-days per person, and
age-specific preferences. This section describes these
approaches and discusses issues that arise in their
application to estimate the value of mortality risk
reduction benefits.
The most commonly applied approaches for
mortality valuation are the value of statistical life and
value of statistical life year. Both of these approaches
6 In a recent article by Ireland and Gilbert (1998), the authors
evaluate value of life estimates used in tort recovery cases. The
article discusses the concept that for an individual there can be
finite utility (or determined value) to life and at the same time no
monetary equivalent. The authors do, however, build on this
argument to demonstrate that existing value of life estimates are in
fact lower bounds to the true value. By "lower bound," the
authors refer to a value representative of a specific individual, not
of a statistical life. In citing a reasonable value of life range, they
use a range similar to that of the 812 retrospective analysis,
although the authors do not cite the source of this range. Ireland
and Gilbert write, "A decedent has lost something of immense
value, for which estimates in the $4-$6 million range is clearly a
low market value estimate".
H-6
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
directly address the value of premature death and
health impairment. The YSL method measures the
value of a given reduction in risk and an individual's
WTP to reduce that risk, relying on wage and
occupational risk tradeoff data or the results of
contingent valuation surveys. Individual WTP
amounts for small reductions in mortality risk are
"standardized" to reflect reduction of population risk
of one statistical life saved. The result of applying this
method is not the value of an identifiable life, but
instead the value of reducing fatal risks in a population
(Viscusi 1992).
Yiscusi (1992) summarizes the value of life
literature, including almost forty studies providing
VSL estimates relevant for policy application. For the
section 812 retrospective analysis, EPA identified 26
studies from that review that reflect the application of
the most sound and defensible methodological
elements (see Table H-l). Five of the 26 studies are
contingent valuation (CV) studies, which directly
solicit WTP information from subjects; the rest are
wage-risk studies, which base WTP on estimates of
the additional compensation demanded in the labor
market for riskier jobs. Using a Weibull distribution
to describe the distribution of the mean mortality risk
valuation estimates from these studies, the mean
estimate of the distribution is $4.8 million with a
standard deviation of $3.2 million (1990$).
Since EPA's retrospective analysis, Desvousges et
al. (1998) has conducted a meta-analysis of twenty-
nine mortality studies presented in Yiscusi (1993) and
Fisher, Chestnut, and Yiolette (1989).7 Desvousges et
al.'smeta-analysis yields $3.3. million (1990 dollars) as
a value of statistical life, with a 90 percent confidence
interval between $0.4 and $6.3 million.8 Their
estimate, $3.3 million, falls well within the range
generated by EPA's uncertainty analysis of YSL
estimates. The selection of studies accounts for much
of the difference between their analysis and EPA's.
The Desvousges et al. analysis includes thirteen studies
that EPA did not use and EPA includes ten studies
omitted by Desvousges et al.
1 In addition to the Viscusi (1993) study, the 812 retrospective
examined two other studies, Miller et al. (1990) and the Fisher,
Chestnut, and Violette (1989). We opted to not use the Miller et
al. study given our concerns regarding the appropriateness of the
selection of studies for valuing reductions in environment-related
mortality risk and concerns about the adjustments made to the
underlying data. The Fisher, Chestnut, and Violette (1989) study
was not used because the data was not as current or
comprehensive as the data in the Viscusi study.
8 Desvousges etal do not adjust the value of statistical life to
account for age differences. They do note that a single estimate
for the value of statistical life may not be a good representation of
the differences between willingness-to-pay of the elderly and
young, healthy workers. They state that Moore and Viscusi (1988)
demonstrate that willingness-to-pay is higher for people with more
life years to lose while Desvousges et al. (1996) and Johnson et al.
(1998) indicate that willingness-to-pay is lower for people with
limited abilities to engage in activities and care for themselves.
H-7
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table H-1
Summary of Mortality Valuation Estimates
Study Valuation (millions 1990$)
Kneisner and Leeth (1991) (US)
Labor Market
0.6
Smith and Gilbert (1984)
Labor Market
0.7
Dillingham (1985)
Labor Market
0.9
Butler (1983)
Labor Market
1.1
Miller and Guria (1991)
Cont. Value
1.2
Moore and Viscusi (1988a)
Labor Market
2.5
Viscusi, Magat, and Huber (1991b)
Cont. Value
2.7
Gegax et al. (1985)
Cont. Value
3.3
Marin and Psacharopoulos (1982)
Labor Market
2.8
Kneisner and Leeth (1991) (Australia)
Labor Market
3.3
Gerking, de Haan, and Schulze (1988)
Cont. Value
3.4
Cousineau, Lacroix, and Girard (1988)
Labor Market
3.6
Jones-Lee (1989)
Cont. Value
3.8
Dillingham (1985)
Labor Market
3.9
Viscusi (1978, 1979)
Labor Market
4.1
R.S. Smith (1976)
Labor Market
4.6
V.K. Smith (1976)
Labor Market
4.7
Olson (1981)
Labor Market
5.2
Viscusi (1981)
Labor Market
6.5
R.S. Smith (1974)
Labor Market
7.2
Moore and Viscusi (1988a)
Labor Market
7.3
Kneisner and Leeth (1991) (Japan)
Labor Market
7.6
Herzog and Schlottman (1987)
Labor Market
9.1
Leigh and Folson (1984)
Labor Market
9.7
Leigh (1987)
Labor Market
10.4
Garen (1988)
Labor Market
13.5
SOURCE: Viscusi, 1992 and EPA analysis.
When applying YSL estimates to estimate
mortality benefits, it is important to determine the
differences between the nature of air pollution risk
and risks faced by persons whose risk-dollar tradeoff
decisions have been addressed in the literature. First,
several studies indicate that the value people place on
mortality risk reduction may depend on the nature of
the risk (e.g., Fisher et al. 1989; Beggs 1984). Current
YSL estimates do not account for a number of the
important factors that affect risk perception. For
example, premature mortality risks from air pollution
are experienced on an involuntary basis and are
generally uncompensated, while job-related risks are
assumed by individuals who presumably have some
choice as to occupation and are compensated for
taking a riskier job. Second, the demographics of the
population at risk from air pollution, particularly in
H-8
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
terms of age, income, and health state, may differ
from the demographics of individuals surveyed in the
literature. For a more detailed discussion of how
these factors can affect the economic valuation of
premature mortality, and specifically estimates derived
from the YSL approach, see the discussion, "Benefits
Transfer and YSL," presented in the section titled,
"Uncertainties in the Valuation Estimates."
The YSLY method values life-years that would be
lost if an individual were to die prematurely. Most
commonly, VSLY estimates are an annualized
equivalent of YSL estimates (Moore and Viscusi 1988,
French and Mauskopf 1992). A VSLY estimate may
imply a stream of constant values per year. The
annualized VSLY estimate depends on three factors:
the underlying VSL estimate; a discount rate; and the
number of remaining life years implied by the
underlying VSL estimate.
We develop an estimate of the value of a statistical
life-year lost (VSLY) based on an approach suggested
by Moore and Viscusi (1988). They assume that the
willingness to pay to save a statistical life is the value
of a single year of life times the expected number of
years of life remaining for an individual. They also
suggest that the typical respondent in a mortality risk
study may have a life expectancy of an additional 35
years. Using the 35-year life expectancy and VSL
estimate of $4.8 million, their approach yields an
estimate of $137,000 per life-year lost or saved. In the
prospective analysis, we also assume that an individual
discounts future additional years. This implies that the
value of each life-year lost must be greater than the
non-discounted value. Assuming a five percent
discount rate and adopting the above outlined
approach, the implied value of each life year lost used
in the prospective analysis is $293,000 (in 1990
dollars).
Critics note several disadvantages to using this
type of VSLY method, most notably that the value of
avoiding premature death depends on more than just
lifespan. With the VSLY approach, the benefit
attributed to avoiding a premature death depends
directly on how premature it is — resulting in smaller
values for older people, who have shorter life
expectancies, and larger values for younger people.
While this approach attempts to derive age-
adjusted values of expected life remaining using VSL
estimates, it does not address potential differences in
the value of a statistical life due to differences in the
average age of the affected population or the average
age at which an effect is experienced. Studies have
shown that simple progressive declines in value as
estimated with the VSLY method may be an
oversimplification; in many cases, values for health
peak several times throughout a lifetime (e.g., after
having children, after retirement). In addition, in
many cases, data restrictions limit researchers' ability
to estimate VSLY because it is difficult to obtain
estimates of age-specific risks and the number of life-
years lost.
Life Quality Adjustments
Another way to make adjustments to account for
heterogeneity in value of life estimates is an approach
that incorporates health status by applying a VSLY
estimate (generated from the VSL literature) to an
estimate of quality-adjusted life years (QALY). The
resulting value estimates measure improvements in
health based on individuals' attitudes toward
symptoms or different levels of pain or physical
impairment (Tolley et al. 1994). This approach utilizes
survey techniques to rate different health conditions
and adjust the number of life years lost to represent
lost quality-adjusted life years. As a result, this
approach aims to develop a value for a single QALY
that is the same regardless of individual characteristics.
In other words, the approach tries to standardize the
measure of mortality risk reduction that emerges from
a health effects analysis, making valuation more
straightforward.
The Life Quality Adjustment approach may
implicitly incorporate morbidity impacts to assess
values for various causes of death, and is often used in
health economics to assess the cost effectiveness of
medical spending programs, to value morbidity
avoidance, and to value mortality avoidance. Using a
QALY rating system, health quality ranges from 0 to
H-9
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
1, where 1 may represent full health, 0 death, and
some number in between (e.g., 0.8) an impaired
condition. If an individual lives with a health quality
index of 0.8, then the implied value of avoiding a year
with this condition and having full health in its place
would be 0.2 x VSLY. By the same token, the value
of gaining an additional life year in this condition is 80
percent of the value of gaining a year in full health
(i.e., 0.8 x VSLY) and represents an annual value for
mortality risk avoidance for a person with the
condition.
Tolley, et al. (1994) estimate values for a variety of
health conditions using numerous techniques,
including, in some cases, valuation of quality-adjusted
life years. For example, when estimating values for
acute and chronic symptoms using QALYs, the
authors calculate low, medium and high value
estimates based on a range of YSL estimates.
Specifically, the authors use the following three YSLY
estimates (1991$) for QALY valuation:
• Low Estimate = $70.000 VSLY: Derived
from Miller, Calhoun and Arthur (1990) —
VSL of $1.95 million, two percent discount
rate.
• Medium Estimate = $120.000 VSLY:
Derived from Miller, Calhoun and Arthur
(1990) — VSL of $1.95 million, six percent
discount rate.
• High Estimate = $175.000 VSLY: Derived
from Moore and Viscusi (1988) - VSL of
$6.0 million, 0 percent discount rate.
The authors multiply the VSLY estimate by the
estimate of QALYs to calculate a value for each
symptom. It is not clear from the analysis discussion
which symptom values represent the application of
this approach.
Cutler and Richardson (1998) apply a VSL
estimate to an estimate of QALYs to measure the
value of health improvements between 1970 and 1990
for ten health conditions. To do this, the authors use
an VSLY estimate of $100,000, derived as the
intermediate value of results reported in studies by
Viscusi (1993) and Tolley et al. (1994). In addition,
the authors estimate QALYs using information on
disease prevalence in the US from 1970 to 1990,
weighted by a factor that represents how quality of life
for a given condition has changed over time (e.g.,
more buildings have ramps and elevators for
individuals who have mobility problems, thus raising
quality of life over time).
Murray and Lopez (1996) modify the above
theoretical approach by deriving an estimate of
disability-adjusted life years (DALYs). DALY
estimates consider the years of life lost and years lived
with disability, adjusted for the severity of the
disability. The approach to estimate DALYs is similar
to that used to estimate QALYs in that both
incorporate judgments about the value of time spent
in different health states. However, DALY and
QALY estimation methods differ in that the methods
to estimate DALYs are elicited from preferences for
particular value choices using a specific standardized
set of value choices.
The Life Quality Adjustment approach scales
WTP values (VSL estimates) using a measure of life
years that reflect heterogeneity in quality of health
(QALYs). In many cases, the applied VSLY estimates
do not reflect consistent use of VSL estimates or
discount rates. In addition, in each of these valuation
analyses health economists have constructed a scale or
index that ranks health outcomes in terms of how
adverse individuals believe them to be. Often, the
extreme points on the scale are "perfect health" and
"immediate death," but some applications allow for
health outcomes that might be viewed as worse than
death. These ranking methods do not yield estimates
of WTP, and therefore are not linked to utility theory.
It is not clear that the ranking of health outcomes
obtained by these indices would match the ranking
obtained by knowing individuals' WTP for various
health effects. As discussed by Johansson (1995),
these scales or indices rely on much more restrictive
assumptions about the nature of individual
preferences than are normally made in WTP studies.
H-10
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Longevity
Several recent efforts estimate values for an
identifiable life by estimating the WTP for own life
extension. Johannesson andjohansson (1996, 1997)
estimate the WTP to increase one's life expectancy by
one additional year (i.e., extending men's life
expectancy from age 75 to 76, and women's from age
80 to 81, conditional on reaching age 75 or 80).
Johannesson, Johansson, and Lofgren (1997) estimate
the value of an immediate small reduction in mortality
risk (a "blip" or one year of fatal risk prevention).
While this methodology represents a utility-theory
based value, the value estimate for a single year of
longevity does not exactly correspond to what is
needed for an assessment of air pollution benefits.
Johannesson andjohansson (1996, 1997) estimate a
value for a single year of life extension near the end of
one's lifetime — values at this age are likely to be low
because of a low expectation of quality of life at this
advanced age. It is likely that mortality values will vary
within an individual's lifetime and with probability of
survival. In addition, mortality associated with
pollutant exposure will likely yield a longevity loss
greater than one year (e.g., mortality associated with
particulate matter yields an average longevity loss of
approximately 14 life years among those who are
afflicted). Moreover, because of the hypothetical
nature of the contingent valuation method, it is
unclear whether respondents accept the scenarios
presented and whether enough context was provided
to understand the risk and the budget implications of
the scenario and the response.
Cost Effectiveness
Garber and Phelps (1997) present a methodology
for valuing a discounted life year that is determined
by income and risk aversion in a life-cycle model. To
calculate the optimal cost effectiveness cut-off for
medical intervention, the authors assume values of a
utility function, health production function, income,
discount rate, and baseline mortality to derive a value
equivalent to WTP for a discounted life year. In this
model, utility is a function of income (less medical
expenditures), and future income is a function of
survival and medical expenditures. As a result, the
authors use mortality rates to calculate expected
income. Changes in these mortality rates result in
changes in survival probabilities, and hence income.
The model estimates an individual's willingness to
trade income from one period to another; the
discounted change in income is equivalent to WTP for
a change in risk.
Although this methodology is based on a life-
cycle model using survival probabilities, it is simplistic
in its assumptions and is based on assumed
preferences, rather than on revealed preferences or
those stated by an individual. In effect, the model
estimates values based largely on one empirical input:
individual income. For example, the YSL for a 40
year-old cannot exceed $250,000 because that amount
exceeds the discounted expected income. The largest
value of discounted life-year obtained by the authors
is approximately $37,000.
Valuation Strategy Chosen for this
Analysis
To estimate the economic value of mortality
benefits associated with air pollution reductions,
economic theorists prefer estimates that reflect ex ante
values of reducing the risk of mortality across the
population (i.e., for individuals having different health
states and other characteristics such as income level
and risk perception). This requires an estimate of an
individual WTP for a reduction in an involuntary risk
that will change individuals' survival probabilities for
a lifetime. Developing a valuation estimate based on
this theoretically ideal approach, however, is currently
subject to significant data and methodological
problems. Moreover, many of the valuation methods
that are frequently presented as an alternative to the
VSL approach rely on YSL estimates and calculate
values that depend on lifespan data, which may be
difficult to measure given the current health data
limitations. Consequently, EPA's current interpreta-
tion of the state-of-the-art in premature mortality
valuation leads to adoption of the VSL approach for
development of the primary benefit estimate.
H-11
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
As discussed above, several different approaches
for estimating a mortality-related value have been
developed. Each, however, has either methodological
inconsistencies with the preferred utility-based
approach, or does not provide a value estimate for a
commodity comparable to that provided by reduced
air pollution. We summarize the potential problems
of these alternatives below and in Table H-2:
• Life Quality Adjustment: This approach
relies on YSL estimates applied to survey
estimates of life-years (i.e., QALYs or
DALYs) for the economic valuation.
Currently, no generally accepted estimate or
range of estimates of VSLY have been
established, instead these values derive from
various VSL studies and reflect numerous
discount rates. In addition, the life years
estimates require data sets that can account
for the health states or utilities specific to a
wide variety of health effects associated with
air pollution. In many cases, these estimates
are not available or are based on health
professionals' perceptions of various health
outcomes, and not necessarily based in
economic utility theory.
• Longevity: The longevity valuation
approach of Johannesson and Johansson
(1996 and 1997) provides an estimate of the
value for an identifiable one-year life
extension. While the contingent valuation
approach used may be consistent with utility
theory, the commodity valued does not
represent the commodity gained through
improvement of ambient air quality.
• Cost Effectiveness: While the approach
taken by Garber and Phelps relies on survival
probabilities throughout an individual's
lifetime, the methodology is based on a utility
function that makes specific assumptions
about individual preferences to measure WTP
rather than eliciting value from either a
revealed or stated preference approach.
Moreover, this approach measures a WTP
that is constrained by income. Where
H-12
individual risks are small (perhaps one in ten
thousand) relative to certain loss of life,
individual WTP may also be small relative to
income.
-------
Exhibit H-2
Summary of Alternative Methods for Assessing the Value of Reduced Mortality Risk
Method
Description
Strengths
Weaknesses
References
Value of Statistical
Life (VSL) - hedonic
wage studies
Uses wage and risk data to
estimate WTP to avoid risk in the
workplace
- Revealed preference
-Well-established approach: more
than 60 primary studies
- Workplace risk context; working-age
subjects and voluntary risk
- VSL may imply ex post risk
Summaries by Viscusi
(1992) and others;
many primary studies
VSL - contingent
valuation studies
Uses survey responses to
estimate WTP to avoid risks
- Flexible approach; some studies
use environmental risk context
- Good data on WTP by
respondent
- Risk information not well-understood
by subjects; questions may be unfamiliar
- VSL may imply ex post risk
Summaries by Viscusi
(1992) and others
VSL - consumer
market studies
Uses consumer expense and risk
data (e.g., smoke detectors) to
estimate WTP to avoid risks
- Revealed preference
- Flexible approach
- Major difficulties estimating both risk
and expense variables
- VSL may imply ex post risk
Summaries by Viscusi
(1992) and others
Value of Statistical
Life Year (VSLY)
Annual equivalent of VSL
estimates
- Provides financially accurate
adjustment for age at death
- Adjustment may not reflect how
individuals consider life-years; assumes
they have equal value for all remaining
life-years
Viscusi and Moore
(1988); French and
Mauskopf (1992)
Quality Adjusted Life
Year (QALY)
Applies quality adjustment to life-
extension data, uses cost-
effectiveness data to value
-Widely used in public health
literature that assess different
private medical interventions
- Lack of data on health state indices
and life quality adjustments that are
applicable to an air pollution context
Tolley (1994); Cutler
and Richardson
(1998)
WTP for change in
survival curve
Reflects WTP for change in risk,
potentially incorporates age-
specific nature of risk reduction
- Theoretically preferred approach
that most accurately reflects
nature of risk reductions from air
pollution control
- Almost no current literature
- Lack of available data due to the
severe methodological difficulties in
presenting complex risk data to subjects
and eliciting reliable values
Cropper and Sussman
(1990)
WTP for change in
longevity
Uses stated preference approach
to generate WTP for longevity or
longer life expectancy
- Life expectancy is familiar term to
most individuals
- Life expectancy is a simplifed term that
does not incorporate age-specific risk
information
-Methodological and data problems in
attempting to adapt to air pollution
context
Johannesson and
Johansson (1997);
Health Canada (1998)
Cost-Effectiveness
Develops a standard of
comparison to measure the
efficiency of various treatments in
achieving a given health outcome
-Widely used in public health
contexts
- Public health context may be for
private goods (i.e., treatment)
- Dollar values do not necessarily reflect
patient preferences
Garber and Phelps
(1997)
H-13
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Valuation of Hospital Admissions
Avoided
The valuation of this benefits category reflects the
value of reduced incidences of hospital admissions
due to respiratory or cardiovascular conditions. We
measure avoided hospital admissions as opposed to
the number of avoided cases of respiratory or
cardiovascular conditions, because of the availability
of C-R relationships for the hospital admissions
endpoint. Hospital admissions reflect a class of health
effects linked to air pollution which are acute in nature
but more severe than the symptom-day measures
discussed below.
As described in Chapter 5, our approach to
estimating the number of incidences for this category
involves reliance on several concentration-response
(C-R) functions. Each concentration-response
function provides an alternative definition of either
respiratory effects or cardiovascular effects, and
defines alternative relationships between a single
health affect and different pollutants. For the
valuation of these incidences, the current literature
provides well-developed and detailed cost estimates of
hospitalization by health effect or illness. Using
illness-specific estimates of avoided medical costs and
avoided costs of lost work-time, developed by
Elixhauser (1993), we construct cost of illness (COI)
estimates that are specific to the suite of health effects
defined by each C-R function. For example, we use
twelve distinct C-R functions to quantify the expected
change in respiratory admissions.9 Consequently in
this analysis, we develop twelve separate COI
estimates, each reflecting the unique composition of
health effects considered in the individual studies.
Because each epidemiology study defines a health
effect by a group of ICD codes, we construct COI
estimates for each study by aggregating estimates that
are specific to an ICD code. These estimates use the
following information reported by Elixhauser (1993):
9For more detailed discussion of the various health effects
considered by each C-R function and methodology for estimating
the number of avoided hospital admissions, see Appendix D.
average hospital costs, average length of stay, and
baseline incidences.10 We use this ICD code
information to develop valuation estimates that have
two components, hospital charges and lost earnings
due to the hospital stay. Our estimate of lost earnings
due to time spent in the hospital is based on valuing
the average length of hospital stay at a daily rate of
$83. This daily rate is the median weekly wage divided
by five work days and is based on U.S. Department of
Commerce figures (1992). After developing values for
each relevant ICD code (i.e., hospital costs plus lost
earnings), we weight these values based on their
prevalence in the baseline. The final COI estimate,
specific to each study, is the sum of the weighted
value of ICD code-specific estimates.
We use a Monte Carlo approach to combines the
valuation and physical effects modeling to generate a
benefits estimate for hospital admissions. This
approach also allows us to account for the variability
in costs due to alternative definitions of respiratory
and cardiovascular conditions that result in a hospital
admission. The Monte Carlo process for integrating
the C-R function and its COI value involves first
randomly selecting an estimated change in incidences
from the suite of applicable C-R functions. For
example, we use five epidemiology studies for the
endpoint hospital admissions due to cardiovascular
effects, and develop COI estimates specific to each
study. The Monte Carlo modeling then selects the
COI estimate specifically developed for that C-R
function. These values are multiplied to generate a
single benefits estimate for reduced hospital
admissions. This process is repeated so that the value
from each iteration is collected to generate a
distribution that characterizes the range and
probability of possible benefits estimates. The
primary benefit estimates of avoided cardiovascular-
related hospital admissions reflect the central value of
this distribution.
The use of COI estimates suggests we are likely to
significantly underestimate the WTP to avoid hospital
10Potential illnesses associated with respiratory and
cardiovascular admissions were identified by ICD-9 code.
H-14
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
admission. The valuation of any given health effect,
such as hospitalization, should reflect the value of
avoiding associated pain and suffering and lost leisure
time, in addition to medical costs and lost work time.
While the probability distributions in this analysis
characterize a range of potential costs associated with
hospitalization, they do not account for the omission
of factors from the COI estimates, such as pain and
suffering. Consequently, the valuations for these
endpoints most likely understate the true social values
for avoiding hospital admissions due to respiratory or
cardiovascular conditions.
Valuation of Chronic Bronchitis
Avoided
In this analysis, chronic bronchitis is one of the
two monetized morbidity endpoints whose effects
may be expected to last from the initial onset of the
illness throughout the rest of the individual's life.
WTP to avoid chronic bronchitis therefore
incorporates the present discounted value of a
potentially long stream of costs (e.g., medical
expenditures and lost earnings) and reduced health-
state utility.11
Two studies, Yiscusi et al. (1991) and Krupnick
and Cropper (1992) provide estimates of WTP to
avoid a case of chronic bronchitis. While alternative
estimates exist, many are derived from these two
primary studies.12 The study by Yiscusi et al. uses a
sample that is larger and more representative of the
general population, while the Krupnick and Cropper
study solicits values only from individuals who have a
relative with the disease. As a result, the valuation of
"The severity of cases of chronic bronchitis valued in some
studies approaches that of chronic obstructive pulmonary disease.
To maintain consistency with the existing literature, we do not
treat those cases separately in this analysis.
12For examples of alternative estimates see Desvousges et al
(1998) and Tolley et al (1994). Both studies present estimates of
avoiding one year of chronic bronchitis that are based on adjusting
values from either Viscusi et al (1991) or Krupnick and Cropper
(1992).
chronic bronchitis is based on the distribution of
WTP responses from Yiscusi et al. (1991).
Both the Viscusi et al. and the Krupnick and
Cropper studies estimate the WTP to avoid a severe
case of chronic bronchitis (CB). The incidence of
pollution-related chronic bronchitis, however, is based
on three studies which consider only new incidences
of the illness and the resulting severity is unknown.13
In response to the uncertainty regarding how the
severity of a new case may progress, the prospective
analysis adjusts Yiscusi et alls WTP estimates
downward. This adjustment reflects the decrease in
severity of a case of pollution-related CB relative to
the case in the Viscusi study and the elasticity of WTP
with respect to severity. The elasticity of WTP to
avoid CB is a marginal value and not unit elastic (i.e.,
not equal to one). Consequently, WTP adjustments
are made in one percent increments. At each step, the
WTP specific to a given CB severity level (sev), is
adjusted to derive the WTP to avoid a case with a one
percent lower level of severity by calculating (
0.99*.sw).14 In this analysis, we derive an estimate of
WTP for a case of chronic bronchitis that represents
a 50 percent reduction in the severity described in the
Viscusi study. The iterative procedure continues until
the severity is half of the of the Viscusi value.
With the downward adjustment to Viscusi et alls
WTP estimate, calculating the WTP to avoid a case of
13 The three studies are Abbey etal. (1993), Abbey etal. (1995)
and Schwartz (1993). For more discussion of estimating the
number of avoided cases of chronic bronchitis see Appendix D,
Human Health Effect of Criteria Pollutants. Incidences are
predicted separately for each year during the period 1990-2010.
It is important that only new cases of chronic bronchitis are
considered in this analysis because WTP estimates reflect lifetime
expenditures and lower utility associated with the illness. If the
total prevalence of chronic bronchitis, rather than the incidence of
only new chronic bronchitis were predicted each year, valuation
estimates reflecting lifetime losses could be repeatedly applied to
the same individual for many years, resulting in a severe
overestimation of the value of avoiding pollution-related chronic
bronchitis.
14 Note that the elasticity changes at each iteration because
the elasticity of WTP with respect to severity is a function of
severity.
H-15
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
pollution-related chronic bronchitis has three
components, each introducing some uncertainty. The
components are (1) WTP to avoid a case of severe
CB, (2) the severity level of an average pollution-
related case of CB relative to that of the severe case,
and (3) the elasticity of WTP with respect to severity.
Based on assumptions about the distributions of each
component's value, a distribution of WTP to avoid a
pollution-related case of CB is derived by Monte Carlo
methods. Each of the three underlying distributions
is described briefly below.
The distribution of WTP to avoid a severe case of
CB is based on the distribution of WTP responses in
the Yiscusi study. Yiscusi et al. derived an implicit
WTP to avoid a statistical case of chronic bronchitis
from respondents' WTP for a specified reduction in
risk. The mean response implied a WTP of about $1
million (1990 dollars); the median response implied a
WTP of about $530,000 (1990 dollars).15 Yiscusi et al.
report the mean and median of their distribution of
WTP responses and the decile points. The
distribution of reliable WTP responses from the
Viscusi study can therefore be approximated by a
discrete distribution, assigning equal probability to
each of the first nine decile points (or one-ninth
probability to each decile). This method omits five
percent of the responses from each end of the
distribution (i.e., the extreme tails which are
considered unreliable). Our present study uses this
trimmed distribution of Yiscusi et al's WTP
responses, for which the mean is $720,000 (1990
dollars), as the distribution of WTPs to avoid a severe
case of CB.
The distribution of the severity level of an average
case of pollution-related CB is based on the severity
levels used in Krupnick and Cropper's study, which
estimates the relationship between severity level and
the natural log of WTP. The distribution is triangular
with a mean of 6.5 and endpoints at 1.0 and 12,
15There is an indication in the Viscusi paper that the dollar
values in the paper are in 1987 dollars. Under this assumption, the
dollar values were converted to 1990 dollars.
although the most severe case of CB in that study is
assigned a severity level of 13.16
The elasticity of WTP to avoid a case of CB with
respect to the severity of the case equals a constant
times the severity level. This constant, estimated in
Krupnick and Cropper's study of the relationship
between severity and the natural log of WTP, is
normally distributed with mean of 0.18 and standard
deviation of 0.0669.
Using distributions of the three WTP
components described above, the Monte Carlo
analysis generates a distribution with a mean of
$260,000 for WTP to avoid a pollution-related case of
CB. Consistent with economic theory, the COI
estimates generated by Cropper and Krupnick (1990)
are lower than the mean WTP estimate (i.e., COI does
not reflects the desire to avoid pain and suffering).17
These COI estimates are approximately $86,000 for a
30 year old, $84,000 for a 40 year old, $76,000 for a 50
year old, and $43,000 for a 60 year old (in 1990
dollars). The prospective's WTP estimate is 3 to 6
times greater than the full COI estimate for 30 year
olds and 60 year olds, respectively.
Valuation of Chronic Asthma Avoided
Chronic asthma is the other morbidity endpoint
that is valued as a health condition lasting throughout
an individual's lifetime. The number of new cases of
chronic asthma is based on a study by McDonnell et
al. (1999), and specifically examines the effects of
ozone as a potential cause of the illness among adult
males (i.e., ages 27 and older). Similar to the valuation
of chronic bronchitis, WTP to avoid chronic asthma
16The Krupnick and Cropper study bases its most severe case
of CB (i.e., severity level equal to 13) on that used in the Viscusi
study.
17 Using a 5 percent discount rate and assuming that 1) lost
earnings continue until age 65, 2) medical expenditures are
incurred until death, and 3) life expectancy is unchanged by
chronic bronchitis, Cropper and Desvousges calculate several
estimates of the present value of the stream of medical
expenditures and lost earnings associated with an average case of
chronic bronchitis.
H-16
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
is presented as the net present value of what would
potentially be a stream of costs and lower well-being
incurred over a lifetime.
Estimates of WTP to avoid asthma are provided
in two studies, one by Blumenschein and Johannes son
(1998) and one by O'Conor and Blomquist (1997).
Both studies use the contingent valuation method to
solicit annual WTP estimates from individuals who
have been diagnosed as asthmatics. Each study,
however, applies a different valuation approach.
Blumenschein and Johannesson solicit WTP values by
asking dichotomous choice and open-ended bidding
game questions. They report an average monthly
WTP of $162 which amounts to an annual value of
approximately $1,900 (1990 dollars). Alternatively,
O'Conor and Blomquist apply a risk-risk tradeoff
approach similar to that used in the chronic bronchitis
studies. They calculate $1,200 (1990 dollars) as the
average annual WTP to avoid asthma.
To maintain consistency between the health
effects modeling and the valuation, the WTP estimates
were adjusted to account for two factors. As
mentioned earlier, valuation of chronic morbidity
endpoints should approximate the costs and lowered
health-state utility that are incurred over an
individual's lifetime. We assume that the health
condition does not affect the average life expectancy
of an individual (i.e., does not cause premature
mortality). Recognizing that the average life
expectancy will vary with different age groups and that
each age group does not represent an equal portion of
the population, the present discounted stream of WTP
is calculated for seven different age cohorts (between
the ages 27 and 85). In turn, the net present value for
each age group is weighted by that age category's
representative share of the total population. This
calculation was performed for the mean WTP
estimates presented in the two studies. The central
estimate of WTP to avoid a case of chronic asthma
among adult males, approximately $25,000, is the
average of the present discounted value from the two
studies. The analysis characterizes the uncertainty
around this estimate by applying upper and lower
values based on the present discounted value derived
from each study, $19,000 derived from O'Conor and
Blomquist study and $29,000 from the Blumenschein
and Johannesson study.
Valuation of Other Morbidity Endpoints
Avoided
The valuation of a specific short-term morbidity
endpoint is generally solicited by representing the
illness as a cluster of acute symptoms. For each
symptom, the WTP is calculated. These values, in
turn, are aggregated to arrive at the WTP to avoid a
specific short term condition. For example, the
endpoint lower respiratory symptoms (LRS) is
represented by two or more of the following
symptoms: runny or stuffy nose; coughing; and eye
irritation. The WTP to avoid one day of LRS is the
sum of values associated with these symptoms. The
primary advantage of this approach is that is provides
some flexibility in constructing estimates to represent
a variety of health effects.
At the time of the Section 812 retrospective
analysis there were only a small number of available
studies on which to base estimates (two or three
studies, for some endpoints; only one study for
others). Since the retrospective analysis, much of the
literature suggests there are developing approaches
that may eventually lead to the refinement of estimates
and the overcoming of some limitations to the current
approach to constructing values. For example there
is extensive progress in developing valuation
techniques that reflect an individual's current health
state and more accurately account for a symptoms's
attributes (i.e., duration and severity).
There are several aspects of the short-term
morbidity valuation estimates worth noting. First,
estimates of WTP may be understated for at least two
reasons. If exposure to pollution has any cumulative
or lagged effects, then a given reduction in pollution
concentrations in one year may confer benefits not
only in that year but in future years as well. Benefits
achieved in later years are not included. In addition,
the possible effects of altruism are not considered in
any of the economic value derivations. Individuals'
WTP for reductions in health risks for others are
implicitly assumed to be zero. The second point
H-17
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
worth noting is that the total benefit attributed to the
reduction of particular pollutant's concentration is
determined largely by the benefit associated with its
corresponding reduction in mortality risk. This is
largely due to the dollar value associated with mortality
which is significantly greater than any other valuation
estimate. More detailed explanations for valuation of
specific morbidity endpoints are given in Table H-3.
The table summarizes the sources and derivation of
the economic values used in the analysis.
Valuation of Welfare Effects
Economic valuations for welfare effects quantified
in the analysis (i.e., visibility and worker productivity)
are documented in Table H-3.18 Worker productivity,
unlike the avoidance of work loss days or restricted
activity days, reflects productivity benefits due to
improvements in work conditions (i.e., reduced
ambient ozone) rather than health improvements (i.e.,
reduced risk of hospitalization). It is measured in
terms of the reduction in daily income of the average
worker engaged in strenuous outdoor labor and
estimated at $1 per ten percent increase in ozone
concentration. (Crocker and Horst, 1981). We discuss
the derivation of the visibility valuation further below.
18 In valuing welfare effects, the retrospective analysis
included the benefits of reduced household soiling. This valuation
was based on 1972 data that projected expenditure patterns from
1972 to 1985 (Manuel et al, 1982). While this study was
appropriate for the twenty year time period of the retrospective
(1970 to 1990), it is of questionable applicability for the current
study. Since the original study, there have been alternative
estimates of benefits due to reduced soiling. These estimates,
however, continue to be based on the original study and its
underlying data (e.g., Desvousges et al., 1998). Consequently, these
valuation coefficients do not reflect more recent information on air
pollution composition and potentially significant changes in
patterns of household expenditure and allocation. Progress in the
valuation of this category's benefits is further limited by the
challenges of developing dose-response functions that accurately
assess the level and rate of materials damage and soiling. Recent
literature does suggest there is progress in refining approaches,
although it has not quite advanced to the level necessary for
credible quantification or monetization of benefits associated with
reduced materials damage and soiling.
Visibility Valuation
Since the late 1970s, a number of contingent
valuation (CV) studies of visibility changes have been
published in the economics literature. These studies
often classify visibility benefits as either residential or
recreational. CY studies of residential visibility
generally survey individuals in urban and suburban
settings. The valuation is also applicable to
households in rural areas. Residential values relate to
the impact of visibility changes on an individual's daily
life (e.g., at home, at work, and while engaged in
routine recreational activities). Benefits of recreational
visibility relate to the impact of visibility changes
manifested at parks and wilderness areas that are
expected to be experienced by its visitors.
Recreational visibility benefits may, however, reflect
the value an individual places on visibility
improvements regardless of whether or not the
person plans to visit the park.19
The reported estimates, expressed as household
willingness to pay (WTP) for a hypothesized
improvement in visibility, have a wide range of values.
For examples, studies of visibility values from western
cities have reported somewhat lower values than those
from eastern cities. This difference raises the question
of how visibility benefits should be evaluated with
respect to location (e.g., eastern U.S. versus western
U.S.), commodity definition (e.g., changes in
recreational areas versus residential areas), and units of
measurement (e.g., visual range, light extinction, and
deciview). While the differing values reported in the
literature may appear to imply that visibility is valued
differently in the eastern and western U.S., other
evidence suggests that eastern and western visibility
are not fundamentally different commodities. For
example, NAPAP data indicates that California's
South Coast Air Basin, which encompasses Los
Angeles and extends northward to the vicinity of San
Francisco, has median baseline visibility more
characteristic of the eastern U.S. than of other areas of
the west (NAPAP 1991; IEc 1992, 1993a). These
19This type of valuation is typically labeled "existence value."
For more discussion see Chestnut and Rowe, 1990.
H-18
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
results suggest that the valuation of marginal visibility
changes is dependent on baseline conditions and
proximity to the commodity being valued (e.g.,
improved visibility in a region with an abundance of
National Parks such as the Pacific Northwest).
Returning to the NAPAP example, the similarity in
values may reflect the similarities between baseline
visibility in eastern and western coastal zones (i.e.,
coastal areas typically have higher humidity, while
areas of the west tend to have lower humidity and
hence a greater baseline visibility).
For the purposes of this report, we interpret
recreational settings applicable for this category of
effects to include National Parks throughout the
nation. Other recreational settings may also be
applicable, for example National Forests, state parks,
or even hiking trails or roadside areas with scenic
vistas. In those cases, a lack of suitable economic
valuation literature to identify these other areas and/or
a lack of visitation data prevents us from generating
estimates for those recreational vista areas. Moreover,
we develop estimates of recreational visibility changes
that account for the tendency of individuals to value
visibility changes based on proximity to the National
Park.
We estimate visibility benefits based on a derived
visibility valuation function. In both cases, residential
and recreational visibility, the valuation function takes
the following form:
HHWTP = B * ln(VR1/VR2)
where:
HHWTP = annual WTP per household for
visibility changes
VR1 = the starting annual average visual
range
YR2 = the annual average visual range after
the change in air quality
B = the estimated visibility coefficient.
The form of this valuation function is designed to
reflect the way individuals perceive and express value
for changes in visibility. In other words, the expressed
WTP for visibility changes varies with the percentage
change in visual range, a measure that is closely related
to, though not exactly analogous to, the Deci View
index used in Chapter 4.
We develop estimates of the visibility coefficients
for residential and recreational visibility from two
studies.20 We use figures reported in Chestnut and
Dennis (1997) for the valuation of residential visibility.
This study publishes estimates of visibility benefits
for the Eastern U.S that are based on original research
conducted in two Eastern cities (Atlanta and Chicago)
by McClelland et al. (1990). We use a central B
coefficient for residential visibility of $141, as reported
in Chestnut and Dennis (1997). For the valuation of
recreational visibility benefits, we use a study by
Chestnut and Rowe (1990). This study reports WTP
estimates of recreational visibility in three park
regions, the Western, Southwestern, and Eastern U.S.
For recreational visibility, the coefficients vary based
on the study region and whether the household is
within or outside of the National Park region of
concern. "In-region" coefficients are higher than
those for "out-of-region" households. The "in-
region" estimates for California, the Southwest, and
Southeast are $105, $137, and $65, respectively; the
corresponding "out-of-region" estimates are $73,
$110, and $40, respectively.
Our valuation of visibility changes is largely based
on unpublished, but peer-reviewed work. For
example, we use the secondary analysis of Chestnut
and Dennis (1997) to value residential visibility
benefits. This article is published in the Journal of Air
and Waste Management Association, but relies on the
unpublished results reported by McClelland et al.
(1990). The source of our recreational visibility
estimates, Chestnut and Rowe (1990), is also
unpublished. Both studies were originally developed
as part of the National Acid Precipitation Assessment
Program (NAPAP) and, therefore, have been subject
to peer-review as part of that program. Moreover,
these two studies are frequently cited and
20The unit of measure for the visibility coefficients is dollars.
However, these coefficients are scaled by the small incremental
changes in visibility to generate our WTP estimates.
H-19
-------
recommended for use in published analyses of
visibility valuation.21
Concerns about the method used in the
McClelland et al. study, however, suggest their results
may not incorporate two potentially important
adjustments. First, their study does not account for
the "warm glow" effect, in which respondents may
provide higher willingness to pay estimates simply
because they favor "good causes" such as
environmental improvement. Second, while the study
accounts for non-response bias, it may not employ the
best available methods. The effect of both these
factors is to suggest an overestimate of WTP. As a
result, we exclude residential visibility estimates from
the overall primary benefits estimate.
21For example see Desvousges et al. (1998).
H-
-------
Table H-3
Unit Values Used for Economic Valuation of Health and Welfare Endpoints
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Central Estimate Uncertainty Distribution
Derivation of Estimates
Mortality
$4.8 million per
statistical life
$293,000 per
statistical life-year
Weibull distribution,
mean = $4.8 million
std. dev. = 3,240,000
Weibull distribution,
mean = $293,000
std. dev. = 198,000
Central Estimate: Value is the mean of value-of-statistical-life estimates
from 26 studies (5 contingent valuation and 21 labor market studies).
Uncertainty: Best-fit distribution to the 26 sample means. The Weibull
distribution prevents selection of negative WTP values.
Central Estimate: Value is the mean of the distribution of the value of a
statistical life-year, derived from the distribution of the value of a
statistical life (see below).
Uncertainty: Assuming the discount rate is five percent, and assuming
an expected 35 years remaining to the average worker in the wage-risk
studies (see above), the value of a statistical life-year is just a constant,
0.061, multiplied by the value of a statistical life. The distribution of the
value of a life-year is derived from the distribution of the value of a
statistical life. Because the VSL is expressed as a Weibull distribution,
as indicated above, the value of a statistical life-year is also expressed
as a Weibull distribution, with mean equal to 0.061 multiplied by the
mean of the original Weibull distribution (0.061 x $4.8 million =
$293,000) and standard deviation equal to 0.061 multiplied by the
standard deviation of the original distribution (0.061 x $3.24 =
$198,000).
H-21
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Central Estimate Uncertainty Distribution
Derivation of Estimates
Chronic Bronchitis (CB)
$260,000
A Monte Carlo-generated
distribution, based on three
underlying distributions, as
described more fully under
"Derivation of Estimates"
and in the text.
Central Estimate: Value is the mean of a Monte Carlo distribution of
WTP to avoid a case of pollution-related CB. WTP to avoid a case of
pollution-related CB is derived by adjusting WTP (as described in
Viscusi et al., 1991) to avoid a severe case of CB for the difference in
severity and taking into account the elasticity of WTP with respect to
severity of CB. The mean of the resulting distribution is $260,000.
Uncertainty: The distribution of WTP to avoid a case of pollution-related
CB was generated by Monte Carlo methods, drawing from each of three
distributions: (1) WTP to avoid a severe case of CB is assigned a 1/9
probability of being each of the first nine deciles of the distribution of
WTP responses in Viscusi et al., 1991; (2) the severity of a pollution-
related case of CB (relative to the case described in the Viscusi study)
is assumed to have a triangular distribution, centered at severity level
6.5 with endpoints at 1.0 and 12.0 (see text for further explanation); and
(3) the constant in the elasticity of WTP with respect to severity is
normally distributed with mean = 0.18 and standard deviation = 0.0669
(from Krupnick and Cropper, 1992). See text for further explanation.
Chronic Asthma $25,000 Triangular distribution, Central Estimate: Based on results reported in two studies
centered at $25,000 on the (Blumenschein and Johannesson, 1998 and O'Conor and Blumquist,
interval [$19,000, $30,000] 1997). Assumes a 5% discount rate and reflects adjustments for age
distribution among adults (ages 27 and older) and projected life years
remaining.
Uncertainty: Reflects the range in central estimate values reported in
the two studies.
H-22
-------
Health or Welfare
Estimated Value Per Incidence (1990$)
Endpoint
Central Estimate
Uncertainty Distribution
Derivation of Estimates
Hospital Admissions
1. All Respiratory
- ICD codes: 460-519
variable—
function of the
analysis
See Derivation of Estimates
Central Estimate: Central estimate is the result of the analysis. The
analysis uses 12 distinct C-R functions. A COI estimate is constructed
for each. The COI estimates are based on ICD-9 code level information
(e.g., average hospital care costs, average length of hospital stay, and
weighted share of total respiratory illnesses) reported in Elixhauser
(1993).
Uncertainty: Probability distribution is a result of the analysis and
reflects: (1) uncertainty range of C-R function outcome; and (2) variation
in study-specific COI estimates.
2. All Cardiovascular
- ICD codes: 390-429
variable—
function of the
analysis
See Derivation of Estimates
Central Estimate: Central estimate is the result of the analysis. The
analysis uses five distinct C-R functions. A COI estimate is constructed
for each. The COI estimates are based on ICD-9 code level information
(e.g., average hospital care costs, average length of hospital stay, and
weighted share of total respiratory illnesses) reported in Elixhauser
(1993).
Uncertainty: Probability distribution is a result of the analysis and
reflects: (1) uncertainty range of C-R function outcome; and (2) variation
in study-specific COI estimates.
3. Emergency room visits for
asthma
$194
Triangular distribution,
centered at $194 on the
interval [$144, $269]
Central Estimate: COI estimate based on data reported by Smith et al.
(1997).
Uncertainty: Based on reported 95% confidence intervals for annual
estimates of the number and costs of ER visits.
H-23
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Central Estimate Uncertainty Distribution
Derivation of Estimates
Respiratory Ailments Not Requiring Hospitalization
1. Upper Resp. Symptoms
(URS)
(defined as one or more
of the
following:
runny or
stuffy nose,
wet cough,
burning,
aching, or
red eyes)
$19 Continuous uniform
distribution over the interval
[$7, $33]
Central Estimate: Combinations of the 3 symptoms for which WTP
estimates are available that closely match those listed by Pope et al.
result in 7 different "symptom clusters," each describing a "type" of
URS. A dollar value was derived for each type of URS, using lEc mid-
range estimates of WTP to avoid each symptom in the cluster and
assuming additivity of WTPs. The dollar value for URS is the average
of the dollar values for the 7 different types of URS.
Uncertainty: Assumed to be a continuous uniform distribution across
the range of values described by the 7 URS types.
2. Lower Resp. Symptoms
(LRS)
(defined in the study as
two or more of the
following: cough, chest
pain, phlegm, and
wheeze.)
$12 Continuous uniform
distribution over the interval
[$5, $19]
Central Estimate: Combinations of the 4 symptoms for which WTP
estimates are available that closely match those listed by Schwartz et
al. result in 11 different "symptom clusters," each describing a "type" of
LRS. A $ value was derived for each type of LRS, using lEc mid-range
estimates of WTP to avoid each symptom in the cluster and assuming
additivity of WTPs. The $ value for LRS is the average of the $ values
for the 11 different types of LRS.
Uncertainty: Taken to be a continuous uniform distribution across the
range of values described by the 11 LRS types.
3. Acute Bronchitis $45 Continuous uniform
distribution over the ii
[$13, $77]
Central Estimate: Average of low and high values recommended by
IEC for use in section 812 analysis (Neumann et al., 1994).
Uncertainty: Continuous distribution between low and high values
(Neumann et al., 1994) assigns equal likelihood of occurrence of any
value within the range.
H-24
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Central Estimate Uncertainty Distribution
Derivation of Estimates
4. Acute Respiratory
Symptoms and Illnesses
- Presence of any of 19
acute respiratory
symptoms
-Any Resp. Symptom
- Respiratory Illness
$18 1. URS, probability = 40%
LRS, probability = 40%
URS+LRS, prob. = 20%
2. If URS, use URS $ dist.
If LRS, use LRS $ dist.
If URS+LRS, randomly
select one value each from
URS and LRS $
distributions; sum the two
Central Estimate: Assuming that respiratory illness and symptoms can
be characterized as some combination of URS and LRS, namely: URS
with 40% probability, LRS with 40% probability, and both URS and LRS
with 20% probability. The $ value for these endpoints is the weighted
average (using the weights 0.40, 0.40, and 0.20) of the $ values derived
for URS, LRS, and URS + LRS.
Uncertainty: Based on variability assumed for central estimate, and
URS and LRS uncertainty distributions presented previously.
5. Asthma Attack
$32 Continuous uniform
distribution over the interval
[$12, $54]
Central Estimate: Mean of average WTP estimates for the four severity
definitions of a "bad asthma day." Source: Rowe and Chestnut (1986), a
study which surveyed asthmatics to estimate WTP for avoidance of a
"bad asthma day," as defined by the subjects.
Uncertainty: Based on the range of values estimated for each of the
four severity definitions.
6. Moderate or worse $32 Continuous uniform
asthma distribution over the i
[12, 54]
Central Estimate: Reflects the mean WTP to avoid a "bad asthma day"
as reported by Rowe and Chestnut (1986).
Uncertainty: Taken to be a continuous uniform distribution across the
range of values obtained from the study.
7. Shortness of breath, $5.30 Continuous uniform Central Estimate: From Ostro et al.. 1995. This is the mean of the
chest tightness or distribution over the interval median estimates from two studies of WTP to avoid a day of shortness
wheeze [$0, $10.60] of breath: Dickie et al., 1991 ($0.00), and Loehman et al., 1979
($10.60).
Uncertainty: Taken to be a continuous uniform distribution across the
range of values obtained from the two studies.
H-25
-------
Health or Welfare Estimated Value Per Incidence (1990$)
Endpoint Central Estimate Uncertainty Distribution Derivation of Estimates
Restricted Activity and Work Loss Days
1. WLDs $83 none available Central Estimate: Median weekly wage for 1990 divided by 5 (U.S.
Department of Commerce, 1992)
Uncertainty: Insufficient information to derive an uncertainty estimate.
2. MRADs $38 Triangular distribution
centered at $38 on the
interval [$16, $61]
Central Estimate: Median WTP estimate to avoid 1 MRRAD — minor
respiratory restricted activity day — from Tolley et al. (1986)
(recommended by lEc as the mid-range estimate).
Uncertainty: Range is based on assumption that value should exceed
WTP for a single mild symptom (the highest estimate for a single
symptom—for eye irritation-is $16.00) and be less than that for a WLD.
The triangular distribution acknowledges that the actual value is likely to
be closer to the point estimate than either extreme.
Welfare Effects
1. Visibility
Residential Visibility
"in-region"
"out-of-region"
Valuation function:
HHWTP= B * ln(VR1/VR2)
where:
HHWTP = annual WTP per household
B = estimated visibility coefficient
VR1 = starting annual average visual range
VR2 = the annual average visual range after the
change in air quality
Central Estimate: Estimated WTP for valuation of visibility changes
depend upon two factors: (i) visibility coefficient, B, and (ii) incremental
change in visual range. Visibility coefficients applied in the primary
analysis vary by category of visibility change and region.
Recreational visibility valuation is based on Chestnut and Rowe (1990).
For "in region" recreational visibility, the coefficients are $105, $137,
$65, for California, the Southwest, and the Southeast, respectively. For
"out-of-region" recreational visibility, the coefficients are $73, $110, $40,
for California, the Southwest, and the Southeast, respectively.
2. Worker Productivity
Change in daily
wages: $1 per
worker per 10%
change in 03
none available
Central Estimate: Based on elasticity of income with respect to 03
concentration derived from study of California citrus workers (Crocker
and Horst, 1981 and U.S. EPA, 1994). Elasticity applied to the average
daily income for workers engaged in strenuous outdoor labor, $73 (U.S.
1990 Census).
Note: All WTP estimates converted to 1990 dollars using the Consumer Price Index (CPI); COI estimates converted using the CPI-Medical.
H-26
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Results of Valuation of Health
and Welfare Effects
We estimate total human health and welfare
benefits by combining the economic valuations
described in this Appendix with the health and welfare
effects results presented in Appendix D for projection
years 2000 and 2010. The valuation results reflect the
annual estimates of benefits for the 48 contiguous
States, or "all U.S. population," which provides a more
accurate depiction of the trend of economic benefits
over the 20-year study period 22 For our Primary
Central estimate we attribute to Titles I through V of
the CAAA total annual human health benefits of $68
billion in 2000 and $118 billion in 2010.
As noted in Appendix D, we also include
alternative estimates for some health and welfare
impacts, which form the basis of several alternative
benefit estimates. For each of the health effects
estimates, we quantify statistical uncertainty. The
range of estimated health and welfare effects, along
with the uncertain economic unit valuations, are
combined to estimate a range of possible results. We
use the Monte Carlo method presented in Chapter 8
to combine the health and economic information.
Both tables show the mean estimate results, as well as
the measured credible range (upper and lower five
percentiles of the results distribution), of economic
benefits for each of the quantified health and welfare
categories. We summarize our primary estimates of
2000 and 2010 monetized benefits in Table H-3 and
22In Appendix D, we present physical effect estimates for
affected population in the contiguous 48 States and for affected
populations within 50 kilometers of a monitor. We present those
results as a sensitivity test that characterizes the possible magnitude
of human health effects. For the purpose of assessing the total
benefit of the CAAA, the results affecting populations in 48 states
provide a better characterization of the total direct benefits than do
the "monitored area only" results. The results of only monitored
areas does not account for the benefits of air quality improvements
affecting approximately 25 percent of the population. The "all
U.S. population" results, however, rely on uncertain extrapolations
of pollution concentrations, and subsequent exposures, from
distant monitoring sites to provide coverage for the 25 percent or
so of the population living far from air quality monitors.
Table H-5, respectively. The tables provide our
Primary Central estimate, in addition to our Primary
Low estimate, 5th percentile values, and our Primary
High estimate, 95th percentile estimates, for each
benefit category.
We also apply the Monte Carlo method when
generating aggregate monetized benefit results. The
Monte Carlo method used in the analysis assumes that
each health and welfare endpoint is independent of
the others. We adopt this approach in response to the
very low probability that the aggregate benefits will
equal the sum of the fifth percentile benefits from
each of the ten endpoints. Consequently, the upper
and lower fifth percentiles of the estimated benefits
from the individual endpoints does not equal the
estimated totals for the Primary High and Primary
Low estimates.
There are two additional aspects of our results
that warrant discussion. The first is the valuation of
premature mortality due to PM exposure. The second
is our strategy to avoid double-counting when
aggregating health benefits. As discussed in Chapter
5, premature mortality is estimated based on PM
exposure. Our primary estimates reflect a lag between
PM exposure and the timing of premature mortality.
While this lag does not alter the number of estimated
incidences, it does alter the monetization of benefits.
Because we value the "event" rather than the present
change in risk, the value of avoided future premature
mortality should be discounted. Therefore, the type
of lag structure employed plays a direct role in the
valuation of this endpoint.
The primary analysis reflects a five-year lag
structure. Under this scenario, 50 percent of the
estimated cases of avoided mortality occur within the
first two years. The remaining 50 percent are then
distributed across the next three years. Our valuation
of avoided premature mortality applies a five percent
discount rate to the lagged estimates over the periods
2000 to 2005 and 2010 to 2015. We discount over the
period between the initial PM exposure change (either
2000 or 2010) and timing of the projected incidences.
H-27
-------
Many of the monetized health benefit categories
include overlapping health endpoints, creating the
potential for double-counting. In an effort to avoid
overstating the benefits, we do not aggregate all of the
quantified health effects. For example, "asthma
attacks" and "moderate to worse asthma", are all
considered components of the endpoint, "Any of 19
Respiratory Symptoms". Consequently, we present
the results but do not include them in our reported
total benefits figures. In other cases, there are
endpoints included in our aggregation of benefits that
appear to have overlapping health effects. For those
benefit categories that describe similar health effects,
it is important to keep in mind that estimated
incidences are based on unique portions of the
population.
H-
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table H-4
Primary Estimates of Health and Welfare Benefits Due to Criteria Pollutants in 2000
Monetary Benefits (in millions 1990$)
5th %ile
Mean
95th %ile
Mortality
Aqes 30+
$ 8,600
$ 63,000
$ 150,000
Chronic Illness
Chronic Bronchitis
$ 220
$ 3,600
$ 11,000
Chronic Asthma
29
140
240
Hospitalization
All Respiratory
$ 46
$ 78
$ 120
Total Cardiovascular
53
200
430
Asthma-Related ER Visits
0.1
0.6
1.8
Minor Illness
Acute Bronchitis
$ 0
$ 1.3
$ 3.3
URS
2.8
12
26
LRS
1.4
3.9
7.2
Respiratory Illness
0.4
2.5
6.1
Mod/Worse Asthma1
1.2
8.5
19
Asthma Attacks1
13
35
66
Chest tightness, Shortness of
Breath, or Wheeze
0
0.5
2.4
Shortness of Breath
0
0.3
0.7
Work Loss Days
180
210
240
MRAD/Any-of-19
420
760
1,100
Welfare
Decreased Worker Productivity
$ 460
$ 460
$ 460
Visibility
Recreational
1,700
2,000
2,300
Agriculture
46
450
860
Total Benefits2
$71,000
Note:
1 Moderate to worse asthma and asthma attacks are endpoints included in the definition of MRAD/Any of 19 respiratory effects.
Although valuation estimates are presented for these categories, the values are not included in total benefits to avoid the potential
for double-counting.
2 Summing 5th and 95th percentile values would yield a misleading estimate of the 5th and 95th percentile estimate of total health
benefits. For example, the likelihood that the 5th percentile estimates for each endpoint would simultaneously be drawn from a
Monte Carlo procedure is much less than 5 percent. As a result, we present only the total mean.
H-29
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Table H-5
Primary Estimates of Health and Welfare Benefits Due to Criteria Pollutants in 2010
Monetary Benefits (in millions 1990$)
Benefits Category 5th %ile Mean 95th %ile
Mortality
Ages 30+ $ 14,000 $ 100,000 $250,000
Chronic Illness
Chronic Bronchitis
Chronic Asthma
Hospitalization
All Respiratory
Total Cardiovascular
Asthma-Related ER Visits
Minor Illness
Acute Bronchitis
URS
LRS
Respiratory Illness
Mod/Worse Asthma1
Asthma Attacks1
Chest tightness, Shortness of
Breath, or Wheeze
Shortness of Breath
Work Loss Days
MRAD/Any-of-19
Welfare
Decreased Worker Productivity $710 $710 $710
Visibility
Recreational 2,500 2,900 3,300
Agriculture 7J 550 1,100
Total Benefits2 $110,000
Note:
1 Moderate to worse asthma, asthma attacks, and shortness of breath are endpoints included in the definition of MRAD/Any of 19
respiratory effects. Although valuation estimates are presented for these categories, the values are not included in total
benefits to avoid the potential for double-counting.
2 Summing 5th and 95th percentile values would yield a misleading estimate of the 5th and 95th percentile estimate of total health
benefits. For example, the likelihood that the 5th percentile estimates for each endpoint would simultaneously be drawn from a
Monte Carlo procedure is much less than 5 percent. As a result, we present only the total mean.
$ 360
40
$ 5,600
180
$ 18,000
300
$76 $130 $200
93 390 960
0.1 1 2.8
$ 0
4.2
2.2
0.9
1.9
20
0
0
300
680
$ 2.1
19
6.2
6.3
13
55
0.6
0.5
340
1,200
$ 5.2
39
12
15
29
100
3.1
1.2
380
1,800
H-30
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Uncertainties in the Valuation
Estimates
The uncertainty ranges for the results on the
present value of the aggregate measured monetary
benefits reported in Table H-4 and Table H-5 reflect
two important sources of measured uncertainty:
• Uncertainty about the avoided incidence of
health and welfare effects deriving from
the concentration-response functions,
including both selection of scientific
studies and statistical uncertainty from the
original studies;
• Uncertainty about the economic value of
each quantified health and welfare effect.
These aggregate uncertainty results incorporate many
decisions about analytical procedures and specific
assumptions discussed in the Appendices to this
report.
In order to provide a more complete
understanding of the economic benefit results, we
conduct sensitivity analyses which examine several
additional important aspects of the main analysis. We
begin with an analysis of the sources of the measured
aggregate uncertainty, identifying which of the
measured uncertainty components of incidence and
valuation for individual health effects categories drive
the overall uncertainty results. We then follow with
an examination of several issues involving the
estimated economic benefits of mortality. In the third
section, we provide some insight into the potential
effects of income growth on the valuation of health
effects.
Relative Importance of Different
Components of Uncertainty
The estimated uncertainty ranges in our primary
results tables, Table H-4 and Table H-5, reflect the
measured uncertainty associated with both avoided
incidence and economic valuation. A better
understanding of the relative influence of individual
Figure H-4
Analysis of Contribution of Key Parameters to
Quantified Uncertainty
V5
O
S; $300
C
o
$250-
S $200
£ $150
C
ffl $100
o
O
H
$50
$0
f 95th %ile
Mean
<
) (
> (
> ffl a e e e
5th %ile
i
i i i i i
e
o
&
o
U
o
11
& 8
13 §
« V
5 y
5 c
3 §
3 J3 '-g
o o =2
U c ,3
8 8|
03 >
CL)
O o
e
•a a
o o
£ 0
o
m
£
•t; o
^3 '-3
£ 3
tsi
E? o
^ o\ ^
(D
3 ^ 3
< O o
Ph «
S ~
Contribution to Variance from the Mean
H-31
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
variables on the overall uncertainty in the analysis can
be gained by isolating the individual effects of
important variables on the range of estimated total
benefits. This can be accomplished by holding all the
inputs to the Monte Carlo uncertainty analysis
constant (at their mean values), and allowing only one
variable — for example, the economic valuation of
mortality — to vary across the range of that variable's
quantified uncertainty. The sensitivity analysis then
isolates how this single source of variability
contributes to the variation in the primary estimates of
total benefits. The results are summarized in Figure
H-4. The nine individual uncertainty factors that
contribute the most to the overall uncertainty are
shown in Figure H-4, ordered by the relative
significance of their contribution to overall
uncertainty. Each of the additional sources of
quantified uncertainty in the overall analysis not
shown contribute a smaller amount of uncertainty to
the estimates of monetized benefits than the sources
that are shown.
Economic Benefits Associated with
Reducing Premature Mortality
Because the economic benefits associated with
premature mortality are the largest source of
monetized benefits in the analysis, and because the
uncertainties in both the incidence and value of
premature mortality are the most important sources of
uncertainty in the overall analysis, it is useful to
examine the mortality benefits estimation in greater
detail. We begin with a discussion of the uncertainties
and possible biases related to the "benefits transfer"
approach employed to develop our YSL estimate. We
then discuss an alternative method for the valuation of
reduced premature mortality, value of statistical life
year (VSLY). We conclude this section with a
sensitivity test that compare the benefit estimates
using a YSL approach and a VSLY approach. Given
the lag structure employed in estimating reduced
premature mortality, we also provide alternative
calculations for the valuation of this benefits category
using two additional discount rates, three and seven
percent.
Benefits Tranfer and VSL
The analytical procedure used in the main analysis
to estimate the monetary benefits of avoided
premature mortality assumes that the appropriate
economic value for each incidence is a value from the
currently accepted range of the value of a statistical
life. As discussed above, the estimated value per
predicted incidence of excess premature mortality is
modeled as a Weibull distribution, with a mean value
of $4.8 million and a standard deviation of $3.2
million. This estimate is based on 26 studies of the
value of mortal risks.
There is considerable uncertainty as to whether
the 26 studies on the value of a statistical life provide
adequate estimates of the value of a statistical life
saved by air pollution reduction. Although there is
considerable variation in the analytical designs and
data used in the 26 underlying studies, the majority of
the studies involve the value of risks to a middle-aged
working population. Most of the studies examine
differences in wages of risky occupations, using a
wage-hedonic approach. Certain characteristics of
both the population affected and the mortality risk
facing that population are believed to affect the
average willingness to pay (WTP) to reduce the risk.
The appropriateness of a distribution of WTP
estimates from the 26 studies for valuing the
mortality-related benefits of reductions in air pollution
concentrations therefore depends not only on the
quality of the studies (i.e., how well they measure what
they are trying to measure), but also on (1) the extent
to which the risks being valued are similar, and (2) the
extent to which the subjects in the studies are similar
to the population affected by changes in pollution
concentrations. As discussed below, there are
possible sources of both upward and downward bias
in the estimates provided by the 26 studies when
applied to the population and risk being considered in
this analysis.
Although there may be several ways in which job-
related mortality risks differ from air pollution-related
mortality risks, the most important difference may be
that job-related risks are incurred voluntarily whereas
air pollution-related risks are incurred involuntarily.
H-32
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
There is some evidence (see, for example, Violette and
Chestnut, 1983) that people will pay more to reduce
involuntarily incurred risks than risks incurred
voluntarily. If this is the case, WTP estimates based
on wage-risk studies may be downward biased
estimates of WTP to reduce involuntarily incurred air
pollution-related mortality risks.
Another possible difference related to the nature
of the risk may be that some workplace mortality risks
tend to involve sudden, catastrophic events (e.g.,
workplace accidents), whereas air pollution-related
risks tend to involve longer periods of disease and
suffering prior to death. Several studies indicate that
the value people place on mortality risk reduction may
depend on the nature of the risk (e.g., Fisher et al.
1989; Beggs 1984). Some evidence suggests that WTP
to avoid a risk of a protracted death involving
prolonged suffering and loss of dignity and personal
control is greater than the WTP to avoid a risk (of
identical magnitude) of sudden death. Some
workplace risks, such as risks from exposure to toxic
chemicals, may be more similar to pollution-related
risks. It is not clear, however, what proportion of the
workplace risks in the wage-risk studies were related to
workplace accidents and what proportion were risks
from exposure to toxic chemicals. To the extent that
the mortality risks addressed in this assessment are
associated with longer periods of illness or greater
pain and suffering than are the risks addressed in the
valuation literature, the WTP measurements employed
in the present analysis would reflect a downward bias.
If the individuals who die prematurely from air
pollution are consistently older than the population in
the valuation studies, the mortality valuations based on
middle-aged people may provide a biased estimate of
the willingness to pay of older individuals to reduce
mortal risk. There is some evidence to suggest that
the people who die prematurely from exposure to
ambient particulate matter tend to be older than the
populations in the valuation studies. In the general
U.S. population far more older people die than
younger people; 88 percent of the deaths are among
people over 64 years old. It is difficult to establish the
proportion of the pollution-related deaths that are
among the older population because it is impossible to
isolate individual cases where one can say with even
reasonable certainty that a specific individual died
because of air pollution.
There is considerable uncertainty whether older
people will have a greater willingness to pay to avoid
risks than younger people. There is reason to believe
that those over 65 are, in general, more risk averse
than the general population, while workers in
wage-risk studies are likely to be less risk averse than
the general population. More risk averse people
would have a greater willingness to pay to avoid risk
than less risk averse people. Although the list of
recommended studies excludes studies that consider
only much-higher-than-average occupational risks,
there is nevertheless likely to be some selection bias in
the remaining studies — that is, these studies are likely
to be based on samples of workers who are, on
average, more risk-loving than the general population.
In contrast, older people as a group exhibit more risk
averse behavior.
The direction of bias resulting from the age
difference is unclear, particularly because age is
confounded by risk aversion (relative to the general
population). It could be argued that, because an older
person has fewer expected years left to lose, his WTP
to reduce mortality risk would be less than that of a
younger person. This hypothesis is supported by one
empirical study, Jones-Lee et al. (1985), that found the
value of a statistical life at age 65 to be about 90
percent of what it is at age 40. Citing the evidence
provided by Jones-Lee et al. (1985), a recent sulfate-
related health benefits study conducted for EPA (U.S.
EPA, 1995) assumes that the value of a statistical life
for those 65 and over is 75 percent of what it is for
those under 65. In addition, it might be argued that
because the elderly have greater average wealth than
those younger, the affected population is also
wealthier, on average, than wage-risk study subjects,
who tend to be blue collar workers. It is possible,
however, that among the elderly it is largely the poor
elderly who are most vulnerable to pollution-related
mortality risk (e.g., because of generally poorer health
care). If this is the case, the average wealth of those
affected by a pollution reduction relative to that of
subjects in wage-risk studies is uncertain. In addition,
H-33
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
the workers in the wage-risk studies will have
potentially more years remaining in which to acquire
streams of income from future earnings.
There is substantial evidence that the income
elasticity of WTP for health risk reductions is positive
(see, for example, Alberini et al., 1994; Mitchell and
Carson, 1986; Loehman and Yo Hu De, 1982;
Gerking et al., 1988; and Jones-Lee et al., 1985),
although there is uncertainty about the exact value of
this elasticity. Individuals with higher incomes (or
greater wealth) should be willing to pay more to
reduce risk, all else equal, than individuals with lower
incomes or wealth. This does not imply that
individuals with higher incomes are willing to pay
proportionally higher values. While many analyses
assume income elasticity of willingness to pay is unit
elastic (i.e., ten percent higher income level implies a
ten percent higher willingness to pay to reduce risk
changes), empirical evidence suggests that income
elasticity is substantially less than one.
The effects of income changes on WTP estimates
can influence benefit estimates in two different ways:
(i) as longitudinal changes that reflect estimates of
income change in the affected population over time,
and (ii) as cross-sectional changes based on differences
in income between study populations and the attracted
populations. Empirical evidence of the effect of
income on WTP gathered to date is based on studies
examining cross-sectional data. Income elasticity
adjustments to better account for changes over time,
therefore, will necessarily be based on potentially
inappropriate data.23
The need to adjust wage-risk-based WTP
estimates downward because of the likely upward bias
introduced by the age discrepancy has received
significant attention (see Chestnut, 1995; IEc, 1992).
If the age difference were the only difference between
the population affected by pollution changes and the
subjects in the wage-risk studies, there might be some
23For more information on the potential impact of income
elasticity on the valuation of health benefits, see the following
section, "Sensitivity Test for Impact of Income Changes Over
Time."
justification for trying to adjust the point estimate of
$4.8 million downward. Even in this case, however,
the degree of the adjustment would be unclear. There
is good reason to suspect, however, that there are
biases in both directions. Because in each case the
extent of the bias is unknown, the overall direction of
bias in the mortality values is similarly unknown.
Adjusting the estimate upward or downward to
compensate for any one source of bias could therefore
increase the degree of bias. Therefore, the range of
values from the 26 studies is used in the primary
analysis without adjustment.
VSLY
An alternative valuation of avoided premature
mortality is to use the YSLY. This approach uses life-
years lost as the unit of measure, rather than
estimating a single value of a statistical life lost
(applicable to all ages). With statistical life-years lost
as the unit of measure, the valuation depends on (1)
how many years of expected life are lost, (2) the
individual's discount rate, and (3) whether the value of
an undiscounted statistical life-year is the same no
matter which life-year it is (e.g., the undiscounted
value of the seventy-fifth year of life is the same as the
undiscounted value of the fortieth year of life).
We estimate the value of a statistical life-year
assuming that the value of a statistical life is directly
related to remaining life expectancy and a constant
value for each life-year. Such an approach results in
smaller values of a statistical life for older people, who
have shorter life expectancies, and larger values for
younger people. For example, if the $4.8 million
mean value of avoiding death for people with a 35
year life expectancy is assumed to be the discounted
present value of 35 equal-valued statistical life-years,
the implied value of each statistical life-year is
$293,000. This values assumes a five percent discount
rate and that the undiscounted value of a life-year is
the same no matter when it occurs in an individual's
life.
To obtain estimates of the number of air
pollution-related deaths in each age cohort, it is
preferable to have age-specific relative risks. Many of
H-34
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
the epidemiological studies, however, do not provide
any estimate of such age-specific risks. In this case,
the age-specific relative risks must be assumed to be
identical. Some epidemiology studies on PM do
provide some estimates of relative risks specific to
certain age categories. The limited information that is
available suggests that relative risks of mortality
associated with exposure to PM are greater for older
people. Most of the available information comes
from short-term exposure studies. There is
considerable uncertainty in applying the evidence from
short-term exposure studies to results from long-term
(chronic exposure) studies. However, using the
available information on the relative magnitudes of the
relative risks, it is possible to form a preliminary
assessment of the relative risks by different age
classes.
The analysis presented below uses two alternative
assumptions about age-specific risks: (1) there is a
constant relative risk (obtained directly from the
health literature) that is applicable to all age cohorts,
and (2) the relative risks differ by age, as estimated
from the available literature. Estimates of age-specific
PM-10 coefficients (and, from these, age-specific
relative risks) were derived from the few age-specific
PM-10 or TSP coefficients reported in the
epidemiological literature. These estimates in the
literature were used to estimate the ratio of each age-
specific coefficient to a coefficient for "all ages" in
such a way that consistency among the age-specific
coefficients is preserved — that is, that the sum of the
health effects incidences in the separate, non-
overlapping age categories equals the health effects
incidence for "all ages." These ratios were then
applied to the coefficient from Pope et al. (1995).
Details of this approach are provided in Post and
Deck (1996). Because Pope et al. considered only
individuals age 30 and older (instead of all ages), the
resulting age-specific PM coefficients may be slightly
different from what they would have been if the ratios
had been applied to an "all ages" coefficient. The
differences, however, are likely to be minimal and well
within the error bounds of this exercise. The age-
specific relative risks used in the example below
assume that the relative risks for people under 65 are
only 16 percent of the population-wide average
relative risk, the risks for people from 65 to 74 are 83
percent of the population-wide risk, and people 75
and older have a relative risk 55 percent greater than
the population average. Details of this approach are
provided in Post and Deck (1996).
The life-years lost approach also requires an
estimate of the number of life-years lost by a person
dying prematurely at each given age. The approach
developed for this analysis assumes that exposure to
elevated levels of PM increases the probability of
dying at a specific age. Increasing the probability of
dying at each age lowers the life expectancy for each
age cohort. The average number of life-years lost will
depend on the distribution of ages in the population
in a location. In addition, this analysis incorporates
the five-year PM mortality lag structure described in
Chapter 5 and Appendix D. It distributes the
mortality for each cohort across a five-year period (25
percent in each of the first two years, 16.7 percent in
each of the remaining years) and adjusts the loss of
life expectancy accordingly. That is, when applying
the lag assumption within a given cohort, individuals
who die later are expected to lose fewer life years than
those who die earlier. Further, this analysis applies a
five percent discount rate when calculating the present
discounted value of the avoided losses of life
expectancy in each cohort over the five-year lag
period.
The life-years lost approach used here assumes
that people who die from air pollution are typical of
people in their age group. The estimated value of the
quantity of life lost assumes that the people who die
from exposure to air pollution had an average life
expectancy. However, it is possible that the people
who die from air pollution are already in ill health, and
that their life expectancy is less than a typical person
of their age. If this is true, then the number of life
years lost per PM-related death would be lower than
calculated here, and the economic value would be
smaller.
The extent to which adverse effects of particulate
matter exposure are differentially imposed on people
of advanced age and/or poor health is one of the
most important current uncertainties in air pollution-
related health studies. There is limited information,
H-35
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
primarily from the short-term exposure studies, which
suggests that at least some of the estimated premature
mortality is imposed disproportionately on people
who are elderly and/or of poor health. Rowlatt, et al.
(1998) indicate that at risk individuals include those
who have suffered strokes or are suffering from
cardiovascular disease and angina. The Criteria
Document for Particulate Matter (U.S. EPA, 1996),
however, identifies only two studies which attempt to
evaluate the disproportionality in premature mortality
among people who are elderly and/or sickly. Spix et
al. (1994) suggests that a small portion of the PM-
associated mortality occurs in individuals who would
have died in a short time anyway. Cifuentes and Lave
(1996) found that 37 to 87 percent of the deaths from
short-term exposure could have been premature by
only a few days, although their evidence is
inconclusive.
Prematurity of death on the order of only a few
days is likely to occur largely among individuals with
pre-existing illnesses. Such individuals might be
particularly susceptible to a high PM day. To the
extent that the pre-existing illness is itself caused by or
exacerbated by chronic exposure to elevated levels of
PM, however, it would be misleading to define the
prematurity of death as only a few days. In the
absence of chronic exposure to elevated levels of PM,
the illness would either not exist (if it was caused by
the chronic exposure to elevated PM) or might be at
a less advanced stage of development (if it was not
caused by but was exacerbated by elevated PM levels).
The prematurity of death should be calculated as the
difference between when the individual died in the
"elevated PM" scenario and when he would have died
in the 'low PM" scenario. If the pre-existing illness
was entirely unconnected with chronic exposure to
PM in the "elevated PM" scenario, and if the
individual who dies prematurely because of a peak PM
day would have lived only a few more days, then the
prematurity of that PM-related death is only those few
days. If, however, in the absence of chronic exposure
to elevated levels of PM, the individual's illness would
have progressed more slowly, so that, in the absence
of a particular peak PM day the individual would have
lived several years longer, the prematurity of that PM-
related death would be those several years.
Long-term studies provide evidence that a portion
of the loss of life associated with long-term exposure
is independent of the death from short-term
exposures, and that the loss of life-years measured in
the long-term studies could be on the order of years.
If much of the premature mortality associated with
PM represents short term prematurity of death
imposed on people who are elderly and/or of ill
health, the estimates of the monetary benefits of
avoided mortality may overestimate society's total
willingness to pay to avoid particulate matter-related
premature mortality. On the other hand, if the
premature mortality measured in the chronic exposure
studies is detecting excess premature deaths which are
largely independent of the deaths predicted from the
short term studies, and the disproportionate effect on
the elderly and/or sick is modest, the benefits
measured in this report could be underestimates of
the total value. At this time there is insufficient
information from both the medical and economic
sciences to satisfactorily resolve these issues from a
theoretical/analytical standpoint. Until there is
evidence from the physical and social sciences which
is sufficiently compelling to encourage broad support
of age-specific values for reducing premature
mortality, EPA will continue to use for its primary
analyses a range of values for mortality risk reduction
which assumes society values reductions in pollution-
related premature mortality equally regardless of who
receives the benefit of such protection.
Sensitivity Test of Benefits Due to
Reduced Premature Mortality Valuation
Examining the sensitivity of the total benefits of
reduced premature mortality to alternative valuation
techniques does provide some illumination to the
potential impacts of alternative approaches. This
section presents alternative results to our primary
estimate of mortality valuation using the life-years lost
approach, and also examine the effects of alternative
discount rates.
The life-years lost approach also requires an
estimate of the number of life-years lost by a person
dying prematurely at each given age. The approach
developed for this analysis assumes that exposure to
H-36
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
elevated levels of PM increases the probability of
dying at a specific age. Increasing the probability of
dying at each age lowers the life expectancy for each
age cohort. The average number of life-years lost will
depend on the distribution of ages in the population
in a location. In addition, this analysis incorporates
the five-year PM mortality lag structure described in
Chapter 5 and Appendix D. It distributes the
mortality for each cohort across a five-year period (25
percent in each of the first two years, 16.7 percent in
each of the remaining years) and adjusts the loss of life
expectancy accordingly. That is, when applying the lag
assumption within a given cohort, individuals who die
later are expected to lose less live expectancy than
those who die earlier. Further, this analysis applies a
five percent discount rate when summing the value of
the avoided losses of life expectancy in each cohort
over the five-year lag period.
The alternative central estimates for avoided PM-
related premature mortality using a five percent dis-
count rate are $33 billion in 2000 and $53 billion in
2010. The YSLY approach results in estimates that
are almost 50 percent lower than our primary est-
imates of benefits due to avoided pre-mature mor-
tality. The sensitivity analysis, however, indicates that
the pattern of monetized mortality benefits with each
valuation procedure is essentially invariant to the dis-
count rate. We summarize these results in Table H-6.
We emphasize that the results of the YSLY
approach to valuing avoided mortality benefits
represent a crude estimate of the value of changes in
age-specific life expectancy. These results should be
interpreted cautiously, due to the several significant
assumptions required to generate a monetized
estimate of life years lost from the relative risks
reported in the Pope et al., 1995 study and the
available economic literature. These assumptions
include, but are not limited to: extrapolation of the
age distribution of the U.S. population in future years;
assumptions about the age-specificity of the relative
risk reported by Pope et al., 1995; assumptions about
the life expectancy of different age groups;
assumption of a particular lag structure; assumptions
about the age-specificity of the lag period (if any);
derivation of VSLY estimates from VSL estimates;
assumptions about the variation in VSLY with age;
and selection of an appropriate rate at which to
discount the lagged estimates of life years lost.
Changes in any of these assumptions could
significantly affect the VSLY benefit estimate. For
example, if we were to assume no lag period for PM-
related mortality effects instead of the five-year lag
structure, VSLY benefit estimates would increase
from $53 billion to $61 billion.
Table H-6
Sensitivity Analysis of Alternative Discount Rates on the Valuation of Reduced Premature
Mortality
Benefit Category &
Discount Rate
2000 (in millions, 1990$) 2010 (in millions, 1990$)
5th %ile Central 95th %ile 5th %ile Central 95th %ile
VSL Approach
3% Discount Rate
5% Discount Rate
7% Discount Rate
$ 8,900 $65,000 $ 150,000 $ 14,000 $ 100,000 $ 250,000
8,600 63,000 150,000 14,000 100,000 250,000
8,300 61,000 150,000 14,000 97,000 240,000
VSLY Approach
3% Discount Rate
5% Discount Rate
7% Discount Rate
$4,600 $ 30,000 $68,000 $ 7,400 $48,000 $110,000
5,000 33,000 74,000 8,100 53,000 120,000
5,400 35,000 80,000 8,800 57,000 130,000
Note: The discount rate affects the benefits estimates of VSL and VSLY approach differently. With the VSL approach, higher
discount rates lead to lower estimates because of the lag structure. With the VSLY approach, the higher discount rates lead to
higher estimates because of its affect on the annualized values.
H-37
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Sensitivity Test for Impact of Income
Changes Over Time
As an illustrative calculation, we adjust
willingness-to-pay (WTP) measures to reflect the
expected increase in real income over the full period
of the analysis, 1990 to 2010. Our procedure results
in an upward adjustment to more accurately reflect the
valuation of improved health as income increases over
time. In this section, we describe the procedure we
use and the results of our illustrative calculation.
Background and Methodology
Economists use income elasticity to evaluate how
private and public goods are valued based on the
interaction between income changes and demand. A
negative relationship between income and demand for
a good implies that the good is an inferior good. An
individual demands less of a good as income rises. A
positive relationship between income and the demand
for a good implies that the good is normal (i.e.,
income elasticity is greater than zero). As income rises
an individual demands more of a good. Depending on
the relative responsiveness of demand to income
changes, normal goods are characterized as a necessity
or a luxury. When income elasticity is between 0 and
+ 1, the good is considered a necessity (i.e., demand is
not significantly responsive to income). In contrast,
when income elasticity exceeds +1, the good is
considered a luxury (i.e., the relative increase in the
good's demand exceeds the increase in income).
The determination of a public good as inferior or
normal based on income elasticity is complicated by
its nonrival nature. In the case of a private good,
varying the level of consumption is measured as a
marginal change and implies that an individual will
adjust his or her consumption level of other good(s).
Consequently, income elasticity of demand estimates
a change in quantity consumed, and not necessarily a
change in utility (or the individual's well-being). With
public goods, the conceptual logic is different.
Income elasticity of WTP for public goods measures
changes in consumer surplus. For example, one
person enjoying the benefits of cleaner air does not
reduce the probability of another person enjoying the
same benefits. There are no apparent mechanisms for
regulating who specifically will enjoy the benefits. In
other words, there is no direct relationship between an
individual's WTP and level of consumption.24 The
consumption level of public goods is exogenous to
the individual's budget constraint. At the same time,
WTP for a public good is not exogenous. An
individual, therefore, must consider how his or her
WTP affects the allocation of income among private
and public goods.25
Flores and Carson (1997) provide examples of
how income elasticity can change depending on how
the good is defined (i.e., private or public). Given the
divergence between private and public goods, they
conclude that income elasticity of WTP and income
elasticity of demand are related. The relationship does
not imply that knowledge of income elasticity of
demand is sufficient to estimate income elasticity of
WTP given that the income elasticity of WTP depends
on factors that cannot be observed.
In addition to the theoretical issues associated
with WTP for public goods, there are important
empirical issues. We are interested in how WTP
changes with respect to increases in U.S. median
income. Measuring changes due to growth in median
income reflect shifts in overall preferences and utility
(or in the case of public goods, social welfare). This
type of analysis requires time series data.
Unfortunately, there are very few relevant studies that
use this approach to estimate income elasticity.26
Consequently, we must rely on income elasticities
estimated from cross-sectional data. The estimates
24The nonrival nature of public goods implies that the
marginal social cost of consuming an additional unit of benefit is
zero.
25CV studies solicit WTP estimates that are subject to the
respondent's current budget constraint. The budget share factor
requires that the income elasticities (for all consumed goods) sum
to one. This generally implies that income elasticity of any single
good is substantially less than one.
26Available studies using time series data estimate income
elasticity of public health care expenditures by analyzing changes
in government spending relative to gross domestic product
(GDP). These studies are not particularly applicable to the
valuation methodology used in the present study.
H-38
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
reflect differences in willingness to pay for improved
health among various income levels. They are
measures of an individual's preferences and expected
utility given the person's current state (i.e., in the
present).
There are several issues associated with the
application of cross-sectional results to estimate
longitudinal changes (i.e., changes over time). Most
important is the potential for misinterpretation of our
recommended application of income elasticity
adjustment. Although we outline an approach that
uses income elasticities derived from cross-sectional
data, the adjustment is solely a proxy for how
preferences and utility may change as projected overall
average income (i.e., real GDP per capita) increases
from 1990 to 2010. Application of these income
elasticity estimates does not imply a strategy for
adjusting benefits valuation by level of household
income in any given year.
Derivation of Elasticity Estimates
Based on our review of the available income
elasticity literature, we conducted sensitivity analyses
that characterize how the valuation of human health
benefits may increase with a rise in real U.S. income.
Given the range of different methodological
approaches and limited available research, we calculate
a range of illustrative values. Table H-7 summarizes
the income elasticities we used to conduct the
sensitivity analysis.
Table H-7
Elasticity Values for Conducting Sensitivity Analysis
Health Endpoint
Lower Estimate
Central Estimate
Upper Estimate
Minor Health Effect
0.04
0.14
0.30
Severe and Chronic
Health Effects
0.25
0.45
0.60
Premature Mortality
0.08
0.40
1.00
Note: Sources for the derivation of these values can be found in Industrial Economics 1999.
Reported income elasticities suggest that the
severity of the morbidity endpoint is a primary
determinant of the strength of the relationship
between changes in income and the willingness to pay.
Without accounting for severity, there is a fairly wide
range of values for income elasticity, 0.04 to 0.60.
Estimates are more closely clustered if we account for
the seriousness of the health effect. For the purposes
of a sensitivity analysis, we use two different ranges
based on whether morbidity endpoints are minor or
severe. With respect to minor health effects, we use
lower and upper values of 04 and 0.30, respectively.
The central estimate is 0.14. For conducting a
sensitivity test of the income elasticity effect on WTP
to avoid severe health effects, we use a lower and
upper estiamtes of 0.25 and 0.60, with 0.45 as the
central estimate. The lower and upper estimates
reflect the lowest and highest estimates derived from
our literature review. The central estimate is the
midpoint of the averages from each study.
With respect to YSL, estimates of income
elasticity range from 0.08 to 1.10. We use lower and
upper estimates that reflect the full range of values.
The central estimate, 0.40, represents the midpoint
between the average low value and the average high
value of the studies we reviewed.
H-39
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Illustrative Calculations —
Morbidity Benefits Estimates
Table H-8 provides a simplified example of how
application of the elasticity ranges we derive could
affect benefits estimates. For illustrative purposes, we
use the WTP to avoid an asthma attack to represent a
minor health effect and WTP to avoid a case of
chronic bronchitis to represent a severe health effect.
By the year 2010, the effect of income growth on
WTP for a minor health effect can increase between
one and eight percent, with the central estimate
indicating three percent growth. The WTP to avoid a
severe health effect grows faster with 2010 estimates,
ranging between seven and sixteen percent and with
the central estimate increasing by thirteen percent.
Table H-8
Illustrative Adjustment to Estimates of WTP to Avoid Morbidity
WTP Estimate (1990 Dollars)1
Year
US Population
(in millions)
Real GDP
(in millions)
Income
Lower
Estimate
Central
Estimate
Upper
Estimate
Minor Health Effect-Asthma
Ey=0.04
Ey=0.14
Ey=0.30
1990
249,440
5,744
23,026
$32
$32
$32
2000
274,634
7,123
25,936
$32.20
$32.50
$33.20
2010
297,716
8,959
30,092
$32.30
$33.20
$34.70
Severe Health Effect- Chronic Bronchitis
Ey=0.25
Ey=0.45
Ey=0.60
1990
249,440
5,744
23,026
$260,000
$260,000
$260,000
2000
274,634
7,123
25,936
$267,850
$274,300
$279,240
2010
297,716
8,959
30,092
$277,990
$293,280
$305,290
Note:
1 WTP estimates are reported in undiscounted 1990 dollars and represent value per case.
H-40
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Illustrative Calculations —
VSL Estimate
We characterize the potential effect of income
elasticity on the YSL estimate in Table H-9. An
income elasticity of 0.08 demonstrates the effect of a
slight adjustment to the YSL estimates as median
income gradually rises. As shown in the figure,
between 1990 and 2010, the YSL estimates increase by
approximately two percent. The central estimate, 0.40,
demonstrates that by 2010, a thirty percent increase in
median income would result in VSL increasing by
approximately eleven percent. The upper bound value
demonstrates the effect of assuming one as the value
of income elasticity. In this twenty year period of the
prospective analysis, the YSL estimate would increase
from $4.8 to $6.3 million if income elasticity equals
one.
Table H-9
Illustrative Adjustment to Estimates of The Value of Statistical Life
Value of Life Estimate (in thousands)1
Year
US Population
(in millions)
Real GDP
(in millions)
Income
Lower
Estimate
Ey=0.08
Central
Estimate
Ey=0.40
Upper
Estimate
Ey=1.0
1990
249,440
5,744
23,026
$4,800
$4,800
$4,800
2000
274,634
7,123
25,936
$4,848
$5,036
$5,410
2010
297,716
8,959
30,092
$4,905
$5,345
$6,271
Note:
1 Value of life estimates reported in undiscounted 1990 dollars.
H-41
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
References
Abbey, D.E., F. Petersen, P. K. Mills, and W. L. Beeson. 1993. "Long-Term Ambient Concentrations of
Total Suspended Particulates, Ozone, and Sulfur Dioxide and Respiratory Symptoms in a
Nonsmoking Population." Archives of Environmental Health 48(1): 33-46.
Abt Associates, Inc. 1992. The Medical Costs of Five Illnesses Related to Exposure to Pollutants. Prepared
for U.S. EPA, Office of Pollution Prevention and Toxics, Washington, DC.
Abt Associates, Inc. 1996. Section 812 Retrospective Analysis: Quantifying Health and Welfare Benefits.
Draft. Prepared for U.S. EPA, Office of Policy Planning and Evaluation, Washington DC. May.
Alberini, A., A. Krupnick, M. Cropper, and W. Harrington. 1994. "Air Quality and the Value of Health in
Taiwan." Paper presented at the annual meeting of the Eastern Economics Association, Boston,
Massachusetts, March.
Beggs, Steven D. 1984. Diverse Risks and the Relative Worth of Government Health and Safety Programs:
An Experimental Survey. U.S. Environmental Protection Agency, EPA-223-04-85-005. June.
Blumenschein, Karen and Magnus Johannesson. 1998. "Relationship Between Quality of Life Instruments,
Health State Utilities, and Willingness to Pay in Patients with Asthma." Annals of Allergy, Asthma,
and Immunology 80:189-194.
Brookshire, David S., Ralph C. dArge, William D. Schulze and Mark A. Thayer. 1979. Methods
Development for Assessing Air Pollution Control Benefits, Vol. II: Experiments in Valuing Non-
Market Goods: A Case Study of Alternative Benefit Measures of Air Pollution Control in the South
Coast Air Basin of Southern California. Prepared for the U.S. Environmental Protection Agency,
Office of Research and Development.
Chestnut, Lauraine G. 1995. Dollars and Cents: The Economic and Health Benefits of Potential Particulate
Matter Reductions in the United States. Prepared for the American Lung Association.
Chestnut, Lauraine G. and Robert D. Rowe. 1989. "Economic Valuation of Changes in Visibility: A State of
the Science Assessment for NAPAP," as cited in National Acid Precipitation Assessment Program,
Methods for Valuing Acidic Deposition and Air Pollution Effects. NAPAP State of Science and
State of Technology Report No. 27, Part B. December.
Chestnut, Lauraine G. and Robin Dennis. 1997. "Economic Benefits of Improvements in Visibility: Acid
Rain Provisions of the 1990 Clean Air Act Ammendments." Journal of Air and Waste Management
Association 47:395-402.
Cifuentes, L. and L.B. Lave. 1996. "Association of Daily Mortality and Air Pollution in Philadelphia, 1983-
1988." J. Air Waste Manage. Assoc.: in press.
Crocker T. D. and R. L. Horst, Jr. 1981. "Hours of Work, Labor Productivity, and Environmental
Conditions: a Case Study." The Review of Economics and Statistics 63:361-368.
H-42
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Cropper, M.L. and A.J. Krupnick. 1990. "The Social Costs of Chronic Heart and Lung Disease," Resources
for the Future Discussion Paper QE 89-16-REV.
Cropper, Maureen L. and Frances G. Sussman. 1990. "Valuing Future Risks to Life." Journal of
Environmental Economics and Management. 19:160-174.
Cutler, David M. and Elizabeth Richardson. 1998. "The Value of Health: 1970—1990." The American
Economic Review. 88(2):97-100. May.
Desvouges, William H., F. Reed Johnson, and H. Spencer Banzhaf. 1998. Environmental Policy Analysis with
Limited Information, Principles and Applications of the Transfer Method. Edward Elgar Publishing
Limited: Northampton.
Dickie, M. et al. 1991. Reconciling Averting Behavior and Contingent Valuation Benefit Estimates of
Reducing Symptoms of Ozone Exposure (draft), as cited in Neumann, J.E., Dickie, M.T., and R.E.
Unsworth. 1994. Industrial Economics, Incorporated. Memorandum to Jim DeMocker, U.S. EPA,
Office of Air and Radiation. March 31.
Elixhauser, A., R.M. Andrews, and S. Fox. 1993. Clinical Classifications for Health Policy Research:
Discharge Statistics by Principal Diagnosis and Procedure. Agency for Health Care Policy and
Research (AHCPR), Center for General Health Services Intramural Research, U.S. Department of
Health and Human Services.
Fisher, Ann, Lauraine G. Chestnut, and Daniel M. Violette. 1989. "The Value of Reducing Risks of Death:
A Note on New Evidence." Journal of Policy Analysis and Management. 8(1):88-100.
French, Michael T. and Josephine A. Mauskopf. 1992. "A Quality-of-Life Method for Estimating the Value
of Avoided Morbidity." American Journal of Public Health. 82(11) :1553-1555. November.
Garber, Alan M. and Charles E. Phelps. 1997. "Economic Foundations of Cost-Effectiveness Analysis."
Journal of Health Economics. 16:1-31.
Gerking, S., M. DeHaan, and W. Schulze. 1988. "The Marginal Value of Job Safety: A Contingent Valuation
Study." Journal of Risk and Uncertainty 1: 185-199.
Industrial Economics, Incorporated (TEc). 1992. Approaches to Environmental Benefits Assessment to
Support the Clean Air Act Section 812 Analysis. Prepared by Robert E. Unsworth, James E.
Neumann, and W. Eric Browne, for Jim DeMocker, Office of Policy Analysis and Review, Office of
Air and Radiation, U.S. Environmental Protection Agency. 6 November.
Industrial Economics, Incorporated (TEc). 1993a. "Analysis of Visibility Valuation Issues for the Section
812 Study," Memorandum to Jim DeMocker, Office of Policy Analysis and Review, Office of Air
and Radiation, U.S. Environmental Protection Agency, prepared by Jim Neumann, Lisa Robinson,
and Bob Unsworth. September 30.
H-43
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Industrial Economics, Incorporated (TEc). 1997. "Visibility Valuation for the CAA Section 812
Retrospective Analysis," Memorandum to Jim DeMocker, Office of Policy Analysis and Review,
Office of Air and Radiation, U.S. Environmental Protection Agency, prepared by Michael H. Hester
and James E. Neumann. 18 February.
Industrial Economics, Incorporated (TEc). 1999. "Recommended Approach to Adjusting WTP Estimates to
Reflect Changes in Real Income," Memorandum to Jim DeMocker, Office of Policy Analysis and
Review, Office of Air and Radiation, U.S. Environmental Protection Agency, prepared by Naomi
Kleckner and James E. Neumann. 3 June.
Ireland, Thomas R. and Roy Gilbert. 1998. "Supramonetary Values, the Value of Life, and the Utility
Theory Meanings of Tort Recovery." Journal of Forensic Economics. 11:189-201.
Irwin, Julie, William Schulze, Gary McClelland, Donald Waldman, David Schenk, Thomas Stewart, Leland
Deck, Paul Slovic, Sarah Lictenstein, and Mark Thayer. 1990. Valuing Visibility: A Field Test of the
Contingent Valuation Method. Prepared for Office of Policy, Planning and Evaluation, U.S.
Environmental Protection Agency. March.
Johannesson, Magnus and Per-Olov Johansson. 1996. "To Be or Not to Be, That is the Question: An
Empirical Study of the WTP for an Increased Life Expectancy at an Advanced Age." Journal of
Risk and Uncertainty. 13:163-174.
Johannesson, Magnus and Per-Olov Johansson. 1997. "The Value of Life Extension and the Marginal Rate of
Time Preference: A Pilot Study." Applied Economic Letters. 4:53-55.
Johannesson, Magnus, Per-Olov Johansson, and Karl-Gustav Lofgren. 1997. "On the Value of Changes in Life
Expectancy: Blips Versus Parametric Changes." Journal of Risk and Uncertainty. 15:221-239.
Johansson, Per-Olov. 1995. Evaluating Health Risks: An Economic Approach. Cambridge University
Press: Great Britain.
Jones-Lee, M.W., et al. 1985. "The Value of Safety: Result of a National Sample Survey." Economic Journal
95(March): 49-72.
Jones-Lee, M.W. 1992. "Paternalistic Altruism and the Value of Statistical Life." The Economic Journal.
102:80-90. January.
Jones-Lee, M.W., M. Hammerton, and P.R. Philips. 1985. "The Value of Safety: Results of a National Sample
Survey. Economic Journal. 95: 49-72. March.
Krupnick, A.J. and M.L. Cropper. 1992. "The Effect of Information on Health Risk Valuations," Journal of
Risk and Uncertainty 5(2): 29-48.
Krupnick, A.J. and R.J. Kopp. 1988. Health and Agricultural Benefits of Reductions in Ambient Ozone in
the United States. Resources for the Future.
H-44
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Loehman, E.T., S.V. Berg, A.A. Arroyo, R.A. Hedinger, J.M. Schwartz, M.E. Shaw, R.W. Fahien, V.H. De,
R.P. Fishe, D.E. Rio, W.F. Rossley, and A.E.S. Green. 1979. "Distributional Analysis of Regional
Benefits and Cost of Air Quality Control." Journal of Environmental Economics and Management
6: 222-243.
Loehman, E.T. and Yo Hu De. 1982. "Application of Stochastic Choice Modeling to Policy Analysis of
Public Goods: A Case Study of Air Quality Improvements." The Review of Economics and
Statistics 64(3): 474-480.
Manuel, E.H., R.L. Horst, KM. Brennan, W.N. Lanen, M.C. Duff, andJ.KTapiero. 1982. Benefits Analysis
of Alternative Secondary National Ambient Air Quality Standards for Sulfur Dioxide and Total
Suspended Particulates, Volumes I-IV. Prepared for U.S. Environmental Protection Agency, Office
of Air Quality Planning and Standards, Research Triangle Park, NC.
McClelland, Gary, William Schulze, Donald Waldman, Julie Irwin, David Schenk, Thomas Stewart, Leland
Deck and Mark Thayer. 1991. Valuing Eastern Visibility: A Field Test of the Contingent Valuation
Method. Prepared for Office of Policy, Planning and Evaluation, U.S. Environmental Protection
Agency. June.
Miller, T.R., C. Calhoun, and W.B. Arthur. 1990. Utility-adjusted Impairment Years: A Low-Cost Approach
to Morbidity Valuation. Federal Highway Administration. March.
Mitchell, R.C. and R.T. Carson. 1986. "Valuing Drinking Water Risk Reductions Using the Contingent
Valuation Methods: A Methodological Study of Risks from THM and Giardia." Paper prepared for
Resources for the Future, Washington, DC.
Moore, M.J. and W.K Viscusi. 1988. "The Quantity-Adjusted Value of Life". Economic Inquiry 26(3): 369-
388.
Murray, Christopher J.L. and Alan D. Lopez, eds. 1996 The Global Burden of Disease. Harvard School of
Public Health on behalf of the World Health Organization and the World Bank.
National Acid Precipitation Assessment Program (NAPAP). 1991. Acidic Deposition: State of Science and
Technology (Summary Report). (Washington, DC: NAPAP). September.
Neumann, J.E., M. T. Dickie, and R.E. Unsworth. 1994. Industrial Economics, Incorporated.
Memorandum to Jim DeMocker, U.S. EPA, Office of Air and Radiation. Linkage Between Health
Effects Estimation and Morbidity Valuation in the Section 812 Analysis — Draft Valuation
Document. March 31.
O'Conor, Richard M. and Glenn C. Blomquist, 1997. "Measurement of Consumer-Patient Preferences
Using a Hybrid Contingent Valuation Method." Journal of Health Economics 16:667-683.
Ostro, B.D., M.J. Lipsett, J.K Mann, H. Braxton-Owens, and M.C. White. 1995. "Air Pollution and Asthma
Exacerbations Among African American Children in Los Angeles." Inhalation Toxicology.
H-45
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
Pope, C.A., III, M.J. Thun, M.M. Namboodiri, D.W. Dockery, J.S. Evans, F.E. Speizer, and C.W. Heath, Jr.
1995. "Particulate Air Pollution as a Predictor of Mortality in a Prospective Study of U.S. Adults."
Am. J. Respir. Crit. Care Med. 151: 669-674.
Post, Ellen and L. Deck. 1996. Abt Associates Inc. Memorandum to Tom Gillis, U.S. EPA, Office of
Office of Policy Planning and Evaluation. September 20.
Rowe, R.D. and L.G. Chestnut. 1986. "Oxidants and Asthmatics in Los Angeles: A Benefits Analysis-
Executive Summary." Prepared by Energy and Resource Consultants, Inc. Report to the U.S. EPA,
Office of Policy Analysis. EPA-230-09-86-018. Washington, D.C. March.
Rowlatt, Penelope, Michael Spackman, Sion Jones, Michael Jones-Lee, and Graham Loomes. 1998. Valuation
of Deaths from Air Pollution. For the Department of Environment, Transport and the Regions and
the Department of Trade and Industry. February.
Salkever, D.S. 1995. "Updated Estimates of Earnings Benefits from Reduced Exposure of Children to
Environmental Lead." Environmental Research 70: 1-6.
Schwartz, J. 1994. "Societal Benefits of Reducing Lead Exposure." Environmental Research 66: 105-124.
Selvin, Steve. 1996. Statistical Analysis of Epidemiologic Data. Second Edition. Monographs in
Epidemiology and Biostatistics, Volume 25. Oxford University Press.
Smith, David H., Daniel C. Malone, Kenneth A. Lawson, Lynn J. Okamoto, Carmelina Battista, and William
B. Saunders, 1997. "A National Estimate of the Economic Costs of Asthma," Am. J. Respir. Crit.
Care Med. 156:787-793.
Smith, V. Kerry, George Van Houtven, Subhrendu Pattanayak, 1999. "Benefit Transfer as a Preference
Calibration." Resources for the Future, Discussion Paper 99-36. May.
Spix, C., J. Heinrich, D. Dockery, J. Schwartz, G. Volksch, K. Schwinkowski, C. Collen, and H.E. Wichmann.
1994. Summary of the Analysis and Reanalysis Corresponding to the Publication Air Pollution and
Daily Mortality in Erfurt, East Germany 1980-1989. Summary report for: Critical Evaluation
Workshop on Particulate Matter—Mortality Epidemiology Studies; November; Raleigh, NC.
Wuppertal, Germany: Bergische Universitat-Gesamthochschule Wuppertal.
Taylor, T.N., P.H. Davis, J.C. Torner, J. Holmes, J.W. Meyer, and M. F. Jacobson. 1996. "Lifetime Cost of
Stroke in the United States." Stroke 27(9): 1459-1466.
Tolley, G.S. et al. 1986. Valuation of Reductions in Human Health Symptoms and Risks. University of
Chicago. Final Report for the U.S. Environmental Protection Agency. January.
Tolley, George, Donald Kenkel, and Robert Fabian, eds. 1994. Valuing Health for Policy: An Economic
Approach. The University of Chicago Press.
U.S. Department of Commerce, Economics and Statistics Administration. 1992. Statistical Abstract of the
United States, 1992: The National Data Book. 112th Edition, Washington, D.C.
H-46
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
U.S. Environmental Protection Agency (U.S. EPA). 1994. Documentation for Oz-One Computer Model
(Version 2.0). Office of Air Quality Planning and Standards. Prepared by: Mathtech, Inc., under
EPA Contract No. 68D30030, WA 1-29. August.
U.S. Environmental Protection Agency (U.S. EPA). 1995. Human Health Benefits From Sulfate Reductions
Under Title IV of the 1990 Clean Air Act Amendments. Prepared by Hagler Bailly Consulting, Inc.
for U.S. EPA, Office of Air and Radiation, Office of Atmospheric Programs. November 10.
U.S. Environmental Protection Agency (U.S. EPA). 1996. Air Quality Criteria for Particulate Matter,
Volume III of III. Office of Research and Development, Washington DC. EPA/600/P-95/001cF
U.S. Environmental Protection Agency (U.S. EPA). 1997 The Benefits and Costs of the Clean Air Act, 1970
to 1990. Prepared for U.S. Congress. October.
Violette, D.M. and L.G. Chestnut. 1983. Valuing Reduction in Risks: A Review of the Empirical Estimates.
Report prepared for the U.S. Environmental Protection Agency, Washington, D.C. EPA-230-05-83-
002.
Viscusi, W.K. 1992. Fatal Tradeoffs: Public and Private Responsibilities for Risk. (New York: Oxford
University Press).
Viscusi, W.K., W. A. Magat, and J. Huber. 1991. "Pricing Environmental Health Risks: Survey Assessments
of Risk-Risk and Risk-dollar Tradeoffs." Journal of Environmental Economics and Management
201: 32-57.
Viscusi, W. Kip. 1992. Fatal Tradeoffs: Public & Private Responsibilities for Risk. Oxford University Press:
New York.
Viscusi, W. Kip. 1993. "The Value of Risks to Life and Health." Journal of Economic Literature. 31(4): 1912-
46. October.
Wittels, E.H., J.W. Hay, and A.M. Gotto, Jr. 1990. "Medical Costs of Coronary Artery Disease in the United
States." The American Journal of Cardiology 65: 432-440.
World Health Organization (WHO). 1996. Final Consultation on Updating and Revision of the Air Quality
Guidelines for Europe. Bilthoven, The Netherlands 28-31 October, 1996 ICP EHH 018 VD96
2.11.
H-47
-------
The Benefits and Costs of the Clean Air Act, 1990 to 2010
[This page left blank intentionally.]
H-48
-------
Appendix I: Valuation of Human Health and
Welfare Effects of Criteria Pollutants
This appendix describes the derivations of the
economic valuations for health and welfare endpoints
considered in the benefits analysis. Valuation esti-
mates were obtained from the literature and reported
in dollars per case avoided for health effects, and dol-
lars per unit of avoided damage for welfare effects.
This appendix first introduces the method for mon-
etizing improvements in health and welfare, followed
by a summary of dollar estimates used to value ben-
efits and detailed descriptions of the derivation of each
estimate. These economic valuations are given both
in terms of a central (point) estimate as well as a prob-
ability distribution which characterizes the uncertainty
about the central estimate. All dollar values are
rounded and are in 1990 dollars. Next, results of the
economic benefits analysis are presented. Finally, un-
certainties in valuing the benefits attributable to the
CAA are explored.
Methods Used to Value Health
and Welfare Effects
Willingness to pay (WTP) and willingness to ac-
cept (WTA) are the two measures commonly used to
quantify the value an individual places on something,
whether it is something that can be purchased in a
market or not. Both WTP and WTA are measures of
the amount of money such that the individual would
be indifferent between having the good (or service)
and having the money. Whether WTP or WTA is the
appropriate measure depends largely on whether an
increase or a decrease of the good is at issue. WTP is
the amount of money an individual would be willing
to pay to have a good (or a specific increase in the
amount of the good) — i.e., the amount such that the
individual would be indifferent between having the
money and having the good (or having the specific
increase in the good). WTA is the amount of money
the individual would have to be compensated in order
to be indifferent to the loss of the good (or a specific
decrease in the amount of the good). WTP is the ap-
propriate measure if the baseline case is that the indi-
vidual does not have the good or when an increase in
the amount of the good is at issue; WTA is the appro-
priate measure if the baseline case is that the indi-
vidual has the good or when a decrease in the amount
of the good is at issue. An important difference be-
tween WTP and WTA is that, in theory, WTP is lim-
ited by the individual's budget, whereas WTA is not.
Nevertheless, while the underlying economic valua-
tion literature is based on studies which elicited ex-
pressions of WTP and/or WTA, the remainder of this
report refers to all valuation coefficients as WTP esti-
mates. In some cases (e.g., stroke-related hospital
admissions), neither WTA nor WTP estimates are
available and WTP is approximated by cost of illness
(COI) estimates, a clear underestimate of the true
welfare change since important value components
(e.g., pain and suffering associated with the stroke)
are not reflected in the out-of-pocket costs for the
hospital stay.
For both market and non-market goods, WTP re-
flects individuals' preferences. Because preferences
are likely to vary from one individual to another, WTP
for both market and non-market goods (e.g., health-
related improvements in environmental quality) is
likely to vary from one individual to another. In con-
trast to market goods, however, non-market goods
such as environmental quality improvements are pub-
lic goods, whose benefits are shared by many indi-
viduals. The individuals who benefit from the envi-
ronmental quality improvement may have different
WTPs for this non-market good. The total social value
of the good is the sum of the WTPs of all individuals
who "consume" (i.e., benefit from) the good.
In the case of health improvements related to pol-
lution reduction, it is not certain specifically who will
receive particular benefits of reduced pollution. For
example, the analysis may predict 100 days of cough
avoided in a given year resulting from CAA reduc-
tions of pollutant concentrations, but the analysis does
not estimate which individuals will be spared those
days of coughing. The health benefits conferred on
individuals by a reduction in pollution concentrations
are, then, actually reductions in the probabilities of
having to endure certain health problems. These ben-
efits (reductions in probabilities) may not be the same
for all individuals (and could be zero for some indi-
viduals). Likewise, the WTP for a given benefit is
likely to vary from one individual to another. In theory,
the total social value associated with the decrease in
risk of a given health problem resulting from a given
1-1
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
reduction in pollution concentrations is
N
(1)
/=l
where Bi is the benefit (i.e., the reduction in prob-
ability of having to endure the health problem) con-
ferred on the ith individual (out of a total of N) by the
reduction in pollution concentrations, and WTP^ET)
is the ith individual's WTP for that benefit. If a re-
duction in pollution concentrations affects the risks
of several health endpoints, the total health-related
social value of the reduction in pollution concentra-
tions is
N J
TP.(B.J (2)
2=1 ;=l
where B;j is the benefit related to the jth health
endpoint (i.e., the reduction in probability of having
to endure the jth health problem) conferred on the ith
individual by the reduction in pollution concentrations,
and WTP^Rj) is the ith individual's WTP for that
benefit.
The reduction in probability of each health prob-
lem for each individual is not known, nor is each
individual's WTP for each possible benefit he or she
might receive known. Therefore, in practice, benefits
analysis estimates the value of a.vto//.s7/6'a/health prob-
lem avoided. For example, although a reduction in
pollutant concentrations may save actual lives (i.e.,
avoid premature mortality), whose lives will be saved
cannot be known ex ante. What is known is that the
reduction in air pollutant concentrations results in a
reduction in mortality risk. It is this reduction in mor-
tality risk that is valued in a monetized benefit analy-
sis. Individual WTPs for small reductions in mortal-
ity risk are summed over enough individuals to infer
the value of a statistical life saved. This is different
from the value of a particular, identified life saved.
Rather than "WTP to avoid a death," then, it is more
accurate to use the term "WTP to avoid a statistical
death," or, equivalently, "the value of a statistical life."
Suppose, for example, that a given reduction in
PM concentrations results in a decrease in mortality
risk of 1/10,000. Then for every 10,000 individuals,
one individual would be expected to die in the ab-
sence of the reduction in PM concentrations (who
would not die in the presence of the reduction in PM
concentrations). IfWTP forthis 1/10,000 decrease in
mortality risk is $500 (assuming, for now, that all in-
dividuals' WTPs are the same), then the value of a
statistical life is 10,000 x $500, or $5 million.
A given reduction in PM concentrations is un-
likely, however, to confer the same risk reduction (e.g.,
mortality risk reduction) on all exposed individuals
in the population. (In terms of the expressions above,
Bi is not necessarily equal to B , for i j). In addition,
different individuals may not be willing to pay the
same amount for the same risk reduction. The above
expression for the total social value associated with
the decrease in risk of a given health problem result-
ing from a given reduction in pollution concentrations
may be rewritten to more accurately convey this. Us-
ing mortality risk as an example, for a given unit risk
reduction (e.g., 1/1,000,000), the total mortality-re-
lated benefit of a given pollution reduction can be
written as
N
number of units of risk reduction).
2=i x (WTPper unit risk reduction). (3)
where (number of units of risk reduction^ is the
number of units of risk reduction conferred on the ith
exposed individual as a result of the pollution reduc-
tion, (WTP per unit risk reduction^ is the ith
individual's willingness to pay for a unit risk reduc-
tion, and N is the number of exposed individuals.
If different subgroups of the population have sub-
stantially different WTPs for a unit risk reduction and
substantially different numbers of units of risk reduc-
tion conferred on them, then estimating the total so-
cial benefit by multiplying the population mean WTP
to save a statistical life times the predicted number of
statistical lives saved could yield a biased result. Sup-
pose, for example, that older individuals' WTP per
unit risk reduction is less than that of younger indi-
viduals (e.g., because they have fewer years of ex-
pected life to lose). Then the total benefit will be less
than it would be if everyone's WTP were the same. In
addition, if each older individual has a larger number
of units of risk reduction conferred on him (because a
given pollution reduction results in a greater absolute
reduction in risk for older individuals than for younger
individuals), this, in combination with smaller WTPs
of older individuals, would further reduce the total
benefit.
While the estimation of WTP for a market good
(i.e., the estimation of a demand schedule) is not a
simple matter, the estimation of WTP for a non-mar-
ket good, such as a decrease in the risk of having a
particular health problem, is substantially more diffi-
cult. Estimation of WTP for decreases in very spe-
cific health risks (e.g., WTP to decrease the risk of a
day of coughing or WTP to decrease the risk of ad-
mission to the hospital for respiratory illness) is fur-
ther limited by a paucity of information. Derivation
of the dollar value estimates discussed below was of-
ten limited by available information.
1-2
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Valuation of Specific Heaith Endpoints
Valuation of Premature Mortality Avoided
As noted above, it is actually reductions in mor-
tality risk that are valued in a monetized benefit
analysis. Individual WTPs for small reductions in
mortality risk are summed over enough individuals
to infer the value of a statistical life saved. This is
different from the value of a particular, identified
life saved. The "value of a premature death
avoided," then, should be understood as shorthand
for "the value of a statistical premature death
avoided."
The value of a premature death avoided is based
on an analysis of 26 policy-relevant value-of-life
studies (see Table 1-1). Five of the 26 studies are
contingent valuation (CV) studies, which directly
solicit WTP information from subjects; the rest are
wage-risk studies, which base WTP estimates on
estimates of the additional compensation demanded
in the labor market for riskier jobs. Each of the 26
studies provided an estimate of the mean WTP to
avoid a statistical premature death. Several plau-
sible standard distributions were fit to the 26 esti-
mates of mean WTP. A Weibull distribution, with
a mean of $4.8 million and standard deviation of
$3.24 million, provided the best fit to the 26 esti-
mates. The central tendency estimate of the WTP
to avoid a statistical premature death is the mean of
this distribution, $4.8 million. The considerable un-
certainty associated with this approach is discussed
in detail below, in the subsection titled "The Eco-
nomic Benefits Associated with Mortality," within
the section titled "Uncertainties."
Life-years lost is a possible alternative measure
of the mortality-related effect of pollution, as dis-
cussed in Appendix D. If life-years lost is the mea-
sure used, then the value of a statistical life-year lost,
rather than the value of a statistical life lost would be
needed. Moore and Viscusi (1988) suggest one ap-
proach for determining the value of a statistical life-
year lost. They assume that the willingness to pay to
save a statistical life is the value of a single year of
life times the expected number of years of life remain-
ing for an individual. They suggest that a typical re-
spondent in a mortal risk study may have a life ex-
pectancy of an additional 35 years. Using a mean es-
timate of $4.8 million to save a statistical life, their
approach would yield an estimate of $ 137,000 per life-
year lost or saved. If an individual discounts future
additional years using a standard discounting proce-
dure, the value of each life-year lost must be greater
than the value assuming no discounting. Using a 35
year life expectancy, a $4.8 million value of a statisti-
cal life, and a 5 percent discount rate, the implied value
Tabic 1-1. Summary of Mortality Valuation Estimates
(millions of 1990 dollars).
Sttausllj
TTyperf
Ksliin ate
Valuation
(in illions
!Ȥ$))
Kneisner and Leeth (1991) (US)
Labor Market
0.6
Smith and Gilbert (1984)
Labor Market
0.7
Dillingham (1985)
Labor Market
0.9
Butler (1983)
Labor Market
1.1
Miller and Guria (1991)
Cont. Value
1.2
Moore and Viscusi (1988a)
Labor Market
2.5
Viscusi, Magat, and Huber (1991b)
Cont. Value
2.7
Gegax et al. (1985)
Cont. Value
3.3
Marin and Psacharopoulos (1982)
Labor Market
2.8
Kneisner and Leeth (1991)
(Australia)
Labor Market
3.3
Gerking, de Haan, and Schulze
(1988)
Cont. Value
3.4
Cousineau, Lacroix, and Girard
(1988)
Labor Market
3.6
Jones-Lee (1989)
Cont. Value
3.8
Dillingham (1985)
Labor Market
3.9
Viscusi (1978, 1979)
Labor Market
4.1
R.S. Smith (1976)
Labor Market
4.6
V.K. Smith (1976)
Labor Market
4.7
Olson (1981)
Labor Market
5.2
Viscusi (1981)
Labor Market
6.5
R.S. Smith (1974)
Labor Market
7.2
Moore and Viscusi (1988a)
Labor Market
7.3
Kneisner and Leeth (1991) (Japan)
Labor Market
7.6
Herzog and Schlottman (1987)
Labor Market
9.1
Leigh and Folson (1984)
Labor Market
9.7
Leigh (1987)
Labor Market
10.4
Gaten (1988)
Labor Market
13.5
SOURCE: Viscusi, 1992
of each life-year lost is $293,000. The Moore and
Viscusi procedure is identical to this approach, but
uses a zero discount rate.
Using the value of a life-year lost and the expected
number of years remaining (obtained from life ex-
pectancy tables), and assuming a given discount rate,
values of age-specific premature mortality can be de-
rived. Examples of valuations of pollution-related
mortality using the life-years lost approach are given
below, in the subsection titled "The Economic Ben-
efits Associated with Mortality," within the section
titled "Uncertainties."
Valuation of Hospital Admissions Avoided
In the case of hospital admissions, cost-of-illness
(COI), or "costs avoided," estimates were used in lieu
of WTP because of the lack of other information re-
1-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
garding willingness to pay to avoid illnesses that ne-
cessitate hospital admissions. For those hospital ad-
missions which were specified to be the initial hospi-
tal admission (in particular, hospital admissions for
coronary heart disease (CHD) events and stroke), COI
estimates include, where possible, all costs of the ill-
ness, including the present discounted value of the
stream of medical expenditures related to the illness,
as well as the present discounted value of the stream
of lost earnings related to the illness. (While an esti-
mate of present discounted value of both medical ex-
penditures and lost earnings was available for stroke,
the best available estimate for CHD did not include
lost earnings. The derivations of the COI estimates
for CHD and stroke, both lead-induced effects, are
discussed in Appendix G.)
In those cases for which it is unspecified whether
the hospital admission is the initial one or not (that is,
for all hospital admissions endpoints other than CHD
and stroke), it is unclear what portion of medical ex-
penditures and lost earnings after hospital discharge
can reasonably be attributed to pollution exposure and
what portion might have resulted from an individual's
pre-existing condition even in the absence of a par-
ticular pollution-related hospital admission. In such
cases, the COI estimates include only those costs as-
sociated with the hospital stay, including the hospital
charge, the associated physician charge, and the lost
earnings while in the hospital. The derivations of these
costs are discussed in Abt Associates Inc., 1996.
These COI estimates are likely to substantially
understate total WTP to avoid an illness that began
with a hospital admission or to avoid a particular hos-
pital admission itself. First, most of the COI estimates
fall short of being full COI estimates either because
of insufficient information or because of ambiguities
concerning what portion of post-hospital costs should
be attributed to pollution exposure. Even full COI es-
timates will understate total WTP, however, because
they do not include the value of avoiding the pain and
suffering associated with the illness for which the in-
dividual entered the hospital.
Valuation of Chronic Bronchitis Avoided
Although the severity of cases of chronic bron-
chitis valued in some studies approaches that of
chronic obstructive pulmonary disease, to maintain
consistency with the existing literature we do not treat
those cases separately for the purposes of this analy-
sis. Chronic bronchitis is one of the only morbidity
endpoints that may be expected to last from the initial
onset of the illness throughout the rest of the
individual's life. WTP to avoid chronic bronchitis
would therefore be expected to incorporate the present
discounted value of a potentially long stream of costs
(e.g., medical expenditures and lost earnings) associ-
ated with the illness. Two studies, Viscusietal. (1991)
and Krupnick and Cropper (1992) provide estimates
of WTP to avoid a case of chronic bronchitis. The
study by Viscusi et al., however, uses a sample that is
larger and more representative of the general popula-
tion than the study by Krupnick and Cropper (which
selects people who have a relative with the disease).
The valuation of chronic bronchitis in this analysis is
therefore based on the distribution of WTP responses
from Viscusi et al. (1991).
Both Viscusi et al. (1991) and Krupnick and Crop-
per (1992), however, defined a case of severe chronic
bronchitis. It is unclear what proportion of the cases
of chronic bronchitis predicted to be associated with
exposure to pollution would turn out to be severe cases.
The incidence of pollution-related chronic bronchitis
was based on Abbey et al. (1993), which considered
only new cases of the illness.1 While a new case may
not start out being severe, chronic bronchitis is a
chronic illness which may progress in severity from
onset throughout the rest of the individual's life. It is
the chronic illness which is being valued, rather than
the illness at onset.
The WTP to avoid a case of pollution-related
chronic bronchitis (CB) is derived by starting with
the WTP to avoid a severe case of chronic bronchitis,
as described by Viscusi et al. (1991), and adjusting it
downward to reflect (1) the decrease in severity of a
case of pollution-related CB relative to the severe case
described in the Viscusi study, and (2) the elasticity
of WTP with respect to severity. Because elasticity is
a marginal concept and because it is a function of se-
verity (as estimated from Krupnick and Cropper,
1992), WTP adjustments were made incrementally,
in one percent steps. At each step, given a WTP to
avoid a case of CB of severity level sev, the WTP to
avoid a case of severity level 0.99*sev was derived.
This procedure was iterated until the desired severity
level was reached and the corresponding WTP was
derived. Because the elasticity of WTP with respect
to severity is a function of severity, this elasticity
changes at each iteration. If, for example, it is believed
that a pollution-related case of CB is of average se-
1 It is important that only new chronic bronchitis be considered in this analysis because WTP estimates reflect lifetime expendi-
tures and/or losses associated with this chronic illness, and incidences are predicted separately for each year during the period 1970-
1990. If the total prevalence of chronic bronchitis, rather than the incidence of only new chronic bronchitis were predicted each year,
valuation estimates reflecting lifetime expenditures could be repeatedly applied to the same individual for many years, resulting in a
severe overestimation of the value of avoiding pollution-related chronic bronchitis.
1-4
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
verity, that is, a 50 percent reduction in severity from
the case described in the Viscusi study, then the itera-
tive procedure would proceed until the severity level
was half of what it started out to be.
The derivation of the WTP to avoid a case of pol-
lution-related chronic bronchitis is based on three com-
ponents, each of which is uncertain: (1) the WTP to
avoid a case of severe CB, as described in the Viscusi
study, (2) the severity level of an average pollution-
related case of CB (relative to that of the case de-
scribed by Viscusi), and (3) the elasticity of WTP with
respect to severity of the illness. Because of these three
sources of uncertainty, the WTP is uncertain. Based
on assumptions about the distributions of each of the
three uncertain components, a distribution of WTP to
avoid a pollution-related case of CB was derived by
Monte Carlo methods. The mean of this distribution,
which was about $260,000, is taken as the central ten-
dency estimate of WTP to avoid a pollution-related
case of CB. Each of the three underlying distributions
is described briefly below.
The distribution of WTP to avoid a severe case of
CB was based on the distribution of WTP responses
in the Viscusi study. Viscusi et al. derived respon-
dents' implicit WTP to avoid a statistical case of
chronic bronchitis from their WTP for a specified re-
duction in risk. The mean response implied a WTP of
about $1,000,000 (1990 dollars)2; the median response
implied a WTP of about $530,000 (1990 dollars).
However, the extreme tails of distributions of WTP
responses are usually considered unreliable. Because
the mean is much more sensitive to extreme values,
the median of WTP responses is often used rather than
the mean. Viscusi et al. report not only the mean and
median of their distribution of WTP responses, how-
ever, but the decile points as well. The distribution of
reliable WTP responses from the Viscusi study could
therefore be approximated by a discrete uniform dis-
tribution giving a probability of one ninth to each of
the first nine decile points. This omits the first five
and the last five percent of the responses (the extreme
tails, considered unreliable). This trimmed distribu-
tion of WTP responses from the Viscusi study was
assumed to be the distribution of WTPs to avoid a
severe case of CB. The mean of this distribution is
about $720,000 (1990 dollars).
The distribution of the severity level of an aver-
age case of pollution-related CB was modeled as a
triangular distribution centered at 6.5, with endpoints
at 1.0 and 12.0. These severity levels are based on the
severity levels used in Krupnick and Cropper, 1992,
which estimated with relationship between ln(WTP)
and severity level, from which the elasticity is derived.
The most severe case of CB in that study is assigned a
severity level of 13. The mean of the triangular distri-
bution is 6.5. This represents a 50 percent reduction
in severity from a severe case.
The elasticity of WTP to avoid a case of CB with
respect to the severity of that case of CB is a constant
times the severity level. This constant was estimated
by Krupnick and Cropper, 1992, in the regression of
ln(WTP) on severity, discussed above. This estimated
constant (regression coefficient) is normally distrib-
uted with mean = 0.18 and standard deviation = 0.0669
(obtained from Krupnick and Cropper, 1992).
The distribution of WTP to avoid a case of pollu-
tion-related CB was generated by Monte Carlo meth-
ods, drawing from the three distributions described
above. On each of 16,000 iterations (1) a value was
selected from each distribution, and (2) a value for
WTP was generated by the iterative procedure de-
scribed above, in which the severity level was de-
creased by one percent on each iteration, and the cor-
responding WTP was derived. The mean of the re-
sulting distribution of WTP to avoid a case of pollu-
tion-related CB was $260,000.
This WTP estimate is reasonably consistent with
full COI estimates derived for chronic bronchitis, us-
ing average annual lost earnings and average annual
medical expenditures reported by Cropper and
Krupnick, 1990. Using a 5 percent discount rate and
assuming that (1) lost earnings continue until age 65,
(2) medical expenditures are incurred until death, and
(3) life expectancy is unchanged by chronic bronchi-
tis, the present discounted value of the stream of medi-
cal expenditures and lost earnings associated with an
average case of chronic bronchitis is estimated to be
about $77,000 for a 30 year old, about $58,000 for a
40 year old, about $60,000 for a 50 year old, and about
$41,000 for a 60 year old. A WTP estimate would be
expected to be greater than a full COI estimate, re-
flecting the willingness to pay to avoid the pain and
suffering associated with the illness. The WTP esti-
mate of $260,000 is from 3.4 times the full COI esti-
mate (for 30 year olds) to 6.3 times the full COI esti-
mate (for 60 year olds).
2 There is an indication in the Viscusi paper that the dollar values in the paper are in 1987 dollars. Under this assumption, the
dollar values were converted to 1990 dollars.
1-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Valuation of Other Morbidity Endpoints Avoided
WTP to avoid a day of specific morbidity end-
points, such as coughing or shortness of breath, has
been estimated by only a small number of studies (two
or three studies, for some endpoints; only one study
for other endpoints). The estimates for health end-
points involving these morbidity endpoints are there-
fore similarly based on only a few studies. However,
it is worth noting that the total benefit associated with
any reduction in pollutant concentrations is determined
largely by the benefit associated with the correspond-
ing reduction in mortality risk because the dollar value
associated with mortality is significantly greater than
any other valuation estimate. More detailed explana-
tions for valuation of specific morbidity endpoints is
given in Table 1-2.
Estimates of WTP may be understated for a couple
of reasons. First, if exposure to pollution has any cu-
mulative or lagged effects, then a given reduction in
pollution concentrations in one year may confer ben-
efits not only in that year but in future years as well.
Benefits achieved in later years are not included. Sec-
ond, the possible effects of altruism are not consid-
ered in any of the economic value derivations. Indi-
viduals' WTP for reductions in health risks for others
are implicitly assumed to be zero.
Table 1-2 summarizes the derivations of the eco-
nomic values used in the analysis. More detailed de-
scriptions of the derivations of lead-related endpoints
(hospital admissions for CHD and stroke, Lost IQ
points, IQ below 70, and hypertension) are discussed
in Appendix G.
Valuation of Welfare Effects
With the exception of agricultural benefits, eco-
nomic valuations for welfare effects quantified in the
analysis (i.e., household soiling damage, visibility and
worker productivity) are documented in Table 1-2. For
agricultural benefits, estimated changes in crop yields
were evaluated with an agricultural sector model,
AGSIM. This model incorporates agricultural price,
farm policy, and other data for each year. Based on
expected yields, the model estimates the production
levels for each crop, the economic benefits to con-
sumers, and the economic benefits to producers asso-
ciated with these production levels. To the extent that
alternative exposure-response relationships were
available, a range of potential benefits was calculated.
Appendix F documents the derivation of the monetary
benefits associated with improved agricultural pro-
duction. The derivation of the residential visibility
valuation estimate is discussed further below.
Visibility Valuation
Residential visibility has historically been valued
through the use of contingent valuation studies, which
employ surveys and questionnaires to determine the
economic value respondents place on specified
changes in visibility. A number of such studies have
been published in the economics literature since the
late 1970s, and have reported a wide range of result-
ing values for visibility, expressed as household will-
ingness to pay (WTP) for a hypothesized improve-
ment in visibility. Those studies were carefully re-
viewed for their applicability to the retrospective
analysis.
One limitation of many existing contingent valu-
ation studies of visibility is that they are local or re-
gional in scope, soliciting values for visibility from
residents of only one or two cities in a single region
of the country. Studies of visibility values from west-
ern cities, the most recent of which was published in
1981, have reported somewhat lower values than those
from eastern cities, raising the question of whether
eastern and western visibility are different commodi-
ties and should be valued differently in this analysis.
While the different visibility values reported in
the literature may appear to imply that visibility is
not valued equally by survey respondents in the east-
ern and western U.S., other evidence suggests that
eastern and western visibility are not fundamentally
different commodities, and that the retrospective ben-
efits calculations should not be based on separate east-
ern and western visibility values. For example,
NAPAP data indicate that California's South Coast
Air Basin, which encompasses Los Angeles and ex-
tends northward to the vicinity of San Francisco, has
median baseline visibility more characteristic of the
eastern U.S. than of other areas of the west (NAPAP
1991; IEc 1992, 1993a), reflecting the influence of
the higher humidity typical of coastal areas. While
inland areas of the west will tend to have lower hu-
midity, and hence greater baseline visibility than ei-
ther the eastern region or the western coastal zone,
such baseline visibility differences are accounted for
in the conversion from the visual range metric to
DeciView.
1-6
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Perhaps the most compelling rationale for employ-
ing a single nationwide visibility valuation strategy
in the retrospective benefits analysis, however, relates
to the air quality modeling output used to calculate
the control and no-control scenario visibility profiles,
and its implications for the valuation of visibility as a
commodity. The RADM model and linear scaling
technique used for the retrospective analysis model
visibility improvements nationwide as changes in re-
gional atmospheric haze. In other words, although the
magnitude of visibility effects may vary between re-
gions, the model output does not distinguish between
a change in eastern visibility and a change in western
visibility as distinct phenomena. Thus, there is no clear
reason to value those same visibility changes differ-
ently in calculating the benefits of visibility improve-
ments. Consequently, a single, consistent valuation
basis has been applied to residential visibility improve-
ments nationwide for this analysis.
In light of advances in the state of the art of con-
tingent valuation over the last decade, the age of many
of the existing studies raised questions regarding their
suitability to serve as the primary basis for the vis-
ibility benefits estimates. A review of the survey and
data analysis methods used in the available studies
indicated that a study conducted for EPA by
McClelland etal. (1991) addressed many of the meth-
odological flaws of earlier studies, employing survey
methods and analytical techniques designed to mini-
mize potential biases (IEc 1992). Although this study
is unpublished, given its methodological improve-
ments over earlier studies it was chosen as the basis
for the central tendency of the visibility benefits esti-
mate, yielding an estimated value of $ 14 per unit im-
provement in DeciView as the annual household WTP
for visibility improvements (IEc 1997), as specified
in Table 1-2.
The difficulty of accurately defining the expected
statistical distribution of WTP values for visibility
improvements on the basis of published studies of
uneven reliability, along with the considerable varia-
tion in reported visibility values, led to the selection
of a hypothesized triangular distribution of values to
characterize the uncertainty in the visibility benefits
estimate. Reliance on any single study to estimate the
uncertainty range would be unlikely to adequately
characterize variations in visibility values that might
exist across cities, and in any case would fail to cap-
ture the full variability of visibility values reported in
the literature. Therefore, to ensure that the retrospec-
tive study characterizes the full range of uncertainty
in visibility values nationwide, the upper and lower
bounds of the triangular distribution were derived by
combining results from appropriate eastern and west-
ern residential visibility valuation studies.
Most of the existing residential visibility valua-
tion studies were found to suffer from part-whole bias,
which results from the failure to differentiate values
for visibility from those for other air quality ameni-
ties, such as reductions in adverse health effects. Of
the studies reviewed for this analysis, only the
McClelland study and Brookshire et al. (1979) have
attempted to obtain bids explicitly for visibility im-
provements (IEc 1992). Since part-whole bias will tend
to produce overstated values for visibility, reported
values from all studies that do not correct for part-
whole bias were adjusted prior to calculating the lower
bound of the uncertainty range. The upper bound of
the uncertainty range was calculated using the unad-
justed values from all studies, which is equivalent to
assuming that the entire value of respondents' stated
WTP for improved air quality can be attributed to in-
creased visibility.
The uncertainty range specified in Table 1-2 cal-
culated using a consensus function derived from a
regression analysis, incorporates a 25 percent adjust-
ment for part-whole bias (i.e., reported values were
multiplied by 0.25) in calculating the lower bound.
This represents an approximate midpoint of the range
defined by the McClelland study's finding that respon-
dents allocated, on average, 18.6 percent of their total
WTP to improvements in visibility, and Chestnut and
Rowe's (1989) conclusion that visibility improvement
accounted for 34 percent of the total WTP reported in
the Brookshire et al. study. Similarly, the "Denver
Brown Cloud" study results indicate that respondents
allocated 27.2 percent of their total WTP to visibility
improvements (Irwin et al. 1990). Therefore, the ap-
plication of a 25 percent adjustment for part-whole
bias to all but the McClelland and Brookshire values
would appear to be supported by the recent literature,
with the resulting consensus value representing a plau-
sible lower bound for the uncertainty range of visibil-
ity values. The consensus function approach, incor-
porating the part-whole bias adjustment, yields esti-
mated upper and lower bound values of $21 and $8,
respectively, for annual household WTP per unit im-
provement in DeciView.
1-7
-------
Tabic 1-2. Unit Values for Economically Valuing Health and Welfare Endpoints.
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
Mortality
$4.8 million per
statistical life
Weibull distribution,
mean = $4.8 million
std. dev. = $3.24 million
Central Est: $ value is the mean of value-of-statistical-life estimates
from 26 studies (5 contingent valuation and 21 labor market studies).
Uncertainty: Best-fit distribution to the 26 sample means. The Weibull
distribution prevents selection of negative WTP values.
$293,000 per
statistical life-
year
Weibull distribution,
mean = $293,000
std. dev. = $198,000
Central Est: $ value is the mean of the distribution of the value of a
statistical life-year, derived from the distribution of the value of a
statistical life (see below).
Uncertainty: Assuming the discount rate is five percent, and assuming
an expected 35 yrs. remaining to the avg. worker in the wage-risk
studies (see above), the value of a statistical life-year is just a constant,
0.061, multiplied by the value of a statistical life. The distribution of
the value of a life-year is derived from the distribution of the value of a
statistical life. Given that this is a Weibull distribution, as indicated
above, the value of a statistical life-year is also a Weibull distribution,
with mean equal to 0.061 multiplied by the mean of the original
Weibull distribution (0.061x$4.8 million = $293,000) and standard
deviation equal to 0.061 multiplied by the standard deviation of the
original distribution (0.061 x $3.24 = $198,000). (If the discount rate
were considered to also be uncertain, then the distribution of a
statistical life-year would depend on this distribution as well and would
have to be generated by Monte Carlo methods.)
-------
Health or Welfare
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Endpoint
Central Estimate
Uncertainty Distribution
Chronic Bronchitis (CB)
$260,000
A Monte Carlo-generated
distribution, based on three
underlying distributions, as
described more fully under
"Derivation of Estimates"
and in the text.
Central Est: $ value is the mean of a Monte Carlo distribution of WTP
to avoid a case of pollution-related CB. WTP to avoid a case of
pollution-related CB is derived by adjusting WTP to avoid a severe
case of CB (as described in Viscusi et al., 1991) for the difference in
severity and taking into account the elasticity of WTP with respect to
severity of CB. The mean of the resulting distribution is $260,000.
Uncertainty: The distribution of WTP to avoid a case of pollution-
related CB was generated by Monte Carlo methods, drawing from each
of three distributions: (1) WTP to avoid a severe case of CB is assigned
a 1/9 probability of being each of the first nine deciles of the
distribution of WTP responses in Viscusi et al., 1991; (2) the severity of
a pollution-related case of CB (relative to the case described in the
Viscusi study) is assumed to have a triangular distribution, centered at
severity level 6.5 with endpoints at 1.0 and 12.0 (see text for further
explanation); and (3) the constant in the elasticity of WTP with respect
to severity is normally distribution with mean = 0.18 and standard
deviation = 0.0669 (from Krupnick and Cropper, 1992). See text for
further explanation.
IQ Changes
1. Lost IQ Points
$3,000 per lost IQ
point
none available
Central Est: $ value is the mean of estimates based on results of 2
studies. With an assumed 5% discount rate, the results in Schwartz
(1994) yield an estimate of $2,500 per IQ point; the results of Salkever
(1995) yield an estimate of $3,400. These estimates include the
combined effects on lifetime earnings: (1) directly based on IQ
decrement, and (2) indirectly based on lower educational attainment
and reduced labor force participation (subtracting from indirect benefits
the costs of additional education and associated opportunity cost).
2. Incidence of IQ < 70
$42,000
none available
Central Est: $ value measures reduction in education costs in terms of
special needs for lower IQ students (in mainstream schools).
3
B-
x'
s"
a
S'
s
§
£
a
3
o
^0
-------
Health or Welfare
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Endpoint
Central Estimate
Uncertainty Distribution
Hypertension
$680 per case per
year
none available
Central Est: $ value quantifies costs associated with phvsician care,
medications, and hospital charges, in addition to opportunity cost of
lost work time due to the disability.
Hospital Admissions
1. Strokes
- initial
cerebrovascular
accidents (ICD code
436)
- initial
atherothrombotic
brain infarctions
(ICD code 434)
$200,000 for
males;
$150,000 for
females
none available
Central Est: $ values for males and females are based on ase- and
gender-specific estimates of lifetime cost of stroke from Taylor et al.,
1996. Estimates include both direct costs (medical expenditures) and
indirect costs (reduced productivity) and assume a five percent discount
rate.
Uncertainty: Although there is uncertainty surrounding the central
estimates presented, there is insufficient information to characterize this
uncertainty.
2. Coronary Heart Disease
(CHD)
$52,000
A Monte Carlo-generated
distribution, based on the
uncertainty about what
proportion of pollution-
related CHD events is acute
myocardial infarction, what
proportion is angina
pectoris, and what
proportion is unstable
angina pectoris (see
"Derivation of Estimates").
Central Est: $ value is the mean of the Monte Carlo-generated
distribution of WTP to avoid a pollution-related case of CHD,
described below.
Uncertainty: The distribution was based on the estimates of the total
medical costs within 5 years of diagnosis of each of three types of CHD
events examined in the Framingham Study, including acute myocardial
infarction, angina pectoris, and unstable angina pectoris (Wittels et al.,
1990). It is unknown what proportion of pollution-related CHD events
are of each type. On each iteration, three proportions were drawn from
three continuous uniform distributions, such that the three proportions
summed to 1.0. The $ value for an iteration is the weighted average of
the $ values for the three types of CHD event (from Wittels et al.,
1990), weighted by the corresponding proportions selected.
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
3. "Respiratory Illness"
$6,100
Normal distribution,
mean = $6,100
std. dev. = $55
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients of all ages, and the cost of associated
physician care, with the opportunity cost of time spent in the hospital.
Source of hospital charge estimate: Elixhauser et al., 1993. Source of
physician charge estimates: Abt Associates Inc., 1992.
Uncertainty: variation about the central estimate based on the standard
error reported for the hospital charge component (greater than the other
two components by an order of magnitude).
4. COPD
(ICD codes 490-496)
$8,100
Normal distribution, with
mean = $8,100
std. dev. = $190
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients 65 and older, and the cost of
associated physician care, with the opportunity cost of time spent in the
hospital. Source of cost-of-illness estimates: Abt Associates Inc., 1992.
Uncertainty: variation about the central estimate derived from a
standard error estimated for the hospital charge component measured
by another study (Elixhauser et al., 1993). The reported standard error
for hospital charge was applied to the combined cost-of-illness and
opportunity cost estimate by assuming that relative variabilities
surrounding the respective means were similar (i.e., coefficients of
variation are equal). The hospital charge represents the vast majority of
the total value to avoid a hospital admission for COPD.
5. Pneumonia
(ICD codes 480-487)
$7,900
Normal distribution, with
mean = $7,900
std. dev. = $110
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients of all ages, and the cost of associated
physician care, with the opportunity cost of time spent in the hospital.
Source of hospital charge estimate: Elixhauser et al., 1993. Source of
physician charge estimates: Abt Associates Inc., 1992.
Uncertainty: Applied the standard error associated with the hospital
charge component to the central estimate of $7,900. The hospital
charge represents the vast majority of the total value to avoid a hospital
admission for pneumonia.
3
B-
x'
s"
a
S'
s
§
£
a
3
o
^0
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
6. Congestive Heart Failure
(ICD code 428)
$8,300
Normal distribution, with
mean = $8,300
std. dev. = $120
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients of all ages, and the cost of associated
physician care, with the opportunity cost of time spent in the hospital.
Source of hospital charge estimate: Elixhauser et al., 1993. Source of
physician charge estimates: Abt Associates Inc., 1992.
Uncertainty: Applied the standard error associated with the hospital
charge component to the central estimate of $8,300. The hospital
charge represents the vast majority of the total value to avoid a hospital
admission for congestive heart failure.
7. Ischemic Heart Disease
(ICD codes 410-414)
$10,300
Normal distribution, with
mean = $10,300
std. dev. = $88
Central Est: $ value combines a cost-of-illness estimate, including the
hospital charge, based on patients of all ages, and the cost of associated
physician care, with the opportunity cost of time spent in the hospital.
Source of hospital charge estimate: Elixhauser et al., 1993. Source of
physician charge estimates: Abt Associates Inc., 1992.
Uncertainty: Applied the standard error associated with the hospital
charge component to the central estimate of $10,300. The hospital
charge represents the vast majority of the total value to avoid a hospital
admission for ischemic heart disease.
Respiratory Ailments Not Requiring Hospitalization
1. Upper Resp. Symptoms
(URS)
(defined as one or
more of the
following: runny or
stuffy nose, wet
cough, burning,
aching, or red eyes)
$19
Continuous uniform
distribution over the
interval [$7, $33]
Central Est: Combinations of the 3 svmptoms for which WTP
estimates are available that closely match those listed by Pope et al.
result in 7 different "symptom clusters," each describing a "type" of
URS. A $ value was derived for each type of URS, using IEc mid-
range estimates of WTP to avoid each symptom in the cluster and
assuming additivity of WTPs. The $ value for URS is the average of
the $ values for the 7 different types of URS.
Uncertainty: taken to be a continuous uniform distribution across the
range of values described by the 7 URS types.
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
2. Lower Resp. Symptoms
(LRS)
(defined in the study
as two or more of the
following: cough,
chest pain, phlegm,
and wheeze.)
$12
Continuous uniform
distribution over the
interval [$5, $19]
Central Est: Combinations of the 4 svmptoms for which WTP
estimates are available that closely match those listed by Schwartz et
al. result in 11 different "symptom clusters," each describing a "type"
of LRS. A $ value was derived for each type of LRS, using IEc mid-
range estimates of WTP to avoid each symptom in the cluster and
assuming additivity of WTPs. The $ value for LRS is the average of
the $ values for the 11 different types of LRS.
Uncertainty: taken to be a continuous uniform distribution across the
range of values described by the 11 LRS types.
3. Acute Bronchitis
$45
Continuous uniform
distribution over the
interval [$13, $77]
Central Est: Average of low and hieh values recommended by TF.C for
use in section 812 analysis (Neumann et al., 1994).
Uncertainty: continuous distribution between low and hiph values
(Neumann et al., 1994) assigns equal likelihood of occurrence of any
value within the range.
4. Acute Respiratory
Symptoms and Illnesses
- Presence of any of
19 acute respiratory
symptoms
- Any Resp.
Symptom
- Increase in Resp.
Illness
$18
1. URS, probability = 40%
LRS, probability = 40%
URS+LRS, prob. = 20%
2. If URS, use URS $ dist.
If LRS, use LRS $ dist.
If URS+LRS, randomly
select one value each from
URS and LRS $
distributions; sum the two
Central Est: Assuming that respiratorv illness and svmptoms can be
characterized as some combination of URS and LRS, namely: URS
with 40% probability, LRS with 40% probability, and both URS and
LRS with 20% probability. The $ value for these endpoints is the
weighted average (using the weights 0.40, 0.40, and 0.20) of the $
values derived for URS, LRS, and URS + LRS.
Uncertainty: based on variability assumed for central estimate, and
URS and LRS uncertainty distributions presented previously.
3
B-
x'
s"
a
S'
s
§
£
a
3
o
^0
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
5. Asthma - Acute
$32
Continuous uniform
distribution over the
interval [$12, $54]
Central Est: Mean of averaee WTP estimates for the four severitv
definitions of a "bad asthma day." Source: Rowe and Chestnut (1986),
a study which surveyed asthmatics to estimate WTP for avoidance of a
"bad asthma day," as defined by the subjects.
Uncertainty: based on the ranee of values estimated for each of the
four severity definitions.
6. Shortness of breath
$5.30
Continuous uniform
distribution over the
interval [$0, $10.60]
Central Est: From Ostro et al.. 1995. This is the mean of the median
estimates from two studies of WTP to avoid a day of shortness of
breath: Dickie et al., 1991 ($0.00), and Loehman et al., 1979 ($10.60).
Uncertainty: taken to be a continuous uniform distribution across the
range of values obtained from the two studies.
Restricted Activity and Work Loss Days
1. WLDs
$83
none available
Central Est: Median weeklv wage for 1990 divided bv 5 (U.S.
Department of Commerce, 1992)
Uncertainty: Insufficient information to derive an uncertainty estimate.
2. RADs
not monetized3
--
-
3. MRADs
$38
triangular distribution
centered at $38 on the
interval [$16, $61]
Central Est: Median WTP estimate to avoid 1 MRRAD ~ minor
respiratory restricted activity day — from Tolley et al. (1986)
(recommended by IEc as the mid-range estimate).
Uncertainty: ranee is based on assumption that value should exceed
WTP for a single mild symptom (the highest estimate for a single
symptom—for eye irritation—is $16.00) and be less than that for a WLD.
The triangular distribution acknowledges that the actual value is likely
to be closer to the point estimate than either extreme.
4. RRADs
not monetized3
--
-
bo
•3,
O
Q)
IS"
Q
a
3
C5
*0
<^>
*0
*0
<^>
-------
Health or Welfare
Endpoint
Estimated Value Per Incidence (1990$)
Derivation of Estimates
Central Estimate
Uncertainty Distribution
Welfare Effects
Household Soiling Damage
$2.50 per
household per
Hg/m3 PM10
(annual cost)
Beta distribution with
mean=$2.50, standard
deviation=$ 1 on the interval
[$1.30, $10.00], The shape
parameters of this
distribution are a=1.2 and
P=7.3.
Central Est: Source: ESEERCO (1994). ESEERCO uses $1.26 as its
low estimate of annual cost of soiling and materials damage per
household (assuming 2.63 persons per household), taken from Manuel
et al. (1982). The Manuel study measured particulate matter as TSP
rather than PM-10. Hypothesizing that at least half of the costs of
household cleaning are for the time value of do-it-yourselfers, which
was not included in the Manuel estimate, ESEERCO multiplied the
Manuel estimate by 2 to get a point estimate of about $2.50, in 1990 $
(reported by ESEERCO as $2.70 in 1992 dollars).
Uncertainty: The Beta distribution selected is a smooth, continuous
function with its probability mass near the mean and it covers the range
of reported estimates.
Visibility
Annual household
WTP = $14 per
unit decrease in
DeciView
(decrease in
DeciView
corresponds to
increase in
visibility)
Triangular distribution
centered at $14 on the
interval [$8, $21]
Central Est: Source: IEc 1997. Calculated bv dividing the household
WTP reported in the McClelland et al. study (1991) by the
corresponding hypothesized change in DeciView.
Uncertainty: Source: IEc 1997. Calculated bv regressing reported
household WTP values on the corresponding change in DeciView
(converted from reported visual range changes) for all relevant city-
scenario combinations posed to respondents in the original studies. The
uncertainty range reflects the 25 percent adjustment for part-whole bias
applied to reported values prior to calculating the lower bound.
Worker Productivity
change in daily
wages: $1 per
worker per 10%
change in 03
none available
Central Est: Based on the elasticity of income with respect to O,
concentration derived from study of California citrus workers (Crocker
and Horst, 1981 and U.S. EPA, 1994). Elasticity applied to the average
daily income for workers engaged in strenuous outdoor labor, $73 (U.S.
1990 Census).
NOTES:
a This endpoint was not monetized because including it in the aggregation of economic benefits would result in double-counting (overlap with WLDs
and MRADs).
3
B-
x'
s"
a
S'
s
§
£
a
3
o
^0
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Results of Valuation of Health
and Welfare Effects
Table 1-3 presents the results of combining the
economic valuations described in this Appendix with
the health and welfare effects results presented in
Appendix D. As noted in Appendix D, there are alter-
native estimates for some health and welfare impacts,
which form the basis of several alternative benefit
estimates. Each of the health effects estimates also
has quantified statistical uncertainty. The range of
estimated health and welfare effects, along with the
uncertain economic unit valuations, were combined
to estimate a range of possible results. The combin-
ing of the health and economic information used the
Monte Carlo method presented in Chapter 7. Table I-
3 shows the mean estimate results, as well as the mea-
sured credible range (upper and lower five percen-
tiles of the results distribution), of economic benefits
for each of the quantified health and welfare catego-
ries.
The results for aggregate monetized benefits were
also calculated using a Monte Carlo method. The re-
sults of the Monte Carlo simulations for the economic
values for each of the major endpoint categories are
presented in Table 1-4. Note that for the upper and
lower fifth percentiles the sum of the estimated ben-
efits from the individual endpoints does not equal the
estimated total. The Monte Carlo method used in the
analysis assumes that each health and welfare end-
point is independent of the others. There is a very low
probability that the aggregate benefits will equal the
sum of the fifth percentile benefits from each of the
ten endpoints.
Table 1-5 shows the estimated total benefits ranges
for the four modeled target years of this study: 1975,
1980, 1985, and 1990. The results of the Monte Carlo
simulations of the aggregate economic benefits for
these four target years are depicted in Figure I-1.
Table 1-6 examines the impact of limiting the
scope of the analysis to locations with more certain
air quality estimates. The main analysis (as shown in
Tables 1-3 through 1-5) covers almost the entire popu-
lation of the 48 States.3 However, the air quality in-
formation is less certain for locations far from a moni-
tor. Table 1-6 presents the results of limiting the analy-
sis to people living within 50 km of an ozone, N02,
S02, or CO monitor, or in counties with a PM moni-
tor. The availability of monitors changes over time.
Hence the proportion of the population included in
this analysis changes overtime as well. Table 1-6 in-
dicates that approximately a quarter of the total ben-
efits estimated in the main analysis comes from areas
with less certain air quality information.
The results of the "all U.S. population" analysis
provides a more accurate depiction of the pattern of
economic benefits across years. The accuracy of the
scale of incidence is less certain. These results pro-
vide a better characterization of the total direct ben-
efits of the Clean Air Act in the lower 48 states than
do the "monitored area only" results because the lat-
ter completely omits historical air quality improve-
ments for about 25 percent of the population. How-
ever, the "all U.S. population" results rely on uncer-
tain extrapolations of pollution concentrations, and
subsequent exposures, from distant monitoring sites
to provide coverage for the 25 percent or so of the
population living far from air quality monitors. Thus,
the main results presented in Tables 1-3 through 1-5
include important uncertainties.
Uncertainties
The uncertainty ranges for the results on the
present value of the aggregate measured monetary
benefits reported in Table 1-3 reflect two important
sources of measured uncertainty:
uncertainty about the avoided incidence of
health and welfare effects deriving from the
concentration-response functions, including
both selection of scientific studies and statis-
tical uncertainty from the original studies; and
uncertainty about the economic value of each
quantified health and welfare effect.
These aggregate uncertainty results incorporate many
decisions about analytical procedures and specific
assumptions discussed in the Appendices to this re-
port.
In order to provide a more complete understand-
ing of the economic benefit results in Table 1-3, sen-
sitivity analyses examine several additional important
aspects of the main analysis. First, this section ex-
3 Except for lead, two to five percent (depending on pollutant) of the population who live in sparsely populated areas are
excluded from the main analysis to maximize computer efficiency. All of the population of the 48 states is included in the lead
analysis.
1-16
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Table 1-3. Criteria Pollutants Health and Welfare Benefits -- Extrapolated to Entire 48 State
Population Present Value (in 1990 using 5% discount rate) of Benefits from 1970 - 1990 (in billions of
1990 dollars).
Eimdl|p«MTrntt
IP«»II Hull torn
PircMnmtt Vallroic ((ItmlllliiifMn* lf lTOIOS))
5th "oilk
Mat
9Sttltn %i31e
Mortality
Mortality (long-term PM-10 exposure)
PM
$2,369
$16,632
$40,597
Mortality (Lead exposure)
Lead
$121
$1,339
$3,910
Chronic Bronchitis
PM
$409
$3,313
$10,401
Other Lead-induced Ailments
Lost IQ Points
Lead
$248
$377
$528
IQ< 70
Lead
$15
$22
$29
Hypertension
Lead
$77
$98
$120
Coronary Heart Disease
Lead
$0
$13
$40
Atherothrombotic brain infarction
Lead
$1
$10
$30
Initial cerebrovascular accident
Lead
$2
$16
$45
Hospital Admissions
*A11 Respiratory
PM & 03
$8
$9
$11
*COPD + Pneumonia
PM & 03
$8
$9
$10
Ischemic He art Disease
PM
$1
$4
$6
Congestive Heart Failure
PM & CO
$3
$5
$7
Other Respiratory-Related Ailments
Children
Shortness of breath, days
PM
$0
$6
$17
** Acute Bronchitis
PM
$0
$7
$18
**Upper & Lower Respiratory Symptoms
PM
$1
$2
$4
Adults
Any of 19 Acute Symptoms
PM&03
$4
$46
$117
All
Asthma Attacks
PM&03
$0
$0
$1
Increase inRespiratory Illness
NO 2
$1
$2
$4
Any Symptom
S02
$0
$0
$0
Restricted Activity and Work Loss Days
MR AD
PM&03
$50
$85
$123
Work Loss Days (WLD)
PM
$30
$34
$39
Human Wei la rc
HouseholdSoiling Damage
PM
$6
$74
$192
Visibility - EasternU.S.
particulates
$38
$54
$71
Decreased Worker Productivity
03
$3
$3
$3
Agriculture (Net Surplus)
03
$11
$23
$35
To avoid double-counting of benefits, the following endpoints weie treated as alternatives:
* Hospital admissions for COPD combined with those for pneumonia are treated as an equally-weighted alternative to hospital
admissions for allrespiratoiy illnesses.
**The definitions of acute bronchitis and upper and lower respiratory illness overlap; both studies count trouble breathing,
diy cough, and wheezing in their estimates. These two studies are treated as alternatives, which reflects the variability of
pollution-induced respiratory effects in children.
1-17
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Table 1-4. Present Value of 1970 to 1990 Monetized Benefits by Endpoint Category for 48 State
Population (billions of $1990, discounted to 1990 at 5 percent).
Present Value
Endpoint
Pollutant(s)
5th %ilc
Mean
95th %ile
Mortality
PM
$2,369
$16,632
$40,597
Mortality
Pb
$121
$1,339
$3,910
Chronic Obstructive Pulmonary Disease
PM
$409
$3,313
$10,401
IQ(LostIQ Pts. + Children w/IQ<70)
Pb
$271
$399
$551
Hypertension
Pb
$77
$98
$120
Hospital Admissions
PM, 03, Pb, & CO
$27
$57
$120
Respiratory-Related Symptoms, Restricted PM, 03, N02, & S02
$123
$182
$261
Activity, & Decreased Productivity
Soiling Damage
PM
$6
$74
$192
Visibility
particulates
$38
$54
$71
Agriculture (Net Surplus)
03
$11
$23
$35
Tabic 1-5. Monte Carlo Simulation Model Results for Target Years. Plus Present Value in 1990
Terms ofTotal Monetized Benefits for Entire 1970 to 1990 Period (in billions of 1990-valuc dollars).
Total Benefits By Year ($Billions)
1975
1980
1985
1990
Present Value (5%)
5th percentile
$87
$235
$293
$329
$5,600
Mean
$355
$930
$1,155
$1,248
$22,200
95th percentile
$799
$2,063
$2,569
$2,762
$49,400
Notes:
Present value reflects compounding of benefits from 1971 to 1990.
"Uncertainty Estimates" are results of Monte Carlo analysis combining economic and physical effects uncertainty (i.e., using both
between- and within-study variability).
Full uncertainty analysis done only for years shown. Uncertainty estimates for intermediate years computed based on ratios of 5th
to 50th percentile and 95th to 50th percentile for years shown. Ratios interpolated between years shown and applied to point
estimates for intermediate years.
1-18
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Figure 1-1. Monte Carlo Simulation Model Results for Target Years (in billions of 1990 dollars).
c
o
PQ
&
$3,000
$2,500
$2,000
$1,500
n $1,000
$500
$0
tc
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
plores the effect of selecting alternative discount rates
on the aggregate present value benefits estimation.
Second, this section examines the sources of the mea-
sured aggregate uncertainty, identifying which of the
measured uncertainty components of incidence and
valuation for individual health effects categories drive
the overall uncertainty results. Third, this section ex-
amines several issues involving the estimated eco-
nomic benefits of mortality.
The Effect of Discount Rates
The main analysis reflected in present value re-
sults shown in Table 1-3 uses a five percent discount
rate. The discount rate primarily enters the calcula-
tions when compounding the economic benefits esti-
mates from individual years between 1970 and 1990
to estimate the present value of the benefits in 1990.
The discount rate also directly enters in the calcula-
tions of the economic values of an IQ point and an
initial case of coronary heart disease.4 There is con-
siderable controversy in the economics and policy lit-
erature about the appropriate discount rate to use in
different settings. Major alternatives recommended by
various authors include a discount rate based on the
social discount rate (typical estimates are in the 2 to 3
percent range), and a discount rate based on the risk-
free rate of return on capital (typically in the 7 to 10
percent range). Table 1-7 presents the aggregate un-
certainty results using three different discount rates:
3 percent, 5 percent and 7 percent. While the aggre-
gate benefits estimates are sensitive to the discount
rate, selecting one of these alternative discount rates
affects the aggregate benefits estimates by only about
15 percent.
The Relative Importance of Different
Components of Uncertainty
The estimated uncertainty ranges in Table 1-3 re-
flect the measured uncertainty associated with both
avoided incidence and economic valuation. A better
understanding of the relative influence of individual
uncertain variables on the overall uncertainty in the
analysis can be gained by isolating the individual ef-
fects of important variables on the range of estimated
benefits. This can be accomplished by holding all the
inputs to the Monte Carlo uncertainty analysis con-
stant (at their mean values), and allowing only one
variable ~ for example, the economic valuation of
mortality ~ to vary across the range of that variable's
uncertainty. The sensitivity analysis then isolates how
this single source of variability contributes to the varia-
tion in estimated total benefits. The results are sum-
marized in Figure 1-2. The nine individual uncertainty
factors that contribute the most to the overall uncer-
tainty are shown in Figure 1-2, ordered by the relative
significance of their contribution to overall uncer-
tainty. Each of the additional sources of quantified
uncertainty in the overall analysis not shown contrib-
ute a smaller amount of uncertainty to the estimates
of monetized benefits than the sources that are shown.
Tabic 1-7. Eflcci of Alternative Discount Rates on Present Value of Total Monetized Benefits lor
1970 to 1990 (in trillions of 1990 dollars).
Present Value in 1990 of TotalBenefits
(Trillions of 1990 Dollars)
3%
5%
7%
5 th percentile
$4.9
$5.6
$6.5
Mean
$19.2
$22.2
$25.8
95th percentile
$42.7
$49.4
$57.5
No tes •
Present value reflects compounding of benefits from 1971 to 1990.
4 The estimated economic value of lost IQ points due to lead exposure is based on the present value of the impact on lifetime
earnings. A discount rate is required to calculate that present value. The impact on income primarily occurs during adulthood, which
is 20 to 70 years after the initial lead exposure. This significant lag results in the discount rate having a significant impact on the
estimated economic benefits of the IQ loss. Similarly, the cost of illness estimate for an initial case of CHD includes the present value
of the annual stream of medical costs incurred after the event, the calculation of which requires an estimate of the discount rate.
1-20
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Figure 1-2. Uncertainty Ranges Deriving From Individual Uncertainty Factors.
a
o
s
.
(/)
—
—
0
TO
¦c
o
0
O
C
0
'sz
o
c
o
c
o
'-4—'
TO
TO
O
_Q
CL
O
c
0
¦o
LO
O
c
£
—
_£=
Q_
o
J= 0
O O
c c
O 0
o
0
O o
^ 0
(/) "O
S O
T3
o
o
X
o
'-4—'
TO
TO
>
O)
o
'-4—'
TO
TO
>
o
9 £
< TO
2 «
o
CO
Because of the multiple uncertainties in the ben-
efits estimation, the total estimated present value of
the monetary benefits of the 1970 to 1990 Clean Air
Act range from a low of about $5.6 trillion to a high
of about $49.4 trillion (in 1990 dollars, discounted at
five percent). Most of the uncertainty in the total esti-
mated benefit levels comes from uncertainty in the
estimate of the economic valuation of mortality, fol-
lowed by the uncertainty in the incidence of mortality
from PM (as a surrogate for all non-lead air pollu-
tion). The incidence of lead-induced mortality also
has a significant influence on the overall uncertainty.
The importance of mortality is not surprising, because
the benefits associated with reduced mortality are such
a large share of overall monetized benefits.
The uncertainty in both the incidence and valua-
tion of chronic bronchitis are the two other signifi-
cant factors driving the overall uncertainty range. The
modeled uncertainty in the other remaining health and
welfare endpoints in the analysis contribute relatively
small amounts to the overall uncertainty in the esti-
mate of total monetary benefits of the Clean Air Act.
Most of these other endpoints account for a relatively
small proportion of the overall benefits estimates,
making it unlikely that they could contribute signifi-
cantly to the overall uncertainty. Estimates of either
the mean values or standard errors of these variables
are generally very small relative to estimated total
monetary benefits.
Economic Benefits Associated with
Reducing Premature Mortality
Because the economic benefits associated with
premature mortality are the largest source of mon-
etized benefits in the analysis, and because the uncer-
tainties in both the incidence and value of premature
mortality are the most important sources of uncertainty
in the overall analysis, it is useful to examine the
mortality benefits estimation in greater detail.
The analytical procedure used in the main analy-
sis to estimate the monetary benefits of avoided pre-
mature mortality assumes that the appropriate eco-
nomic value for each incidence is a value from the
currently accepted range of the value of a statistical
life. As discussed above, the estimated value per pre-
dicted incidence of excess premature mortality is
modeled as a Weibull distribution, with a mean value
of $4.8 million and a standard deviation of $3.2 mil-
lion. This estimate is based on 26 studies of the value
of mortal risks.
1-21
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
There is considerable uncertainty as to whether
the 26 studies on the value of a statistical life provide
adequate estimates of the value of a statistical life
saved by air pollution reduction. Although there is
considerable variation in the analytical designs and
data used in the 26 underlying studies, the majority of
the studies involve the value of risks to a middle-aged
working population. Most of the studies examine dif-
ferences in wages of risky occupations, using a wage-
hedonic approach. Certain characteristics of both the
population affected and the mortality risk facing that
population are believed to affect the average willing-
ness to pay (WTP) to reduce the risk. The appropri-
ateness of a distribution of WTP estimates from the
26 studies for valuing the mortality-related benefits
of reductions in air pollution concentrations therefore
depends not only on the quality of the studies (i.e.,
how well they measure what they are trying to mea-
sure), but also on (1) the extent to which the subjects
in the studies are similar to the population affected by
changes in pollution concentrations, and (2) the ex-
tent to which the risks being valued are similar. As
discussed below, there are possible sources of both
upward and downward bias in the estimates provided
by the 26 studies when applied to the population and
risk being considered in this analysis.
If the individuals who die prematurely from air
pollution are consistently older than the population in
the valuation studies, the mortality valuations based
on middle-aged people may provide a biased estimate
of the willingness to pay of older individuals to re-
duce mortal risk. There is some evidence to suggest
that the people who die prematurely from exposure to
ambient particulate matter tend to be older than the
populations in the valuation studies. In the general
U.S. population far more older people die than younger
people; 88 percent of the deaths are among people
over 64 years old. It is difficult to establish the pro-
portion of the pollution-related deaths that are among
the older population because it is impossible to iso-
late individual cases where one can say with even rea-
sonable certainty that a specific individual died be-
cause of air pollution.
There is considerable uncertainty whether older
people will have a greater willingness to pay to avoid
risks than younger people. There is reason to believe
that those over 65 are, in general, more risk averse
than the general population, while workers in
wage-risk studies are likely to be less risk averse than
the general population. More risk averse people would
have a greater willingness to pay to avoid risk than
less risk averse people. Although the list of recom-
mended studies excludes studies that consider only
much-higher-than- average occupational risks, there
is nevertheless likely to be some selection bias in the
remaining studies ~ that is, these studies are likely to
be based on samples of workers who are, on average,
more risk-loving than the general population. In con-
trast, older people as a group exhibit more risk averse
behavior.
In addition, it might be argued that because the
elderly have greater average wealth than those
younger, the affected population is also wealthier, on
average, than wage-risk study subjects, who tend to
be blue collar workers. It is possible, however, that
among the elderly it is largely the poor elderly who
are most vulnerable to pollution-related mortality risk
(e.g., because of generally poorer health care). If this
is the case, the average wealth of those affected by a
pollution reduction relative to that of subjects in
wage-risk studies is uncertain. In addition, the work-
ers in the wage-risk studies will have potentially more
years remaining in which to acquire streams of in-
come from future earnings.
Although there may be several ways in which job-
related mortality risks differ from air pollution-related
mortality risks, the most important difference may be
that job-related risks are incurred voluntarily whereas
air pollution-related risks are incurred involuntarily.
There is some evidence (see, for example, Violette
and Chestnut, 1983) that people will pay more to re-
duce involuntarily incurred risks than risks incurred
voluntarily. If this is the case, WTP estimates based
on wage-risk studies may be downward biased esti-
mates of WTP to reduce involuntarily incurred air
pollution-related mortality risks.
Finally, another possible difference related to the
nature of the risk may be that some workplace mor-
tality risks tend to involve sudden, catastrophic events
(e.g., workplace accidents), whereas air pollution-re-
lated risks tend to involve longer periods of disease
and suffering prior to death. Some evidence suggests
that WTP to avoid a risk of a protracted death involv-
ing prolonged suffering and loss of dignity and per-
sonal control is greater than the WTP to avoid a risk
(of identical magnitude) of sudden death. Some work-
place risks, such as risks from exposure to toxic chemi-
cals, may be more similar to pollution-related risks. It
is not clear, however, what proportion of the work-
place risks in the wage-risk studies were related to
workplace accidents and what proportion were risks
1-22
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
from exposure to toxic chemicals. To the extent that
the mortality risks addressed in this assessment are
associated with longer periods of illness or greater pain
and suffering than are the risks addressed in the valu-
ation literature, the WTP measurements employed in
the present analysis would reflect a downward bias.
The direction of bias resulting from the age dif-
ference is unclear, particularly because age is con-
founded by risk aversion (relative to the general popu-
lation). It could be argued that, because an older per-
son has fewer expected years left to lose, his WTP to
reduce mortality risk would be less than that of a
younger person. This hypothesis is supported by one
empirical study, Jones-Lee et al. (1985), that found
the value of a statistical life at age 65 to be about 90
percent of what it is at age 40. Citing the evidence
provided by Jones-Lee et al. (1985), a recent sulfate-
related health benefits study conducted for EPA (U.S.
EPA, 1995) assumes that the value of a statistical life
for those 65 and over is 75 percent of what it is for
those under 65.
There is substantial evidence that the income elas-
ticity of WTP for health risk reductions is positive
(see, for example, Alberini et al., 1994; Mitchell and
Carson, 1986; Loehman and VoHuDe, 1982;Gerking
etal., 1988; and Jones-Lee etal., 1985), although there
is uncertainty about the exact value of this elasticity.
Individuals with higher incomes (or greater wealth)
should, then, be willing to pay more to reduce risk, all
else equal, than individuals with lower incomes or
wealth. Whether the average income or level of wealth
of the population affected by PM reductions is likely
to be significantly different from that of subjects in
wage-risk studies, however, is unclear, as discussed
above.
The need to adjust wage-risk-based WTP esti-
mates downward because of the likely upward bias
introduced by the age discrepancy has received sig-
nificant attention (see Chestnut, 1995; IEc, 1992). If
the age difference were the only difference between
the population affected by pollution changes and the
subjects in the wage-risk studies, there might be some
justification for trying to adjust the point estimate of
$4.8 million downward. Even in this case, however,
the degree of the adjustment would be unclear. There
is good reason to suspect, however, that there are bi-
ases in both directions. Because in each case the ex-
tent of the bias is unknown, the overall direction of
bias in the mortality values is similarly unknown.
Adjusting the estimate upward or downward to com-
pensate for any one source of bias could therefore in-
crease the degree of bias. Therefore, the range of val-
ues from the 26 studies is used in the primary analy-
sis without adjustment.
Examining the sensitivity of the overall results to
the mortality values can help illuminate the potential
impacts of alternative mortality valuations. As men-
tioned above, a contractor study performed for EPA
used one approach to evaluate the economic value of
sulfate-related human health improvements resulting
from 1990 Clean Air Act Amendments Title IV acid
rain controls. That study assumed that 85 percent of
the people dying from sulfates (an important compo-
nent of particulate matter) were over 65, and that
people over 65 have a willingness to pay to avoid a
mortal risk that is 75 percent of the values that middle-
aged people have. Using this approach, the value of
an average statistical life (using a weighted average)
is reduced to 79 percent of the previous value.
If statistical life-years lost are used as the unit of
measure, rather than statistical lives lost, the benefit
attributed to avoiding a premature death depends di-
rectly on how premature it is. One way to estimate
the value of a statistical life-year assumes that the value
of a statistical life is directly related to remaining life
expectancy and a constant value for each life-year.
Such an approach results in smaller values of a statis-
tical life for older people, who have shorter life ex-
pectancies, and larger values for younger people. For
example, if the $4.8 million mean value of avoiding
death for people with a 35 year life expectancy is as-
sumed to be the discounted present value of 35 equal-
valued statistical life-years, the implied value of each
statistical life-year is $293,000 (using a 5% discount
rate). The average number of life-years lost by indi-
viduals dying prematurely from exposure to PM is 14
years. This average is obtained by multiplying the
predicted number of PM-related premature deaths in
each age category by the life expectancy for that age
category and dividing by the total number of PM-re-
lated premature deaths.) Using $293,000 per life-year,
the discounted present value of a statistical life for a
person with 14 years of expected life remaining (e.g.,
a 70 year old) is $2.9 million). If statistical life-years
lost are used to value fatal risks, however, other
sources of uncertainty are introduced in the valuation
process.
1-23
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
If statistical life-years lost is the unit of measure,
the value of a statistical life lost depends on (1) how
many years of expected life are lost, (2) the
individual's discount rate, and (3) whether the value
of an undiscounted statistical life-year is the same no
matter which life-year it is (e.g., the undiscounted
value of the seventy-fifth year of life is the same as
the undiscounted value of the fortieth year of life).
Each of these is uncertain. The uncertainty surround-
ing the expected years of life lost by an individual
involves the uncertainty about whether individuals
who die from exposure to air pollution are average
individuals in the demographic (e.g., age-gender-race)
classification to which they belong. The uncertainty
surrounding individuals' discount rates is well docu-
mented. Finally, even if it is assumed that all life-years
are valued the same (apart from differences due to
discounting), the value of a statistical life-year is de-
rived from the value of a statistical life (of a 40 year
old) and the discount rate, each of which is uncertain.
Using life-years lost as the unit of measure means
that, rather than estimating a single value of a statisti-
cal life lost (applicable to all ages), the analysis would
instead estimate age-specific values of statistical lives
lost. It is unclear whether the variability of estimates
of age-specific values of statistical lives lost (in par-
ticular, for ages greater than the average age of work-
ers in the wage-risk studies) would be less than or
greater than the variability of the original estimate of
the value of a statistical life lost from which they would
be derived. If there is an age-related upward bias in
the central tendency value of a statistical life that is
larger than any downward bias, then valuing life-years
rather than lives lost may decrease the bias. Even this,
however, is uncertain.
In spite of the substantial uncertainties and pau-
city of available information, this section presents an
example of a preliminary estimate of the present value
of avoided premature mortality using the life-years
lost approach. The basic approach is to (1) estimate
the number of pollution-related premature deaths in
each age category, (2) estimate the average number
of life-years lost by an individual in a given age cat-
egory dying prematurely, and (3) using the value of a
statistical life-year of $293,000, described above (as-
suming that the undiscounted value of a life-year is
the same no matter when in an individual's life it is)
and assuming a five percent discount rate, calculate
the value of a statistical life lost in each age category.
To obtain estimates of the number of air pollu-
tion-related deaths in each age cohort, it is preferable
to have age-specific relative risks. Many of the epide-
miological studies, however, do not provide any esti-
mate of such age-specific risks. In this case, the age-
specific relative risks must be assumed to be identi-
cal.
Some epidemiology studies on PM do provide
some estimates of relative risks specific to certain age
categories. The limited information that is available
suggests that relative risks of mortality associated with
exposure to PM are greater for older people. Most of
the available information comes from short-term ex-
posure studies. There is considerable uncertainty in
applying the evidence from short-term exposure stud-
ies to results from long-term (chronic exposure) stud-
ies. However, using the available information on the
relative magnitudes of the relative risks, it is possible
to form a preliminary assessment of the relative risks
by different age classes.
The analysis presented below uses two alterna-
tive assumptions about age-specific risks: (1) there is
a constant relative risk (obtained directly from the
health literature) that is applicable to all age cohorts,
and (2) the relative risks differ by age, as estimated
from the available literature. Estimates of age-spe-
cific PM coefficients (and, from these, age-specific
relative risks) were derived from the few age-specific
PM coefficients reported in the epidemiological lit-
erature. These estimates in the literature were used to
estimate the ratio of each age-specific coefficient to a
coefficient for "all ages" in such a way that consis-
tency among the age-specific coefficients is preserved
— that is, that the sum of the health effects incidences
in the separate, non-overlapping age categories equals
the health effects incidence for "all ages." These ra-
tios were then applied to the coefficient from Pope et
al. (1995). Details of this approach are provided in
Post and Deck (1996). Because Pope etal. considered
only individuals age 30 and older (instead of all ages),
the resulting age-specific PM coefficients may be
slightly different from what they would have been if
the ratios had been applied to an "all ages" coeffi-
cient. The differences, however, are likely to be mini-
mal and well within the error bounds of this exercise.
The age-specific relative risks used in the example
below assume that the relative risks for people under
65 are only 16 percent of the population-wide aver-
age relative risk, the risks for people from 65 to 74
are 83 percent of the population-wide risk, and people
1-24
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
75 and older have a relative risk 55 percent greater
than the population average. Details of this approach
are provided in Post and Deck (1996).
The life-years lost approach also requires an esti-
mate of the number of life-years lost by a person dy-
ing prematurely at each given age. The average num-
ber of life-years lost will depend not only on whether
relative risks are age-specific or uniform across all
age groups, but also on the distribution of ages in the
population in a location. As noted above, using the
same relative risk for all age categories, the average
number of life-years lost in PM-related premature
deaths in the United States was estimated to be 14
years. Using the age-specific relative risk estimates
developed for this analysis, the average number of
life-years lost becomes 9.8 years. In a location with a
population that is younger than average in the United
States, the same age-specific relative risks will pro-
duce a larger estimated average number of life-years
lost. For example, using the same age-specific rela-
tive risks, the average number of life-years lost in PM-
related premature deaths in Los Angeles County,
which has a younger population, is estimated to be
15.6 years.
The present value benefits estimates for PM-re-
lated mortality using the alternative approaches dis-
cussed above are shown in Table 1-8. Table 1-8 is based
on a single health study: Pope et al., 1995. Alterna-
tive studies, or the uncertainty approach used in the
primary analysis, would result in a similar pattern of
the relationship between valuation approaches. The
pattern of monetized mortality benefits across the dif-
ferent valuation procedures shown in Table 1-8 is es-
sentially invariant to the particular relative risk and
the particular dollar value used.
As noted above, the life-years lost approach used
here assumes that people who die from air pollution
are typical of people in their age group. The estimated
value of the quantity of life lost assumes that the people
who die from exposure to air pollution had an aver-
age life expectancy. However, it is possible that the
people who die from air pollution are already in ill
health, and that their life expectancy is less than a
typical person of their age. If this is true, then the num-
ber of life years lost per PM-related death would be
lower than calculated here, and the economic value
would be smaller.
The extent to which adverse effects of particulate
matter exposure are differentially imposed on people
of advanced age and/or poor health is one of the most
important current uncertainties in air pollution-related
health studies. There is limited information, prima-
rily from the short-term exposure studies, which sug-
gests that at least some of the estimated premature
mortality is imposed disproportionately on people who
are elderly and/or of poor health. The Criteria Docu-
ment for Particulate Matter (U.S. EPA, 1996) identi-
fies only two studies which attempt to evaluate this
disproportionality. Spix et al. (1994) suggests that a
small portion of the PM-associated mortality occurs
in individuals who would have died in a short time
anyway. Cifuentes and Lave (1996) found that 37 to
87 percent of the deaths from short-term exposure
could have been premature by only a few days, al-
though their evidence is inconclusive.
Table 1-8. Alternative Esti mates of the Present Value of Mortality Associated With PM
(based on Pope et al., 1996, in trillions of 1990 dollars).
Present Value of PM
Valuation Procedure
Mortality Benefits
Primary Analysis Method ( $4.8 million per statistical life saved)
$16.6
Life Years Lost approaches
Single relative risk, valuation using 5% discounting
$9.1
Annroximate as>c-sr>ccific relative risk, valuation usins> 5% discounting
$8 3
Notes:
Present value reflects compounding of benefits from 1971 to 1990, usinga 5 percent discount rate.
1-25
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Prematurity of death on the order of only a few
days is likely to occur largely among individuals with
pre-existing illnesses. Such individuals might be par-
ticularly susceptible to a high PM day. To the extent
that the pre-existing illness is itself caused by or ex-
acerbated by chronic exposure to elevated levels of
PM, however, it would be misleading to define the
prematurity of death as only a few days. In the ab-
sence of chronic exposure to elevated levels of PM,
the illness would either not exist (if it was caused by
the chronic exposure to elevated PM) or might be at a
less advanced stage of development (if it was not
caused by but exacerbated by elevated PM levels).
The prematurity of death should be calculated as the
difference between when the individual died in the
"elevated PM" scenario and when he would have died
in the "low PM" scenario. If the pre-existing illness
was entirely unconnected with chronic exposure to
PM in the "elevated PM" scenario, and if the indi-
vidual who dies prematurely because of a peak PM
day would have lived only a few more days, then the
prematurity of that PM-related death is only those few
days. If, however, in the absence of chronic exposure
to elevated levels of PM, the individual's illness would
have progressed more slowly, so that, in the absence
of a particular peak PM day the individual would have
lived several years longer, the prematurity of that PM-
related death would be those several years.
Long-term studies provide evidence that a por-
tion of the loss of life associated with long-term ex-
posure is independent of the death from short-term
exposures, and that the loss of life-years measured in
the long-term studies could be on the order of years.
If much of the premature mortality associated with
PM represents short term prematurity of death im-
posed on people who are elderly and/or of ill health,
the estimates of the monetary benefits of avoided
mortality may overestimate society's total willingness
to pay to avoid particulate matter-related premature
mortality. On the other hand, if the premature mortal-
ity measured in the chronic exposure studies is de-
tecting excess premature deaths which are largely in-
dependent of the deaths predicted from the short term
studies, and the disproportionate effect on the elderly
and/or sick is modest, the benefits measured in this
report could be underestimates of the total value. At
this time there is insufficient information from both
the medical and economic sciences to satisfactorily
resolve these issues from a theoretical/analytical stand-
point. Until there is evidence from the physical and
social sciences which is sufficiently compelling to
encourage broad support of age-specific values for
reducing premature mortality, EPA will continue to
use for its primary analyses a range of values for mor-
tality risk reduction which assumes society values re-
ductions in pollution-related premature mortality
equally regardless of who receives the benefit of such
protection.
1-26
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Economic Valuation References
Abbey, D.E., F. Petersen, P. K. Mills, and W. L. Beeson.
1993. "Long-Term Ambient Concentrations of
Total Suspended Particulates, Ozone, and Sul-
fur Dioxide and Respiratory Symptoms in a
Nonsmoking Population. "Archives of Environ-
mental Health 48(1): 33-46.
Abt Associates, Inc. 1992. The Medical Costs of Five
Illnesses Related to Exposure to Pollutants.
Prepared for U.S. EPA, Office of Pollution Pre-
vention and Toxics, Washington, DC.
Abt Associates, Inc. 1996. Section 812 Retrospective
Analysis: Quantifying Health and Welfare Ben-
efits. Draft. Prepared for U.S. EPA, Office of
Policy Planning and Evaluation, Washington
DC. May.
Alberini, A., A. Krupnick, M. Cropper, and W.
Harrington. 1994. "Air Quality and the Value
of Health in Taiwan." Paper presented at the
annual meeting of the Eastern Economics As-
sociation, Boston, Massachusetts, March.
Brookshire, David S., Ralph C. d'Arge, William D.
Schulze and Mark A. Thayer. 1979. Methods
Development for Assessing Air Pollution Con-
trol Benefits, Vol. IP. Experiments in Valuing
Non-Market Goods: A Case Study of Alterna-
tive Benefit Measures of Air Pollution Control
in the South Coast Air Basin of Southern Cali-
fornia. Prepared for the U.S. Environmental
Protection Agency, Office of Research and De-
velopment.
Chestnut, Lauraine G. 1995. Dollars and Cents: The
Economic and Health Benefits of Potential
Particulate Matter Reductions in the United
States. Prepared for the American Lung Asso-
ciation.
Chestnut, Lauraine G. and Robert D. Rowe. 1989. "Eco-
nomic Valuation of Changes in Visibility: A
State of the Science Assessment for NAPAP,"
as cited in National Acid Precipitation Assess-
ment Program, Methods for Valuing Acidic
Deposition and Air Pollution Effects. NAPAP
State of Science and State of Technology Re-
port No. 27, Part B. December.
Cifuentes, L. and L.B. Lave. 1996. "Association of
Daily Mortality and Air Pollution in Phila-
delphia, 1983-1988." J. Air Waste Manage.
Assoc.: in press.
Crocker T. D. and R. L. Horst, Jr. 1981. "Hours of
Work, Labor Productivity, and Environmen-
tal Conditions: a Case Study." The Review of
Economics and Statistics 63:361-368.
Cropper, M.L. and A.J. Krupnick. 1990. "The Social
Costs of Chronic Heart and Lung Disease,"
Resources for the Future Discussion Paper QE
89-16-REV.
Dickie, M. et al. 1991. Reconciling Averting Behav-
ior and Contingent Valuation Benefit Esti-
mates of Reducing Symptoms of Ozone Ex-
posure (draft), as cited in Neumann, J.E.,
Dickie, M.T., and R.E. Unsworth. 1994. In-
dustrial Economics, Incorporated. Memoran-
dum to Jim DeMocker, U.S. EPA, Office of
Air and Radiation. March 31.
Elixhauser, A., R.M. Andrews, and S. Fox. 1993.
Clinical Classifications for Health Policy
Research: Discharge Statistics by Principal
Diagnosis and Procedure. Agency for Health
Care Policy and Research (AHCPR), Center
for General Health Services Intramural Re-
search, U.S. Department of Health and Hu-
man Services.
Empire State Electric Energy Research Corporation
(ESEERCO). 1994. New York State Environ-
mental Externalities Cost Study. Report 2:
Methodology. Prepared by: RCG/Hagler,
Bailly, Inc., November.
Gerking, S., M. DeHaan, and W. Schulze. 1988. "The
Marginal Value of Job Safety: A Contingent
Valuation Study." Journal of Risk and Un-
certainty 1: 185-199.
Industrial Economics, Incorporated (IEc). 1992. Ap-
proaches to Environmental Benefits Assess-
ment to Support the Clean Air Act Section 812
Analysis. Prepared by Robert E. Unsworth,
James E. Neumann, and W. Eric Browne, for
Jim DeMocker, Office of Policy Analysis and
Review, Office of Air and Radiation, U.S.
Environmental Protection Agency. 6 Novem-
ber.
1-27
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
Industrial Economics, Incorporated (IEc). 1993a.
"Analysis of Visibility Valuation Issues for
the Section 812 Study," Memorandum to Jim
DeMocker, Office of Policy Analysis and
Review, Office of Air and Radiation, U.S.
Environmental Protection Agency, prepared
by Jim Neumann, Lisa Robinson, and Bob
Unsworth. September 30.
Industrial Economics, Incorporated (IEc). 1997. "Vis-
ibility Valuation for the CAA Section 812
Retrospective Analysis," Memorandum to
Jim DeMocker, Office of Policy Analysis and
Review, Office of Air and Radiation, U.S.
Environmental Protection Agency, prepared
by Michael H. Hester and James E. Neumann.
18 February.
Irwin, Julie, William Schulze, Gary McClelland,
Donald Waldman, David Schenk, Thomas
Stewart, Leland Deck, Paul Slovic, Sarah
Lictenstein, and Mark Thayer. 1990. Valuing
Visibility: A Field Test of the Contingent Valu-
ation Method. Prepared for Office of Policy,
Planning and Evaluation, U.S. Environmen-
tal Protection Agency. March.
Jones-Lee, M.W., et al. 1985. "The Value of Safety:
Result of a National Sample Survey." Eco-
nomic Journal 95(March): 49-72.
Krupnick, A.J. and M.L. Cropper. 1992. "The Effect
of Information on Health Risk Valuations,"
Journal of Risk and Uncertainty 5(2): 29-48.
Loehman, E.T., S.V. Berg, A.A. Arroyo, R.A.
Hedinger, J.M. Schwartz, M.E. Shaw, R.W.
Fahien, V.H. De, R.P. Fishe, D.E. Rio, W.F.
Rossley, and A.E.S. Green. 1979. "Distribu-
tional Analysis of Regional Benefits and Cost
of Air Quality Control." Journal of Environ-
mental Economics and Management 6: 222-
243.
Loehman, E.T. and Vo Hu De. 1982. "Application of
Stochastic Choice Modeling to Policy Analy-
sis of Public Goods: A Case Study of Air
Quality Improvements." The Review of Eco-
nomics and Statistics 64(3): 474-480.
Manuel, E.H., R.L. Horst, K.M. Brennan, W.N. Lanen,
M.C. Duff, and J.K Tapiero. 1982. Benefits
Analysis of Alternative Secondary National
Ambient Air Quality Standards for Sulfur Di-
oxide and Total Suspended Particulates, Vol-
umes I-IV Prepared for U.S. Environmental
Protection Agency, Office of Air Quality
Planning and Standards, Research Triangle
Park, NC.
McClelland, Gary, William Schulze, Donald
Waldman, Julie Irwin, David Schenk, Tho-
mas Stewart, Leland Deck and Mark Thayer.
1991. Valuing Eastern Visibility: A Field Test
of the Contingent Valuation Method. Prepared
for Office of Policy, Planning and Evaluation,
U.S. Environmental Protection Agency. June.
Mitchell, R.C. and R.T. Carson. 1986. "Valuing Drink-
ing Water Risk Reductions Using the Con-
tingent Valuation Methods: A Methodologi-
cal Study of Risks from THM and Giardia."
Paper prepared for Resources for the Future,
Washington, DC.
Moore, M.J. and W.K. Viscusi. 1988. "The Quantity-
Adjusted Value of Life". Economic Inquiry
26(3): 369-388.
National Acid Precipitation Assessment Program
(NAPAP). 1991. Acidic Deposition: State of
Science and Technology (Summary Report).
(Washington, DC: NAPAP). September.
Neumann, J.E., M. T. Dickie, and R.E. Unsworth.
1994. Industrial Economics, Incorporated.
Memorandum to Jim DeMocker, U.S. EPA,
Office of Air and Radiation. Linkage Between
Health Effects Estimation and Morbidity
Valuation in the Section 812 Analysis —
Draft Valuation Document. March 31.
Ostro, B.D., M.J. Lipsett, J.K. Mann, H. Braxton-
Owens, and M.C. White. 1995. "Air Pollu-
tion and Asthma Exacerbations Among Afri-
can American Children in Los Angeles." In-
halation Toxicology.
Pope, C.A., III, M.J. Thun, M.M. Namboodiri, D.W.
Dockery, J.S. Evans, F.E. Speizer, and C.W.
Heath, Jr. 1995. "Particulate Air Pollution as
a Predictor of Mortality in a Prospective Study
of U.S. Adults." Am. J. Respir. Crit. Care
Med. 151: 669-674.
1-28
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Post, Ellen and L. Deck. 1996. Abt Associates Inc.
Memorandum to Tom Gillis, U.S. EPA, Of-
fice of Office of Policy Planning and Evalua-
tion. September 20.
Rowe, R.D. and L.G. Chestnut. 1986. "Oxidants and
Asthmatics in Los Angeles: A Benefits Analy-
sis—Executive Summary." Prepared by En-
ergy and Resource Consultants, Inc. Report
to the U.S. EPA, Office of Policy Analysis.
EPA-230-09-86-018. Washington, D.C.
March.
Salkever, D.S. 1995. "Updated Estimates of Earnings
Benefits from Reduced Exposure of Children
to Environmental Lead." Environmental Re-
search 70: 1-6.
Schwartz, J. 1994. "Societal Benefits of Reducing
Lead 'Exposure" Environmental Research 66:
105-124.
Spix, C., J. Heinrich, D. Dockery, J. Schwartz, G.
Volksch, K. Schwinkowski, C. Collen, and
H.E. Wichmann. 1994. Summary of the
Analysis and Reanalysis Corresponding to the
Publication Air Pollution and Daily Mortal-
ity in Erfurt, East Germany 1980-1989. Sum-
mary report for: Critical Evaluation Work-
shop on Particulate Matter—Mortality Epi-
demiology Studies; November; Raleigh, NC.
Wuppertal, Germany: Bergische Universitat-
Gesamthochschule Wuppertal.
Taylor, T.N., P H. Davis, J.C. Torner, J. Holmes, J.W.
Meyer, and M. F. Jacobson. 1996. "Lifetime
Cost of Stroke in the United States." Stroke
27(9): 1459-1466.
Tolley, G.S. et al. 1986. Valuation of Reductions in
Human Health Symptoms and Risks. Univer-
sity of Chicago. Final Report for the U.S.
Environmental Protection Agency. January.
U.S. Department of Commerce, Economics and Sta-
tistics Administration. 1992. Statistical Ab-
stract of the United States, 1992: The National
Data Book. 112th Edition, Washington, D.C.
U.S. Environmental Protection Agency (U.S. EPA).
1994. Documentation for Oz-One Computer
Model (Version 2.0). Office of Air Quality
Planning and Standards. Prepared by:
Mathtech, Inc., under EPA Contract No.
68D30030, WA 1-29. August.
U.S. Environmental Protection Agency (U.S. EPA).
1995. Human Health Benefits From Sulfate
Reductions Under Title IV of the 1990 Clean
Air Act Amendments. Prepared by Hagler
Bailly Consulting, Inc. for U.S. EPA, Office
of Air and Radiation, Office of Atmospheric
Programs. November 10.
U.S. Environmental Protection Agency (U.S. EPA).
1996. Air Quality Criteria for Particulate
Matter, Volume III of III. Office of Research
and Development, Washington DC. EPA/600/
P-95/001cF
Violette, D.M. and L.G. Chestnut. 1983. Valuing Re-
duction in Risks: A Review of the Empirical
Estimates. Report prepared for the U.S. En-
vironmental Protection Agency, Washington,
D.C. EPA-230-05-83-002.
Viscusi, W.K. 1992. Fatal Tradeoffs: Public andPri-
vate Responsibilities for Risk. (New York:
Oxford University Press).
Viscusi, W.K., W. A. Magat, and J. Huber. 1991.
"Pricing Environmental Health Risks: Survey
Assessments of Risk-Risk and Risk-dollar
Tradeoffs." Journal of Environmental Eco-
nomics and Management 201: 32-57.
Wittels, E.H., J.W. Hay, and A.M. Gotto, Jr. 1990.
"Medical Costs of Coronary Artery Disease
in the United States." The American Journal
of Cardiology 65: 432-440.
World Health Organization (WHO). 1996. Final Con-
sultation on Updating and Revision of the Air
Quality Guidelines for Europe. Bilthoven, The
Netherlands 28-31 October, 1996 ICP EHH
018VD96 2.il.
1-29
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
1-30
-------
Appendix J: Future Directions
Research Implications
In virtually any benefit analysis of environmen-
tal issues, the state of scientific information limits the
degree of coverage possible and the confidence in
benefit estimation. For most benefit categories, fur-
ther scientific research would allow for a better quanti-
fication of benefits. One of the major outcomes of the
retrospective analysis is a clear delineation of the
major limitations in the scientific and economics lit-
erature in carrying out an analysis of this scope. Of-
ten, a list of research needs is generated in studies
such as this, but there is no clear internal mechanism
to address these needs. With this study (and the ongo-
ing section 812 program), a process has been initiated
where identified research needs are to be integrated
into EPA's overall extramural research grants pro-
gram, administered by the Office of Research and De-
velopment. It is hoped that the research projects that
flow from this process will enable future analyses to
be less uncertain and more comprehensive.
Certain of the limitations in the retrospective
analysis are directly related to the historical nature of
the analysis, such as sparse information about air qual-
ity in the early 1970's in many areas in the country.
Other important limitations are related to the effects
of elevated airborne lead concentrations, which has
been virtually eliminated by the removal of lead from
gasoline. A better understanding of these relationships
would improve our understanding of the historical
impact of the Clean Air Act, but would only indirectly
contribute to developing future air pollution policy.
However, most of the research that will reduce the
major gaps and uncertainties needed to improve the
section 812 analyses will be directly relevant to EPA's
primary ongoing mission of developing and imple-
menting sound environmental policies to meet the
national goals established in the Clean Air Act and
other legislation.
There are a number of biological, physical and
economic research areas which the EPA Project Team
identified as particularly important for improving fu-
ture section 812 analyses. These research topics can
be divided into two principal categories: (1) those
which might reduce uncertainties in cost and benefit
estimates with significant potential for influencing
estimated net benefits of the Clean Air Act, and (2)
those which might improve the comprehensiveness
of section 812 assessments by facilitating quantifica-
tion and/or monetization of currently excluded cost
or benefit endpoints. The following subsections pro-
vide examples of research topics which, if pursued,
might improve the certainty and/or comprehensive-
ness of future section 812 studies.
Research Topics to Reduce Uncertainty
Scientific information about the effects of long-
term exposure to air pollutants is just beginning to
emerge, but continues to be the subject of intense sci-
entific inquiry. The relationship between chronic PM
exposure and excess premature mortality included in
the quantified results of the present analysis is one
example of such research. However, many other po-
tential chronic effects that are both biologically plau-
sible and suggested by existing research are not in-
cluded. Research to identify the relationship linking
certain known or hypothesized physical effects (e.g.,
ozone's effects on lung function or fibrosis) with the
development of serious health effects (e.g., cardiop-
ulmonary diseases or premature mortality), and the
appropriate economic valuation of the willingness to
pay to avoid the risks of such diseases, would reduce
the uncertainty caused by a major category of excluded
health effects which could have a significant impact
on the aggregate benefits estimates.
As described in Chapter 7 and Appendix I, pre-
mature mortality is both the largest source of benefits
and the major source of quantified uncertainty in the
J-l
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
retrospective analysis. In addition to the quantified
uncertainty, there is considerable additional
unquantified uncertainty about premature mortality
associated with air pollution. Much of the informa-
tion base about these relationships is relatively new,
more is coming out virtually daily, and there is sub-
stantial disagreement in the scientific community
about many of the key issues. EPA's Research Strat-
egy and Research Needs document for particulate
matter, currently under development, will address
many of these scientific issues as they relate to PM.
The following selection of highly uncertain issues
could have a significant impact on both the aggregate
mortality benefits estimates and the measured uncer-
tainty range:
the relationship of specific pollutants in the
overall premature mortality effect, including
the individual or interactive relationships be-
tween specific chemicals (e.g., ozone, sul-
fates, nitrates, and acid aerosols), and particle
sizes (i.e., coarse, fine and ultra-fine particles);
the degree of overlap (if any) between the
measured relationships between effects asso-
ciated with short term exposures and effects
from long term exposure;
the confounding effect of changes in historic
air pollution, including changes over time in
both pollution levels and the composition of
the pollutant mix;
the extent to which life spans are shortened
by exposure to the pollutants, and the distri-
bution of ages at the time of death;
the willingness to pay to avoid the risks of
shortened life spans; and
the extent to which total PM2 exposure in-
crementally augments the variability of out-
door PM25 and increases the dose that would
cause excess morbidity or mortality.
After premature mortality, chronic bronchitis is
the next largest health effect benefit category included
in the retrospective analysis. There is considerable
measured uncertainty about both the incidence esti-
mation and the economic valuation. Additional re-
search could reduce uncertainties about the level of
the pollutants associated with the observed effects,
the baseline incidence used to model the changes in
the number of new cases, and the correspondence be-
tween the definition of chronic bronchitis used in the
health effects studies and the economic valuation stud-
ies.
Another area of potentially useful research would
be further examination of the effects of criteria pol-
lutants on cardiovascular disease incidence and mor-
tality. Considering available epidemiological evidence
and the potential economic cost of cardiovascular dis-
ease, the value of avoiding these outcomes may sig-
nificantly influence the overall benefit estimates gen-
erated in future assessments.
Further research on the willingness to pay to avoid
the risk of hospital admissions for specific conditions
would reduce a potentially significant source of non-
measured uncertainty. The Project Team used
"avoided costs" for the value of an avoided hospital
admission, based on the avoided direct medical cost
of hospitalization (including lost wages for the em-
ployed portion of the hospitalized population).
Avoided costs are likely to be a substantial underesti-
mate of the appropriate willingness to pay, especially
for such serious health effects as hospitalization for
strokes and congestive heart failure, particularly be-
cause they omit the value of avoided pain, suffering,
and inconvenience. Furthermore, in addition to hos-
pitalization, there is evidence that some people seek
medical assistance as outpatients. It is also likely that
there are additional people adversely affected by short-
term air pollution levels who seek physician services
(but stop short of hospital admissions). Revised esti-
mates of the appropriate economic value of avoided
hospitalization and other primary care medical ser-
vices could increase the total economic benefits of
this cluster of health effects sufficiently that it could
be a much larger portion of the aggregate benefit to-
tal.
Finally, one of the challenges in preparing the
retrospective analysis was modeling the integrated
relationships between emissions of many different
chemicals, the subsequent mixture of pollutants in the
ambient air, and the resulting health and welfare ef-
fects of simultaneous exposure to multiple pollutants.
One element of the uncertainty in the analysis derives
from the limited current understanding of any inter-
active (synergistic or antagonistic) effects of multiple
pollutants. The need to better understand these com-
plex issues is not a limited scientific question to im-
prove section 812 analyses, but is the primary focus
of EPA's current activities, organized under the Fed-
J-2
-------
Appendix J: Future Directions
eral Advisory Council Act (FACA) process, to de-
velop an integrated set of attainment policies dealing
with ozone, particulate matter, sulfur and nitrogen
oxides, and visibility. Further research on multi-pol-
lutant issues may both (a) reduce a source of unmea-
sured uncertainty in the section 812 analyses and (b)
allow for effective apportionment of endpoint reduc-
tion benefits to specific pollutants or pollutant mixes.
Research Topics to Improve
Comprehensiveness
Even though research efforts falling in this cat-
egory may not result in significant changes in net mon-
etary benefit estimates, one of the goals of the section
812 studies is to provide comprehensive information
about Clean Air Act programs. For example, programs
to control hazardous air pollutants (HAPs) tend to
impose costs and yield benefits which are relatively
small compared to programs of pervasive national
applicability such as those aimed at meeting National
Ambient Air Quality Standards. Nevertheless, there
are significant social, political, financial, individual
human health, and specific ecosystem effects associ-
ated with emissions of HAPs and the programs imple-
mented to control them. Under these circumstances,
continued efforts to understand these consequences
and evaluate their significance in relation to other pro-
grammatic and research investment opportunities
might be considered reasonable, particularly in the
context of comprehensive program assessments such
as the present study.
Some cost and benefit effects could not be fully
assessed and incorporated in the net monetary benefit
estimate developed for the present study for a variety
of reasons. Various effects were excluded due to (a)
inadequate historical data (e.g., lack of data on his-
torical ambient concentrations of HAPs), (b) inad-
equate scientific knowledge (e.g., lack of concentra-
tion-response information for ecological effects of
criteria and hazardous air pollutants), or (c) resource-
intensity or limited availability of analytical tools
needed to assess specific endpoints (e.g., indirect ef-
fects resulting from deposition and subsequent expo-
sure to HAPs). Other specific examples of presently
omitted or underrepresented effect categories include
health effects of hazardous air pollutants, ecosystem
effects, any long-term impact of displaced capital on
productivity slowdown, and redirected technological
innovation.
Although the primary focus of 1970 to 1990 CAA
programs was reduction of criteria pollutants to
achieve attainment of national ambient air quality stan-
dards, emissions of air toxics were also substantially
reduced. Some air toxics were deliberately controlled
because of their known or suspected carcinogenicity,
while other toxic emissions were reduced indirectly
due to control procedures aimed at other pollutants,
particularly ozone and particulate matter. The current
analysis was able to present only limited information
on the effects of changes in air toxic emissions. These
knowledge gaps may be more serious for future sec-
tion 812 analyses, however, since the upcoming pro-
spective study will include evaluation of the effects
of an expanded air toxic program under the CAA Title
III. Existing knowledge gaps that prevented a more
complete consideration of toxics in the present study
include (a) methods to estimate changes in acute and
chronic ambient exposure conditions nationwide, (b)
concentration-response relationships linking exposure
and health or ecological outcomes, (c) economic valu-
ation methods for a broad array of potential serious
health effects such as renal damage, reproductive ef-
fects and non-fatal cancers, and (d) potential ecologi-
cal effects of air toxics.
In addition to research to improve the understand-
ing of the consequences of changes in air pollution on
human health and well-being, further research on non-
health effects could further improve the comprehen-
siveness of future assessments. Improvements in air
quality have likely resulted in improvements in the
health of aquatic and terrestrial ecosystems and the
myriad of ecological services they provide, but knowl-
edge gaps prevented them from being included in the
current analysis. Additional research in both scien-
tific understanding and appropriate modeling proce-
dures could facilitate inclusion of additional benefits
such as improvements in water quality stemming from
a reduction in acid deposition-related air pollutants.
Water quality improvements would benefit human
welfare through enhancements in certain consump-
tive services such as commercial and recreational fish-
ing, in addition to non-consumptive services such as
wildlife viewing, maintenance of biodiversity, and
nutrient cycling. Similarly, increased growth, produc-
tivity and overall health of U.S. forests could occur
from reducing ozone, resulting in benefits from in-
creased timber production, greater opportunities for
recreational services such as hunting, camping, wild-
life observation, and nonuse benefits such as nutrient
cycling, temporary C02 sequestration, and existence
J-3
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
value. Finally, additional research using a watershed
approach to examine the potential for ecological ser-
vice benefits which emerge only at the watershed scale
might be useful and appropriate given the broad geo-
graphic scale of the section 812 assessments.
While there are insufficient data and/or analyti-
cal resources to adequately model the short-run eco-
logical and ecosystem effects of air pollution reduc-
tion, even less is known about the long-run effects of
prolonged exposure. Permanent species displacement
or altered forest composition are examples of poten-
tial ecosystem effects that are not reflected in the cur-
rent monetized benefit analysis, and could be a source
of additional benefits. In addition to these ecological
research needs, an equally large, or larger, gap in the
benefit-cost analysis is the lack of adequate tools to
monetize the benefits of such ecosystem services.
Future Section 812 Analyses
This retrospective study of the benefits and costs
of the Clean Air Act was developed pursuant to sec-
tion 812 ofthe 1990 Clean Air Act Amendments. Sec-
tion 812 also requires EPA to generate an ongoing
series of prospective studies of the benefits and costs
of the Act, to be delivered as Reports to Congress every
two years.
Design of the first section 812 prospective study
commenced in 1993. The EPA Project Team devel-
oped a list of key analytical design issues and a
"strawman" analytical design reflecting notional de-
cisions with respect to each of these design issues.1
The analytical issues list and strawman design were
presented to the Science Advisory Board Advisory
Council on Clean Air Compliance Analysis (Coun-
cil), the same SAB review group which has provided
review of the retrospective study. Subsequently, the
EPA Project Team developed a preliminary design
for the first prospective study. Due to resource limita-
tions, however, full-scale efforts to implement the first
prospective study did not begin until 1995 when ex-
penditures for retrospective study work began to de-
cline as major components of that study were com-
pleted.
As for the retrospective, the first prospective study
is designed to contrast two alternative scenarios; how-
ever, in the prospective study the comparison will be
between a scenario which reflects full implementa-
tion of the CAAA90 and a scenario which reflects
continued implementation only of those air pollution
control programs and standards which were in place
as of passage of the CAAA90. This means that the
first prospective study will provide an estimate of the
incremental benefits and costs of the CAAA90.
The first prospective study is being implemented
in two phases. The first phase involves development
of a screening study, and the second phase will in-
volve a more detailed and refined analysis which will
culminate in the first prospective study Report to Con-
gress. The screening study compiles currently avail-
able information on the costs and benefits of the imple-
mentation of CAAA90 programs, and is intended to
assist the Project Team in the design of the more de-
tailed analysis by providing insights regarding the
quality of available data sources and analytical mod-
els, and the relative importance of specific program
areas; emitting sectors; pollutants; health, welfare, and
ecological endpoints; and other important factors and
variables.
In developing and implementing the retrospective
study, the Project Team developed a number of im-
portant modeling systems, analytical resources, and
techniques which will be directly applicable and use-
ful for the ongoing series of section 812 Prospective
Studies. Principal among these are the Criteria Air
Pollutant Modeling System (CAPMS) model devel-
oped to translate air quality profile data into quantita-
tive measures of physical outcomes; and the economic
valuation models, coefficients, and approaches devel-
oped to translate those physical outcomes to economic
terms.
The Project Team also learned valuable lessons
regarding analytical approaches or methods which
were not as productive or useful. In particular, the
Project Team plans not to perform macroeconomic
modeling as an integral part of the first prospective
analysis. In fact, there are currently no plans to con-
duct a macroeconomic analysis at all. Essentially, the
Project Team concluded, with confirmation by the
SAB Council, that the substantial investment of time
and resources necessary to perform macroeconomic
modeling would be better invested in developing high
quality data on the likely effects of the CAA on key
emitting sectors, such as utilities, on-highway vehicles,
refineries, etc. While the intended products of a mac-
1 Copies of the prospective study planning briefing materials are available from EPA.
J-4
-------
Appendix J: Future Directions
roeconomic modeling exercise - such as overall ef-
fects on productivity, aggregate employment effects,
indirect economic effects- are of theoretical interest,
the practical results of such exercises in the context
of evaluating environmental programs may be disap-
pointing for several reasons.
First, the CAA has certainly had a significant ef-
fect on several industrial sectors. However, the coarse
structure of a model geared toward simulating effects
across the entire economy requires crude and poten-
tially inaccurate matching of these polluting sectors
to macroeconomic model sectors. For example, the J/
W model used for the retrospective study has only 35
sectors, with electric utilities comprising a single sec-
tor. In reality, a well-structured analysis of the broader
economic effects of the CAA would provide for sepa-
rate and distinct treatment of coal-fired utility plants,
oil-fired plants, and so on. Furthermore, the outputs
of the macroeconomic model are too aggregated to
provide useful and accurate input information for the
sector-specific emission models used to project the
emissions consequences of CAA programs. Again, the
critical flaw is the inability to project important de-
tails about differential effects on utilities burning al-
ternative fuels.
The second critical problem with organizing a
comprehensive analysis of the CAA around a macro-
economic modeling approach is that the effect infor-
mation produced by the macroeconomic model is rela-
tively unimportant with respect to answering the fun-
damental, target variable: "How do the overall health,
welfare, ecological, and economic benefits of Clean
Air Act programs compare to the costs of these pro-
grams? " The Project Team believes that any adverse
effect, no matter how small in a global context, should
not be deemed "insignificant" if even one individual
is seriously harmed. However, the retrospective study
results themselves have shown that, when analytical
resources are limited, the resources invested in the
macroeconomic modeling would have been better
spent to provide a more complete and less uncertain
assessment of the benefit side of the equation. Even
on the cost side of the equation, it is far more impor-
tant to invest in developing accurate and reliable esti-
mates of sector-specific compliance strategies and the
direct cost implications of those strategies. This will
be even more true in the prospective study context
when the Project Team will be faced with forecasting
compliance strategies and costs rather than simply
compiling survey data on actual, historical compli-
ance expenditures.
The third and most important limitation of mac-
roeconomic modeling analysis of environmental pro-
grams is that, unlike the economic costs of protection
programs, the economic benefits are not allowed to
propagate through the economy. For example, while
productivity losses associated with reduced capital
investment due to environmental regulation are
counted, the productivity gains resulting from reduced
pollution-related illness and absenteeism of workers
are not counted. The resulting imbalance in the treat-
ment of regulatory consequences raises serious con-
cerns about the value of the macroeconomic model-
ing evaluation of environmental programs. In the fu-
ture, macroeconomic models which address this and
other concerns may be developed; however, until such
time EPA is likely to have limited confidence in the
value of macroeconomic analysis of even broad-scale
environmental protection programs.
Based on these findings and other factors, the de-
sign of the first prospective study differs in important
ways from the retrospective study design. First, rather
than relying on broad-scale, hypothetical, macroeco-
nomic model-based scenario development and analy-
sis, the first prospective study will make greater use
of existing information from EPA and other analyses
which assess compliance strategies and costs, and the
emission and air quality effects of those strategies.
After developing as comprehensive a data set as pos-
sible of regulatory requirements, compliance strate-
gies, compliance costs, and emissions consequences,
the data set will be reviewed, refined, and extended
as feasible and appropriate. In particular, a number of
in-depth sector studies will be conducted to develop
up-to-date, detailed projections of the effects of new
CAA requirements on key emitting sectors. Candi-
date sectors for in-depth review include, among oth-
ers, utilities, refineries, and on-highway vehicles.
The first prospective study will also differ from
the retrospective study in that analytical resources will
be directed toward development of a more complete
assessment of benefits. Efforts will be made to ad-
dress the deficiencies which prevailed in the retro-
spective study relating to assessment of the benefits
of air toxics control. In addition, the Project Team
will endeavor to provide a more complete and effec-
tive assessment of the ecological effects of air pollu-
tion control.
J-5
-------
The Benefits and Costs of the Clean Air Act, 1970 to 1990
J-6
------- |