jfll
Agency
Regulatory Impact Analysis for the
Final Clean Air Interstate Rule
-------
EPA-452/R-05-002
March 2005
Regulatory Impact Analysis for the Final
Clean Air Interstate Rule
U.S. Environmental Protection Agency
Office of Air and Radiation
Air Quality Strategies and Standards Division,
Emission, Monitoring, and Analysis Division
and
Clean Air Markets Division
-------
CONTENTS
Section Page
1. EXECUTIVE SUMMARY 1-1
1.1 Results 1-1
1.1.1 Health Benefits 1-3
1.1.2 Welfare Benefits 1-3
1.1.3 Uncertainty in the Benefits Estimates 1-6
1.2 Not All Benefits Quantified 1-9
1.3 Costs and Economic Impacts 1-9
1.4 Limitations 1-12
1.5 References 1-13
2. INTRODUCTION AND BACKGROUND 2-1
2.1 Introduction 2-1
2.2 Background 2-1
2.3 Regulated Entities 2-2
2.4 Baseline and Years of Analysis 2-3
2.5 Control Scenario 2-3
2.6 Benefits of Emission Controls 2-4
2.7 Cost of Emission Controls 2-4
2.8 Organization of the Regulatory Impact Analysis 2-4
3. EMISSIONS AND AIR QUALITY IMPACTS 3-1
3.1 Emissions Inventories and Estimated Emissions Reductions 3-1
iii
-------
3.2 Air Quality Impacts 3-4
3.2.1 PM Air Quality Estimates 3-5
3.2.1.1 Modeling Domain 3-6
3.2.1.2 Simulation Periods 3-6
3.2.1.3 Model Inputs 3-7
3.2.1.4 CMAQ Model Evaluation 3-8
3.2.1.5 Converting CMAQ Outputs to Benefits Inputs .... 3-12
3.2.1.6 PM Air Quality Results 3-14
3.2.2 Ozone Air Quality Estimates 3-15
3.2.2.1 Modeling Domain 3-17
3.2.2.2 Simulation Periods 3-17
3.2.2.3 Nonemissions Modeling Inputs 3-17
3.2.2.4 Model Performance for Photochemical Ozone 3-18
3.2.2.5 Converting CAMx Outputs to Full-Season
Profiles for Benefits Analysis 3-20
3.2.2.6 Ozone Air Quality Results 3-21
3.2.3 Visibility Degradation Estimates 3-21
3.3 References 3-25
4. BENEFITS ANALYSIS AND RESULTS 4-1
4.1 Benefit Analysis—Data and Methods 4-12
4.1.1 Valuation Concepts 4-14
4.1.2 Growth in WTP Reflecting National Income Growth
Over Time 4-16
4.1.3 Methods for Describing Uncertainty 4-19
4.1.4 Demographic Projections 4-24
4.1.5 Health Benefits Assessment Methods 4-25
4.1.5.1 Selecting Health Endpoints and Epidemiological
Effect Estimates 4-26
4.1.5.2 Uncertainties Associated with Health Impact
Functions 4-40
4.1.5.3 Baseline Health Effect Incidence Rates 4-46
4.1.5.4 Selecting Unit Values for Monetizing Health
Endpoints 4-50
4.1.6 Human Welfare Impact Assessment 4-65
4.1.6.1 Visibility Benefits 4-65
4.1.6.2 Agricultural, Forestry and other Vegetation-
Related Benefits 4-70
4.1.6.3 Benefits from Reductions in Materials Damage .... 4-72
4.1.6.4 Benefits from Reduced Ecosystem Damage 4-73
4.2 Benefits Analysis—Results 4-73
4.2.1 Potential Benefits of the New Jersey and Delaware Proposal . 4-77
IV
-------
4.3 Probabilistic Analysis of Uncertainty in the Benefits Estimates 4-77
4.4 Discussion 4-84
4.5 References 4-85
5. QUALITATIVE ASSESSMENT OF NONMONETIZED BENEFITS 5-1
5.1 Introduction 5-1
5.2 Atmospheric Deposition of Sulfur and Nitrogen—Quantification
of Impacts for the Rule 5-1
5.3 Atmospheric Deposition of Sulfur and Nitrogen—Impacts on
Aquatic, Forest, and Coastal Ecosystems 5-3
5.3.1 Freshwater Acidification 5-4
5.3.1.1 Water/Watershed Modeling 5-6
5.3.1.2 Description of the MAGIC Model and Methods ....5-7
5.3.1.3 Model Structure 5-7
5.3.1.4 Model Implementation 5-8
5.3.1.5 Calibration Procedure 5-9
5.3.1.6 MAGIC Modeling Results 5-11
5.3.1.7 Study of B enefits of Natural Resource
Improvements in the Adirondacks 5-12
5.3.2 Forest Ecosystems 5-13
5.3.3 Coastal Ecosystems 5-15
5.3.4 Potential Other Impacts 5-17
5.4 References 5-17
6. ELECTRIC POWER SECTOR PROFILE 6-1
6.1 Power-Sector Overview 6-1
6.1.1 Generation 6-1
6.1.2 Transmission 6-3
6.1.3 Distribution 6-3
6.2 Deregulation and Restructuring 6-4
6.3 Pollution and EPA Regulation of Emissions 6-5
6.4 Pollution Control Technologies 6-6
6.5 Regulation of the Power Sector 6-7
6.6 Cap and Trade 6-9
-------
6.7 Clean Air Interstate Rule 6-9
6.8 Price Elasticity of Electricity 6-11
7. COST, ECONOMIC, AND ENERGY IMPACTS 7-1
7.1 Modeling Background 7-1
7.2 Projected SO2 and NOX Emissions and Reductions 7-4
7.3 Projected Costs 7-5
7.4 Projected Control Technology Retrofits 7-5
7.5 Projected Generation Mix 7-9
7.6 Projected Capacity Additions 7-11
7.7 Projected Coal Production for the Electric Power Sector 7-11
7.8 Projected Retail Electricity Prices 7-11
7.9 Projected Fuel Price Impacts 7-15
7.10 Key Differences in EPA Model Runs for Final CAIR Modeling .... 7-15
7.11 Projected Primary PM Emissions from Power Plants 7-16
7.12 Limitations of Analysis 7-17
7.13 Significant Energy Impact 7-21
7.14 Industry-Sector Impacts 7-21
7.15 References 7-22
8. STATUTORY AND EXECUTIVE ORDER IMPACT ANALYSES 8-1
8.1 Small Entity Impacts 8-1
8.1.1 Identification of Small Entities 8-3
8.1.2 Overview of Analysis and Results 8-4
8.1.2.1 Methodology for Estimating Impacts of CAIR
on Small Entities 8-5
8.1.2.2 Results 8-7
vi
-------
8.1.3 Summary of Small Entity Impacts 8-10
8.2 Unfunded Mandates Reform Act (UMRA) Analysis 8-11
8.2.1 Identification of Government-Owned Entities 8-13
8.2.2 Overview of Analysis and Results 8-14
8.2.2.1 Methodology for Estimating Impacts of CAIR on
Government Entities 8-14
8.2.2.2 Results 8-16
8.2.3 Summary of Government Entity Impacts 8-18
8.3 Paperwork Reduction Act 8-20
8.4 Children's Health 8-21
8.5 Tribal Impacts 8-21
8.6 Environmental Justice 8-22
8.7 Reference 8-23
9. COMPARISON OF BENEFITS AND COSTS 9-1
9.1 References 9-3
Appendix A: Benefits and Costs of the Clean Air Interstate Rule, the Clean Air Visibility
Rule, and the Clean Air Interstate Rule plus the Clean Air Visibility Rule . . A-l
Appendix B: Supplemental Analyses Addressing Uncertainties in the Benefits Analyses . . B-l
Appendix C: Sensitivity Analyses of the Key Parameters in the Benefits Analysis C-l
Appendix D: Sensitivity Analyses of Key Parameters in the Cost and Economic Impact
Analysis and a Listing of IPM Runs in Support of CAIR D-l
Appendix E: CAIR Industry-Sector Impacts E-l
Appendix F: Additional Technical Information Supporting the Benefits Analysis F-l
Appendix G: Health-Based Cost-Effectiveness of Reductions in Ambient PM2 5
Associated with CAIR G-l
vn
-------
LIST OF FIGURES
Number
1-1 Comparative Assessment of Relative Range of Uncertainty in Estimated
Avoided Incidence of Premature Mortality Using Classical Statistical Error
and the Pilot Expert-Based Characterizations of Uncertainty 1-7
2-1 Final CAIR Region 2-2
3-1 CMAQ Modeling Domain 3-7
3-2 CAMx Eastern U.S. Modeling Domain 3-18
4-1 Key Steps in Air Quality Modeling Based Benefits Analysis 4-7
4-2 CAIR Final Rule Visibility Improvements in Class I Areas in the
Southeast 4-70
4-3 Results of Illustrative Application of Pilot Expert Elicitation: Annual
Reductions in Premature Mortality in 2015 Associated with the Clean
Air Interstate Rule 4-81
4-4 Results of Illustrative Application of Pilot Expert Elicitation: Dollar
Value of Annual Reductions in Premature Mortality in 2015 Associated
with the Clean Air Interstate Rule 4-83
5-1 Percentage Reduction of All Forms of Sulfur Deposition for the Years
2010 and 2015 5-2
5-2 Percentage Reduction of All Forms of Nitrogen Deposition for the Years
2010 and 2015 5-2
5-3 CAIR Nitrogen and Sulfur Deposition Reduction in the Adirondacks,
New England, and the Blue Ridge 5-3
5-4 CAIR Nitrogen Deposition Reductions in Hydrologic Regions 5-16
6-1 Status of State Electricity Industry Restructuring Activities
(as of February 2003) 6-5
6-2 Emissions of SO2 and NOX from the Power Sector (2003) 6-6
7-1 CAIR Modeled Region 7-1
7-2 SO2 Emissions from the Power Sector in 2010 and 2015 With and
Without CAIR 7-6
7-3 NOX Emissions from the Power Sector in 2010 and 2015 With and
Without CAIR 7-7
Vlll
-------
7-4 Ozone Season NOX Emissions from the Power Sector in 2010 and 2015
With and Without CAIR 7-8
7-5 Generation Mix with and without CAIR 7-10
7-6 Current Coal Production Levels and Projected Production with CAIR 7-12
7-7 Regional Electricity Prices with and without CAIR 7-13
7-8 NERC Power Regions 7-14
7-9 Final CAIR Region 7-16
IX
-------
LIST OF TABLES
Number
1-1 Summary of Annual Benefits, Costs, and Net Benefits of the Clean Air
Interstate Rule (billions of 1999$) 1-2
1-2 Clean Air Interstate Rule: Estimated Reduction in Incidence of Adverse
Health Effects 1-4
1-3 Estimated Monetary Value of Reductions in Incidence of Health and
Welfare Effects (in millions of 1999$) 1-5
1-4 Unquantified and Nonmonetized Benefits of the Clean Air Interstate Rule .. 1-10
1-5 Unquantified Costs of the Clean Air Interstate Rule 1-12
3-1 Emissions Sources and Basis for Current and Future-Year Inventories 3-2
3-2 Summary of Modeled Baseline Emissions for Lower 48 States 3-3
3-3 Summary of Modeled Emissions Changes for the Clean Air Interstate
Rule: 2010 and 2015 3-4
3-4 Model Performance Statistics for CAIR CMAQ 2001 3-11
3-5 Selected Performance Evaluation Statistics from the CMAQ 2001
Simulation 3-12
3-6 Summary of Base Case PM Air Quality and Changes Due to Clean
Air Interstate Rule: 2010 and 2015 3-15
3-7 Distribution of PM2 5 Air Quality Improvements Over Population Due
to Clean Air Interstate Rule: 2010 and 2015 3-16
3-8 Model Performance Statistics for Hourly Ozone in the Eastern U.S.
CAMx Ozone Simulations: 1995 Base Case 3-19
3-9 Summary of CAMx Derived Population-Weighted Ozone Air Quality
Metrics for Health Benefits Endpoints Due to Clean Air Interstate Rule:
Eastern U.S 3-22
3-10 Summary of Deciview Visibility Impacts at Class I Areas in the
CAIR Region 3-23
4-1 Estimated Monetized Benefits of the Final CAIR 4-2
4-2 Human Health and Welfare Effects of Pollutants Affected by the
Final CAIR 4-3
4-3 Elasticity Values Used to Account for Projected Real Income Growth 4-18
4-4 Adjustment Factors Used to Account for Projected Real Income Growth ... 4-19
4-5 Primary Sources of Uncertainty in the Benefits Analysis 4-21
-------
4-6 Summary of Considerations Used in Selecting C-R Functions 4-27
4-7 Endpoints and Studies Used to Calculate Total Monetized Health Benefits . . 4-30
4-8 Studies Examining Health Impacts in the Asthmatic Population Evaluated
for Use in the Benefits Analysis 4-41
4-9 Baseline Incidence Rates and Population Prevalence Rates for Use in
Impact Functions, General Population 4-48
4-10 Asthma Prevalence Rates Used to Estimate Asthmatic Populations in
Impact Functions 4-50
4-11 Unit Values Used for Economic Valuation of Health Endpoints (1999$) . . . 4-52
4-12 Expected Impact on Estimated Benefits of Premature Mortality Reductions
of Differences Between Factors Used in Developing Applied VSL and
Theoretically Appropriate VSL 4-57
4-13 Alternative Direct Medical Cost of Illness Estimates for Nonfatal
Heart Attacks 4-62
4-14 Estimated Costs Over a 5-Year Period (in 2000$) of a Nonfatal Myocardial
Infarction 4-62
4-15 Women with Children: Number and Percentage in the Labor Force,
2000, and Weighted Average Participation Rate 4-64
4-16 Clean Air Interstate Rule: Estimated Reduction in Incidence of
Adverse Health Effects 4-74
4-17 Estimated Monetary Value in Reductions in Incidence of Health and
Welfare Effects (in millions of 1999$) 4-75
5-1 Acidification Changes in Water Bodies as a Result of CAIR 5-12
6-1 Existing Electricity Generating Capacity by Energy Source, 2002 6-1
6-2 Total U.S. Electric Power Industry Retail Sales in 2003 (Billion kWh) 6-2
6-3 Electricity Net Generation in 2003 (Billion kWh) 6-2
6-4 Emissions of SO2 and NOX in 2003 and Percentage of Emissions in the
CAIR Affected Region (tons) 6-10
6-5 Current Electricity Net Generation and EPA Projections for 2010 and
2015 (Billion kWh) 6-10
7-1 CAIR Annual Emissions Caps (Million Tons) 7-2
7-2 Projected Emissions of SO2 and NOX with the Base Case (No Further
Controls) and with CAIR (Million Tons) 7-5
7-3 Annualized Regional Cost of CAIR and Marginal Cost of SO2 and NOX
Reductions with CAIR ($1999) 7-9
7-4 Pollution Controls by Technology with the Base Case (No Further Controls)
and with CAIR (GW) 7-9
7-5 Generation Mix with the Base Case (No Further Controls) and with CAIR
(Thousand GWhs) 7-10
7-6 Total Coal and Natural Oil/Gas-Fired Capacity by 2020 (GW) 7-11
XI
-------
7-7 Coal Production for the Electric Power Sector with the Base Case (No
Further Controls) and with CAIR (Million Tons) 7-12
7-8 Projected Regional Retail Electricity Prices with the Base Case (No
Further Controls) and with CAIR (Mills/kWh) 7-13
7-9 Retail Electricity Prices by NERC Region with the Base Case (No
Further Controls) and with CAIR (Mills/kWh) 7-14
7-10 Henry Hub Natural Gas Prices and Average Minemouth Coal Prices with the
Base Case (No Further Controls) and with CAIR ($1999) 7-15
8-1 Potentially Regulated Categories and Entities 8-2
8-2 Projected Impact of CAIR on Small Entities 8-4
8-3 Summary of Distribution of Economic Impacts of CAIR on Small Entities . . . 8-9
8-4 Incremental Annualized Costs under CAIR Summarized by Ownership
Group and Cost Category ($1999 millions) 8-10
8-5 Summary of Potential Impacts on Government Entities under CAIR 8-13
8-6 Distribution of Economic Impacts on Government Entities under CAIR .... 8-18
8-7 Incremental Annualized Costs under CAIR Summarized by Ownership
Group and Cost Category ($1,000,000) 8-19
9-1 Summary of Annual Benefits, Costs, and Net Benefits of the Clean
Air Interstate Rule (billions of 1999 dollars) 9-2
xn
-------
CHAPTER 1
EXECUTIVE SUMMARY
This Regulatory Impact Analysis (RIA) presents the health and welfare benefits and
the costs of the Clean Air Interstate Rule (CAIR) and compares the benefits to the costs of
implementing CAIR in 2010 and 2015.
1.1 Results
Synopsis
EPA has estimated the benefits and costs of the Clean Air Interstate Rule and finds
that the rule results in estimated annual net benefits of $71.4 or $60.4 in 2010 and $98.5 or
$83.2 billion in 2015. These alternate net benefit estimates reflect differing assumptions
about the social discount rate used to estimate the social benefits and costs of the rule. The
lower estimates reflect a discount rate of 7 percent and the higher estimates a discount rate
of 3 percent. In 2015, the total annual quantified benefits are $101 or $86.3 billion and the
annual social costs are $2.6 or $3.1 billion—benefits outweigh social costs in 2015 by a
ratio of 39 to 1 or 28 to 1 (3 percent and 7 percent discount rates respectively). An
alternative comparison of the annual benefits of the rule to the estimated private costs to the
electric generating industry in 2015 result in benefits outweighing costs by a ratio of 25 to 1
(benefits of $101 billion compared to costs of $3.6 billion). These estimates do not include
the value of benefits or costs that we cannot monetize. Upon consideration of the
uncertainties and limitations in the analysis, it remains clear that the benefits of CAIR are
substantial and far outweigh the costs.
A comparison of the benefits and costs of the rule in 2010 and 2015 is shown in
Table 1-1. The benefits and costs reported for CAIR in Table 1-1 represent estimates for a
complete CAIR program that includes the CAIR promulgated rule and the concurrent
proposal to include annual sulfur dioxide (SO2) and nitrogen oxide (NOX) controls for New
Jersey and Delaware. The modeling used to provide these estimates also assumes annual
SO2 and NOX controls for Arkansas that are not a part of the complete CAIR program
resulting in a slight overstatement of the reported benefits and costs.
1-1
-------
Table 1-1. Summary of Annual Benefits, Costs, and Net Benefits of the Clean Air
Interstate Rule3 (billions of 1999$)
Description
Social costsb
3 percent discount rate
7 percent discount rate
Social benefitsc'd'e
3 percent discount rate
7 percent discount rate
Health-related benefits:
3 percent discount rate
7 percent discount rate
Visibility benefits
Net benefits (benefits-costs)e>f
3 percent discount rate
7 percent discount rate
2010
$1.91
$2.14
73.3 +B
62.6 + B
72.1
61.4
1.14
$71.4 + B
$60.4 + B
2015
$2.56
$3.07
101+B
86.3 +B
99.3
84.5
1.78
$98.5 +B
$83.2 + B
a All estimates are rounded to three significant digits and represent annualized benefits and costs anticipated
for the year 2010 and 2015. Estimates relate to the complete CAIR program including the CAIR
promulgated rule and the proposal to include SO2 and annual NOX controls for New Jersey and Delaware.
Modeling used to develop these estimates assumes annual SO2 and NOX controls for Arkansas resulting in a
slight overstatement of the reported benefits and costs for the complete CAIR program.
b Note that costs are the annualized total costs of reducing pollutants including NOX and SO2 for the EGU
source category in the CAIR region.
0 As this table indicates, total benefits are driven primarily by PM-related health benefits. The reduction in
premature fatalities each year accounts for over 90 percent of total monetized benefits in 2015. Benefit
estimates in this table are nationwide (with the exception of ozone and visibility) and reflect NOX and SO2
reductions. The analysis assumes that States will choose to achieve CAIR caps solely from the EGU source
category. Ozone benefits represent benefits in the eastern United States. Visibility benefits represent
benefits in Class I areas in the southeastern United States.
d Not all possible benefits or disbenefits are quantified and monetized in this analysis. B is the sum of all
unqualified benefits and disbenefits. Potential benefit categories that have not been quantified and
monetized are listed in Table 1-4.
e Valuation assumes discounting over the SAB-recommended 20-year segmented lag structure described in
Chapter 4. Results reflect 3 percent and 7 percent discount rates consistent with EPA and OMB guidelines
for preparing economic analyses (U.S. EPA, 2000; OMB, 2003).
f Net benefits are rounded to the nearest $100 million. Columnar totals may not sum due to rounding.
1-2
-------
Details of the important analysis assumptions, including entities regulated, baseline,
analysis year, control scenario, and other relevant analysis assumptions are discussed in
Chapter 2 of this report.
1.1.1 Health Benefits
CAIR is expected to yield significant health benefits by reducing emissions of two
key contributors to fine particle and ozone formation. Sulfur dioxide contributes to the
formation of fine particle pollution (PM25), and nitrogen oxide contributes to the formation
of both PM25 and ground-level ozone.1
Our analyses suggest CAIR would yield benefits in 2015 of $101 billion (based on a
3 percent discount rate) and $86.3 billion (based on a 7 percent discount rate) that includes
the value of avoiding approximately 17,000 premature deaths, 22,000 nonfatal heart attacks,
12,300 hospitalizations for respiratory and cardiovascular diseases, 1.7 million lost work
days, 500,000 school absences, and 10.6 million days when adults restrict normal activities
because of respiratory symptoms exacerbated by PM25 and ozone pollution.2
We also estimate substantial additional health improvements for children from
reductions in upper and lower respiratory illnesses, acute bronchitis, and asthma attacks. See
Table 1-2 for a list of the annual reduction in health effects expected in 2010 and 2015 and
Table 1-3 for the estimated value of those reductions.
1.1.2 Welfare Benefits
The term welfare benefits covers both environmental and societal benefits of reducing
pollution, such as reductions in damage to ecosystems, improved visibility and
improvements in recreational and commercial fishing, agricultural yields, and forest
1 Although well over 90 percent of the expected benefits of this rule are derived from reductions in SO2 and
NOX, a small portion of EPA's projected benefits are a result of reductions in primary PM from power
plants. Although this reduction is not required by the rule, it is a potential ancillary benefit of installing
certain SO2 control technologies.
2 These estimates account for growth in the public's willingness to pay for reductions in health and
environmental risks and account for growth in real gross domestic product per capita between the present
and 2015. Benefit estimates reflect the use of 3 percent and 7 percent discount rates consistent with the U.S.
Environmental Protection Agency (EPA) and the Office of Management and Budget (OMB) guidelines for
preparing economic analyses (U.S. EPA, 2000; OMB, 2003).
1-3
-------
Table 1-2. Clean Air Interstate Rule: Estimated Reduction in Incidence of Adverse
Health Effects3
Health Effect
2010
Incidence
2015
Reduction
PM-Related Endpoints:
Premature mortality15'0
Adult, age 30 and over
Infant, age <1 year
Chronic bronchitis (adult, age 26 and over)
Non-fatal myocardial infarction (adults, age 18 and older)
Hospital admissions — respiratory (all ages)d
Hospital admissions — cardiovascular (adults, age >18)e
Emergency room visits for asthma (age 18 years and younger)
Acute bronchitis (children, age 8-12)
Lower respiratory symptoms (children, age 7-14)
Upper respiratory symptoms (asthmatic children, age 9-18)
Asthma exacerbation (asthmatic children, age 6-18)
Work loss days (adults, age 18-65)
Minor restricted-activity days (MRADs) (adults, age 18-65)
13,000
29
6,900
17,000
4,300
3,800
10,000
16,000
190,000
150,000
240,000
1,400,000
8,100,000
17,000
36
8,700
22,000
5,500
5,000
13,000
19,000
230,000
180,000
290,000
1,700,000
9,900,000
Ozone-Related Endpoints
Hospital admissions — respiratory causes (adult, 65 and older/
Hospital admissions — respiratory causes (children, under 2)
Emergency room visit for asthma (all ages)
Minor restricted-activity days (MRADs) (adults, age 18-65)
School absence days
610
380
100
280,000
180,000
1,700
1,100
280
690,000
510,000
a Incidences are rounded to two significant digits. These estimates represent benefits from CAIR nationwide. The
modeling used to derive these incidence estimates are reflective of those expected for the final CAIR program including
the CAIR promulgated rule and the proposal to include SO2 and annual NO,, controls for New Jersey andDelaware.
Modeling used to develop these estimates assumes annual SO2 and NOX controls for Arkansas resulting in a
slight overstatement of the reported benefits and costs for the complete CAIR program.
b Premature mortality benefits associated with ozone are not analyzed in the primary analysis.
c Adult premature mortality based upon studies by Pope et al., 2002. Infant premature mortality is based upon studies by
Woodruff, Grillo, and Schoendorf, 1997.
d Respiratory hospital admissions for PM include admissions for chronic obstructive pulmonary disease (COPD),
pneumonia, and asthma.
e Cardiovascular hospital admissions for PM include total cardiovascular and subcategories for ischemic heart disease,
dysrhythmias, and heart failure.
f Respiratory hospital admissions for ozone include admissions for all respiratory causes and subcategories for COPD and
pneumonia.
1-4
-------
Table 1-3. Estimated Monetary Value of Reductions in Incidence of Health and
Welfare Effects (in millions of 1999$)a b
Health Effect
Premature mortalityc'd
Adult >30 years
3% discount rate
7% discount rate
Child <1 year
Chronic bronchitis (adults, 26 and over)
Nonfatal acute myocardial infarctions
3% discount rate
7% discount rate
Hospital admissions for respiratory causes
Hospital admissions for cardiovascular causes
Emergency room visits for asthma
Acute bronchitis (children, age 8-12)
Lower respiratory symptoms (children, 7-14)
Upper respiratory symptoms (asthma, 9-11)
Asthma exacerbations
Work loss days
Minor restricted-activity days (MRADs)
School absence days
Worker productivity (outdoor workers, 18-65)
Recreational visibility, 81 Class I areas
Monetized Total6
Base Estimate:
3% discount rate
7% discount rate
Pollutant
PM25
PM25
PM25
PM25, 03
PM25
PM25,03
PM25
PM25
PM25
PM25
PM25,
PM25,O3
03
03
PM25
PM25,O3
2010
Estimated
$67,300
$56,600
$168
$2,520
$1,420
$1,370
$45.2
$80.7
$2.84
$5.63
$2.98
$3.80
$10.3
$180
$422
$12.9
$7.66
$1,140
$73,300 +
$62,600 +
2015
Value of Reductions
$92,800
$78,100
$222
$3,340
$1,850
$1,790
$78.9
$105
$3.56
$7.06
$3.74
$4.77
$12.7
$219
$543
$36.4
$19.9
$1,780
B $101,000 + B
B $86,300 + B
Monetary benefits are rounded to three significant digits. These estimates represent benefits from CAIR nationwide for
NO and SO2 emission reductions from electricity-generating units (EGU) sources (with the exception of ozone and
visibility benefits). Ozone benefits relate to the eastern United States. Visibility benefits relate to Class I areas in the
southeastern United States. The benefit estimates reflected relate to the final CAIR program that includes the CAIR
promulgated rule and the proposal to include SO2 and annual NO,, controls for New Jersey and Delaware. Modeling
used to develop these estimates assumes annual SO2 and NOX controls for Arkansas resulting in a slight
overstatement of the reported benefits and costs for the complete CAIR program.
Monetary benefits adjusted to account for growth in real GDP per capita between 1990 and the analysis year (2010 or
2015).
Valuation assumes discounting over the SAB-recommended 20-year segmented lag structure described in Chapter 4.
Results show 3 percent and 7 percent discount rates consistent with EPA and OMB guidelines for preparing economic
analyses (U.S. EPA, 2000; OMB, 2003).
Adult premature mortality based upon studies by Pope et al, 2002. Infant premature mortality based upon studies by
Woodruff, Grille, and Schoendorf, 1997.
B represents the monetary value of health and welfare benefits not monetized. A detailed listing is provided in
Table 1-4.
1-5
-------
productivity. Although we are unable to monetize all welfare benefits, EPA estimates CAIR
will yield welfare benefits of $1.8 billion in 2015 (1999$) for visibility improvements in
southeastern Class I (national park) areas.
1.1.3 Uncertainty in the Benefits Estimates
As part of an overall program to improve the Agency's characterization of
uncertainties in health benefits analyses, we present two types of probabilistic approaches to
characterize uncertainty. The first approach generates a distribution of benefits based on the
classical statistical error expressed in the underlying health effects and economic valuation
studies used in the benefits modeling framework. The second approach uses the results from
a pilot expert elicitation project designed to characterize key aspects of uncertainty in the
ambient PM2 5/mortality relationship, and augments the uncertainties in the mortality
estimate with the statistical error reported for other endpoints in the benefit analysis. Both
approaches provide insights into the likelihood of different outcomes and about the state of
knowledge regarding the benefits estimates.
The uncertainty estimates have the strength of presenting a statistical measure of the
uncertainty in the underlying studies serving as the basis for the estimates used in the
analysis. However, this approach captures only a limited portion of the uncertainty about the
parameters. The 5th and 95th percentile points of the distributions are based on statistical
error and cross-study variability and provide some insight into how uncertain our estimate is
with regard to those sources of uncertainty. However, it does not capture other sources of
uncertainty regarding the model specification and other inputs to the model, including
emissions, air quality, and aspects of the health science not captured in the studies, such as
the likelihood that PM is causally related to premature mortality and other serious health
effects.
Figure 1-1 presents box plots of the distributions of the reduction in PM2 5-related
premature mortality based on the C-R distributions provided by each expert, as well as that
for our primary estimate, based on the statistical error associated with Pope et al. (2002).
The distributions are depicted as box plots with the diamond symbol (+) showing the mean,
the dash (-) showing the median (50th percentile), the box defining the interquartile range
(bounded by the 25th and 75th percentiles), and the whiskers defining the 90 percent
confidence interval (bounded by the 5th and 95th percentiles of the distribution). Our
primary estimate based on the Pope et al. (2002) study shows that the average number of
premature deaths avoided in 2015 is 17,000 (at a value of $100 billion) with the 5th and 95th
percentiles of the distribution ranging from 6,000 to 27,000 fewer mortalities. The figure
1-6
-------
Distribution of
Primary
Estimate
Pcpe et al (2002)
Satistical Error
Distributions Based on Application of Pilot Expert Elicitation Results
13,700
-
Expert A
Experts
Expert C
Expert D
Expert E
INote: Distributions labeled Expert A- Expert E are based on indvidual expert responses. Trie distribution labeled Pope et al. (2002) Statistical Error is based on the mean and standard
error of the C-R function from the study.
Note: The results of the Pilot Expert Elicitation are presented here as an illustration of EPA's initial
efforts to characterize the uncertainties associated with the estimate of benefits from the PM2 /mortality
relationship. The Pilot was limited in scope and does not address inherent differences in the thought
processes and background information used by each expert to express their distribution. Based on
findings from the Pilot and a favorable peer review of the Pilot, EPA is conducting a full-scale expert
elicitation to better characterize uncertainty in the mortality estimate for future regulatory analyses. See
Appendix B for a full description of the two approaches used to characterize uncertainty.
Figure 1-1. Comparative Assessment of Relative Range of Uncertainty in Estimated
Avoided Incidence of Premature Mortality Using Classical Statistical Error and the
Pilot Expert-Based Characterizations of Uncertainty
shows that the average annual number of premature deaths avoided in 2015 based on the
pilot expert elicitation ranges from approximately 3,000 (based on the judgments of Expert
C), which is valued at $23 billion to 23,000 (based on the judgments of Expert E), which is
valued at $140 billion. The confidence intervals vary across experts with all experts
estimating zero at the 5th percentile and the 95th percentile ranging from 10,000 to 54,000
fewer mortalities.
As part of the CAIR analysis, we conducted a variety of supplemental analyses
designed to provide the reader with an understanding of the degree of uncertainty that may
be associated with the benefits resulting from implementation of this regulation. Because
estimates of premature mortality contribute the most to the monetized benefits, our efforts
focused on the sensitivity of the final benefits estimate to analytic judgments regarding this
1-7
-------
relationship. Specifically, we conducted analyses designed to characterize the degree of
uncertainty in the slope (magnitude) of the PM25 concentration-response function, the form
of the PM2 5 concentration-response function (i.e., the potential for a threshold), and the
cessation lag (i.e., temporal relationship between cessation of exposure and reduction in
adverse health effects). Both discrete and probabilistic approaches were used to characterize
the uncertainty associated with the concentration response function.
These supplemental analyses yield the following insights:
• Use of statistical error associated with the ACS (Pope et al., 2002) estimate for
the concentration response function for PM2 5—premature mortality as well as the
statistical error associated with the concentration response functions for each of
the other health endpoints to describe the probability distribution of total benefits
yields a distribution in which the 95th percentile is nearly twice the mean ($100
billion in 2015) and 5th percentile is one fourth the mean. The overall range from
5th to 95th percentile on the total benefits estimate represents one order of
magnitude ($26 billion to $210 billion).
• Description of the probability distribution of the concentration response function
for PM2 5—premature mortality using the results from the pilot expert elicitation
(rather than the estimate based on the statistical error associated with the ACS
cohort) yields a larger degree of uncertainty because the elicitation exercise was
designed to encompass a broader set of model uncertainty. The mean annual
benefits for each expert elicited during the pilot project range from approximately
$16 billion to $130 billion in 2015
• Substitution of the steeper concentration response function for PM2 5—premature
mortality from the Six Cities study increases the value of the total benefits from
$101 billion to $208 billion in 2015.
• Substitution of the most plausible alternative lag structures has little overall
impact on the estimate of total benefits (reductions are on the order of 5 to 15
percent).
• The assessment of alternative assumptions regarding the existence (and level) of a
threshold in the PM2 5 premature mortality concentration response function
highlights the sensitivity of the analysis to this assumption. Only Spercent of the
estimated premature mortality is due to changes in exposure above 15mg/m3,
while over 84 percent of the premature morality related benefits are due to
changes in PM25 concentrations occurring above 10ug/m3.
1-8
-------
• Estimates of premature mortality from ozone exposure may result in an additional
500 premature deaths avoided and an increase in the estimated health benefits of
CAIR by approximately $3 billion annually.
In addition to these mortality related supplemental analyses, we also conducted analyses
related to non-health (welfare) effects, including visibility and household cleaning costs.
Based on these analyses, expanded coverage of welfare effects could increase benefits by
over $500 million. Other welfare effects have been quantified such as nitrogen and sulfur
deposition reductions in the CAIR region, acidification reductions in lakes the Adirondacks
and the northeastern US, and reduced nitrogen deposition to the Chesapeake Bay. While
monetized estimates of these benefits could not be examined even in sensitivity analyses, it is
likely these benefit categories are significant in terms of the total ecological endpoints.
1.2 Not All Benefits Quantified
EPA was unable to quantify or monetize all of the health and environmental benefits
associated with CAIR. EPA believes these unquantified benefits are substantial, including
the value of increased agricultural crop and commercial forest yields, visibility
improvements, reductions in nitrogen and acid deposition and the resulting changes in
ecosystem functions, and health and welfare benefits associated with reduced mercury
emissions. Table 1-4 provides a list of these benefits.
1.3 Costs and Economic Impacts
For the affected region, the projected annual incremental private costs of CAIR to the
power industry are $2.36 billion in 2010 and $3.57 billion in 2015. These costs represent the
total cost to the electricity-generating industry of reducing NOX and SO2 emissions to meet
the caps set out in the rule. Estimates are in 1999 dollars. Costs of the rule are estimated
using the Integrated Planning Model and assume firms make decisions using costs of capital
ranging from 5.34 percent to 6.74 percent.
In estimating the net benefits of regulation, the appropriate cost measure is "social
costs." Social costs represent the welfare costs of the rule to society. These costs do not
consider transfer payments (such as taxes) that are simply redistributions of wealth. The
social costs of this rule are estimated to be $1.91 billion in 2010 and $2.56 billion in 2015
assuming a 3 percent discount rate. These costs become $2.14 billion in 2010 and $3.07
billion in 2015, if one assumes a 7 percent discount rate.
1-9
-------
Table 1-4. Unquantified and Nonmonetized Effects of the Clean Air Interstate Rule
Pollutant/Effect
Effects Not Included in Primary Estimates—Changes in:
Ozone—Health3
Ozone—Welfare
PM—Health0
PM—Welfare
Nitrogen and Sulfate
Deposition—Welfare
Mercury Health
Mercury Deposition
Welfare
Premature mortality15
Chronic respiratory damage
Premature aging of the lungs
Nonasthma respiratory emergency room visits
Increased exposure to UVb
Yields for:
- commercial forests,
- fruits and vegetables, and
- commercial and noncommercial crops
Damage to urban ornamental plants
Recreational demand from damaged forest aesthetics
Ecosystem functions
Increased exposure to UVb
Premature mortality: short-term exposures'1
Low birth weight
Pulmonary function
Chronic respiratory diseases other than chronic bronchitis
Nonasthma respiratory emergency room visits
Exposure to UVb (+/-)e
Visibility in many Class I areas
Residential and recreational visibility in non-Class I areas
Soiling and materials damage
Ecosystem functions
Exposure to UVb (+/-)e
Commercial forests due to acidic sulfate and nitrate deposition
Commercial freshwater fishing due to acidic deposition
Recreation in terrestrial ecosystems due to acidic deposition
Existence values for currently healthy ecosystems
Commercial fishing, agriculture, and forests due to nitrogen deposition
Recreation in estuarine ecosystems due to nitrogen deposition
Ecosystem functions
Passive fertilization due to nitrogen deposition
Incidence of neurological disorders
Incidence of learning disabilities
Incidence of developmental delays
Potential reproductive effectsf
Potential cardiovascular effectsf, including:
- Altered blood pressure regulationf
- Increased heart rate variabilityf
- Incidence of myocardial infarctionf
Impacts on birds and mammals (e.g., reproductive effects)
Impacts to commercial, subsistence, and recreational fishing
(continued)
1-10
-------
Table 1-4. Unquantified and Nonmonetized Effects of the Clean Air Interstate Rule
(continued)
a In addition to primary economic endpoints, there are a number of biological responses that have been associated with
ozone health effects including increased airway responsiveness to stimuli, inflammation in the lung, acute inflammation
and respiratory cell damage, and increased susceptibility to respiratory infection. The public health impact of these
biological responses may be partly represented by our quantified endpoints.
b Premature mortality associated with ozone is not currently included in the primary analysis. Recent evidence suggests
that short-term exposures to ozone may have a significant effect on daily mortality rates, independent of exposure to PM.
EPA is currently conducting a series of meta-analyses of the ozone mortality epidemiology literature. EPA will consider
including ozone mortality in primary benefits analyses once a peer-reviewed methodology is available.
c In addition to primary economic endpoints, there are a number of biological responses that have been associated with PM
health effects including morphological changes and altered host defense mechanisms. The public health impact of these
biological responses may be partly represented by our quantified endpoints.
d While some of the effects of short term exposures are likely to be captured in the estimates, there may be premature
mortality due to short term exposure to PM not captured in the cohort study upon which the primary analysis is based.
e May result in benefits or disbenefits. See discussion in Section 5.3.4 for more details.
f These are potential effects as the literature is insufficient.
Retail electricity prices are projected to increase roughly 2.0 to 2.7 percent with
CAIR in the 2010 and 2015 time frame and then drop below 2.0 percent thereafter. The
effects of CAIR on natural gas prices and the power-sector generation mix is also small, with
a 1.6 percent or less increase in gas prices projected from 2010 to 2020. There will be a
continued reliance on coal-fired generation, which is projected to remain at roughly
50 percent of total electricity generated. A relatively small amount of coal-fired capacity,
about 5.3 GW (1.7 percent of all coal-fired capacity and 0.5 percent of all generating
capacity), is projected to be uneconomic to maintain. In practice units projected to be
uneconomic to maintain may be "mothballed," retired, or kept in service to ensure
transmission reliability in certain parts of the grid. For the most part, these units are small
and infrequently used generating units that are dispersed throughout the CAIR region. As
demand grows in the future, additional coal-fired generation is projected to be built under
CAIR and utilization of coal-fired units will increase. Because of this, coal production is
projected to increase from 2003 levels by about 15 percent in 2010 and by 25 percent by
2020, and we expect greater coal production in Appalachia and the Interior coal regions of
the country with CAIR. Overall, the impacts of CAIR are modest, particularly in light of the
large projected benefits of CAIR.
1-11
-------
1.4 Limitations
Every analysis examining the potential benefits and costs of a change in
environmental protection requirements is limited to some extent by data gaps, limitations in
model capabilities (such as geographic coverage), and variability or uncertainties in the
underlying scientific and economic studies used to configure the benefit and cost models.
Despite these uncertainties, we believe this benefit-cost analysis provides a reasonable
indication of the expected economic benefits and costs of C AIR in future years.
For this analysis, such uncertainties include possible errors in measurement and
projection for variables such as population growth and baseline incidence rates; uncertainties
associated with estimates of future-year emissions inventories and air quality; variability in
the estimated relationships between changes in pollutant concentrations and the resulting
changes in health and welfare effects; and uncertainties in exposure estimation. We have
used sensitivity analyses to address these limitations where possible.
EPA's cost estimates assume that all States in the CAIR region fully participate in the
cap and trade programs that reduce SO2 and NOX emissions from EGUs. The cost
projections also do not take into account the potential for advancements in the capabilities of
pollution control technologies for SO2 and NOX removal and other compliance strategies,
such as fuel switching or the reductions in their costs over time. EPA projections also do not
take into account demand response (i.e., consumer reaction to electricity prices) because the
consumer response is likely to be relatively small, but the effect on lowering private
compliance costs may be substantial. Costs may be understated since an optimization model
was employed and the regulated community may not react in the same manner to comply
with the rules. The Agency also did not factor in the costs and/or savings for the government
to operate the CAIR program as opposed to other air pollution compliance programs and
transact!onal costs and savings from CAIR's effects on the labor supply. A listing of
possible unquantified costs associated with the CAIR program are shown in Table 1-5.
Table 1-5. Unquantified Costs of the Clean Air Interstate Rule
Effects Not Quantified
Employment shifts as workers are retrained at the same company or re-employed elsewhere in the economy.
Costs of running and administering the program to State and Federal Government.
Certain relatively small permitting costs associated with Title IV that new program entrants face.
1-12
-------
1.5 References
Pope, C.A., III, R.T. Burnett, MJ. Thun, E.E. Calle, D. Krewski, K. Ito, and G.D. Thurston.
2002. "Lung Cancer, Cardiopulmonary Mortality, and Long-term Exposure to Fine
Particulate Air Pollution." Journal of the American Medical Association
287:1132-1141.
U.S. Environmental Protection Agency (EPA). September 2000. Guidelines for Preparing
Economic Analyses. EPA 240-R-00-003.
U.S. Office of Management and Budget (OMB). 2003. Circular A-4 Guidance to Federal
Agencies on Preparation of Regulatory Analysis.
Woodruff, T.J., J. Grille, and K.C. Schoendorf. 1997. "The Relationship Between Selected
Causes of Postneonatal Infant Mortality and Particulate Infant Mortality and
Particulate Air Pollution in the United States." Environmental Health Perspectives
105(6):608-612.
1-13
-------
CHAPTER 2
INTRODUCTION AND BACKGROUND
2.1 Introduction
For this rulemaking, the U.S. Environmental Protection Agency (EPA) has assessed
the role that transported emissions from upwind states play in contributing to unhealthy
levels of PM25 and 8-hour ozone in downwind states. Based on this assessment, the Clean
Air Interstate Rule (CAIR) requires air emissions reductions from upwind states. This
document presents the health and welfare benefits of CAIR and compares the benefits of this
rule to the estimated costs of implementing the rule in 2010 and 2015. This chapter contains
background information relative to the rule and an outline of the chapters of the report.
2.2 Background
Congress recognized that interstate pollution transport from upwind states can
contribute to unhealthy pollution levels in downwind states. Therefore, the Clean Air Act
(CAA) contains provisions in Section 110(a)(2)(D) that require upwind states to eliminate
emissions that contribute significantly to nonattainment downwind. Under Section
110(a)(2), states are required to submit plans to EPA within 3 years of issuance of a revised
National Ambient Air Quality Standard (NAAQS). Among other requirements, these plans
are required to prohibit emissions in the state that contribute significantly to nonattainment
downwind.
EPA's final rule finds that 28 states and the District of Columbia contribute
significantly to nonattainment, or interfere with maintenance, of the NAAQS for PM25
and/or 8-hour ozone in downwind states. EPA requires these upwind states to revise their
State Implementation Plans (SIPs) to include control measures to reduce emissions of SO2
and/or NOX. SO2 is a precursor to PM2 5 formation, and NOX is a precursor to both ozone and
PM2 5 formation. Reducing upwind precursor emissions will assist the downwind PM2 5 and
8-hour ozone nonattainment areas in achieving the NAAQS. Moreover, attainment would be
achieved in a more equitable, cost-effective manner than if each nonattainment area
attempted to achieve attainment by implementing local emissions reductions alone. The
relevant regions for PM25 and ozone significant contribution are depicted in Figure 2-1.
2-1
-------
States controlled for both SO2 and NOX
States controlled for Ozone Season NOX
States not covered by CAIR
Figure 2-1. Final CAIR Region
The estimates presented in this report represent the benefits and costs for a final CAIR
program that includes the final promulgated CAIR and the proposal to include SO2 and
annual NOX controls for New Jersey and Delaware in CAIR. The modeling used to provide
these estimates also assumes annual SO2 and NOX controls for Arkansas that are not a part of
the complete CAIR program resulting in a slight overstatement of the reported benefits and
costs.
2.3 Regulated Entities
This action does not directly regulate emissions sources. Instead, it requires states to
revise their SIPs to include control measures to reduce emissions of NOX and SO2. The
emissions reduction requirements that would be assigned to the states are based on controls
that are known to be highly cost-effective for EGUs. EPA modeled emission cap-and-trade
programs phased in over time beginning with SO2 and NOX caps in 2010 and lowering these
emission caps in 2015. The timing of emission caps was decided on the basis of when
control actions would be needed to help the states in their NAAQS attainment efforts,
feasibility of installing emission controls, and other factors. However, states would have the
flexibility to choose the sources to control and how to control them. Although states have
the flexibility to control pollution from sources other than EGUs, the analysis conducted
assumes controls for EGUs only.
2-2
-------
2.4 Baseline and Years of Analysis
The final rule on which this analysis is based sets forth the requirements for states to
eliminate their significant contribution to downwind nonattainment of ozone and PM2 5
NAAQS. To reduce this significant contribution, EPA requires that certain states reduce
their emissions of SO2 and NOX. The Agency considered all promulgated CAA requirements
and known state actions in the baseline used to develop the estimates of benefits and costs
for this rule. EPA did not consider actions states may take to implement the ozone and PM2 5
NAAQS standards in the baseline for this analysis. The years 2010 and 2015 are used in this
analysis. The year of 2010 was chosen as one of the analysis years, because this year
represents the date of Phase I of the rule and is in close proximity to the 2009 phase date for
NOX. The year 2015 represents the year in which Phase II of the rule is anticipated to be
implemented. All estimates presented in this report represent annualized estimates of the
benefits and costs of C AIR in 2010 and 2015 rather than the net present value of a stream of
benefits and costs in these particular years of analysis.
2.5 Control Scenario
The analysis conducted assumes that a cap-and-trade program will be used to achieve
the emission reduction requirements from the electric power industry. All fossil-fuel electric
generating units (EGUs) over 25 megawatt (MW) capacity within the CAIR region are
covered by the program. With the complete CAIR program (CAIR final plus the New Jersey
and Delaware proposal), EPA would establish regional emission budgets (caps) for SO2 and
NOX to address the transport problem. In this final rule, these requirements would effectively
establish annual emission caps in 2010 for SO2 and NOX of 3.7 million tons and 1.5 million
tons, respectively. These emission budgets (caps) would be lowered in 2015 to provide
annual SO2 and NOX emission caps of 2.6 million tons and 1.3 million tons, respectively, in
the control region. Banking of emissions is allowed in the program. These caps were
derived by determining the amount of emissions of SO2 and NOX that EPA believes can be
controlled from EGUs in a highly cost-effective manner. When fully implemented, this
would result in nationwide SO2 emissions of approximately 3.4 million tons. This is
significantly lower than the 8.95 million tons of SO2 emissions allowed under the current
Title IV Acid Rain SO2 Trading Program. For the final CAIR promulgated rule (exclusive of
the New Jersey and Delaware proposal) emission caps are 3.6 million tons for SO2 and 1.5
million tons for NOX in 2010. These estimates become 2.5 million tons and 1.3 million tons
in 2015.
2-3
-------
2.6 Benefits of Emission Controls
The benefits of CAIR are discussed in Chapters 4 and 5 of this report. Annual
monetized benefits of $101 billion (3 percent discount rate) or $86.3 billion (7 percent
discount rate) are expected for CAIR in 2015. Despite the fact that the final CAIR program
is comparable in most respects to the proposed rule, the benefits reported for the final rule
exceed the estimates reported in the Benefits of the Proposed Inter state Air Quality Rule
(OAR-2003-0053-0175, January 2004). There are several reasons for the increase in
monetized benefit estimates for the final rule. These reasons include increased SO2 emission
reductions, geographical changes in the location of emission reductions with great reductions
occurring near population centers, and additional direct PM2 5 emissions reductions for the
final rule.
2.7 Cost of Emission Controls
EPA analyzed the costs of CAIR using the Integrated Planning Model (IPM). EPA
has used this model in the past to analyze the impacts of regulations on the power sector.
EPA estimates the private industry costs of the rule to the power sector to be $3.57 billion in
2015 (1999 dollars). In estimating the net benefits of the rule, EPA uses social costs of the
rule that represent the costs to society of this rule. The social costs of the rule are estimated
to be $2.56 or $3.07 billion in 2015 (3 percent and 7 percent discount rates, respectively). A
description of the methodology used to model the costs and economic impacts to the power
sector is discussed in Chapter 7 of this report.
2.8 Organization of the Regulatory Impact Analysis
This report presents EPA's analysis of the benefits, costs, and other economic effects
of the final CAIR to fulfill the requirements of a Regulatory Impact Analysis (RIA). This
RIA includes the following chapters:
• Chapter 3, Emissions and Air Quality Impacts, describes emission inventories and
air quality modeling that are essential inputs into the benefits assessment.
• Chapter 4, Benefits Analysis and Results, describes the methodology and results
of the benefits analysis.
• Chapter 5, Qualitative Assessment of Nonmonetized Benefits, describes benefits
that are not monetized for this rulemaking.
• Chapter 6, Electric Power Sector Profile, describes the industry potentially
affected by the rule.
2-4
-------
Chapter 7, Cost, Economic, and Energy Impacts of the Rule, describes the
modeling conducted to estimate the cost, economic, and energy impacts to the
affected sources.
Chapter 8, Statutory and Executive Order Impact Analyses, describes the small
business, unfunded mandates, paperwork reduction act, and other analyses
conducted for the rule to meet statutory and Executive Order requirements.
Chapter 9, Comparison of Benefits and Costs, shows a comparison of the social
benefits to social costs of the rule.
Appendix A, Benefits and Costs of the Clean Air Interstate Rule, Clean Air
Visibility Rule, and the Clean Air Interstate Rule Plus the Clean Air Visibility
Rule
Appendix B, Supplemental Analyses Addressing Uncertainties in the Benefits
Analyses
Appendix C, Sensitivity Analyses of Key Parameters in the Benefits Analysis
Appendix D, Sensitivity Analysis of Key Parameters in the Cost and Economic
Impact Analysis and Listing of IPM Runs in Support of C AIR
Appendix E, CAIR Industry Sector Impacts
Appendix F, Additional Technical Information Supporting the Benefits Analysis
Appendix G, Health-Based Cost-Effectiveness of Reductions in Ambient PM2 5
Associated With CAIR
2-5
-------
SECTION 3
EMISSIONS AND AIR QUALITY IMPACTS
This chapter summarizes the emissions inventories and air quality modeling that
serve as the inputs to the benefits analysis of this rule as detailed in Chapter 4. EPA uses
sophisticated photochemical air quality models to estimate baseline and post-control ambient
concentrations of ozone and PM and deposition of nitrogen and sulfur for each year. The
estimated changes in ambient concentrations are then combined with monitoring data to
estimate population-level exposures to changes in ambient concentrations for use in
estimating health effects. Modeled changes in ambient data are also used to estimate
changes in visibility and changes in other air quality statistics that are necessary to estimate
welfare effects.
Section 3.1 of this chapter provides a summary of the baseline emissions inventories
and the emissions reductions that were modeled for this rule. Section 3.2 provides a
summary of the methods for and results of estimating air quality for the 2010 and 2015 base
cases and control scenarios for the purposes of the benefits analysis. There are separate
sections for PM, ozone, and visibility.
3.1 Emissions Inventories and Estimated Emissions Reductions
The technical support document for emissions inventories discusses the development
of the 2001, 2010, and 2015 baseline emissions inventories for the benefits analysis of this
final rule. The emission sources and the basis for current and future-year inventories are
listed in Table 3-1. Tables 3-2 and 3-3 summarize the baseline emissions of NOX and SO2
and the change in the emissions from EGUs that were used in modeling air quality changes.
For details on EPA's projected emissions for the EGU sector, see Chapter 7 of this RIA.
3-1
-------
Table 3-1. Emissions Sources and Basis for Current and Future-Year Inventories
Sector
Emissions Source 2001 Base Year
Future-Year Base Case Projections
ECU
Non-EGU
Average
Fire
Average
Fire
Ag
Ag
Area
On-road
Nonroad
Nonroad
Power industry
electric generating
units (EGUs)
Non-Utility Point,
including point-
source fugitive
dust
Wildfire,
prescribed burning
Agricultural
burning, open
burning
Livestock NH3
Fertilizer NH3
All other stationary
area sources,
including area-
source fugitive
dust
Highway vehicles
Locomotives,
commercial marine
vessels, and
aircraft
All other nonroad
vehicles
2001 data from
Acid Rain
Trading Program
2001 National
Emission
Inventory (NET)
Same as future
year
2001 NEI
2002 Preliminary
NEP
2001 NEI
1999 NEI, version
3 grown to 2001
MOBILE6.2
model
2001 NEI; CMV
adjusted to new
national totals
from OTAQ
NONROAD2004
model
Integrated Planning Model (IPM)
(1) Department of Energy (DOE) fuel use
projections, (2) Regional Economic Model, Inc.
(REMI) Policy Insight® model, (3) decreases to
REMI results based on trade associations, Bureau
of Labor Statistics (BLS) projections and Bureau of
Economic Analysis (BEA) historical growth from
1987 to 2002, (4) control assumptions
Average fires from 1996 through 2002 (based on
state-total acres burned), with the same emissions
rates and county distributions of emissions as in the
2001 NEI
2001 NEI
2010 and 2015 emissions estimated with the same
approach as was used for the 2002 preliminary NEP
2001 NEI
(1) DOE fuel use projections, (2) REMI Policy
Insight model, (3) decreases to REMI results based
on trade associations, BLS projections and BEA
historical growth from 1987-2002
Projected vehicle miles traveled same as IAQR
proposal; emissions fromMOBILE6.2 model
Grown based on national totals from OTAQ, using
state/county distribution of emissions from the 2001
NEI
NONROAD 2004 model
This table documents only the sources of data for the U.S. inventory. The sources of data used for Canada
and Mexico are explained in the technical support document (TSD) and were held constant from the base
year to the future years.
All fugitive dust emissions were adjusted downward using county-specific transportable fractions needed as
part of the current state of the art in air quality modeling.
ftp://ftp.epa.gov/EmisInventory/prelim2002nei/nonpoint/documentation/nh3inventorydraftJan2004.pdf.
3-2
-------
Table 3-2. Summary of Modeled Baseline Emissions for Lower 48 States
Source"
2001 Baseline
EGUs
Non-EGUs
Average Fire
Area
On-road
Nonroad
Total, All Sources
2010 Base Case
EGUs
Non-EGUs
Average Fire
Area
Mobile
Nonroad
Total, All Sources
2015 Base Case
EGUs
Non-EGUs
Average Fire
Area
Mobile
Nonroad
Total, All Sources
Pollutant
NOX
4,937,398
2,942,618
238,931
1,462,276
8,064,067
4,050,655
21,695,945
3,672,929
2,931,360
238,931
1,630,411
4,683,085
3,282,338
16,439,055
3,708,658
3,183,499
238,931
1,702,154
3,152,562
2,912,382
14,898,186
Emissions (tons)
SO2
10,901,127
2,958,692
49,108
1,295,146
271,026
433,250
15,908,349
9,903,882
3,189,864
49,108
1,408,990
27,435
219,029
14,798,308
9,079,214
3,422,915
49,108
1,480,348
30,824
232,627
14,295,035
The "ag" sector does not have emissions of NOX and SO2.
3-3
-------
Table 3-3. Summary of Modeled Emissions Changes for the Clean Air Interstate Rule:
2010 and 2015
Pollutant
Item NOX SO2
2010 Emission Reductions3
Absolute Tons 1,245,038 3,620,280
Percentage of Base EGU Emissions 33.9% 36.6%
Percentage of All Manmade Emissions 7.6% 24.5%
2015 Emission Reductions3
Absolute Tons 1,535,821 3,967,777
Percentage of Base EGU Emissions 41.4% 43.7%
Percentage of All Manmade Emissions 10.3% 27.8%
a Note that the emission changes only occur within the affected transport region; however, the percentage
reductions reflect the change as a share of baseline emissions for the lower 48 states as presented in
Table 3-2.
3.2 Air Quality Impacts
This section summarizes the methods for and results of estimating air quality for the
2010 and 2015 base cases and control scenarios for the purposes of the benefits analysis.
EPA has focused on the health, welfare, and ecological effects that have been linked to air
quality changes. These air quality changes include the following:
1. Ambient particulate matter (PM2 5)—as estimated using a national-scale
applications of the Community Multi-Scale Air Quality (CMAQ) model;
2. Ambient ozone—as estimated using regional-scale applications of the
Comprehensive Air Quality Model with Extensions (CAMx); and
3. Visibility degradation (i.e., regional haze), as developed using empirical
estimates of light extinction coefficients and efficiencies in combination with
CMAQ modeled reductions in pollutant concentrations.
3-4
-------
The air quality estimates in this section are based on the emission changes
summarized in the preceding section. These air quality results are in turn associated with
human populations and ecosystems to estimate changes in health and welfare effects. In
Section 3.2.1, we describe the estimation of PM air quality using CMAQ, and in
Section 3.2.2, we cover the estimation of ozone air quality using CAMx. Lastly, in
Section 3.2.3, we discuss the estimation of visibility degradation.
3.2.1 PMAir Quality Estimates
We use the emissions inputs summarized above with a national-scale application of
the Community Multi-scale Air Quality (CMAQ) modeling system to estimate PM air
quality in the contiguous United States. CMAQ is a three-dimensional grid-based Eulerian
air quality model designed to estimate annual particulate concentrations and deposition over
large spatial scales (e.g., over the contiguous United States). Consideration of the different
processes that affect primary (directly emitted) and secondary (formed by atmospheric
processes) PM at the regional scale in different locations is fundamental to understanding
and assessing the effects of pollution control measures that affect PM, ozone, and deposition
of pollutants to the surface.1 Because it accounts for spatial and temporal variations as well
as differences in the reactivity of emissions, CMAQ is useful for evaluating the impacts of
the rule on U.S. PM concentrations. Our analysis applies the modeling system to the entire
United States for the six emissions scenarios: a 2001 base year, a 2010 baseline projection
and a 2010 projection with controls, and a 2015 baseline projection and a 2015 projection
with controls.
The CMAQ version 4.3 was employed for this CAIR modeling analysis (Byun and
Schere, 2004). This version reflects updates in a number of areas to improve performance
and address comments from the peer review, including (1) the formation of nitrates based on
updated gaseous/heterogeneous chemistry and a current inorganic nitrate partitioning
module, (2) a state-of-the-science Secondary Organic Aerosol (SOA) module that includes a
more comprehensive gas-particle partioning algorithm from both anthropogenic and biogenic
SOA, (3) an in-cloud sulfate chemistry that accounts for the nonlinear sensitivity of sulfate
'Given the focus of this rule on secondarily formed particles (e.g., sulfates) it is important to employ a Eulerian
model such as CMAQ. The formation and fate of secondarily formed pollutants typically involve emissions
of precursor pollutants (e.g., SO2) from a multitude of widely dispersed sources coupled with chemical and
physical processes which are best addressed using an air quality model that employs a Eulerian grid model
design.
3-5
-------
formation to varying pH, and (4) the updated CB-IV gas-phase chemistry mechanism and
aqueous chemistry mechanism that provide a comprehensive simulation of aerosol precursor
oxidants.
CMAQ simulates every hour of every day of the year and, thus, requires a variety of
input files that contain information pertaining to the modeling domain and simulation period.
These include hourly emissions estimates and meteorological data in every grid cell, as well
as a set of pollutant concentrations to initialize the model and to specify concentrations along
the modeling domain boundaries. These initial and boundary concentrations were obtained
from output of a global chemistry model. As discussed below, we use the model predictions
in a relative sense by first determining the ratio of species predictions between the 2001 base
year and each future-year scenario. The calculated relative change is then combined with the
corresponding ambient species measurements to project concentrations for the future case
scenarios. The annual mean PM air quality is used as input to the health and welfare
concentration-response (C-R) functions of the benefits analysis. The following sections
provide a more detailed discussion of each of the steps in this evaluation and a summary of
the results.
3.2.1.1 Modeling Domain
As shown in Figure 3-1, the modeling domain encompasses the lower 48 states and
extends from 126 degrees to 66 degrees west longitude and from 24 degrees north latitude to
52 degrees north latitude. The modeling domain is segmented into rectangular blocks
referred to as grid cells. The model actually predicts pollutant concentrations for each of
these grid cells. For this application the horizontal grid cells are roughly 36 km by 36 km.
In addition, the modeling domain contains 14 vertical layers with the top of the modeling
domain at about 16,200 meters, or 100 mb. Within the domain each vertical layer has 16,576
grid cells.
3.2.1.2 Simulation Periods
For use in this benefits analysis, the simulation periods modeled by CMAQ included
separate full-year application for each of the five emissions scenarios (i.e., 2001 base year
and the 2010 and 2015 base cases and control scenarios).
3-6
-------
Figure 3-1. CMAQ Modeling Domain
3.2.1.3 Model Inputs
CMAQ requires a variety of input files that contain information pertaining to the
modeling domain and simulation period. These include gridded, hourly emissions estimates
and meteorological data and initial and boundary conditions. Separate emissions inventories
were prepared for the 2001 base year and each of the future-year base cases and control
scenarios. All other inputs were specified for the 2001 base year model application and
remained unchanged for each future-year modeling scenario.
3-7
-------
CMAQ requires detailed emissions inventories containing temporally allocated
emissions for each grid cell in the modeling domain for each species being simulated. The
previously described annual emission inventories were preprocessed into model-ready inputs
through the emissions preprocessing system. Details of the preprocessing of emissions are
provided in the Clean Air Interstate Rule Emissions Inventory Technical Support Document
(Emissions Inventory TSD) (EPA, 2005). Meteorological inputs reflecting 2001 conditions
across the contiguous United States were derived from version 5 of the Mesoscale Model
(MM5). These inputs include horizontal wind components (i.e., speed and direction),
temperature, moisture, vertical diffusion rates, and rainfall rates for each grid cell in each
vertical layer.
The lateral boundary and initial species concentrations are provided by a three-
dimensional global atmospheric chemistry and transport model (GEOS-CHEM). The lateral
boundary species concentrations varied with height and time (every 3 hours). Terrain
elevations and land use information were obtained from the U.S. Geological Survey database
at 10 km resolution and aggregated to the roughly 36 km horizontal resolution used for this
CMAQ application.
3.2.1.4 CMAQ Model Evaluation
An operational model performance evaluation for PM2 5 and its related speciated
components (e.g., sulfate, nitrate, elemental carbon, organic carbon) as well as deposition of
ammonium, nitrate, and sulfate for 2001 was performed to estimate the ability of the CMAQ
modeling system to replicate base-year concentrations. This evaluation principally
comprises statistical assessments of model versus observed pairs that were paired in time and
space on a daily or weekly basis, depending on the sampling period of measured data. The
statistics are presented separately for the entire domain, the East, and the West (using the
100th meridian to divide the eastern and western United States). In addition, scatterplots of
seasonal average and annual average predictions versus observations paired by site are
included in the model performance evaluation. A spatial analysis was also performed for
sulfate and nitrate to examine how well the modeling platform (year-specific meteorology,
anthropogenic and biogenic emissions, and boundary conditions representative of 2001)
predicts the spatial patterns and gradients evident from the observations. The details of these
graphical analyses can be found in the Clean Air Interstate Rule Air Quality Modeling
Technical Support Document (Air Quality Modeling TSD).
3-8
-------
For PM25 species, this evaluation includes comparisons of model predictions to the
corresponding measurements from the Clean Air Status and Trends Network (CASTNet)
and the Speciation Trend Network (STN) in addition to measurements from the Interagency
Monitoring of PROtected Visual Environments (IMPROVE). The CASTNet dry deposition
monitoring network contained a total of 79 sites in 2001, with a total number of 56 sites
located in the East and 23 sites located in the West. Sulfate and total nitrate data were used
in the evaluation. CASTNet data are collected and reported as weekly average data. The
data are collected in filter packs that sample the ambient air continuously during the week.
The sulfate data are of high quality because sulfate is a stable compound. However, the
particulate nitrate concentration data collected by CASTNet are known to be problematic and
subject to volatility because of the length of the sampling period. CASTNet also reports a
total nitrate measurement, which is the combination of particulate nitrate and nitric acid.
Because the total nitrate measurement is not affected by this sampling problem, it is
considered a more reliable measurement. Therefore, we chose to use the total nitrate data
and not to use the particulate nitrate data in this evaluation.
The EPA STN network began operation in 1999 to provide nationally consistent
speciated PM2 5 data for the assessment of trends at representative sites in urban areas. STN
reports mass concentrations and PM2 5 constituents, including sulfate, nitrate, ammonium,
and elemental and organic carbon. Most STN sites collect data on a frequency of 1 in every
3 days, (some supplemental sites are collected 1 in every 6 days). For the 2001 analysis,
CMAQ predictions were evaluated against 133 STN sites (105 sites in the East and 28 sites
in the West).
The IMPROVE network is a cooperative visibility monitoring effort between EPA,
federal land management agencies, and state air agencies. Data are collected at Class I areas
across the United States mostly at national parks, national wilderness areas, and other
protected pristine areas. Approximately 134 IMPROVE rural/remote sites had complete
annual PM25 mass and/or PM25 species data for 2001. Eighty-six sites were in the West, and
forty-eight sites were in the East. IMPROVE data are collected once in every 3 days.
The principal evaluation statistics used to evaluate CMAQ performance are the
fractional bias and fractional error. Fractional bias is defined as:
3-9
-------
2 N (Pred' -
FBIAS = — E ^ ^ * 100
A/ .4 / T-» / I f\ J 1 \
•
-------
Table 3-4. Model Performance Statistics for CAIR CMAQ 2001
CAIR CMAQ 2001 Annual
PM25
Total Mass
Sulfate
Nitrate
Total Nitrate
(NO3 + HNO3)
Elemental Carbon
Organic Carbon
STN
IMPROVE
STN
IMPROVE
CASTNet
STN
IMPROVE
CASTNet
STN
IMPROVE
STN
IMPROVE
East
West
East
West
East
West
East
West
East
West
East
West
East
West
East
West
East
West
East
West
East
West
East
West
Fractional Bias
(%)
-12
-51
8
14
8
-32
7
-2
-2
-35
-12
-92
-32
-44
16
-60
32
-20
-23
-13
-3
-31
-10
51
Fractional Error
(%)
44
64
43
57
45
52
40
49
24
50
86
115
107
115
81
105
63
67
51
66
75
66
52
76
• summer sulfate is in the range of-10 percent to +30 percent for fractional bias
and 35 percent to 50 percent for fractional error and
• winter nitrate is in the range of+50 percent to +70 percent for fractional bias and
85 percent to 105 percent for fractional error.
Thus, CMAQ is considered appropriate for use in projecting changes in future year PM2 5
concentrations and the resultant health/economic benefits due to the emissions reductions.
3-11
-------
Table 3-5. Selected Performance Evaluation Statistics from the CMAQ 2001
Simulation
Eastern United States
Sulfate
(Summer)
Nitrate
(Winter)
STN
IMPROVE
CASTNet
STN
IMPROVE
CMAQ 2001
Fractional Bias (%)
14
10
3
15
21
Fractional Error (%)
44
42
22
73
92
3.2.1.5 Converting CMAQ Outputs to Benefits Inputs
CMAQ generates predictions of hourly PM species concentrations for every grid.
The species include a primary coarse fraction (corresponding to PM in the 2.5 to 10 micron
size range), a primary fine fraction (corresponding to PM less than 2.5 microns in diameter),
and several secondary particles (e.g., sulfates, nitrates, and organics). PM2 5 is calculated as
the sum of the primary fine fraction and all of the secondarily formed particles. Future-year
estimates of PM2 5 were calculated using relative reduction factors (RRFs) applied to 2002
ambient PM25 and PM25 species concentrations. A gridded field of PM2 5 concentrations was
created by interpolating Federal Reference Monitor ambient data and IMPROVE ambient
data. Gridded fields of PM25 species concentrations were created by interpolating EPA
speciation network (ESPN) ambient data and IMPROVE data. The ambient data were
interpolated to the CMAQ 36 km grid.
The procedures for determining the RRFs are similar to those in EPA's draft
guidance for modeling the PM2 5 standard (EPA, 2000). This guidance has undergone
extensive peer review and is anticipated to be finalized this year. The guidance recommends
that model predictions be used in a relative sense to estimate changes expected to occur in
each major PM25 species. The procedure for calculating future-year PM2 5 design values is
called the "Speciated Modeled Attainment Test (SMAT)." EPA used this procedure to
estimate the ambient impacts of the CAIR NPR emissions controls. The SMAT procedures
for the No Further Remediation (NFR) have been revised. Full documentation of the revised
SMAT methodology is contained in the Air Quality Modeling TSD.
3-12
-------
The revised SMAT uses an FRM mass construction methodology that results in
reduced nitrates (relative to the amount measured by routine speciation networks, such as
ESPN), higher mass associated with sulfates (reflecting water included in FRM
measurements), and a measure of organic carbonaceous mass that is derived from the
difference between measured PM2 5 and its noncarbon components. This characterization of
PM2 5 mass also reflects crustal material and other minor constituents. The resulting
characterization provides a complete mass balance. It does not have any unknown mass that
is sometimes presented as the difference between measured PM2 5 mass and the characterized
chemical components derived from routine speciation measurements. The revised SMAT
methodology uses the following PM2 5 species components: sulfates, nitrates, ammonium,
organic carbon mass, elemental carbon, crustal, water, and blank mass (a fixed value of 0.5
ug/m3). In each grid cell, the PM2 5 component species mass adds up to interpolated PM2 5
mass.
For the purposes of projecting future PM25 concentrations for input to the benefits
calculations, we applied the SMAT procedure using the base-year 2001 modeling scenario
and each of the future-year scenarios. In our application of SMAT we used temporally
scaled speciated PM2 5 monitor data from 2002 as the set of base-year measured
concentrations. Temporal scaling is based on ratios of model-predicted future case PM25
species concentrations to the corresponding model-predicted 2001 concentrations. Output
files from this process include both quarterly and annual mean PM2 5 mass concentrations,
which are then manipulated within SAS to produce a BenMAP input file containing 364
daily values (created by replicating the quarterly mean values for each day of the appropriate
season).
The SMAT procedures documented in the Air Quality Modeling TSD are applicable
for projecting future nonattainment counties and downwind receptor areas for the transport
analysis. Those procedures are the same as those performed for the PM benefits analysis
with the following exceptions:
1) The benefits analysis uses interpolated PM25 data that cover all of the grid
cells in the modeling domain (covering the entire country), whereas the
nonattainment analysis is performed at each ambient monitoring site in the
East using measured PM2 5 data (only the species data are interpolated).
3-13
-------
2) The benefits analysis is anchored by the interpolated PM2 5 data from the
single year of 2002, whereas the nonattainment analysis uses a 5-year
weighted average (1999-2003) of PM25 design values at each monitoring site.
3.2.1.6 PMAir Quality Results
Table 3-6 summarizes the projected PM2 5 concentrations for the 2010 and 2015 base
cases and changes associated with the rule. The table includes the annual mean
concentration averaged across all model grid cells in the East and West,2 separately, along
with the average change between base and control concentrations. We also provide the
population-weighted average that better reflects the baseline levels and predicted changes for
more populated areas of the nation. This measure, therefore, better reflects the potential
benefits of these predicted changes through exposure changes to these populations. As
shown, the average annual mean concentrations of PM25 across populated eastern U.S. grid
cells declines by roughly 7.1 percent (or 0.73 |ig/m3) and 8.7 percent (or 0.89 |ig/m3) in 2010
and 2015, respectively. The population-weighted average mean concentration declined by
8.1 percent (or 0.96 |ig/m3) in 2010 and 9.8 percent (or 1.15 |ig/m3) in 2015, and this change
is larger in absolute terms than the spatial average for both years. This indicates the rule
generates greater absolute air quality improvements in more populated urban areas.
Table 3-7 provides information on the populations in 2010 and 2015 that will
experience improved PM air quality. Significant populations live in areas with meaningful
reductions in annual mean PM25 concentrations resulting from the rule. As shown, in 2015,
almost 63 percent of the U.S. population located in the eastern 37-state modeling domain is
predicted to experience reductions of greater than 0.5 |ig/m3. This is an increase from the
54 percent of the U.S. population that is expected to experience such reductions in 2010.
Furthermore, over 40 percent of this population will benefit from reductions in annual mean
PM2 5 concentrations of greater than 1 |ig/m3, and almost 23 percent will live in areas with
reductions of greater than 1.5 |ig/m3.
2For the purpose of this analysis "East" is defined as the U.S. portion of the modeling domain east of 100
degrees longitude, and similarly "West" is defined as the U.S. portion of the domain west of 100 degrees
longitude.
3-14
-------
Table 3-6. Summary of Base Case PM Air Quality and Changes Due to Clean Air
Interstate Rule: 2010 and 2015
Region
-------
Table 3-7. Distribution of PM2 5 Air Quality Improvements Over Population Due to
Clean Air Interstate Rule: 2010 and 2015
Change in Annual Mean
PM2 5 Concentrations (jig/m3)3
A PM2 5 Cone < 0.25
0.25 > A PM2 5 Cone < 0.5
0.5 > A PM25 Cone < 0.75
0.75>APM25Conc < 1.0
1.0>APM25Conc < 1.25
1.25>APM25Conc < 1.5
1.5>APM25Conc < 1.75
1.75>APM25Conc < 2.0
A PM2 , Cone > 2.0
2010
Number
(millions)
81.7
58.4
36.2
24.0
41.1
21.8
14.5
8.9
17.7
Population13
Percent (%)
26.8%
19.2%
11.9%
7.9%
13.5%
7.2%
4.8%
2.9%
5.8%
2015
Number
(millions)
80.8
31.9
54.8
26.1
16.3
37.5
26.4
16.8
26.2
Population
Percent (%)
26.5%
10.5%
18.0%
8.6%
5.4%
12.3%
8.7%
5.5%
8.6%
a The change is defined as the control case value minus the base case value.
b Population counts and percentages are for the fraction of the continental U.S. population located in the
modeling domain considered in modeling health benefits for the rule.
we use the relative predictions from the model by combining the 2001 base-year and each
future-year scenario with current ambient air quality observations to determine the expected
change in 2010 or 2015 ozone concentrations due to the rule. These results are used solely in
the benefits analysis.
The CAMx modeling system requires a variety of input files that contain information
pertaining to the modeling domain and simulation period. These include gridded, day-
specific emissions estimates and meteorological fields, initial and boundary conditions, and
land-use information. As applied to the Eastern United States, the model segments the area
into square blocks called grids (roughly equal in size to counties), each of which has several
layers of air conditions that are considered in the analysis. Using these data, the CAMx
model generates predictions of hourly ozone concentrations for every grid. We used the
results of this process to develop 2010 and 2015 ozone profiles at monitor sites by
normalizing the CAMx predictions to the observed ozone concentrations at each monitor
site. For areas (grids) without ozone monitoring data, we interpolated ozone values using
data from monitors surrounding the area. After completing this process, we calculated daily
3-16
-------
and seasonal ozone metrics to be used as inputs to the health and welfare C-R functions of
the benefits analysis. The following sections provide a more detailed discussion of each of
the steps in this evaluation and a summary of the results.
3.2.2.1 Modeling Domain
The modeling domain representing the Eastern United States is the same as that used
previously for Ozone Transport Assessment Group and for the On-highway Tier-2
rulemaking. As shown in Figure 3-2, this domain encompasses most of the Eastern United
States from the East coast to mid-Texas and consists of two grids with differing resolutions.
The modeling domain extends from 99 degrees to 67 degrees west longitude and from 26
degrees to 47 degrees north latitude. The inner portion of the modeling domain shown in
Figure 3-2 uses a relatively fine grid of approximately 12 km consisting of nine vertical
layers. The outer area has less horizontal resolution. The grid cell size in the outer grid is
approximately 36 km with the same nine vertical layers. The vertical height of the modeling
domain is 4,000 meters above ground level for both areas.
3.2.2.2 Simulation Periods
For use in this benefits analysis, the simulation periods modeled by CAMx included
several multiday periods when ambient measurements recorded high ozone concentrations.
A simulation period, or episode, consists of meteorological data characterized over a block of
days that are used as inputs to the air quality model. We modeled three periods during the
summer of 1995: June 12-24, July 5-15, and August 7-21. Collectively, these periods
contain episodes of high ozone in various portions of the East. The six emissions scenarios
(1995 base year, 2001 base year, 2010 base and control, 2015 base and control) were
simulated for all three episodes. The periods modeled include three "ramp-up" days to
initialize the model, but the results for these days are not used in this analysis.
3.2.2.3 Nonemissions Modeling Inputs
The meteorological data required for input into CAMx (e.g., wind, temperature,
vertical mixing) were developed by separate meteorological models. The gridded
meteorological data for the three historical 1995 episodes were developed using the Regional
Atmospheric Modeling System (RAMS), version 3b. This model provided needed data at
every grid cell on an hourly basis. These meteorological modeling results were evaluated
against observed weather conditions before being input into CAMx, and it was concluded
that the model fields were adequate representations of the historical meteorology. A more
3-17
-------
Figure 3-2. CAMx Eastern U.S. Modeling Domain
Note: The inner area represents fine grid modeling at 12 km resolution. The outer area represents the coarse
grid modeling at 36 km resolution.
detailed description of the settings and assorted input files employed in these applications is
provided in the Air Quality Modeling TSD.
The modeling assumed background pollutant levels at the top and along the periphery
of the domain. Initial conditions were also assumed to be relatively clean. Given the ramp-
up days and the expansive domains, it is expected that these assumptions will not affect the
modeling results. The development of model inputs is discussed in greater detail in the Air
Quality Modeling TSD, which is available in the docket for this rule.
3.2.2.4 Model Performance for Photochemical Ozone
A performance evaluation of CAMx for the three 1995 episodes was conducted prior
to CAIR, in support of the Nonroad Diesel Engine Rule. A summary of model performance
3-18
-------
from the study is provided here. In this analysis, a series of performance statistics was
calculated for the Eastern U.S. domain as well as the four quadrants of this domain and
multiple subregions. The model performance evaluation consisted solely of comparisons
against ambient surface ozone data.
Three primary statistical metrics were used to assess the overall accuracy of the base-
year modeling simulations:
• Mean normalized bias is defined as the average difference between the hourly
model predictions and observations (paired in space and time) at each monitoring
location, normalized by the magnitude of the observations.
• Mean normalized gross error is defined as the average absolute difference
between the hourly model predictions and observations (paired in space and time)
at each monitoring location, normalized by the magnitude of the observations.
• Average accuracy of the peak is defined as the average difference between peak
daily model predictions and observations at each monitoring location, normalized
by the magnitude of the observations.
In general, the model tends to slightly underestimate observed ozone. When all
hourly observed ozone values greater than a 60 ppb threshold are compared to their model
counterparts for the 30 episode modeling days in the eastern domain, the mean normalized
bias is -1.1 percent and the mean normalized gross error is 20.5 percent. As shown in
Table 3-8, the model generally underestimates observed ozone values for the June and July
episodes but predicts higher than observed amounts for the August episode.
Table 3-8. Model Performance Statistics for Hourly Ozone in the Eastern U.S. CAMx
Ozone Simulations: 1995 Base Case
Episode
June 1995
July 1995
August 1995
Average Accuracy of
the Peak
-7.3
-3.3
9.6
Mean Normalized
Bias
-8.8
-5.0
8.6
Mean Normalized
Gross Error
19.6
19.1
23.3
3-19
-------
At present, there are no guidance criteria by which one can determine if a regional
ozone modeling exercise is exhibiting adequate model performance. These base-case
simulations were determined to be acceptable based on comparisons to previously completed
model rulemaking analyses (e.g., OTAG, Tier-2, and Heavy-Duty Engine). The modeling
completed for this rule exhibits less bias and error than any past regional ozone modeling
application done by EPA. Thus, the model is considered appropriate for use in projecting
changes in future-year ozone concentrations and the resultant health/economic benefits due
to the emissions reductions.
In addition, the CAMx modeling results were also evaluated at a "local" level to
ensure that areas determined to need the emissions reductions based on projected
exceedances of the ozone standard were not unduly influenced by local overestimation of
ozone in the model base year. As detailed in the Air Quality Modeling TSD, performance
statistics were computed for each of 51 local subregions within the modeling domain. These
performance statistics were compared to the recommended performance ranges for urban
attainment modeling (EPA, 1999). The results indicate that model performance for the June
episode was within the recommended ranges for 69 percent of the local areas examined. For
the July and August episodes, the percentage of local areas with performance within the
recommended ranges was 80 percent and 61 percent, respectively.
3.2.2.5 Converting CAMx Outputs to Full-Season Profiles for Benefits Analysis
This study extracted hourly, surface-layer ozone concentrations for each grid cell
from the standard CAMx output file containing hourly average ozone values. These model
predictions are used in conjunction with the observed concentrations obtained from the
Aerometric Information Retrieval System (AIRS) to generate ozone concentrations for the
entire ozone season.3'4 The predicted changes in ozone concentrations from the future-year
base case to future-year control scenario serve as inputs to the health and welfare C-R
functions of the benefits analysis (i.e., the Environmental Benefits Mapping and Analysis
Program [BenMAP]).
3The ozone season for this analysis is defined as the 5-month period from May to September; however, to
estimate certain crop yield benefits, the modeling results were extended to include months outside the 5-
month ozone season.
"Based on AIRS, there were 961 ozone monitors with sufficient data (i.e., 50 percent or more days reporting at
least nine hourly observations per day [8 am to 8 pm] during the ozone season).
3-20
-------
To estimate ozone-related health and welfare effects for the contiguous United States,
full-season ozone data are required for every BenMAP grid-cell. Given available ozone
monitoring data, we generated full-season ozone profiles for each location in the contiguous
48 states in two steps: (1) we combined monitored observations and modeled ozone
predictions to interpolate hourly ozone concentrations to a grid of 8 km by 8 km population
grid cells, and (2) we converted these full-season hourly ozone profiles to an ozone measure
of interest, such as the daily average.5'6
3.2.2.6 Ozone Air Quality Results
This section provides a summary of the predicted ambient ozone concentrations from
the CAMx model for the 2010 and 2015 base cases and changes associated with the rule.
Table 3-9 provides those ozone metrics for grid cells in the Eastern United States that enter
the C-R functions for health benefits endpoints. The population-weighted average reflects
the baseline levels and predicted changes for more populated areas of the nation. This
measure, therefore, will better reflect the potential benefits of these predicted changes
through exposure changes to these populations.
3.2.3 Visibility Degradation Estimates
Visibility degradation is often directly proportional to decreases in light transmittal in
the atmosphere. Scattering and absorption by both gases and particles decrease light
transmittance. To quantify changes in visibility, our analysis computes a light-extinction
coefficient, based on the work of Sisler (1996), which shows the total fraction of light that is
decreased per unit distance. This coefficient accounts for the scattering and absorption of
light by both particles and gases and accounts for the higher extinction efficiency of fine
particles compared to coarse particles. Fine particles with significant light-extinction
efficiencies include sulfates, nitrates, organic carbon, elemental carbon (soot), and soil
(Sisler, 1996).
5The 8-km grid squares contain the population data used in the health benefits analysis model, BenMAP. See
Chapter 4 for a discussion of this model.
6This approach is a generalization of planar interpolation that is technically referred to as enhanced Voronoi
Neighbor Averaging (EVNA) spatial interpolation.
3-21
-------
Table 3-9. Summary of CAMx Derived Population-Weighted Ozone Air Quality
Metrics for Health Benefits Endpoints Due to Clean Air Interstate Rule: Eastern U.S.
Statistic3
Population-Weighted Average (ppb)d
Daily 1 -Hour Maximum Concentration
Daily 8-Hour Average Concentration
Daily 12-Hour Average Concentration
Daily 24-Hour Average Concentration
Base
Case
51.82
42.59
40.09
30.15
2010
Change"
-0.50
-0.39
-0.37
-0.27
Percent
Change"
-1.0%
-1.0%
-1.0%
-1.0%
2015
Base
Case
50.77
41.84
39.41
29.73
Change"
-1.36
-1.05
-0.97
-0.67
Percent
Change"
-2.7%
-2.5%
-2.5%
-2.3%
a These ozone metrics are calculated at the CAMX grid-cell level for use in health effects estimates based on
the results of spatial and temporal Voronoi Neighbor Averaging. Except for the daily 24-hour average,
these ozone metrics are calculated over relevant time periods during the daylight hours of the "ozone
season" (i.e., May through September). For the 8-hour average, the relevant time period is 9 am to 5 pm,
and for the 12-hour average it is 8 am to 8 pm.
b The change is defined as the control-case value minus the base-case value. The percentage change is the
"Change" divided by the "Base Case" and then multiplied by 100 to convert the value to a percentage.
d Calculated by summing the product of the projected CAMx grid-cell population and the estimated CAMx
grid cell seasonal ozone concentration and then dividing by the total population.
Based on the light-extinction coefficient, we also calculated a unitless visibility
index, called a "deciview," which is used in the valuation of visibility. The deciview metric
provides a scale for perceived visual changes over the entire range of conditions, from clear
to hazy. Under many scenic conditions, the average person can generally perceive a change
of one deciview. The higher the deciview value, the worse the visibility. Thus, an
improvement in visibility is a decrease in deciview value.
Table 3-10 provides visibility improvements expected to occur in specific parks in
the CAIR region. As shown, major parks in the Eastern United States, including the Great
Smokey Mountains and Shenandoah, are expected to see significant improvements in
visibility. By 2015, on the 20 percent of worst visibility days, the Great Smokey Mountains
National Park is expected to see improvements of over 2.5 deciviews (9 percent), and
Shenandoah National Park is expected to see improvements of over 3.3 deciviews (12
percent). Under average light conditions, these represent improvements in visual range by
close to 7 miles in the Great Smokies and over 10 miles in Shenandoah.
3-22
-------
Table 3-10. Summary of Deciview Visibility Impacts at Class I Areas in the CAIR Region
a,b
Federal Class I Area
Acadia, ME
Boundary Waters Canoe Area,
MN
Brigantine, NJ
Caney Creek, AR
*Chassahowitzka, FL
*Dolly Sods, WV
*Everglades, FL
Great Gulf, NH
* Great Smoky Mountains, TN
Isle Royale, MI
*James River Face, VA
* Joyce Kilmer — Slickrock,
TN
*Linville Gorge, NC
Lye Brook, VT
*Mammoth Cave, KY
Mingo, MO
Moosehorn, ME
*Okefenokee, GA
2010
Change in
Average of 20%
Worst Davs
0.88
0.26
1.88
1.08
0.90
2.39
0.42
1.32
1.85
0.31
2.09
1.85
1.71
1.76
1.68
0.82
0.82
0.99
Percent Change
in Average of
20% Worst Davs
3.98
1.36
6.90
4.30
3.76
9.03
2.13
5.87
6.40
1.42
7.55
6.40
6.22
7.20
5.61
2.97
3.83
3.87
Change in
Annual
Average
0.36
0.12
0.93
0.42
0.47
1.44
0.17
0.44
0.97
0.17
1.21
0.97
1.00
0.61
0.94
0.50
0.30
0.63
Percent Change
in Annual
Average
2.59
0.95
4.59
2.17
2.45
7.41
1.16
3.10
4.65
1.29
5.93
4.65
5.18
4.31
4.20
2.41
2.10
3.21
2015
Change in
Average of 20%
Worst Davs
1.00
0.29
2.07
1.32
1.66
2.75
0.49
1.56
2.61
0.38
2.45
2.61
2.14
2.10
2.45
0.95
0.92
1.44
Percent Change
in Average of
20% Worst Davs
4.54
1.52
7.54
5.24
6.92
10.54
2.43
6.95
9.12
1.76
8.96
9.12
7.92
8.64
8.31
3.46
4.30
5.64
Change in
Annual
Average
0.42
0.12
1.04
0.54
1.03
1.68
0.21
0.50
1.36
0.18
1.41
1.36
1.28
0.72
1.32
0.58
0.33
0.93
Percent Change
in Annual
Average
2.94
0.94
4.97
2.80
5.38
8.71
1.35
3.53
6.58
1.41
6.94
6.58
6.65
5.07
5.96
2.85
2.30
4.70
(continued)
> oo
5-23
-------
Table 3-10. Summary of Deciview Visibility Impacts at Class I Areas in the CAIR Region (continued)
Federal Class I Area
*Otter Creek, WV
Presidential Range - Dry, NH
*Cape Remain, SC
Roosevelt Campobello, ME
Seney, MI
*Shenandoah, VA
*Sipsey, AL
*Swanquarter, NC
Upper Buffalo, AR
Voyageurs, MN
* Wolf Island. GA
2010
Change in
Average of 20%
Worst Davs
2.47
1.50
1.01
0.80
0.65
2.81
1.45
1.45
0.67
0.12
0.84
Percent Change
in Average of
20% Worst Davs
9.22
6.70
4.17
3.76
2.64
10.23
5.27
6.04
2.74
0.68
3.28
Change in
Annual
Average
1.45
0.52
0.77
0.29
0.23
1.69
0.81
0.80
0.31
0.09
0.60
Percent Change
in Annual
Average
7.43
3.75
4.07
2.03
1.67
8.54
3.83
4.39
1.72
0.74
3.06
2015
Change in
Average of 20%
Worst Davs
2.94
1.76
1.44
0.94
0.78
3.31
2.06
1.86
0.80
0.12
1.12
Percent Change
in Average of
20% Worst Davs
11.14
7.93
5.98
4.41
3.16
12.27
7.55
7.85
3.27
0.72
4.37
Change in
Annual
Average
1.72
0.60
1.06
0.33
0.25
2.01
1.08
1.00
0.39
0.08
0.79
Percent Change
in Annual
Average
8.90
4.28
5.58
2.30
1.84
10.24
5.11
5.54
2.17
0.65
3.99
a The change is defined as the base case value minus the control case value.
b The percent change is the "Change" divided by the "Base Case" and then multiplied by 100 to convert the value to a percentage.
* Visibility Benefits were monitized for this park.
5-24
-------
3.3 References
Byun, D., and K.L. Schere. March 2004. "Review of the Governing Equations,
Computational Algorithms, and Other Components of the Models-3 Community
Multiscale Air Quality (CMAQ) Modeling System." Submitted to the Journal of
Applied Mechanics Reviews.
Sisler, J.F. July 1996. Spatial and Seasonal Patterns and Long Term Variability of the
Composition of the Haze in the United States: An Analysis of Data from the
IMPROVE Network. Fort Collins, CO: Cooperative Institute for Research in the
Atmosphere, Colorado State University.
U.S. Environmental Protection Agency (EPA). 1999. Draft Guidance on the Use of Models
and Other Analyses in Attainment Demonstrations for the 8-Hour Ozone NAAQS,
Office of Air Quality Planning and Standards, Research Triangle Park, NC.
U.S. Environmental Protection Agency (EPA). 2000. Draft Guidance for Demonstrating
Attainment of the Air Quality Goals for PM25 and Regional Haze; Draft 1.1, Office
of Air Quality Planning and Standards, Research Triangle Park, NC.
U.S. Environmental Protection Agency (EPA). 2005. Clean Air Interstate Rule Air Quality
Modeling Technical Support. Office of Air Quality Planning and Standards.
Research Triangle Park, N.C.
U.S. Environmental Protection Agency (EPA). 2005. Clean Air Interstate Rule Emission
Inventory Technical Support Document. Office of Air Quality Planning and
Standards. Research Triangle Park, NC.
3-25
-------
CHAPTER 4
BENEFITS ANALYSIS AND RESULTS
This chapter reports EPA's analysis of a subset of the public health and welfare
impacts and associated monetized benefits to society of CAIR. EPA is required by
Executive Order (E.O.) 12866 to estimate the benefits and costs of major new pollution
control regulations. Accordingly, the analysis presented here attempts to answer three
questions: (1) what are the physical health and welfare effects of changes in ambient air
quality resulting from reductions in precursors to particulate matter (PM) including NOX and
SO2 emissions? (2) what is the monetary value of the changes in these effects attributable to
the final rule? and (3) how do the monetized benefits compare to the costs? It constitutes one
part of EPA's thorough examination of the relative merits of this regulation.
The analysis presented in this chapter uses a methodology generally consistent with
benefits analyses performed for the recent analysis of Nonroad Diesel Engines Tier 4
Standards and the proposed Clear Skies Act of 2003 (EPA, 2003c). The benefits analysis
relies on three major modeling components:
1) Calculation of the impact of CAIR on EGUs assuming a cap-and-trade program
based on the national inventory of precursors to PM, specifically NOX and SO2.
2) Air quality modeling for 2010 and 2015 to determine changes in ambient
concentrations of ozone and PM, reflecting baseline and postcontrol emissions
inventories.
3) A benefits analysis to determine the changes in human health and welfare, both in
terms of physical effects and monetary value, that result from the projected
changes in ambient concentrations of various pollutants for the modeled
standards.
A wide range of human health and welfare effects are linked to the emissions of NOX
and SO2 from EGUs and the resulting impact on ambient concentrations of ozone and PM.
Potential human health effects associated with PM2 5 range from premature mortality to
morbidity effects linked to long-term (chronic) and shorter-term (acute) exposures (e.g.,
respiratory and cardiovascular symptoms resulting in hospital admissions, asthma
4-1
-------
exacerbations, and acute and chronic bronchitis [CB]). Exposure to ozone has also been
linked to a variety of respiratory effects including hospital admissions and illnesses resulting
in school absences. Some studies, including a recent multi-city analysis of 95 major U.S.
urban areas (Bell et al., 2004), have linked short term ozone exposures with premature
mortality.1 Welfare effects potentially linked to PM include materials damage and visibility
impacts, while ozone can adversely affect the agricultural and forestry sectors by decreasing
yields of crops and forests. Although methods exist for quantifying the benefits associated
with many of these human health and welfare categories, not all can be evaluated at this time
because of limitations in methods and/or data. Table 4-1 summarizes the annual monetized
health and welfare benefits associated with CAIR for 2 years, 2010 and 2015. Table 4-2 lists
the full complement of human health and welfare effects associated with PM and ozone and
identifies those effects that are quantified for the primary estimate and those that remain
unquantified because of current limitations in methods or available data.
Figure 4-1 illustrates the major steps in the benefits analysis. Given baseline and
post-control emissions inventories for the emission species expected to affect ambient air
quality, we use sophisticated photochemical air quality models to estimate baseline and post-
control ambient concentrations of ozone and PM and deposition of nitrogen and sulfur for
each year. The estimated changes in ambient concentrations are then combined with
Table 4-1. Estimated Monetized Benefits of the Final CAIR
Total Benefits3'b (billions 1999$)
2010 2015
Using a 3 % discount rate $73.3 + B $ 101 + B
Using a 7% discount rate $62.6 + B $86.3 + B
a For notational purposes, unquantified benefits are indicated with a "B" to represent the sum of additional
monetary benefits and disbenefits. A detailed listing of unquantified health and welfare effects is provided
in Table 4-2.
b Results reflect the use of two different discount rates: 3 and 7 percent, which are recommended by EPA's
Guidelines for Preparing Economic Analyses (EPA, 2000b) and OMB Circular A-4 (OMB, 2003). Results
are rounded to three significant digits for ease of presentation and computation.
'Short-term exposure to ambient ozone has also been linked to premature death. EPA is currently evaluating the
epidemiological literature examining the relationship between ozone and premature mortality, sponsoring
three independent meta-analyses of the literature. EPA will consider including ozone mortality in primary
benefits analyses once a peer-reviewed methodology is available.
4-2
-------
Table 4-2. Human Health and Welfare Effects of Pollutants Affected by the Final CAIR
Pollutant/Effect
Quantified and Monetized in
Base Estimates3
Quantified and/or Monetized
Effects in Sensitivity Analyses
Unquantified Effects - Changes in:
Ozone/Healthb Hospital admissions: respiratory
Emergency room visits for asthma
Minor restricted-activity days
School loss days
Premature mortality: short term
exposures0
Asthma attacks
Cardiovascular emergency room
visits
Acute respiratory symptoms
Chronic respiratory damage
Premature aging of the lungs
Nonasthma respiratory emergency room visits
Increased exposure to UVb
Ozone/Welfare
Decreased outdoor worker
productivity
Yields for:
- Commercial forests
- Fruits and vegetables, and
- Other commercial and noncommercial crops
Damage to urban ornamental plants
Recreational demand from damaged forest aesthetics
Ecosystem functions
Increased exposure to UVb
(continued)
-------
Table 4-2. Human Health and Welfare Effects of Pollutants Affected by the Final CAIR (continued)
Pollutant/Effect
Quantified and Monetized in
Base Estimates3
Quantified and/or Monetized
Effects in Sensitivity Analyses
Unquantified Effects - Changes in
PM/Healthd Premature mortality based on
cohort study estimates6
Bronchitis: chronic and acute
Hospital admissions: respiratory
and cardiovascular
Emergency room visits for asthma
Nonfatal heart attacks (myocardial
infarction)
Lower and upper respiratory illness
Minor restricted-activity days
Work loss days
Asthma exacerbations (asthmatic
population)
Respiratory symptoms (asthmatic
population)
Infant mortality
PM/Welfare Visibility in Southeastern Class I
areas
Premature mortality: short term
exposuresf
Subchronic bronchitis cases
Visibility in northeastern and
Midwestern Class I areas
Household soiling
Low birth weight
Pulmonary function
Chronic respiratory diseases other than chronic bronchitis
Nonasthma respiratory emergency room visits
UVb exposure (+/-)B
Visibility in western U.S. Class I areas
Visibility in residential and non-Class I areas
UVb exposure (+/-)g
(continued)
4-4
-------
Table 4-2. Human Health and Welfare Effects of Pollutants Affected by the Final CAIR (continued)
Pollutant/Effect
Quantified and Monetized in
Base Estimates3
Quantified and/or Monetized
Effects in Sensitivity Analyses
Unquantified Effects - Changes in:
Nitrogen and
Sulfate
Deposition/
Welfare
SO,/Health
NOv/Health
Commercial forests due to acidic sulfate and nitrate
deposition
Commercial freshwater fishing due to acidic deposition
Recreation in terrestrial ecosystems due to acidic
deposition
Commercial fishing, agriculture, and forests due to
nitrogen deposition
Recreation in estuarine ecosystems due to nitrogen
deposition
Ecosystem functions
Passive fertilization
Hospital admissions for respiratory and cardiac diseases
Respiratory symptoms in asthmatics
Lung irritation
Lowered resistance to respiratory infection
Hospital admissions for respiratory and cardiac diseases
(continued)
4-5
-------
Table 4-2. Human Health and Welfare Effects of Pollutants Affected by the Final CAIR (continued)
Quantified and Monetized in Quantified and/or Monetized
Pollutant/Effect Base Estimates3 Effects in Sensitivity Analyses Unquantified Effects
Mercury Health Incidences of neurological disorders
Incidences of learning disabilities
Incidences in developmental delays
Potential cardiovascular effects'1, including:
- Altered blood pressure regulation11
- Increased heart rate variability11
- Incidences of Myocardial infarction11
Potential reproductive effects'1
Mercury Impact on birds and mammals (e.g., reproductive effects)
Deposition Impacts to commercial, subsistence, and recreational
Welfare fishing
a Primary quantified and monetized effects are those included when determining the primary estimate of total monetized benefits of CAIR. See Appendix
C for a more complete discussion of the benefit estimates.
b In addition to primary economic endpoints, there are a number of biological responses that have been associated with ozone health including increased
airway responsiveness to stimuli, inflammation in the lung, acute inflammation and respiratory cell damage, and increased susceptibility to respiratory
infection. The public health impact of these biological responses may be partly represented by our quantified endpoints.
0 Premature mortality associated with ozone is not currently included in the primary analysis. Recent evidence suggests that short-term exposures to ozone
may have a significant effect on daily mortality rates, independent of exposure to PM. EPA is currently conducting a series of meta-analyses of the ozone
mortality epidemiology literature. EPA will consider including ozone mortality in primary benefits analyses once a peer-reviewed methodology is
available.
d In addition to primary economic endpoints, there are a number of biological responses that have been associated with PM health effects including
morphological changes and altered host defense mechanisms. The public health impact of these biological responses may be partly represented by our
quantified endpoints.
e Cohort estimates are designed to examine the effects of long term exposures to ambient pollution, but relative risk estimates may also incorporate some
effects due to shorter term exposures (see Kunzli, 2001 for a discussion of this issue).
f While some of the effects of short term exposure are likely to be captured by the cohort estimates, there may be additional premature mortality from short
term PM exposure not captured in the cohort estimates included in the primary analysis.
g May result in benefits or disbenefits. See Section 5.3.4 for more details.
h These are potential effects as the literature is insufficient.
4-6
-------
INPUTS
PROCESSES
Emissions inventories
(2001 GEM, 1996 NEI,
MOBILE 5b and 6 PARTS
Model baseline and
quality (REMSAD, CAM-X)
!
!
Air quality monitoring data
AIRS (ozone), FRM (total
PM), STN (speciated PM)
J^
i
Concentration response
functions
Incidence and
health endpoints
Population and
demographic data (with
growth projections)
Valuation functions
T 1 1 ' 'I '
loU L L 'dj L oj
^I-I-VT,
i
i
i
i
i
i
i
i
i
i
i
1—
i
i
cc
Interpolation of projected air
ncentration surfaces (base and control)
•SMAT-denved •BenMAP-denved
(PM2.5) (ozone)
BenMAP
integratec
model
Model population exposure to
changes in ambient concentrations
i
human heal
i
- Estimate mon
^ 1 • 1
changes in h
i
^ Adjust monetary
i
in real income to
i
r
ted changes in
th outcomes
r
etary value of
uman health
r
values for g
year of an
TOWth ^
1 • ^ —
ilysis
r
Sum health and welfare monetary
I values to obtain total monetary benefits
I
Estm
expe
chang
welf
(visib
i
Estiri
mone
valu
chang
welf
effe
INPUTS
Figure 4-1. Key Steps in Air Quality Modeling Based Benefits Analysis
monitoring data to estimate population-level potential exposures to changes in ambient
concentrations for use in estimating health effects. Modeled changes in ambient data are also
used to estimate changes in visibility and changes in other air quality statistics that are
necessary to estimate welfare effects. Changes in population exposure to ambient air
pollution are then input to impact functions2 to generate changes in the incidence of health
2The term "impact function" as used here refers to the combination of a) an effect estimate obtained from the
epidemiological literature, b) the baseline incidence estimate for the health effect of interest in the modeled
population, c) the size of that modeled population, and d) the change in the ambient air pollution metric of
interest. These elements are combined in the impact function to generate estimates of changes in incidence
of the health effect. The impact function is distinct from the C-R function, which strictly refers to the
estimated equation from the epidemiological study relating incidence of the health effect and ambient
pollution. We refer to the specific value of the relative risk or estimated coefficients in the epidemiological
study as the "effect estimate." In referencing the functions used to generate changes in incidence of health
effects for this RIA, we use the term "impact function" rather than C-R function because "impact function"
includes all key input parameters used in the incidence calculation.
4-7
-------
effects, or changes in other exposure metrics are input to dose-response functions to generate
changes in welfare effects. The resulting effects changes are then assigned monetary values,
taking into account adjustments to values for growth in real income out to the year of
analysis (values for health and welfare effects are in general positively related to real income
levels). Finally, values for individual health and welfare effects are summed to obtain an
estimate of the total monetary value of the changes in emissions.
The benefits discussed in this chapter represent the estimates based upon emission
changes anticipated for the final CAIR program (final CAIR plus the proposal to include SO2
and annual NOX controls for New Jersey and Delaware in the final CAIR program) with one
exception. The benefits estimated in this report are slightly overstated due to the inclusion of
emission reductions for SO2 and annual NOX controls for Arkansas. Thus, the analysis
presented reflects the EPA's best estimate of the benefits for a complete CAIR program
assuming New Jersey and Delaware become a part of the CAIR region for PM25 as well as
ozone, but these benefits are slightly overstated due to use of modeling that includes
Arkansas in the CAIR region for SO2 and annual NOX controls.
On September 26, 2002, the National Research Council (NRC) released a report on
its review of the Agency's methodology for analyzing the health benefits of measures taken
to reduce air pollution. The report focused on EPA's approach for estimating the health
benefits of regulations designed to reduce concentrations of ambient PM.
In its report, the NRC said that EPA has generally used a reasonable framework for
analyzing the health benefits of PM-control measures. It recommended, however, that the
Agency take a number of steps to improve its benefits analysis. The current analysis reflects
the following suggestions of that NRC report:
• estimate benefits for intervals, such as every 5 years, rather than a single year;
• clearly state the projected baseline statistics used in estimating health benefits,
including those for air emissions, air quality, and health outcomes;
• when appropriate, use data from non-U.S. studies to broaden age ranges to which
current estimates apply and to include more types of relevant health outcomes;
In addition, the NRC recommended that EPA move the assessment of uncertainties
from its ancillary analyses into its base analyses by conducting probabilistic, multiple-source
uncertainty analyses. However, for this rule, given the limited data available for such a
complex uncertainty assessment, EPA made the decision only to summarize the results of an
ancillary probabilistic uncertainty analysis to provide context for sources of uncertainty
4-8
-------
reflecting statistical error in the base analysis. The EPA followed the NRC
recommendations that the probabilistic assessment should be based on available data and
expert judgment.
The NRC made a number of recommendations for improving EPA's approach and
found that the studies selected by the Agency for use in its benefits analyses were generally
reasonable choices. In particular, the NRC agreed with EPA's decision to use cohort studies
for estimating premature mortality benefits. It also concluded that the Agency's selection of
the American Cancer Society (ACS) study for the evaluation of PM-related premature
mortality was reasonable, although it noted the publication of new cohort studies that the
Agency should evaluate. Since the publication of the NRC report, EPA has reviewed new
cohort studies, including reanalyses of the ACS study data and has carefully considered these
new study data in developing the analytical approach for CAIR (see below).
In addition to the NRC report, EPA received technical guidance and input regarding
its methodology for conducting PM- and ozone-related benefits analysis from the Health
Effects Subgroup (HES) of the SAB Council reviewing the 812 blueprint (SAB-HES, 2004)
and the Office of Management and Budget (OMB) through ongoing discussions regarding
methods used in conducting regulatory impact analyses (RIAs), and developments during the
collaboration on the recent Nonroad Diesel rulemaking. EPA addressed many of the
comments received from the NRC, the SAB-HES, and OMB in developing the analytical
approach for the recent Nonroad Diesel Rule RIA. These improvements are also reflected in
this analysis for the final CAIR.
Recommendations from OMB regarding RIA methods have focused on the approach
used to characterize uncertainty in the benefits estimates generated for RIAs and the
approach used to value mortality estimates. EPA is currently developing a comprehensive
integrated strategy for characterizing the impact of uncertainty in key elements of the
benefits modeling process (e.g., emissions modeling, air quality modeling, health effects
incidence estimation, valuation) on the health impact and monetized benefits estimates that
are generated. A subset of this effort, which has recently been completed and peer reviewed,
was a pilot expert elicitation designed to characterize uncertainty in the estimation of PM-
4-9
-------
related mortality resulting from both short-term and long-term exposure.3 The peer review of
the pilot expert elicitation was generally favorable. We provide a detailed description of the
pilot in Appendix B, along with a summary of results in Section 4.3.
We have also reflected advances in data and methods in air quality modeling,
epidemiology, and economics in developing this analysis. Updates to the assumptions and
methods used in estimating PM2 5-related and ozone-related benefits since the analysis for the
proposed rule include the following:
Air Quality
• Use of CMAQ-based predictions for ambient PM2 5 and component species.
• Use of an updated SMAT approach for developing PM2 5 air modeling results.
For the CAIR proposal analysis, we used temporally scaled speciated PM2 5
monitor data from 2001-2002, reconstructed into total PM2 5 mass based on
2000-2002 design values and kriged to 12 kilometer grids (nested within the
standard 36 km REMSAD grid structure). Temporal scaling was based on ratios
of future modeled REMSAD data to 2001 REMSAD model data, using REMSAD
modeling conducted at the 36 km grid resolution. For this analysis of the final
rule, we used a modified method that is based on the future to 2001 modeled
CMAQ speciated outputs and spatially interpolated speciated monitor data (see
Chapter 3 for more details).
• The CAIR proposal analysis was limited to the Eastern U.S. For the final rule
analysis, PM benefits are estimated for the entire U.S., to account for the transport
of PM precursor emissions from the CAIR domain to the western states. The
ozone benefits assessment is still limited to the eastern U.S. due to limitations in
the models for ozone formation in the western states.
Valuation
In generating the monetized benefits for reductions in premature mortality in the
primary analysis, a 20-year segmented lag structure will be used to characterize
the relationship between the time when exposure to ambient PM2 5 is changed and
the time when reductions in premature mortality are expected to occur.
3Expert elicitation is a formal, highly structured and well documented process whereby expert judgments,
usually of multiple experts, are obtained (Ayyub, 2002).
4-10
-------
Uncertainty
• In addition to the primary estimate of the benefits of reduced premature mortality,
we characterize uncertainty using a probabilistic range of benefits based on
statistical uncertainty as captured in the standard errors associated with the Pope
et al (2002) epidemiological study, and model uncertainties obtained from the
pilot expert elicitation. Uncertainty in some other elements of the model are
characterized by statistical uncertainty as captured in either standard errors on
epidemiological effect estimates or variability in published estimates of valuation
estimates.
The benefits estimates generated for the final CAIR are subject to a number of
assumptions and uncertainties, which are discussed throughout this document. For example,
key assumptions underlying the primary estimate for the mortality category include the
following:
1) Inhalation of fine particles is causally associated with premature death at
concentrations near those experienced by most Americans on a daily basis.
Although biological mechanisms for this effect have not yet been completely
established, the weight of the available epidemiological and experimental
evidence supports an assumption of causality.
2) All fine particles, regardless of their chemical composition, are equally potent
in causing premature mortality. This is an important assumption, because PM
produced via transported precursors emitted from EGUs may differ
significantly from direct PM released from automotive engines and other
industrial sources. However, no clear scientific grounds exist for supporting
differential effects estimates by particle type.
3) The C-R function for fine particles is approximately linear within the range of
ambient concentrations under consideration. Thus, the estimates include
health benefits from reducing fine particles in areas with varied
concentrations of PM, including both regions that are in attainment with the
fine particle standard and those that do not meet the standard.
4) The forecasts for future emissions and associated air quality modeling are
valid. Although recognizing the difficulties, assumptions, and inherent
uncertainties in the overall enterprise, these analyses are based on
peer-reviewed scientific literature and up-to-date assessment tools, and we
believe the results are highly useful in assessing this rule.
4-11
-------
In addition to the quantified and monetized benefits summarized above, a number of
additional categories are not currently amenable to quantification or valuation. These
include reduced acid and particulate deposition damage to cultural monuments and other
materials, reduced ozone effects on forested ecosystems, and environmental benefits due to
reductions of impacts of acidification in lakes and streams and eutrophication in coastal
areas. Additionally, we have not quantified a number of known or suspected health effects
linked with PM and ozone for which appropriate health impact functions are not available or
which do not provide easily interpretable outcomes (i.e., changes in forced expiratory
volume [FEV1]). As a result, monetized benefits generated for the primary estimate may
underestimate the total benefits attributable to the final regulatory option.
Benefits estimates for the final CAIR were generated using BenMAP, a computer
program developed by EPA that integrates a number of the modeling elements used in
previous RIAs (e.g., interpolation functions, population projections, health impact functions,
valuation functions, analysis and pooling methods) to translate modeled air concentration
estimates into health effects incidence estimates and monetized benefits estimates. BenMAP
provides estimates of both the mean impacts and the distribution of impacts (more
information on BenMAP can be found at http://www.epa.gov/ttn/ecas/ benmodels.html).
In general, this chapter is organized around the steps illustrated in Figure 4-1. In
Section 4.1, we provide an overview of the data and methods that were used to quantify and
value health and welfare endpoints and discuss how we incorporate uncertainty into our
analysis. In Section 4.2, we report the results of the analysis for human health and welfare
effects (the overall benefits estimated for the final CAIR are summarized in Table 4-1).
Details on the emissions inventory and air modeling are presented in Chapter 3.
4.1 Benefit Analysis—Data and Methods
Given changes in environmental quality (ambient air quality, visibility, nitrogen, and
sulfate deposition), the next step is to determine the economic value of those changes. We
follow a "damage-function" approach in calculating total benefits of the modeled changes in
environmental quality. This approach estimates changes in individual health and welfare
endpoints (specific effects that can be associated with changes in air quality) and assigns
values to those changes assuming independence of the individual values. Total benefits are
calculated simply as the sum of the values for all nonoverlapping health and welfare
endpoints. This imposes no overall preference structure and does not account for potential
income or substitution effects (i.e., adding a new endpoint will not reduce the value of
changes in other endpoints). The "damage-function" approach is the standard approach for
4-12
-------
most cost-benefit analyses of environmental quality programs and has been used in several
recent published analyses (Banzhaf et al., 2002; Levy et al., 2001; Levy et al., 1999; Ostro
and Chestnut, 1998).
To assess economic value in a damage-function framework, the changes in
environmental quality must be translated into effects on people or on the things that people
value. In some cases, the changes in environmental quality can be directly valued, as is the
case for changes in visibility. In other cases, such as for changes in ozone and PM, a health
and welfare impact analysis must first be conducted to convert air quality changes into
effects that can be assigned dollar values.
For the purposes of this RIA, the health impacts analysis is limited to those health
effects that are directly linked to ambient levels of air pollution and specifically to those
linked to ozone and PM. There may be other, indirect health impacts associated with
implementing emissions controls, such as occupational health impacts for coal miners.
These impacts may be positive or negative, but in general, for this set of control options, they
are expected to be small relative to the direct air pollution-related impacts.
The welfare impacts analysis is limited to changes in the environment that have a
direct impact on human welfare. For this analysis, we are limited by the available data to
examine impacts of changes in visibility. We also provide qualitative discussions of the
impact of changes in other environmental and ecological effects, for example, changes in
deposition of nitrogen and sulfur to terrestrial and aquatic ecosystems, but we are unable to
place an economic value on these changes.
We note at the outset that EPA rarely has the time or resources to perform extensive
new research to measure either the health outcomes or their values for this analysis. Thus,
similar to Kunzli et al. (2000) and other recent health impact analyses, our estimates are
based on the best available methods of benefits transfer. Benefits transfer is the science and
art of adapting primary research from similar contexts to obtain the most accurate measure of
benefits for the environmental quality change under analysis. Adjustments are made for the
level of environmental quality change, the sociodemographic and economic characteristics of
the affected population, and other factors to improve the accuracy and robustness of benefits
estimates.
4-13
-------
4.1.1 Valuation Concepts
In valuing health impacts, we note that reductions in ambient concentrations of air
pollution generally lower the risk of future adverse health effects by a fairly small amount for
a large population. The appropriate economic measure is willingness to pay4 (WTP) for
changes in risk prior to the regulation (Freeman, 1993).5 Adoption of WTP as the measure of
value implies that the value of environmental quality improvements depends on the
individual preferences of the affected population and that the existing distribution of income
(ability to pay) is appropriate. For some health effects, such as hospital admissions, WTP
estimates are generally not available. In these cases, we use the cost of treating or mitigating
the effect as the measure of benefits. These cost of illness (COI) estimates generally
understate the true value of reductions in risk of a health effect, because they do not include
the value of avoided pain and suffering from the health effect (Harrington and Portney, 1987;
Bergeretal., 1987).
One distinction in environmental benefits estimation is between use values and
nonuse values. Although no general agreement exists among economists on a precise
distinction between the two (see Freeman [1993]), the general nature of the difference is
clear. Use values are those aspects of environmental quality that affect an individual's
welfare directly. These effects include changes in product prices, quality, and availability;
changes in the quality of outdoor recreation and outdoor aesthetics; changes in health or life
expectancy; and the costs of actions taken to avoid negative effects of environmental quality
changes.
Nonuse values are those for which an individual is willing to pay for reasons that do
not relate to the direct use or enjoyment of any environmental benefit but might relate to
existence values and bequest values. Nonuse values are not traded, directly or indirectly, in
markets. For this reason, measuring nonuse values has proven to be significantly more
4For many goods, WTP can be observed by examining actual market transactions. For example, if a gallon of
bottled drinking water sells for $1, it can be observed that at least some people are willing to pay $1 for such
water. For goods not exchanged in the market, such as most environmental "goods," valuation is not as
straightforward. Nevertheless, a value may be inferred from observed behavior, such as sales and prices of
products that result in similar effects or risk reductions (e.g., nontoxic cleaners or bike helmets).
Alternatively, surveys can be used in an attempt to directly elicit WTP for an environmental improvement.
5In general, economists tend to view an individual's WTP for an improvement in environmental quality as the
appropriate measure of the value of a risk reduction. An individual's willingness to accept (WTA)
compensation for not receiving the improvement is also a valid measure. However, WTP is generally
considered to be a more readily available and conservative measure of benefits.
4-14
-------
difficult than measuring use values. The air quality changes produced by CAIR cause
changes in both use and nonuse values, but the monetary benefits estimates are almost
exclusively for use values.
More frequently than not, the economic benefits from environmental quality changes
are not traded in markets, so direct measurement techniques cannot be used. There are three
main nonmarket valuation methods used to develop values for endpoints considered in this
analysis: stated preference (or contingent valuation [CV]), indirect market (e.g., hedonic
wage), and avoided cost methods.
The stated preference or CV method values endpoints by using carefully structured
surveys to ask a sample of people what amount of compensation is equivalent to a given
change in environmental quality. There is an extensive scientific literature and body of
practice on both the theory and technique of stated preference-based valuation. Well-
designed and well-executed stated preference studies are valid for estimating the benefits of
air quality regulations.6 Stated preference valuation studies form the basis for valuing a
number of health and welfare endpoints, including the value of mortality risk reductions, CB
risk reductions, minor illness risk reductions, and visibility improvements.
Indirect market methods can also be used to infer the benefits of pollution reduction.
The most important application of this technique for our analysis is the calculation of the
VSL for use in estimating benefits from mortality risk reductions. No market exists where
changes in the probability of death are directly exchanged. However, people make decisions
about occupation, precautionary behavior, and other activities associated with changes in the
risk of death. By examining these risk changes and the other characteristics of people's
choices, it is possible to infer information about the monetary values associated with changes
in mortality risk (see Section 4.1.5).
^Concerns about the reliability of value estimates from CV studies arose because research has shown that bias
can be introduced easily into these studies if they are not carefully conducted. Accurately measuring WTP
for avoided health and welfare losses depends on the reliability and validity of the data collected. There are
several issues to consider when evaluating study quality, including but not limited to 1) whether the sample
estimates of WTP are representative of the population WTP; 2) whether the good to be valued is
comprehended and accepted by the respondent; 3) whether the WTP elicitation format is designed to
minimize strategic responses; 4) whether WTP is sensitive to respondent familiarity with the good, to the
size of the change in the good, and to income; 5) whether the estimates of WTP are broadly consistent with
other estimates of WTP for similar goods; and 6) the extent to which WTP responses are consistent with
established economic principles.
4-15
-------
Avoided cost methods are ways to estimate the costs of pollution by using the
expenditures made necessary by pollution damage. For example, if buildings must be
cleaned or painted more frequently as levels of PM increase, then the appropriately
calculated increment of these costs is a reasonable lower-bound estimate (under most
conditions) of true economic benefits when PM levels are reduced. Avoided costs methods
are also used to estimate some of the health-related benefits related to morbidity, such as
hospital admissions (see Section 4.1.5).
4.1.2 Growth in WTP'ReflectingNationalIncome Growth Over Time
Our analysis accounts for expected growth in real income over time. Economic
theory argues that WTP for most goods (such as environmental protection) will increase if
real incomes increase. There is substantial empirical evidence that the income elasticity7 of
WTP for health risk reductions is positive, although there is uncertainty about its exact value.
Thus, as real income increases, the WTP for environmental improvements also increases.
Although many analyses assume that the income elasticity of WTP is unit elastic (i.e., a 10
percent higher real income level implies a 10 percent higher WTP to reduce risk changes),
empirical evidence suggests that income elasticity is substantially less than one and thus
relatively inelastic. As real income rises, the WTP value also rises but at a slower rate than
real income.
The effects of real income changes on WTP estimates can influence benefits
estimates in two different ways: through real income growth between the year a WTP study
was conducted and the year for which benefits are estimated, and through differences in
income between study populations and the affected populations at a particular time.
Empirical evidence of the effect of real income on WTP gathered to date is based on studies
examining the former. The Environmental Economics Advisory Committee (EEAC) of the
Science Advisory Board (SAB) advised EPA to adjust WTP for increases in real income over
time but not to adjust WTP to account for cross-sectional income differences "because of the
sensitivity of making such distinctions, and because of insufficient evidence available at
present" (EPA-SAB-EEAC-00-013). A recent advisory by another committee associated
with the SAB, the Advisory Council on Clean Air Compliance Analysis, has provided
conflicting advice. While agreeing with "the general principle that the willingness to pay to
reduce mortality risks is likely to increase with growth in real income. The same increase
7Income elasticity is a common economic measure equal to the percentage change in WTP for a 1 percent
change in income.
4-16
-------
should be assumed for the WTP for serious nonfatal health effects
(EPA-SAB-COUNCIL-ADV-04-004, p. 52)," they note that "given the limitations and
uncertainties in the available empirical evidence, the Council does not support the use of the
proposed adjustments for aggregate income growth as part of the primary analysis
(EPA-SAB-COUNCIL-ADV-04-004, p. 53)." Until these conflicting advisories have been
reconciled, EPA will continue to adjust valuation estimates to reflect income growth using
the methods described below.
Based on a review of the available income elasticity literature, we adjusted the
valuation of human health benefits upward to account for projected growth in real U.S.
income. Faced with a dearth of estimates of income elasticities derived from time-series
studies, we applied estimates derived from cross-sectional studies in our analysis. Details of
the procedure can be found in Kleckner and Neumann (1999). An abbreviated description of
the procedure we used to account for WTP for real income growth between 1990 and 2010
and 2015 is presented below.
Reported income elasticities suggest that the severity of a health effect is a primary
determinant of the strength of the relationship between changes in real income and WTP. As
such, we use different elasticity estimates to adjust the WTP for minor health effects, severe
and chronic health effects, and premature mortality. Note that because of the variety of
empirical sources used in deriving the income elasticities, there may appear to be
inconsistencies in the magnitudes of the income elasticities relative to the severity of the
effects (apriori one might expect that more severe outcomes would show less income
elasticity of WTP). We have not imposed any additional restrictions on the empirical
estimates of income elasticity. We also expect that the WTP for improved visibility in Class I
areas would increase with growth in real income. The relative magnitude of the income
elasticity of WTP for visibility compared with those for health effects suggests that visibility
is not as much of a necessity as health, thus, WTP is more elastic with respect to income.
The elasticity values used to adjust estimates of benefits in 2010 and 2015 are presented in
Table 4-3.
In addition to elasticity estimates, projections of real gross domestic product (GDP)
and populations from 1990 to 2010 and 2015 are needed to adjust benefits to reflect real per
capita income growth. For consistency with the emissions and benefits modeling, we used
national population estimates for the years 1990 to 1999 based on U.S. Census Bureau
estimates (Hollman et al., 2000). These population estimates are based on application of a
4-17
-------
Table 4-3. Elasticity Values Used to Account for Projected Real Income Growth"
Benefit Category Central Elasticity Estimate
Minor Health Effect 0.14
Severe and Chronic Health Effects 0.45
Premature mortality 0.40
Visibility 0.90
a Derivation of estimates can be found in Kleckner and Neumann (1999) and Chestnut (1997). COI estimates
are assigned an adjustment factor of 1.0.
cohort-component model applied to 1990 U.S. Census data projections (U.S. Bureau of
Census, 2000). For the years between 2000 and 2015, we applied growth rates based on the
U.S. Census Bureau projections to the U.S. Census estimate of national population in 2000.
We used projections of real GDP provided in Kleckner and Neumann (1999) for the years
1990 to 2010.8 We used projections of real GDP (in chained 1996 dollars) provided by
Standard and Poor's (2000) for the years 2010 to 2015.9
Using the method outlined in Kleckner and Neumann (1999) and the population and
income data described above, we calculated WTP adjustment factors for each of the
elasticity estimates listed in Table 4-4. Benefits for each of the categories (minor health
effects, severe and chronic health effects, premature mortality, and visibility) are adjusted by
multiplying the unadjusted benefits by the appropriate adjustment factor. Table 4-4 lists the
estimated adjustment factors. Note that, for premature mortality, we applied the income
adjustment factor to the present discounted value of the stream of avoided mortalities
occurring over the lag period. Also note that because of a lack of data on the dependence of
COI and income, and a lack of data on projected growth in average wages, no adjustments
are made to benefits based on the COI approach or to work loss days and worker
productivity. This assumption leads us to underpredict benefits in future years because it is
8U.S. Bureau of Economic Analysis, Table 2A (1992$) (available at http://www.bea.doc.gov/bea/dn/0897nip2/
tab2a.htm.) and U.S. Bureau of Economic Analysis, Economics and Budget Outlook. Note that projections
for 2007 to 2010 are based on average GDP growth rates between 1999 and 2007.
9In previous analyses, we used the Standard and Poor's projections of GDP directly. This led to an apparent
discontinuity in the adjustment factors between 2010 and 2011. We refined the method by applying the
relative growth rates for GDP derived from the Standard and Poor's projections to the 2010 projected GDP
based on the Bureau of Economic Analysis projections.
4-18
-------
Table 4-4. Adjustment Factors Used to Account for Projected Real Income Growth"
Benefit Category 2010 2015
Minor Health Effect 1.034 1.073
Severe and Chronic Health Effects 1.113 1.254
Premature Mortality 1.100 1.222
Visibility 1.239 1.581
a Based on elasticity values reported in Table 4-3, U.S. Census population projections, and projections of real
GDP per capita.
likely that increases in real U.S. income would also result in increased COI (due, for
example, to increases in wages paid to medical workers) and increased cost of work loss days
and lost worker productivity (reflecting that if worker incomes are higher, the losses
resulting from reduced worker production would also be higher).
4.1.3 Methods for Describing Uncertainty
In any complex analysis using estimated parameters and inputs from numerous
models, there are likely to be many sources of uncertainty. This analysis is no exception. As
outlined both in this and preceding chapters, many inputs were used to derive the final
estimate of benefits, including emission inventories, air quality models (with their associated
parameters and inputs), epidemiological health effect estimates, estimates of values (both
from WTP and COI studies), population estimates, income estimates, and estimates of the
future state of the world (i.e., regulations, technology, and human behavior). Each of these
inputs may be uncertain and, depending on its role in the benefits analysis, may have a
disproportionately large impact on final estimates of total benefits. For example, emissions
estimates are used in the first stage of the analysis. As such, any uncertainty in emissions
estimates will be propagated through the entire analysis. When compounded with
uncertainty in later stages, small uncertainties in emission levels can lead to large impacts on
total benefits.
Some key sources of uncertainty in each stage of the benefits analysis are the
following:
• gaps in scientific data and inquiry;
• variability in estimated relationships, such as epidemiological effect estimates,
introduced through differences in study design and statistical modeling;
4-19
-------
• errors in measurement and projection for variables such as population growth
rates;
• errors due to misspecification of model structures, including the use of surrogate
variables, such as using PM10 when PM2 5 is not available, excluded variables, and
simplification of complex functions; and
• biases due to omissions or other research limitations.
Some of the key uncertainties in the benefits analysis are presented in Table 4-5.
The NRC report on EPA's benefits analysis methodology highlighted the need for
EPA to conduct rigorous quantitative analysis of uncertainty in its benefits estimates. In
response to these comments, EPA has initiated the development of a comprehensive
methodology for characterizing the aggregate impact of uncertainty in key modeling
elements on both health incidence and benefits estimates. For this analysis of the final
CAIR, EPA has developed a limited probabilistic simulation approach based on Monte Carlo
methods to propagate the impact of a limited set of sources of uncertainty through the
modeling framework. Issues such as correlation between input parameters and the
identification of reasonable upper and lower bounds for input distributions characterizing
uncertainty in additional model elements will be addressed in future versions of the
uncertainty framework.
One component of EPA's uncertainty analysis methodology that is partially reflected
in the CAIR analysis is our work using the results of an expert elicitation to characterize
uncertainty in the effect estimates used to estimate premature mortality resulting from both
short-term and longer-term exposures to PM. This expert elicitation was aimed at evaluating
uncertainty in both the form of the mortality impact function (e.g., threshold versus linear
models) and the fit of a specific model to the data (e.g., confidence bounds for specific
percentiles of the mortality effect estimates). Additional issues, such as the ability of long-
term cohort studies to capture premature mortality resulting from short-term peak PM
exposures, were also addressed in the expert elicitation. In collaboration with OMB, EPA
completed a pilot expert elicitation which is used in the ancillary uncertainty analysis for
CAIR (as discussed in Section 4.3). Based on our experience during the pilot, EPA plans to
conduct a full-scale expert elicitation in 2005 that will provide a more robust characterization
of the uncertainty in the premature mortality function.
4-20
-------
Table 4-5. Primary Sources of Uncertainty in the Benefits Analysis
1. Uncertainties Associated with Impact Functions
— The value of the ozone or PM effect estimate in each impact function.
— Application of a single impact function to pollutant changes and populations in all locations.
— Similarity of future-year impact functions to current impact functions.
— Correct functional form of each impact function.
— Extrapolation of effect estimates beyond the range of ozone or PM concentrations observed in the source epidemiological study.
— Application of impact functions only to those subpopulations matching the original study population.
2. Uncertainties Associated with Ozone andPM Concentrations
— Responsiveness of the models to changes in precursor emissions resulting from the control policy.
— Projections of future levels of precursor emissions, especially ammonia and crustal materials.
— Model chemistry for the formation of ambient nitrate concentrations.
— Lack of ozone monitors in rural areas requires extrapolation of observed ozone data from urban to rural areas.
— Use of separate air quality models for ozone and PM does not allow for a fully integrated analysis of pollutants and their interactions.
— Full ozone season air quality distributions are extrapolated from a limited number of simulation days.
— Comparison of model predictions of particulate nitrate with observed rural monitored nitrate levels indicates that REMSAD
overpredicts nitrate in some parts of the Eastern United States
3. Uncertainties Associated with PM Mortality Risk
— Limited scientific literature supporting a direct biological mechanism for observed epidemiological evidence.
— Direct causal agents within the complex mixture of PM have not been identified.
— The extent to which adverse health effects are associated with low-level exposures that occur many times in the year versus peak
exposures.
— The extent to which effects reported in the long-term exposure studies are associated with historically higher levels of PM rather than
the levels occurring during the period of study.
— Reliability of the limited ambient PM2 5 monitoring data in reflecting actual PM2 5 exposures.
4. Uncertainties Associated with Possible Lagged Effects
— The portion of the PM-related long-term exposure mortality effects associated with changes in annual PM levels that would occur in a
single year is uncertain as well as the portion that might occur in subsequent years.
5. Uncertainties Associated with Baseline Incidence Rates
— Some baseline incidence rates are not location specific (e.g., those taken from studies) and therefore may not accurately represent the
actual location-specific rates.
— Current baseline incidence rates may not approximate well baseline incidence rates in 2015.
— Projected population and demographics may not represent well future-year population and demographics.
6. Uncertainties Associated with Economic Valuation
— Unit dollar values associated with health and welfare endpoints are only estimates of mean WTP and therefore have uncertainty
surrounding them.
— Mean WTP (in constant dollars) for each type of risk reduction may differ from current estimates because of differences in income or
other factors.
7. Uncertainties Associated with Aggregation of Monetized Benefits
— Health and welfare benefits estimates are limited to the available impact functions. Thus, unquantified or unmonetized benefits are
not included.
4-21
-------
For the final CAIR, EPA addressed key sources of uncertainty through Monte Carlo
propagation of uncertainty in the C-R functions and economic valuation functions and
through a series of sensitivity analyses examining the impact of alternate assumptions on the
benefits estimates that are generated. It should be noted that the Monte Carlo-generated
distributions of benefits reflect only some of the uncertainties in the input parameters.
Uncertainties associated with emissions, air quality modeling, populations, and baseline
health effect incidence rates are not represented in the distributions of benefits for CAIR.
Our point estimate of total benefits is uncertain because of the uncertainty in model
elements discussed above (see Table 4-5). Uncertainty about specific aspects of the health
and welfare estimation models is discussed in greater detail in the following sections. The
total benefits estimate may understate or overstate actual benefits of the rule.
In considering the monetized benefits estimates, the reader should remain aware of
the many limitations of conducting the analyses mentioned throughout this RIA. One
significant limitation of both the health and welfare benefits analyses is the inability to
quantify many of the effects listed in Table 4-1. For many health and welfare effects, such as
changes in ecosystem functions and PM-related materials damage, reliable impact functions
and/or valuation functions are not currently available. In general, if it were possible to
monetize these benefit categories, the benefits estimates presented in this analysis would
increase, although the magnitude of such an increase is highly uncertain. Unquantified
benefits are qualitatively discussed in the health and welfare effects sections. Furthermore,
EPA explores the implication of the potential relationship between O3 and premature
mortality in its sensitivity analysis. In addition to unquantified benefits, there may also be
environmental costs (disbenefits) that we are unable to quantify. These endpoints are
qualitatively discussed in the health and welfare effects sections as well. The net effect of
excluding benefit and disbenefit categories from the estimate of total benefits depends on the
relative magnitude of the effects.
Although we are not currently able to estimate the magnitude of these unquantified
and unmonetized benefits, specific categories merit further discussion. The EPA believes
there is considerable value to the public for the benefit categories that could not be
monetized. With regard to unmonetized PM-related health benefit categories listed in
Table 4-2, we feel these benefits may be small relative to those categories we were able to
quantify and monetize.
However, there is one category where new studies suggest the possibility of
significant additional economic benefits. Over the past several years, EPA's SAB has
4-22
-------
expressed the view that there were not sufficient data to show a separate ozone mortality
effect, in essence, saying that any ozone benefits are captured in the PM-related mortality
benefit estimates. However, in their most recent advice, the SAB recommended that EPA
reconsider the evidence on ozone-related mortality based on the publication of several recent
analyses that found statistically significant associations between ozone and mortality. Based
on these studies and the recommendations from the SAB, EPA has sponsored three
independent meta-analyses of the ozone-mortality epidemiology literature to inform a
determination on including this important health endpoint. The studies are complete and
have been accepted for publication in the journal Epidemiology in July 2005 [see Bell et al.,
in press; Ito et al., in press; Levy et al., in press].
The Agency believes that publication of these meta-analyses will significantly
enhance the scientific defensibility of benefits estimates for ozone, that include the benefits
of premature mortality reductions. In addition, a study published in JAMA in November
2004 also confirmed that ozone mortality impacts can be calculated separately from PM
mortality impacts (Bell et al., 2004). EPA's believes that there is sufficient evidence to
return to the SAB to confirm that these studies address their previous concerns. Using effect
estimates similar to those found in these new studies, EPA estimates the monetary value of
the ozone-related premature mortality benefits could be substantial. We estimate ozone
mortality benefits may yield roughly 500 reduced premature mortalities per year and may
increase the benefits of CAIR by approximately $3 billion annually.
In addition to unquantified and unmonetized health benefit categories, Table 4-2
shows a number of welfare benefit categories that are omitted from the monetized benefit
estimates for this rule. Only a subset of the expected visibility benefits-those for Class I
areas in the southeastern United States are included in the monetary benefits estimates we
project for this rule. We believe the benefits associated with these non-health benefit
categories are likely significant. For example, we are able to quantify significant visibility
improvements in Class I areas in the Northeast and Midwest, but are unable at present to
place a monetary value on these improvement. Similarly, we anticipate improvement in
visibility in residential areas within the CAIR region for which we are currently unable to
monetize benefits. For the Class I areas in the southeastern U.S., we estimate annual benefits
of $1.78 billion beginning in 2015 for visibility improvements. The value of visibility
benefits in areas where we were unable to monetize benefits could also be substantial.
4-23
-------
We conduct supplemental analyses related to visibility and household cleaning costs
later in this chapter. Based on these analyses, expanded coverage of these benefit categories
could increase total benefits by over $500 million. (See Appendix C for more details.)
In a recent study, Resources for the Future (RFF) estimates total benefits (i.e., the
sum of use and nonuse values) of natural resource improvements for the Adirondacks
resulting from a program that would reduce acidification in 40 percent of the lakes in the
Adirondacks of concern for acidification (Banzhaf, 2004). While the study requires further
evaluation, the RFF study does suggests that the benefits of acid deposition reductions for
CAIR could be substantial in terms of the total monetized value for ecological endpoints.
Another area of potential benefits not monetized relates to potential reductions in
nitrogen deposition from CAIR for estuaries and coastal waters within the CAIR region.
Nitrogen deposition contributes to eutrophication and water quality degradation in estuaries
and coastal waters. While we are unable to monetize the benefits of such reductions, the
Chesapeake Bay Program estimated the reduced mass of delivered nitrogen loads likely to
result from CAIR, based upon the CAIR proposal deposition estimates published in January
2004. Atmospheric deposition of nitrogen accounts for a significant portion of the nitrogen
loads to the Chesapeake with 28 percent of the nitrogen loads to the watershed coming from
air deposition. Based upon the CAIR proposal nitrogen deposition rates published in the
January 2004 proposal, the Chesapeake Bay Program finds that CAIR will likely reduce the
nitrogen loads to the Bay by 10 million pounds per year by 2010 (Sweeney, 2004). These
substantial nitrogen load reductions more than fulfill the EPA's commitment to reduce
atmospheric deposition delivered to the Chesapeake Bay by 8 million pounds annually. The
benefits of these atmospheric deposition reductions for the Bay are likely to be substantial.
4.1.4 Demographic Projections
Quantified and monetized human health impacts depend on the demographic
characteristics of the population, including age, location, and income. We use projections
based on economic forecasting models developed by Woods and Poole, Inc. The Woods and
Poole (WP) database contains county-level projections of population by age, sex, and race
out to 2025. Projections in each county are determined simultaneously with every other
county in the United States to take into account patterns of economic growth and migration.
The sum of growth in county-level populations is constrained to equal a previously
determined national population growth, based on Bureau of Census estimates (Hollman et
al., 2000). According to WP, linking county-level growth projections together and
constraining to a national-level total growth avoids potential errors introduced by forecasting
4-24
-------
each county independently. County projections are developed in a four-stage process. First,
national-level variables such as income, employment, and populations are forecasted.
Second, employment projections are made for 172 economic areas defined by the Bureau of
Economic Analysis, using an "export-base" approach, which relies on linking industrial-
sector production of nonlocally consumed production items, such as outputs from mining,
agriculture, and manufacturing with the national economy. The export-based approach
requires estimation of demand equations or calculation of historical growth rates for output
and employment by sector. Third, population is projected for each economic area based on
net migration rates derived from employment opportunities and following a cohort-
component method based on fertility and mortality in each area. Fourth, employment and
population projections are repeated for counties, using the economic region totals as bounds.
The age, sex, and race distributions for each region or county are determined by aging the
population by single year of age by sex and race for each year through 2015 based on
historical rates of mortality, fertility, and migration.
The WP projections of county-level population are based on historical population
data from 1969 through 1999 and do not include the 2000 Census results. Given the
availability of detailed 2000 Census data, we constructed adjusted county-level population
projections for each future year using a two-stage process. First, we constructed ratios of the
projected WP populations in a future year to the projected WP population in 2000 for each
future year by age, sex, and race. Second, we multiplied the block-level 2000 Census
population data by the appropriate age-, sex-, and race-specific WP ratio for the county
containing the census block for each future year. This results in a set of future population
projections that is consistent with the most recent detailed Census data.
As noted above, values for environmental quality improvements are expected to
increase with growth in real per capita income. Accounting for real income growth over
time requires projections of both real GDP and total U.S. populations. For consistency with
the emissions and benefits modeling, we used national population estimates based on the
U.S. Census Bureau projections.
4.1.5 Health Benefits Assessment Methods
The largest monetized benefits of reducing ambient concentrations of PM and ozone
are attributable to reductions in health risks associated with air pollution. EPA's Criteria
Documents for ozone and PM list numerous health effects known to be linked to ambient
concentrations of these pollutants (EPA, 1996a; 1996b). As illustrated in Figure 4-1,
quantification of health impacts requires several inputs, including epidemiological effect
4-25
-------
estimates (concentration-response functions), baseline incidence and prevalence rates,
potentially affected populations, and estimates of changes in ambient concentrations of air
pollution. Previous sections have described the population and air quality inputs. This
section describes the effect estimates and baseline incidence and prevalence inputs and the
methods used to quantify and monetize changes in the expected number of incidences of
various health effects.
4.1.5.1 Selecting Health Endpoints and Epidemiological Effect Estimates
Certain quantified health benefits of the rule may be related to ozone only, PM only,
or both pollutants. Based on the available epidemiological data, we quantified decreased
worker productivity, respiratory hospital admissions for children under two years of age, and
school absences related to ozone but not PM. The PM-only health effects we quantified
include premature mortality, nonfatal heart attacks, CB, acute bronchitis, upper and lower
respiratory symptoms, asthma exacerbations, and days of work lost. The health effects that
we quantified relate to both PM and ozone include hospital admissions, emergency room
visits for asthma, and MRADs. Although recent epidemiological evidence points to an
association between short term exposures to ozone and premature mortality, EPA is not
prepared to quantify this impact in the primary analysis for the CAIR due to the need for
additional review of the issue by the Health Effects Subcommittee of the SAB.
We relied on the published scientific literature to ascertain the relationship between
PM and ozone exposure and adverse human health effects. We evaluated studies using the
selection criteria summarized in Table 4-6. These criteria include consideration of whether
the study was peer reviewed, the match between the pollutant studied and the pollutant of
interest, the study design and location, and characteristics of the study population, among
other considerations. The selection of C-R functions for the benefits analysis is guided by
the goal of achieving a balance between comprehensiveness and scientific defensibility.
Some health effects are excluded from this analysis for three reasons: the possibility
of double-counting (such as hospital admissions for specific respiratory diseases),
uncertainties in applying effect relationships based on clinical studies to the affected
population, or a lack of an established relationship between the health effect and pollutant in
the published epidemiological literature. An improvement in ambient PM and ozone air
quality may reduce the number of incidences within each unquantified effect category that
the U.S. population would experience. Although these health effects are believed to be PM
or ozone induced, effect estimates are not available for quantifying the benefits associated
4-26
-------
Table 4-6. Summary of Considerations Used in Selecting C-R Functions
Consideration
Comments
Peer-Reviewed
Research
Study Type
Study Period
Population
Attributes
Study Size
Study Location
Pollutants
Included in
Model
Measure of PM
Economically
Valuable Health
Effects
Nonoverlapping
Endpoints
Peer-reviewed research is preferred to research that has not undergone the peer-review
process.
Among studies that consider chronic exposure (e.g., over a year or longer), prospective
cohort studies are preferred over cross-sectional studies because they control for
important individual-level confounding variables that cannot be controlled for in
cross-sectional studies.
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 lifestyle over time. However, when there are only a few studies available, studies
from all years will be included.
The most technically appropriate measures of benefits would be based on impact
functions that cover the entire sensitive population but allow for heterogeneity across age
or other relevant demographic factors. In the absence of effect estimates specific to age,
sex, preexisting condition status, or other relevant factors, it may be appropriate to select
effect estimates that cover the broadest population to match with the desired outcome of
the analysis, which is total national-level health impacts.
Studies examining a relatively large sample are preferred because they generally have
more power to detect small magnitude effects. A large sample can be obtained in several
ways, either through a large population or through repeated observations on a smaller
population (e.g., through a symptom diary recorded for a panel of asthmatic children).
U.S. studies are more desirable than non-U.S. studies because of potential differences in
pollution characteristics, exposure patterns, medical care system, population behavior,
and lifestyle.
When modeling the effects of ozone and PM (or other pollutant combinations) jointly, it
is important to use properly specified impact functions that include both pollutants.
Using single-pollutant models in cases where both pollutants are expected to affect a
health outcome can lead to double-counting when pollutants are correlated.
For this analysis, impact functions based on PM2 5 are preferred to PM10 because CAIR
will regulate emissions of PM2 5 precursors, and air quality modeling was conducted for
this size fraction of PM. Where PM2 5 functions are not available, PM10 functions are
used as surrogates, recognizing that there will be potential downward (upward) biases if
the fine fraction of PM10 is more (less) toxic than the coarse fraction.
Some health effects, such as forced expiratory volume and other technical measurements
of lung function, are difficult to value in monetary terms. These health effects are not
quantified in this analysis.
Although the benefits associated with each individual health endpoint may be analyzed
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.
4-27
-------
with reducing these effects.10 The inability to quantify these effects lends a downward bias
to the monetized benefits presented in this analysis.
In general, the use of results from more than a single study can provide a more robust
estimate of the relationship between a pollutant and a given health effect. However, there are
often differences between studies examining the same endpoint, making it difficult to pool
the results in a consistent manner. For example, studies may examine different pollutants or
different age groups. For this reason, we consider very carefully the set of studies available
examining each endpoint and select a consistent subset that provides a good balance of
population coverage and match with the pollutant of interest. In many cases, either because
of a lack of multiple studies, consistency problems, or clear superiority in the quality or
comprehensiveness of one study over others, a single published study is selected as the basis
of the effect estimate.
When several effect estimates for a pollutant and a given health endpoint have been
selected, they are quantitatively combined or pooled to derive a more robust estimate of the
relationship. The Benefits TSD completed for the nonroad diesel rulemaking provides
details of the procedures used to combine multiple impact functions (Abt Associates, 2003).
In general, we used fixed or random effects models to pool estimates from different studies
of the same endpoint. Fixed effects pooling simply weights each study's estimate by the
inverse variance, giving more weight to studies with greater statistical power (lower
variance). Random effects pooling accounts for both within-study variance and between-
study variability, due, for example, to differences in population susceptibility. We used the
fixed effects model as our null hypothesis and then determined whether the data suggest that
we should reject this null hypothesis, in which case we would use the random effects
model.11 Pooled impact functions are used to estimate hospital admissions (PM), school
absence days (ozone), lower respiratory symptoms (PM), asthma exacerbations (PM), and
asthma-related emergency room visits (ozone). For more details on methods used to pool
1 "There has been a great deal of research recently on the potential effect of ozone on premature mortality
(Anderson et al, 2004; Bell et al, 2004; Thurston and Ito, 2001). Although the air pollutant most clearly
associated with premature mortality is PM, with dozens of studies reporting such an association, repeated
ozone exposure is a likely contributing factor for premature mortality, causing an inflammatory response in
the lungs that may predispose elderly and other sensitive individuals to become more susceptible. Appendix
C presents a sensitivity analysis showing the potential impacts of CAIR on ozone-related mortality.
"In this analysis, the fixed effects model assumes that there is only one pollutant coefficient for the entire
modeled area. The random effects model assumes that studies conducted in different locations are
estimating different parameters; therefore, there may be a number of different underlying pollutant
coefficients.
4-28
-------
incidence estimates, see the Benefits TSD for the nonroad diesel rulemaking (Abt
Associates, 2003).
Effect estimates for a pollutant and a given health endpoint were applied consistently
across all locations nationwide. This applies to both impact functions defined by a single
effect estimate and those defined by a pooling of multiple effect estimates. Although the
effect estimate may, in fact, vary from one location to another (e.g., because of differences in
population susceptibilities or differences in the composition of PM), location-specific effect
estimates are generally not available.
The specific studies from which effect estimates for the primary analysis are drawn
are included in Table 4-7.
Premature Mortality. Both long- and short-term exposures to ambient levels of air
pollution have been associated with increased risk of premature mortality. The size of the
mortality risk estimates from epidemiological studies, the serious nature of the effect itself,
and the high monetary value ascribed to prolonging life make mortality risk reduction the
most significant health endpoint quantified in this analysis.
Although a number of uncertainties remain to be addressed by continued research
(NRC, 1998), a substantial body of published scientific literature documents the correlation
between elevated PM concentrations and increased mortality rates. Time-series methods
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 concentrations. Cohort
methods examine the potential relationship between community-level PM exposures over
multiple years (i.e., long-term exposures) and community-level annual mortality rates.
Researchers have found statistically significant associations between PM and premature
mortality using both types of studies. In general, the risk estimates based on the cohort
studies are larger than those derived from time-series studies. Cohort analyses are thought to
better capture the full public health impact of exposure to air pollution over time, because
they capture the effects of long-term exposures and possibly some component of short-term
exposures (Kunzli et al., 2001; NRC, 2002). This section discusses some of the issues
surrounding the estimation of premature mortality.
Over a dozen studies have found significant associations between various measures
of long-term exposure to PM and elevated rates of annual mortality, beginning with Lave and
Seskin (1977). Most of the published studies found positive (but not always statistically
4-29
-------
Table 4-7. Endpoints and Studies Used to Calculate Total Monetized Health Benefits
Endpoint
Premature Mortality
Premature mortality
— cohort study, all-
cause
Premature mortality
— all-cause
Chronic Illness
Chronic bronchitis
Nonfatal heart
attacks
Pollutant Study
PM2 5 Pope et al. (2002)
(annual
mean)
PM2 5 Woodruff et al. (1997)
(annual
mean)
PM25 Abbey et al. (1995)
PM25 Peters etal. (2001)
Study
Population
>29 years
Infant (<1 year)
>26 years
Adults
Hospital Admissions
Respiratory
Cardiovascular
Asthma-related ER
visits
Ozone Pooled estimate:
Schwartz (1995)—ICD 460-519 (all resp)
Schwartz (1994a, 1994b)—ICD 480-486
(pneumonia)
Moolgavkaretal. (1997)—ICD 480-487
(pneumonia)
Schwartz (1994b)—ICD 491-492, 494-496 (COPD)
Moolgavkaretal. (1997)—ICD 490-496 (COPD)
Ozone Burnett et al. (2001)
PM2 5 Pooled estimate:
Moolgavkar (2003)—ICD 490-496 (COPD)
Ito (2003)—ICD 490-496 (COPD)
PM2 5 Moolgavkar (2000)—ICD 490-496 (COPD)
PM25 Ito (2003)—ICD 480-486 (pneumonia)
PM25 Sheppard (2003)—ICD 493 (asthma)
PM2 5 Pooled estimate:
Moolgavkar (2003)—ICD 390-429 (all
cardiovascular)
Ito (2003)—ICD 410-414, 427-428 (ischemic heart
disease, dysrhythmia, heart failure)
PM2 5 Moolgavkar (2000)—ICD 3 90-429 (all
cardiovascular)
Ozone Pooled estimate:
Weisel et al. (1995), Cody et al. (1992), Stieb et al.
(1996)
PM25 Norrisetal. (1999)
>64 years
<2 years
>64 years
20-64 years
>64 years
<65 years
>64 years
20-64 years
All ages
0-18 years
(continued)
4-30
-------
Table 4-7. Endpoints and Studies Used to Calculate Total Monetized Health Benefits
(continued)
Endpoint
Pollutant
Study
Study
Population
Other Health Endpoints
Acute bronchitis
Upper respiratory
symptoms
Lower respiratory
symptoms
Asthma
exacerbations
Work loss days
School absence days
PM25
PM10
PM25
PM25
PM25
Ozone
Dockeryetal. (1996)
Pope etal. (1991)
Schwartz and Neas (2000)
8-12 years
Asthmatics,
9-11 years
7-14 years
Pooled estimate: 6-18 years3
Ostro et al. (2001) (cough, wheeze and shortness of
breath)
Vedal et al. (1998) (cough)
Worker productivity Ozone
MRADs PM25,
Ozone
Ostro (1987)
Pooled estimate:
Gillilandetal. (2001)
Chen et al. (2000)
Crocker and Horst (1981)
Ostro and Rothschild (1989)
18-65 years
9-10 years
6-11 years
Outdoor workers,
18-65
18-65 years
The original study populations were 8 to 13 for the Ostro et al. (2001) study and 6 to 13 for the Vedal et al.
(1998) study. Based on advice from the SAB-HES, we extended the applied population to 6 to 18, reflecting
the common biological basis for the effect in children in the broader age group.
significant) associations with available PM indices such as total suspended particles (TSP).
However, exploration of alternative model specifications sometimes raised questions about
causal relationships (e.g., Lipfert, Morris, and Wyzga [1989]). These early "cross-sectional"
studies (e.g., Lave and Seskin [1977]; Ozkaynak and Thurston [1987]) were criticized for a
number of methodological limitations, particularly for inadequate control at the individual
level for variables that are potentially important in causing mortality, such as wealth,
smoking, and diet. More recently, several studies have been published that use improved
approaches and appear to be consistent with the earlier body of literature. These new
"prospective cohort" studies reflect a significant improvement over the earlier work because
they include individual-level information with respect to health status and residence. The
most extensive analyses have been based on data from two prospective cohort groups, often
4-31
-------
referred to as the Harvard "Six-Cities Study" (Dockery et al., 1993) and the "American
Cancer Society or ACS study" (Pope et al., 1995); these studies have found consistent
relationships between fine particle indicators and premature mortality across multiple
locations in the United States. A third major data set comes from the California-based 7th
Day Adventist Study (e.g., Abbey et al., 1999), which reported associations between
long-term PM exposure and mortality in men. Results from this cohort, however, have been
inconsistent, and the air quality results are not geographically representative of most of the
United States. More recently, a cohort of adult male veterans diagnosed with hypertension
has been examined (Lipfert et al., 2000). The characteristics of this group differ from the
cohorts in the Six-Cities, ACS, and 7th Day Adventist studies with respect to income, race,
health status, and smoking status. Unlike previous long-term analyses, this study found
some associations between mortality and ozone but found inconsistent results for PM
indicators. Because of the selective nature of the population in the veteran's cohort, we have
chosen not to include any effect estimates from the Lipfert et al. (2000) study in our benefits
assessment.12
Given their consistent results and broad geographic coverage, the Six-Cities and ACS
data have been particularly important in benefits analyses. The credibility of these two
studies is further enhanced by the fact that they were subject to extensive reexamination and
reanalysis by an independent team of scientific experts commissioned by HEI (Krewski
et al., 2000). The final results of the reanalysis were then independently peer reviewed by a
Special Panel of the HEI Health Review Committee. The results of these reanalyses
confirmed and expanded those of the original investigators. This intensive independent
reanalysis effort was occasioned both by the importance of the original findings and concerns
that the underlying individual health effects information has never been made publicly
available.
12EPA recognizes that the ACS cohort also is not representative of the demographic mix in the general
population. The ACS cohort is almost entirely white and has higher income and education levels relative to
the general population. EPA's approach to this problem is to match populations based on the potential for
demographic characteristics to modify the effect of air pollution on mortality risk. Thus, for the various
ACS-based models, we are careful to apply the effect estimate only to ages matching those in the original
studies, because age has a potentially large modifying impact on the effect estimate, especially when
younger individuals are excluded from the study population. For the Lipfert analysis, the applied population
should be limited to that matching the sample used in the analysis. This sample was all male, veterans, and
diagnosed hypertensive. There are also a number of differences between the composition of the sample and
the general population, including a higher percentage of African Americans (35 percent) and a much higher
percentage of smokers (81 percent former smokers, 57 percent current smokers) than the general population
(12 percent African American, 24 percent current smokers).
4-32
-------
While the HEI reexamination lends credibility to the original studies, it also
highlights sensitivities concerning the relative impact of various pollutants, the potential role
of education in mediating the association between pollution and mortality, and the influence
of spatial correlation modeling. Further confirmation and extension of the overall findings
using more recent air quality and a longer follow-up period for the ACS cohort was recently
published (Pope et al., 2002).
In developing and improving the methods for estimating and valuing the potential
reductions in mortality risk over the years, EPA consulted with the SAB-HES. That panel
recommended using long-term prospective cohort studies in estimating mortality risk
reduction (EPA-SAB-COUNCIL-ADV-99-005, 1999). This recommendation has been
confirmed by a recent report from the National Research Council, which stated that "it is
essential to use the cohort studies in benefits analysis to capture all important effects from air
pollution exposure" (NRC, 2002, p. 108). More specifically, the SAB recommended
emphasis on the ACS study because it includes a much larger sample size and longer
exposure interval and covers more locations (e.g., 50 cities compared to the Six-Cities Study)
than other studies of its kind. As explained in the regulatory impact analysis for the
Heavy-Duty Engine/Diesel Fuel rule (EPA, 2000d), more recent EPA benefits analyses have
relied on an improved specification of the ACS cohort data that was developed in the HEI
reanalysis (Krewski et al., 2000). The latest reanalysis of the ACS cohort data (Pope et al.,
2002) provides additional refinements to the analysis of PM-related mortality by
a) extending the follow-up period for the ACS study subjects to 16 years, which triples the
size of the mortality data set; b) substantially increasing exposure data, including
consideration for cohort exposure to PM25 following implementation of the PM25 standard in
1999; c) controlling for a variety of personal risk factors including occupational exposure
and diet; and d) using advanced statistical methods to evaluate specific issues that can
adversely affect risk estimates including the possibility of spatial autocorrelation of survival
times in communities located near each other. Because of these refinements, the SAB-HES
recommends using the Pope et al. (2002) study as the basis for the primary mortality estimate
for adults and suggests that alternate estimates of mortality generated using other cohort and
time-series studies could be included as part of the sensitivity analysis (SAB-HES, 2004).
The SAB-HES also recommended using the estimated relative risks from the Pope et
al. (2002) study based on the average exposure to PM2 5, measured by the average of two
PM2 5 measurements, over the periods 1979-1983 and 1999-2000. In addition to relative
risks for all-cause mortality, the Pope et al. (2002) study provides relative risks for
cardiopulmonary, lung cancer, and all-other cause mortality. Because of concerns regarding
4-33
-------
the statistical reliability of the all-other cause mortality relative risk estimates, we calculated
mortality impacts for the primary analysis based on the all-cause relative risk. However, we
provide separate estimates of cardiopulmonary and lung cancer deaths to show how these
important causes of death are affected by reductions in PM25.
Recently published studies have strengthened the case for an association between PM
exposure and respiratory inflamation and infection leading to premature mortality in children
under 5 years of age. Specifically, the SAB-HES noted the release of the WHO Global
Burden of Disease Study focusing on ambient air, which cites several recently published
time-series studies relating daily PM exposure to mortality in children (SAB-HES, 2003).
The SAB-HES also cites the study by Belanger et al. (2003) as corroborating findings linking
PM exposure to increased respiratory inflamation and infections in children. Recently, a
study by Chay and Greenstone (2003) found that reductions in TSP caused by the recession
of 1981-1982 were related to reductions in infant mortality at the county level. With regard
to the cohort study conducted by Woodruff et al. (1997), the SAB-HES notes several
strengths of the study, including the use of a larger cohort drawn from a large number of
metropolitan areas and efforts to control for a variety of individual risk factors in infants
(e.g., maternal educational level, maternal ethnicity, parental marital status, and maternal
smoking status). Based on these findings, the SAB-HES recommends that EPA incorporate
infant mortality into the primary benefits estimate and that infant mortality be evaluated
using an impact function developed from the Woodruff et al. (1997) study (SAB-HES,
2004).
Chronic Bronchitis. CB is characterized by mucus in the lungs and a persistent wet
cough for at least 3 months a year for several years in a row. CB affects an estimated 5
percent of the U.S. population (American Lung Association, 1999). A limited number of
studies have estimated the impact of air pollution on new incidences of CB. Schwartz (1993)
and Abbey et al. (1995) provide evidence that long-term PM exposure gives rise to the
development of CB in the United States. Because the CAIR is expected to reduce primarily
PM2 5, this analysis uses only the Abbey et al. (1995) study, because it is the only study
focusing on the relationship between PM25 and new incidences of CB.
NonfatalMyocardialInfarctions (heart attacks). Nonfatal heart attacks have been
linked with short-term exposures to PM25 in the United States (Peters et al., 2001) and other
countries (Poloniecki et al., 1997). We used a recent study by Peters et al. (2001) as the
basis for the impact function estimating the relationship between PM2 5 and nonfatal heart
attacks. Peters et al. is the only available U.S. study to provide a specific estimate for heart
4-34
-------
attacks. Other studies, such as Samet et al. (2000) and Moolgavkar (2000), show a consistent
relationship between all cardiovascular hospital admissions, including those for nonfatal
heart attacks, and PM. Given the lasting impact of a heart attack on long-term health costs
and earnings, we provide a separate estimate for nonfatal heart attacks. The estimate used in
the CAIR analysis is based on the single available U.S. effect estimate. The finding of a
specific impact on heart attacks is consistent with hospital admission and other studies
showing relationships between fine particles and cardiovascular effects both within and
outside the United States. Several epidemiologic studies (Liao et al., 1999; Gold et al., 2000;
Magari et al., 2001) have shown that heart rate variability (an indicator of how much the
heart is able to speed up or slow down in response to momentary stresses) is negatively
related to PM levels. Heart rate variability is a risk factor for heart attacks and other
coronary heart diseases (Carthenon et al., 2002; Dekker et al., 2000; Liao et al., 1997; Tsuji
et al., 1996). As such, significant impacts of PM on heart rate variability are consistent with
an increased risk of heart attacks.
Hospital and Emergency Room Admissions. Because of the availability of detailed
hospital admission and discharge records, there is an extensive body of literature examining
the relationship between hospital admissions and air pollution. Because of this, many of the
hospital admission endpoints use pooled impact functions based on the results of a number of
studies. In addition, some studies have examined the relationship between air pollution and
emergency room visits. Since most emergency room visits do not result in an admission to
the hospital (the majority of people going to the emergency room are treated and return
home), we treat hospital admissions and emergency room visits separately, taking account of
the fraction of emergency room visits that are admitted to the hospital.
The two main groups of hospital admissions estimated in this analysis are respiratory
admissions and cardiovascular admissions. There is not much evidence linking ozone or PM
with other types of hospital admissions. The only type of emergency room visits that have
been consistently linked to ozone and PM in the United States are asthma-related visits.
To estimate avoided incidences of cardiovascular hospital admissions associated with
PM2 5, we used studies by Moolgavkar (2003) and Ito (2003). Additional published studies
show a statistically significant relationship between PM10 and cardiovascular hospital
admissions. However, given that the control options we are analyzing are expected to reduce
primarily PM25, we focus on the two studies that examine PM25. Both of these studies
provide an effect estimate for populations over 65, allowing us to pool the impact functions
for this age group. Only Moolgavkar (2000) provided a separate effect estimate for
4-35
-------
populations 20 to 64.13 Total cardiovascular hospital admissions are thus the sum of the
pooled estimate for populations over 65 and the single study estimate for populations 20 to
64. Cardiovascular hospital admissions include admissions for myocardial infarctions. To
avoid double-counting benefits from reductions in myocardial infarctions when applying the
impact function for cardiovascular hospital admissions, we first adjusted the baseline
cardiovascular hospital admissions to remove admissions for myocardial infarctions.
To estimate total avoided incidences of respiratory hospital admissions, we used
impact functions for several respiratory causes, including chronic obstructive pulmonary
disease (COPD), pneumonia, and asthma. As with cardiovascular admissions, additional
published studies show a statistically significant relationship between PM10 and respiratory
hospital admissions. We used only those focusing on PM2 5. Both Moolgavkar (2000) and
Ito (2003) provide effect estimates for COPD in populations over 65, allowing us to pool the
impact functions for this group. Only Moolgavkar (2000) provides a separate effect estimate
for populations 20 to 64. Total COPD hospital admissions are thus the sum of the pooled
estimate for populations over 65 and the single study estimate for populations 20 to 64. Only
Ito (2003) estimated pneumonia and only for the population 65 and older. In addition,
Sheppard (2003) provided an effect estimate for asthma hospital admissions for populations
under age 65. Total avoided incidences of PM-related respiratory-related hospital
admissions is the sum of COPD, pneumonia, and asthma admissions.
To estimate the effects of PM air pollution reductions on asthma-related ER visits, we
use the effect estimate from a study of children 18 and under by Norris et al. (1999). As
noted earlier, there is another study by Schwartz examining a broader age group (less than
65), but the Schwartz study focused on PM10 rather than PM2 5. We selected the Norris et al.
(1999) effect estimate because it better matched the pollutant of interest. Because children
tend to have higher rates of hospitalization for asthma relative to adults under 65, we will
likely capture the majority of the impact of PM25 on asthma emergency room visits in
populations under 65, although there may still be significant impacts in the adult population
under 65.
13Note that the Moolgavkar (2000) study has not been updated to reflect the more stringent GAM convergence
criteria. However, given that no other estimates are available for this age group, we chose to use the existing
study. Given the very small (<5 percent) difference in the effect estimates for people 65 and older with
cardiovascular hospital admissions between the original and reanalyzed results, we do not expect this choice
to introduce much bias.
4-36
-------
To estimate avoided incidences of respiratory hospital admissions associated with
ozone, we used a number of studies examining hospital admissions for a range of respiratory
illnesses, including pneumonia and COPD. Two age groups, adults over 65 and children
under 2, were examined. For adults over 65, Schwartz (1995) provides effect estimates for
two different cities relating ozone and hospital admissions for all respiratory causes (defined
as ICD codes 460-519). Impact functions based on these studies were pooled first before
being pooled with other studies. Two studies (Moolgavkar et al., 1997; Schwartz, 1994a)
examine ozone and pneumonia hospital admissions in Minneapolis. One additional study
(Schwartz, 1994b) examines ozone and pneumonia hospital admissions in Detroit. The
impact functions for Minneapolis were pooled together first, and the resulting impact
function was then pooled with the impact function for Detroit. This avoids assigning too
much weight to the information coming from one city. For COPD hospital admissions, two
studies are available: Moolgavkar et al. (1997), conducted in Minneapolis, and Schwartz
(1994b), conducted in Detroit. These two studies were pooled together. To estimate total
respiratory hospital admissions for adults over 65, COPD admissions were added to
pneumonia admissions, and the result was pooled with the Schwartz (1995) estimate of total
respiratory admissions. Burnett et al. (2001) is the only study providing an effect estimate
for respiratory hospital admissions in children under 2.
Acute Health Events and School/Work Loss Days. As indicated in Table 4-1, in
addition to mortality, chronic illness, and hospital admissions, a number of acute health
effects not requiring hospitalization are associated with exposure to ambient levels of ozone
and PM. The sources for the effect estimates used to quantify these effects are described
below.
Around 4 percent of U.S. children between the ages of 5 and 17 experience episodes
of acute bronchitis annually (American Lung Association, 2002c). Acute bronchitis is
characterized by coughing, chest discomfort, slight fever, and extreme tiredness, lasting for a
number of days. According to the MedlinePlus medical encyclopedia,14 with the exception
of cough, most acute bronchitis symptoms abate within 7 to 10 days. Incidence of episodes
of acute bronchitis in children between the ages of 5 and 17 were estimated using an effect
estimate developed from Dockery et al. (1996).
Incidences of lower respiratory symptoms (e.g., wheezing, deep cough) in children
aged 7 to 14 were estimated using an effect estimate from Schwartz and Neas (2000).
14See http://www.nlm.nih.gov/medlineplus/ency/article/000124.htm, accessed January 2002.
4-37
-------
Because asthmatics have greater sensitivity to stimuli (including air pollution),
children with asthma can be more susceptible to a variety of upper respiratory symptoms
(e.g., runny or stuffy nose; wet cough; and burning, aching, or red eyes). Research on the
effects of air pollution on upper respiratory symptoms has thus focused on effects in
asthmatics. Incidences of upper respiratory symptoms in asthmatic children aged 9 to 11 are
estimated using an effect estimate developed from Pope et al. (1991).
Health effects from air pollution can also result in missed days of work (either from
personal symptoms or from caring for a sick family member). Days of work lost due to
PM25 were estimated using an effect estimate developed from Ostro (1987). Children may
also be absent from school because of respiratory or other diseases caused by exposure to air
pollution. Most studies examining school absence rates have found little or no association
with PM2 5, but several studies have found a significant association between ozone levels and
school absence rates. We used two recent studies, Gilliland et al. (2001) and Chen et al.
(2000), to estimate changes in absences (school loss days) due to changes in ozone levels.
The Gilliland et al. study estimated the incidence of new periods of absence, while the Chen
et al. study examined absence on a given day. We converted the Gilliland estimate to days of
absence by multiplying the absence periods by the average duration of an absence. We
estimated an average duration of school absence of 1.6 days by dividing the average daily
school absence rate from Chen et al. (2000) and Ransom and Pope (1992) by the episodic
absence rate from Gilliland et al. (2001). This provides estimates from Chen et al. (2000)
and Gilliland et al. (2001), which can be pooled to provide an overall estimate.
MRAD result when individuals reduce most usual daily activities and replace them
with less strenuous activities or rest, yet not to the point of missing work or school. For
example, a mechanic who would usually be doing physical work most of the day will instead
spend the day at a desk doing paper and phone work because of difficulty breathing or chest
pain. The effect of PM25 and ozone on MRAD was estimated using an effect estimate
derived from Ostro and Rothschild (1989).
For CAIR, we have followed the SAB-HES recommendations regarding asthma
exacerbations in developing the primary estimate. To prevent double-counting, we focused
the estimation on asthma exacerbations occurring in children and excluded adults from the
calculation.15 Asthma exacerbations occurring in adults are assumed to be captured in the
''Estimating asthma exacerbations associated with air pollution exposures is difficult, due to concerns about
double-counting of benefits. Concerns over double-counting stem from the fact that studies of the general
population also include asthmatics, so estimates based solely on the asthmatic population cannot be directly
4-38
-------
general population endpoints such as work loss days and MRADs. Consequently, if we had
included an adult-specific asthma exacerbation estimate, we would likely double-count
incidence for this endpoint. However, because the general population endpoints do not cover
children (with regard to asthmatic effects), an analysis focused specifically on asthma
exacerbations for children (6 to 18 years of age) could be conducted without concern for
double-counting.
To characterize asthma exacerbations in children, we selected two studies (Ostro et
al., 2001; Vedal et al., 1998) that followed panels of asthmatic children. Ostro et al. (2001)
followed a group of 138 African-American children in Los Angeles for 13 weeks, recording
daily occurrences of respiratory symptoms associated with asthma exacerbations (e.g.,
shortness of breath, wheeze, and cough). This study found a statistically significant
association between PM25, measured as a 12-hour average, and the daily prevalence of
shortness of breath and wheeze endpoints. Although the association was not statistically
significant for cough, the results were still positive and close to significance; consequently,
we decided to include this endpoint, along with shortness of breath and wheeze, in generating
incidence estimates (see below). Vedal et al. (1998) followed a group of elementary school
children, including 74 asthmatics, located on the west coast of Vancouver Island for 18
months including measurements of daily peak expiratory flow (PEF) and the tracking of
respiratory symptoms (e.g., cough, phlegm, wheeze, chest tightness) through the use of daily
diaries. Association between PM10 and respiratory symptoms for the asthmatic population
was only reported for two endpoints: cough and PEF. Because it is difficult to translate PEF
measures into clearly defined health endpoints that can be monetized, we only included the
cough-related effect estimate from this study in quantifying asthma exacerbations. We
added to the general population numbers without double-counting. In one specific case (upper respiratory
symptoms in children), the only study available is limited to asthmatic children, so this endpoint can be
readily included in the calculation of total benefits. However, other endpoints, such as lower respiratory
symptoms and MRADs, are estimated for the total population that includes asthmatics. Therefore, to simply
add predictions of asthma-related symptoms generated for the population of asthmatics to these total
population-based estimates could result in double-counting, especially if they evaluate similar endpoints.
The SAB-HES, in commenting on the analytical blueprint for 812, acknowledged these challenges in
evaluating asthmatic symptoms and appropriately adding them into the primary analysis (SAB-HES, 2004).
However, despite these challenges, the SAB-HES recommends the addition of asthma-related symptoms
(i.e., asthma exacerbations) to the primary analysis, provided that the studies use the panel study approach
and that they have comparable design and baseline frequencies in both asthma prevalence and exacerbation
rates. Note also, that the SAB-HES, while supporting the incorporation of asthma exacerbation estimates,
does not believe that the association between ambient air pollution, including ozone and PM, and the new
onset of asthma is sufficiently strong to support inclusion of this asthma-related endpoint in the primary
estimate.
4-39
-------
employed the following pooling approach in combining estimates generated using effect
estimates from the two studies to produce a single asthma exacerbation incidence estimate.
First, we pooled the separate incidence estimates for shortness of breath, wheeze, and cough
generated using effect estimates from the Ostro et al. study, because each of these endpoints
is aimed at capturing the same overall endpoint (asthma exacerbations) and there could be
overlap in their predictions. The pooled estimate from the Ostro et al. study is then pooled
with the cough-related estimate generated using the Vedal study. The rationale for this
second pooling step is similar to the first; both studies are attempting to quantify the same
overall endpoint (asthma exacerbations).
Additional epidemiological studies are available for characterizing asthma-related
health endpoints (the full list of epidemiological studies considered for modeling asthma-
related incidence is presented in Table 4-8). However, based on recommendations from the
SAB-HES, we decided not to use these additional studies in generating the primary estimate.
In particular, the Yu et al. (2000) estimates show a much higher baseline incidence rate than
other studies, which may lead to an overstatement of the expected impacts in the overall
asthmatic population. The Whittemore and Korn (1980) study did not use a well-defined
endpoint, instead focusing on a respondent-defined "asthma attack." Other studies looked at
respiratory symptoms in asthmatics but did not focus on specific exacerbations of asthma.
4.1.5.2 Uncertainties Associated with Health Impact Functions
Within-Study Variation. Within-study variation refers to the precision with which a
given study estimates the relationship between air quality changes and health effects. Health
effects studies provide both a "best estimate" of this relationship plus a measure of the
statistical uncertainty of the relationship. The size of this uncertainty depends on factors
such as the number of subjects studied and the size of the effect being measured. The results
of even the most well-designed epidemiological studies are characterized by this type of
uncertainty, though well-designed studies typically report narrower uncertainty bounds
around the best estimate than do studies of lesser quality. In selecting health endpoints, we
generally focus on endpoints where a statistically significant relationship has been observed
in at least some studies, although we may pool together results from studies with both
statistically significant and insignificant estimates to avoid selection bias.
Across-Study Variation. Across-study variation refers to the fact that different
published studies of the same pollutant/health effect relationship typically do not report
identical findings; in some instances the differences are substantial. These differences can
exist even between equally reputable studies and may result in health effect estimates that
4-40
-------
Table 4-8. Studies Examining Health Impacts in the Asthmatic Population Evaluated
for Use in the Benefits Analysis
Endpoint
Definition
Pollutant
Study
Study Population
Asthma Attack Indicators
Shortness of Prevalence of shortness of PM2 5
breath breath; incidence of
shortness of breath
Cough Prevalence of cough; PM2 5
incidence of cough
Wheeze Prevalence of wheeze; PM2 5
incidence of wheeze
Asthma > 1 mild asthma symptom: PM10,
exacerbation wheeze, cough, chest PM{ „
tightness, shortness of breath
Cough Prevalence of cough
Other Symptoms/Illness Endpoints
PM,,
PM,,
Upper >1 of the following: runny
respiratory or stuffy nose; wet cough;
symptoms burning, aching, or red eyes
Moderate or Probability of moderate (or PM2 5
worse asthma worse) rating of overall
asthma status
Ostroetal. (2001)
Ostroetal. (2001)
Ostroetal. (2001)
Yu et al. (2000)
Vedaletal. (1998)
Pope etal. (1991)
African-American
asthmatics, 8-13
African-American
asthmatics, 8-13
African-American
asthmatics, 8-13
Asthmatics, 5-13
Asthmatics, 6-13
Asthmatics, 9-11
Ostro et al. (1991) Asthmatics, all ages
Acute bronchitis
Phlegm
Asthma attacks
> 1 episodes of bronchitis in
the past 12 months
"Other than with colds, does
this child usually seem
congested in the chest or
bring up phlegm?"
Respondent-defined asthma
attack
PM25
PM25
PM25,
ozone
McConnell et al.
(1999)
McConnell et al.
(1999)
Whittemore and
Korn(1980)
Asthmatics, 9-15
Asthmatics, 9-15
Asthmatics, all ages
vary considerably. Across-study variation can result from two possible causes. One
possibility is that studies report different estimates of the single true relationship between a
given pollutant and a health effect because of differences in study design, random chance, or
other factors. For example, a hypothetical study conducted in New York and one conducted
in Seattle may report different C-R functions for the relationship between PM and mortality,
in part because of differences between these two study populations (e.g., demographics,
activity patterns). Alternatively, study results may differ because these two studies are in
fact estimating different relationships; that is, the same reduction in PM in New York and
4-41
-------
Seattle may result in different reductions in premature mortality. This may result from a
number of factors, such as differences in the relative sensitivity of these two populations to
PM pollution and differences in the composition of PM in these two locations. In either
case, where we identified multiple studies that are appropriate for estimating a given health
effect, we generated a pooled estimate of results from each of those studies.
Application of C-R Relationship Nationwide. Regardless of the use of impact
functions based on effect estimates from a single epidemiological study or multiple studies,
each impact function was applied uniformly throughout the United States to generate health
benefit estimates. However, to the extent that pollutant/health effect relationships are region
specific, applying a location-specific impact function at all locations in the United States
may result in overestimates 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 on health benefit estimates introduced by applying a single
impact function to the entire United States. This may be a significant uncertainty in the
analysis, but the current state of the scientific literature does not allow for a region-specific
estimation of health benefits.16
Extrapolation of Impact Functions Across Populations. Epidemiological studies
often focus on specific age ranges, either due to data availability limitations (e.g., most
hospital admission data come from Medicare records, which are limited to populations 65
and older), or to simplify data collection (e.g., some asthma symptom studies focus on
children at summer camps, which usually have a limited age range). We have assumed for
the primary analysis that most impact functions should be applied only to those populations
with ages that strictly match the populations in the underlying epidemiological studies.
However, in many cases, there is no biological reason why the observed health effect would
not also occur in other populations within a reasonable range of the studied population. For
example, Dockery et al. (1996) examined acute bronchitis in children aged 8 to 12. There is
no biological reason to expect a very different response in children aged 6 or 14. By
excluding populations outside the range in the studies, we may be underestimating the health
impact in the overall population. In response to recommendations from the SAB-HES,
where there appears to be a reasonable physiological basis for expanding the age group
associated with a specific effect estimate beyond the study population to cover the full age
""Although we are not able to use region-specific effect estimates, we use region-specific baseline incidence
rates where available. This allows us to take into account regional differences in health status, which can
have a significant impact on estimated health benefits.
4-42
-------
group (e.g., expanding from a study population of 7 to 11 year olds to the full 6- to 18-year
child age group), we have done so and used those expanded incidence estimates in the
primary analysis.
Uncertainties in the PMMortality Relationship. A substantial body of published
scientific literature demonstrates a correlation between elevated PM concentrations and
increased premature mortality. However, much about this relationship is still uncertain.
These uncertainties include the following:
Causality: Epidemiological studies are not designed to definitively prove causation.
For the analysis of the CAIR, we assumed a causal relationship between exposure to elevated
PM and premature mortality, based on the consistent evidence of a correlation between PM
and mortality reported in the substantial body of published scientific literature.
Other Pollutants: PM concentrations are correlated with the concentrations of other
criteria pollutants, such as ozone and CO. To the extent that there is correlation, this analysis
may be assigning mortality effects to PM exposure that are actually the result of exposure to
other pollutants. Recent studies (see Thurston and Ito [2001] and Bell et al [2004]) have
explored whether ozone may have mortality effects independent of PM. EPA is currently
evaluating the epidemiological literature on the relationship between ozone and mortality.
Shape of the C-R Function: The shape of the true PM mortality C-R function is
uncertain, but this analysis assumes the C-R function has a non-threshold log-linear form
throughout the relevant range of exposures. If this is not the correct form of the C-R
function, or if certain scenarios predict concentrations well above the range of values for
which the C-R function was fitted, avoided mortality may be misestimated. Although not
included in the primary analysis, the potential impact of a health effects threshold on avoided
incidences of PM-related premature mortality is explored as a key sensitivity analysis.
The possible existence of an effect threshold is a very important scientific question
and issue for policy analyses such as this one. In 1999, the EPA SAB Advisory Council for
Clean Air Compliance advised EPA that there was currently no scientific basis for selecting
a threshold of 15 |ig/m3 or any other specific threshold for the PM-related health effects
considered in typical benefits analyses (EPA-SAB-Council-ADV-99-012, 1999). In 2000, as
a part of their review of benefits methods, the National Research Council concluded that
there is no evidence for any departure from linearity in the observed range of exposure to
PM10 or PM25, nor any indication of a threshold (NRC, 2002). They cite the weight of
evidence available from both short- and long-term exposure models and the similar effects
4-43
-------
found in cities with low and high ambient concentrations of PM. Most recently, EPA's
updated (2004) Criteria Document states, "In summary, the available evidence does not
either support or refute the existence of thresholds for effects of PM on mortality across the
range of uncertainties in the studies." The PM criteria document identifies the general shape
of exposure-response relationship(s) between PM and/or other pollutants and observed health
effects (e.g., potential indications of thresholds), as an important issue and uncertainty in
interpreting the overall PM epidemiology database.
These recommendations are supported by the recent literature on health effects of
short and longer term PM exposures (Daniels et al., 2000; Pope, 2000; Pope et al, 2002;
Rossi et al., 1999; Schwartz and Zanobetti, 2000; Schwartz, Laden, and Zanobetti, 2002;
Smith et al, 2000) that finds in most cases no evidence of a nonlinear relationship between
PM and health effects and certainly does not find a distinct threshold. Recent cohort
analyses by HEI (Krewski et al., 2000) and Pope et al. (2002) provide additional evidence of
a quasi-linear relationship between long-term exposures to PM2 5 and mortality. According
to the latest draft PM criteria document, Krewski et al. (2000) found a "found a visually
near-linear relationship between all-cause and cardiopulmonary mortality residuals and mean
sulfate concentrations, near-linear between cardiopulmonary mortality and mean PM2.5, but
a somewhat nonlinear relationship between all-cause mortality residuals and mean PM2.5
concentrations that flattens above -20 |ig/m3. The confidence bands around the fitted curves
are very wide, however, neither requiring a linear relationship nor precluding a nonlinear
relationship if suggested by reanalyses" (Krewski et al. (2000), page 8-138). The Pope et al.
(2002) analysis, which represented an extension to the Krewski et al. analysis, found that the
functions relating PM2 5 and mortality are not significantly different from linear associations.
Based on the recent literature and advice from the SAB, we assume there are no
thresholds for modeling health effects. Although not included in the primary analysis, the
potential impact of a health effects threshold on avoided incidences of PM-related premature
mortality is explored as a key sensitivity analysis.
Regional Differences: As discussed above, significant variability exists in the results
of different PM/mortality studies. This variability may reflect regionally specific C-R
functions resulting from regional differences in factors such as the physical and chemical
composition of PM. If true regional differences exist, applying the PM/mortality C-R
function to regions outside the study location could result in misestimation of effects in these
regions.
4-44
-------
Exposure/Mortality Lags: There is a time lag between changes in PM exposures and
the total realization of changes in annual mortality rates. For the chronic PM/mortality
relationship, the length of the lag is unknown and may be dependent on the kind of exposure.
The existence of such a lag is important for the valuation of premature mortality incidence
because economic theory suggests that benefits occurring in the future should be discounted.
There is no specific scientific evidence of the existence or structure of a PM effects lag.
However, current scientific literature on adverse health effects similar to those associated
with PM (e.g., smoking-related disease) and the difference in the effect size between chronic
exposure studies and daily mortality studies suggests that all incidences of premature
mortality reduction associated with a given incremental change in PM exposure probably
would not occur in the same year as the exposure reduction. The smoking-related literature
also implies that lags of up to a few years or longer are plausible. The SAB-HES suggests
that appropriate lag structures may be developed based on the distribution of cause-specific
deaths within the overall all-cause estimate. Diseases with longer progressions should be
characterized by long-term lag structures, while impacts occurring in populations with
existing disease may be characterized by short-term lags.
A key question is the distribution of causes of death within the relatively broad
categories analyzed in the cohort studies used. While we may be more certain about the
appropriate length of cessation lag for lung cancer deaths, it is not clear what the appropriate
lag structure should be for different types of cardiopulmonary deaths, which include both
respiratory and cardiovascular causes. Some respiratory diseases may have a long period of
progression, while others, such as pneumonia, have a very short duration. In the case of
cardiovascular disease, there is an important question of whether air pollution is causing the
disease, which would imply a relatively long cessation lag, or whether air pollution is
causing premature death in individuals with preexisting heart disease, which would imply
very short cessation lags.
The SAB-HES provides several recommendations for future research that could
support the development of defensible lag structures, including the use of disease-specific lag
models, and the construction of a segmented lag distribution to combine differential lags
across causes of death. The SAB-HES recommended that until additional research has been
completed, EPA should assume a segmented lag structure characterized by 30 percent of
mortality reductions occurring in the first year, 50 percent occurring evenly over years 2 to 5
after the reduction in PM2 5, and 20 percent occurring evenly over the years 6 to 20 after the
reduction in PM2 5. The distribution of deaths over the latency period is intended to reflect
the contribution of short-term exposures in the first year, cardiopulmonary deaths in the 2- to
4-45
-------
5-year period, and long-term lung disease and lung cancer in the 6- to 20-year period. For
future analyses, the specific distribution of deaths over time will need to be determined
through research on causes of death and progression of diseases associated with air pollution.
It is important to keep in mind that changes in the lag assumptions do not change the total
number of estimated deaths but rather the timing of those deaths.
Cumulative Effects: We attribute the PM/mortality relationship in the underlying
epidemiological studies to cumulative exposure to PM. However, the relative roles of PM
exposure duration and PM exposure level in inducing premature mortality remain unknown
at this time.
4.1.5.3 Baseline Health Effect Incidence Rates
The epidemiological studies of the association between pollution levels and adverse
health effects generally provide a direct estimate of the relationship of air quality changes to
the relative risk of a health effect, rather than an estimate of the absolute number of avoided
cases. For example, a typical result might be that a 10 |ig/m3 decrease in daily PM25 levels
might decrease hospital admissions by 3 percent. The baseline incidence of the health effect
is necessary to convert this relative change into a number of cases. The baseline incidence
rate provides an estimate of the incidence rate (number of cases of the health effect per year,
usually per 10,000 or 100,000 general population) in the assessment location corresponding
to baseline pollutant levels in that location. To derive the total baseline incidence per year,
this rate must be multiplied by the corresponding population number (e.g., if the baseline
incidence rate is number of cases per year per 100,000 population, it must be multiplied by
the number of 100,000s in the population).
Some epidemiological studies examine the association between pollution levels and
adverse health effects in a specific subpopulation, such as asthmatics or diabetics. In these
cases, it is necessary to develop not only baseline incidence rates, but also prevalence rates
for the defining condition (e.g., asthma). For both baseline incidence and prevalence data,
we use age-specific rates where available. Impact functions are applied to individual age
groups and then summed over the relevant age range to provide an estimate of total
population benefits.
In most cases, because of a lack of data or methods, we have not attempted to project
incidence rates to future years, instead assuming that the most recent data on incidence rates
is the best prediction of future incidence rates. In recent years, better data on trends in
incidence and prevalence rates for some endpoints, such as asthma, have become available.
4-46
-------
We are working to develop methods to use these data to project future incidence rates.
However, for our primary benefits analysis of the final CAIR, we continue to use current
incidence rates.
Table 4-9 summarizes the baseline incidence data and sources used in the benefits
analysis. We use the most geographically disaggregated data available. For premature
mortality, county-level data are available. For hospital admissions, regional rates are
available. However, for all other endpoints, a single national incidence rate is used, due to a
lack of more spatially disaggregated data. In these cases, we used national incidence rates
whenever possible, because these data are most applicable to a national assessment of
benefits. However, for some studies, the only available incidence information comes from
the studies themselves; in these cases, incidence in the study population is assumed to
represent typical incidence at the national level.
Age, cause, and county-specific mortality rates were obtained from the U.S. Centers
for Disease Control and Prevention (CDC) for the years 1996 through 1998. CDC maintains
an online data repository of health statistics, CDC Wonder, accessible at
http://wonder.cdc.gov/. The mortality rates provided are derived from U.S. death records
and U.S. Census Bureau postcensal population estimates. Mortality rates were averaged
across 3 years (1996 through 1998) to provide more stable estimates. When estimating rates
for age groups that differed from the CDC Wonder groupings, we assumed that rates were
uniform across all ages in the reported age group. For example, to estimate mortality rates
for individuals ages 30 and up, we scaled the 25- to 34-year-old death count and population
by one-half and then generated a population-weighted mortality rate using data for the older
age groups. Note that we have not projected any changes in mortality rates over time. We
are aware that the U.S. Census projections of total and age-specific mortality rates used in
our population projections are based on projections of declines in mortality rates for younger
populations and increases in mortality rates for older populations over time. We are
evaluating the most appropriate way to incorporate these projections into our database of
county-level cause-specific mortality rates. In the interim, we have not attempted to adjust
future mortality rates. This will lead to an overestimate of mortality benefits in future years,
with the overestimation bias increasing as benefits are projected into the future. We do not at
this time have a quantified estimate of the magnitude of the potential bias in the years
analyzed for this rule (2010 and 2015).
4-47
-------
Table 4-9. Baseline Incidence Rates and Population Prevalence Rates for Use in Impact
Functions, General Population
Rates
Endpoint
Parameter
Value
Source3
Mortality
Daily or annual mortality
rate
Hospitalizations Daily hospitalization rate
Age-, cause-, and
county-specific
rate
Age-, region-, and
cause-specific rate
CDC Wonder (1996-1998)
1999 NHDS public use data filesb
Asthma ER
Visits
Chronic
Bronchitis
Nonfatal
Myocardial
Infarction (heart
attacks)
Asthma
Exacerbations
Acute Bronchitis
Daily asthma ER visit
rate
Annual prevalence rate
per person
• Aged 18-44
• Aged 45-64
• Aged 65 and older
Annual incidence rate per
person
Daily nonfatal myocardial
infarction incidence rate
per person, 18+
• Northeast
• Midwest
• South
• West
Incidence (and
prevalence) among
asthmatic African-
American children
• daily wheeze
• daily cough
• daily dyspnea
Prevalence among
asthmatic children
• daily wheeze
• daily cough
• daily dyspnea
Annual bronchitis
incidence rate, children
Age- and region-
specific visit rate
0.0367
0.0505
0.0587
0.00378
0.0000159
0.0000135
0.0000111
0.0000100
0.076(0.173)
0.067(0.145)
0.037 (0.074)
0.038
0.086
0.045
0.043
2000 NHAMCS public use data
files0; 1999 NHDS public use data
filesb
1999 NHIS (American Lung
Association, 2002b, Table 4)
Abbey et al. (1993b, Table 3)
1999 NHDS public use data filesb;
adjusted by 0.93 for probability of
surviving after 28 days (Rosamond et
al., 1999)
Ostroetal. (2001)
Vedaletal. (1998)
American Lung Association (2002c,
Table 11)
(continued)
4-48
-------
Table 4-9. Baseline Incidence Rates and Population Prevalence Rates for Use in Impact
Functions, General Population (continued)
Endpoint
Lower
Respiratory
Symptoms
Upper
Respiratory
Symptoms
Work Loss Days
Minor
Restricted-
Parameter
Daily lower respiratory
symptom incidence
among children"1
Daily upper respiratory
symptom incidence
among asthmatic children
Daily WLD incidence
rate per person (18-65)
• Aged 18-24
• Aged 25-44
• Aged 45-64
Daily MRAD incidence
rate per person
Value
0.0012
0.3419
0.00540
0.00678
0.00492
0.02137
Rates
Source3
Schwartz et al. (1994, Table
2)
Pope etal. (1991, Table 2)
1996 HIS (Adams et al., 1999,
Table 41); U.S. Bureau of the
Census (2000)
Ostro and Rothschild (1989,
p. 243)
Activity Days
School Loss
Days6
Daily school absence rate
per person
Daily illness-related
school absence rate per
person6
• Northeast
• Midwest
• South
• Southwest
Daily respiratory illness-
related school absence
rate per person
• Northeast
• Midwest
• South
• West
0.055
0.0136
0.0146
0.0142
0.0206
0.0073
0.0092
0.0061
0.0124
National Center for Education
Statistics (1996)
1996 HIS (Adams et al., 1999,
Table 47); estimate of 180
school days per year
1996 HIS (Adams et al., 1999,
Table 47); estimate of 180
school days per year
The following abbreviations are used to describe the national surveys conducted by the National Center for Health Statistics: HIS refers
to the National Health Interview Survey; NHDS—National Hospital Discharge Survey; NHAMCS—National Hospital Ambulatory
Medical Care Survey.
Seeftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Datasets/NHDS/.
See ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Datasets/NHAMCS/.
Lower respiratory symptoms are defined as two or more of the following: cough, chest pain, phlegm, and wheeze.
The estimate of daily illness-related school absences excludes school loss days associated with injuries to match the definition in the
Gilliland et al. (2001) study.
4-49
-------
For the set of endpoints affecting the asthmatic population, in addition to baseline
incidence rates, prevalence rates of asthma in the population are needed to define the
applicable population. Table 4-9 lists the baseline incidence rates and their sources for
asthma symptom endpoints. Table 4-10 lists the prevalence rates used to determine the
applicable population for asthma symptom endpoints. Note that these reflect current asthma
prevalence and assume no change in prevalence rates in future years. As noted above, we are
investigating methods for projecting asthma prevalence rates in future years.
Table 4-10. Asthma Prevalence Rates Used to Estimate Asthmatic Populations in
Impact Functions
Population Group
All Ages
<18
5-17
18^4
45-64
65+
Male, 27+
African American, 5 to 17
African American, <18
Value
0.0386
0.0527
0.0567
0.0371
0.0333
0.0221
0.021
0.0726
0.0735
Asthma Prevalence Rates
Source
American Lung Association (2002a, Table 7) — based on 1999 HIS
American Lung Association (2002a, Table 7)— based on 1999 HIS
American Lung Association (2002a, Table 7)— based on 1999 HIS
American Lung Association (2002a, Table 7)— based on 1999 HIS
American Lung Association (2002a, Table 7)— based on 1999 HIS
American Lung Association (2002a, Table 7) — based on 1999 HIS
2000 HIS public use data files'
American Lung Association (2002a, Table 9) — based on 1999 HIS
American Lung Association (2002a, Table 9) — based on 1999 HIS
Seeftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Datasets/NHIS/2000/.
4.1.5.4 Selecting Unit Values for Monetizing Health Endpoints
The appropriate economic value for a change in a health effect depends on whether
the health effect is viewed ex ante (before the effect has occurred) or ex post (after the effect
has occurred). Reductions in ambient concentrations of air pollution generally lower the risk
of future adverse health affects by a small amount for a large population. The appropriate
economic measure is therefore ex ante WTP for changes in risk. However, epidemiological
studies generally provide estimates of the relative risks of a particular health effect avoided
due to a reduction in air pollution. A convenient way to use this data in a consistent
framework is to convert probabilities to units of avoided statistical incidences. This measure
is calculated by dividing individual WTP for a risk reduction by the related observed change
4-50
-------
in risk. For example, suppose a measure is able to reduce the risk of premature mortality
from 2 in 10,000 to 1 in 10,000 (a reduction of 1 in 10,000). If individual WTP for this risk
reduction is $100, then the WTP for an avoided statistical premature mortality amounts to $1
million ($100/0.0001 change in risk). Using this approach, the size of the affected
population is automatically taken into account by the number of incidences predicted by
epidemiological studies applied to the relevant population. The same type of calculation can
produce values for statistical incidences of other health endpoints.
For some health effects, such as hospital admissions, WTP estimates are generally
not available. In these cases, we use the cost of treating or mitigating the effect as a primary
estimate. For example, 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 admission. These
COI estimates generally understate the true value of reductions in risk of a health effect.
They tend to reflect the direct expenditures related to treatment but not the value of avoided
pain and suffering from the health effect. Table 4-11 summarizes the value estimates per
health effect that we used in this analysis. Values are presented both for a 1990 base income
level and adjusted for income growth in the two future analysis years, 2010 and 2015. Note
that the unit values for hospital admissions are the weighted averages of the ICD-9 code-
specific values for the group of ICD-9 codes included in the hospital admission categories.
A discussion of the valuation methods for premature mortality and CB is provided here
because of the relative importance of these effects. Discussions of the methods used to value
nonfatal myocardial infarctions (heart attacks) and school absence days are provided because
these endpoints have only recently been added to the analysis and the valuation methods are
still under development. In the following discussions, unit values are presented at 1990
levels of income for consistency with previous analyses. Equivalent future-year values can
be obtained from Table 4-11. COI estimates are converted to constant 1999 dollar
equivalents using the medical CPI.
4.1.5.4.1 Valuing Reductions in Premature Mortality Risk. We estimate the
monetary benefit of reducing premature mortality risk using the VSL approach, which is a
summary measure for the value of small changes in mortality risk experienced by a large
number of people. The mean value of avoiding one statistical death is assumed to be $5.5
million in 1999 dollars. This represents a central value consistent with the range of values
suggested by recent meta-analyses of the wage-risk VSL literature. The distribution of VSL
is characterized by a confidence interval from $1 to $10 million, based on two meta-analyses
of the wage-risk VSL literature. The $1 million lower confidence limit represents the lower
4-51
-------
Table 4-11. Unit Values Used for Economic Valuation of Health Endpoints (1999$)
Central Estimate of Value Per Statistical Incidence
Health Endpoint
1990 Income
Level
2010 Income
Level
2015 Income
Level
Derivation of Estimates
Premature Mortality (Value of a
Statistical Life)
Chronic Bronchitis (CB)
Nonfatal Myocardial Infarction
(heart attack)
3% discount rate
$5,500,000 $6,000,000 $6,400,000 Point estimate is the mean of a normal distribution with a 95 percent
confidence interval between $1 and $10 million. Confidence interval is
based on two meta-analyses of the wage-risk VSL literature: $1 million
represents the lower end of the interquartile range from the Mrozek and
Taylor (2002) meta-analysis and $10 million represents the upper end of the
interquartile range from the Viscusi and Aldy (2003) meta-analysis. The
VSL represents the value of a small change in mortality risk aggregated over
the affected population.
$340,000 $380,000 $400,000 Point estimate is the mean of a generated 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.
Age-specific cost-of-illness values reflect lost earnings and direct medical
costs over a 5-year period following a nonfatal MI. Lost earnings estimates
are based on Cropper and Krupnick (1990). Direct medical costs are based
Age 0-24
Age 25^14
Age 45-54
Age 55-65
Age 66 and over
7% discount rate
Age 0-24
Age 25^14
Age 55-65
Age 66 and over
$66,902
$74,676
$78,834
$140,649
$66,902
$65,293
$73,149
$76,871
$132,214
$65,293
$66,902
$74,676
$78,834
$140,649
$66,902
$65,293
$73,149
$76,871
$132,214
$65,293
$66,902
$74,676
$78,834
$140,649
$66,902
$65,293
$73,149
$76,871
$132,214
$65,293
on simple average of estimates from Russell et al. (1998) and Wittels et al.
(1990).
Lost earnings:
Cropper and Krupnick (1990). Present discounted value of 5 years of lost
earnings:
aae of onset: at 3% at 7%
25-44 $8,774 $7,855
45-54 $12,932 $11,578
55-65 $74,746 $66,920
Direct medical expenses: An average of:
1. Wittels et al. (1990) ($102,658— no discounting)
2. Russell et al. (1998), 5-year period ($22,331 at 3% discount rate; $21,1 13
at 7% discount rate)
(continued)
4-52
-------
Table 4-11. Unit Values Used for Economic Valuation of Health Endpoints (1999$) (continued)
Central Estimate of Value Per Statistical
Incidence
Health Endpoint
1990 Income
Level
2010 Income
Level
2015 Income
Level
Derivation of Estimates
Hospital Admissions
Chronic Obstructive Pulmonary
Disease (COPD)
(ICD codes 490-492,494-496)
Pneumonia
(ICD codes 480-487)
Asthma Admissions
$12,378
$14,693
$6,634
$12,378
$14,693
All Cardiovascular
(ICD codes 390-429)
Emergency Room Visits for Asthma
$18,387
$286
$6,634
$18,387
$286
$12,378 The COI estimates (lost earnings plus direct medical costs) are based on
ICD-9 code-level information (e.g., average hospital care costs, average
length of hospital stay, and weighted share of total COPD category
illnesses) reported in Agency for Healthcare Research and Quality (2000)
(www.ahrq.gov).
$14,693 The COI estimates (lost earnings plus direct medical costs) are based on
ICD-9 code-level information (e.g., average hospital care costs, average
length of hospital stay, and weighted share of total pneumonia category
illnesses) reported in Agency for Healthcare Research and Quality (2000)
(www.ahrq.gov).
$6,634 The COI estimates (lost earnings plus direct medical costs) are based on
ICD-9 code-level information (e.g., average hospital care costs, average
length of hospital stay, and weighted share of total asthma category
illnesses) reported in Agency for Healthcare Research and Quality (2000)
(www.ahrq.gov).
$18,387 The COI estimates (lost earnings plus direct medical costs) are based on
ICD-9 code-level information (e.g., average hospital care costs, average
length of hospital stay, and weighted share of total cardiovascular category
illnesses) reported in Agency for Healthcare Research and Quality (2000)
(www.ahrq.gov).
$286 Simple average of two unit COI values:
(1) $311.55, from Smith et al. (1997) and
(2) $260.67, from Stanford et al. (1999).
(continued)
4-53
-------
Table 4-11. Unit Values Used for Economic Valuation of Health Endpoints (1999$) (continued)
Central Estimate of Value Per Statistical
Incidence
Health Endpoint
1990 Income
Level
2010 Income
Level
2015 Income
Level
Derivation of Estimates
Respiratory Ailments Not Requiring Hospitalization
Upper Respiratory Symptoms (URS) $25
$26
Lower Respiratory Symptoms (LRS)
$16
$17
Asthma Exacerbations
$42
$43
Acute Bronchitis
$360
$26 Combinations of the three symptoms for which WTP estimates are available
that closely match those listed by Pope et al. result in seven different
"symptom clusters," each describing a "type" of URS. A dollar value was
derived for each type of URS, using mid-range estimates of WTP (lEc,
1994) 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
seven different types of URS.
$17 Combinations of the four 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 dollar value was
derived for each type of LRS, using mid-range estimates of WTP (lEc,
1994) to avoid each symptom in the cluster and assuming additivity of
WTPs. The dollar value for LRS is the average of the dollar values for the
11 different types of LRS.
$44 Asthma exacerbations are valued at $42 per incidence, based on the mean of
average WTP estimates for the four severity definitions of a "bad asthma
day," described in Rowe and Chestnut (1986). This study surveyed
asthmatics to estimate WTP for avoidance of a "bad asthma day," as defined
by the subjects. For purposes of valuation, an asthma attack is assumed to
be equivalent to a day in which asthma is moderate or worse as reported in
the Rowe and Chestnut (1986) study.
$380 Assumes a 6-day episode, with daily value equal to the average of low and
high values for related respiratory symptoms recommended in Neumann et
al. (1994).
(continued)
4-54
-------
Table 4-11. Unit Values Used for Economic Valuation of Health Endpoints (1999$) (continued)
Central Estimate of Value Per Statistical Incidence
Health Endpoint
1990 Income
Level
2010 Income
Level
2015 Income
Level
Derivation of Estimates
Restricted Activity and Work/School Loss Days
Work Loss Days (WLDs)
School Absence Days
Variable
(national
median =)
$75
Worker Productivity
Minor Restricted Activity Days
(MRADs)
$0.95 per
worker per 10%
change in ozone
per day
$51
$75
$0.95 per
worker per
10% change in
ozone per day
$53
w
County-specific median annual wages divided by 50 (assuming 2 weeks of
vacation) and then by 5—to get median daily wage. U.S. Year 2000
Census, compiled by Geolytics, Inc.
$75 Based on expected lost wages from parent staying home with child.
Estimated daily lost wage (if a mother must stay at home with a sick child)
is based on the median weekly wage among women age 25 and older in
2000 (U.S. Census Bureau, Statistical Abstract of the United States: 2001,
Section 12: Labor Force, Employment, and Earnings, Table No. 621).
This median wage is $551. Dividing by 5 gives an estimated median daily
wage of $103.
The expected loss in wages due to a day of school absence in which the
mother would have to stay home with her child is estimated as the
probability that the mother is in the workforce times the daily wage she
would lose if she missed a day = 72.85% of $103, or $75.
Based on $68—median daily earnings of workers in farming, forestry and
fishing—from Table 621, Statistical Abstract of the United States ("Full-
Time Wage and Salary Workers—Number and Earnings: 1985 to 2000")
(Source of data in table: U.S. Bureau of Labor Statistics, Bulletin 2307 and
Employment and Earnings, monthly).
$54 Median WTP estimate to avoid one MRAD from Tolley et al. (1986).
$0.95 per
vvorker per 10%
change in ozone
per day
4-55
-------
end of the interquartile range from the Mrozek and Taylor (2002) meta-analysis. The $10
million upper confidence limit represents the upper end of the interquartile range from the
Viscusi and Aldy (2003) meta-analysis. Because the majority of the studies in these meta-
analyses are based on datasets from the early 1990s or previous decades, we continue to
assume that the VSL estimates provided by those meta-analyses are in 1990 income
equivalents. Future research might provide income-adjusted VSL values for individual
studies that can be incorporated into the meta-analyses. This would allow for a more reliable
base-year estimate for use in adjusting VSL for aggregate changes in income over time.
As indicated in the previous section on quantification of premature mortality benefits,
we assumed for this analysis that some of the incidences of premature mortality related to
PM exposures occur in a distributed fashion over the 20 years following exposure. To take
this into account in the valuation of reductions in premature mortality, we applied an annual
3 percent discount rate to the value of premature mortality occurring in future years.17
The economics literature concerning the appropriate method for valuing reductions in
premature mortality risk is still developing. The adoption of a value for the projected
reduction in the risk of premature mortality is the subject of continuing discussion within the
economics and public policy analysis community. Regardless of the theoretical economic
considerations, EPA prefers not to draw distinctions in the monetary value assigned to the
lives saved even if they differ in age, health status, socioeconomic status, gender, or other
characteristic of the adult population.
Following the advice of the EEAC of the SAB, EPA currently uses the VSL approach
in calculating the primary estimate of mortality benefits, because we believe this calculation
provides the most reasonable single estimate of an individual's willingness to trade off
money for reductions in mortality risk (EPA-SAB-EEAC-00-013, 2000). Although there are
several differences between the labor market studies EPA uses to derive a VSL estimate and
the PM air pollution context addressed here, those differences in the affected populations and
the nature of the risks imply both upward and downward adjustments. Table 4-12 lists some
of these differences and the expected effect on the VSL estimate for air pollution-related
17The choice of a discount rate, and its associated conceptual basis, is a topic of ongoing discussion within the
federal government. EPA adopted a 3 percent discount rate for its base estimate in this case to reflect
reliance on a "social rate of time preference" discounting concept. We have also calculated benefits and
costs using a 7 percent rate consistent with an "opportunity cost of capital" concept to reflect the time value
of resources directed to meet regulatory requirements. In this case, the benefit and cost estimates were not
significantly affected by the choice of discount rate. Further discussion of this topic appears in EPA's
Guidelines for Preparing Economic Analyses (EPA, 2000b).
4-56
-------
Table 4-12. Expected Impact on Estimated Benefits of Premature Mortality Reductions
of Differences Between Factors Used in Developing Applied VSL and Theoretically
Appropriate VSL
Attribute
Expected Direction of Bias
Age
Life Expectancy/Health Status
Attitudes Toward Risk
Income
Voluntary vs. Involuntary
Catastrophic vs. Protracted Death
Uncertain, perhaps overestimate
Uncertain, perhaps overestimate
Underestimate
Uncertain
Uncertain, perhaps underestimate
Uncertain, perhaps underestimate
mortality. In the absence of a comprehensive and balanced set of adjustment factors, EPA
believes it is reasonable to continue to use the $5.5 million value while acknowledging the
significant limitations and uncertainties in the available literature.
The SAB-EEAC has reviewed many potential VSL adjustments and the state of the
economics literature. The SAB-EEAC advised EPA to "continue to use a wage-risk-based
VSL as its primary estimate, including appropriate sensitivity analyses to reflect the
uncertainty of these estimates," and that "the only risk characteristic for which adjustments
to the VSL can be made is the timing of the risk" (EPA-SAB-EEAC-00-013, EPA, 2000). In
developing our primary estimate of the benefits of premature mortality reductions, we have
followed this advice and discounted over the lag period between exposure and premature
mortality.
Uncertainties Specific to Premature Mortality Valuation. The economic benefits
associated with premature mortality are the largest category of monetized benefits of the
final CAIR. In addition, in prior analyses, EPA has identified valuation of mortality benefits
as the largest contributor to the range of uncertainty in monetized benefits (see EPA
[1999a]).18 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
18This conclusion was based on a assessment of uncertainty based on statistical error in epidemiological effect
estimates and economic valuation estimates. Additional sources of model error such as those examined in
the pilot PM mortality expert elicitation may result in different conclusions about the relative contribution of
sources of uncertainty.
4-57
-------
economic approaches available for mortality valuation. Such an assessment also requires an
understanding of how alternative valuation approaches reflect that some individuals may be
more susceptible to air pollution-induced mortality or reflect differences in the nature of the
risk presented by air pollution relative to the risks studied in the relevant economics
literature.
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 elderly
and children). Health status prior to exposure also affects susceptibility. An ideal benefits
estimate of mortality risk reduction would reflect these human characteristics, in addition to
an individual's WTP to improve one's own chances of survival plus WTP to improve other
individuals' survival rates. The ideal measure would also 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. 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 depend on such characteristics 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 benefits of reduced risk of premature mortality associated with reducing air
pollution, the approach requires a great deal of data to implement. The economic valuation
literature does not yet include good estimates of the value of this risk reduction commodity.
As a result, in this study we value avoided premature mortality risk using the VSL approach.
Other uncertainties specific to premature mortality valuation include the following:
• Across-study variation: There is considerable uncertainty as to whether the
available literature on VSL provides adequate estimates of the VSL saved by air
pollution reduction. Although there is considerable variation in the analytical
designs and data used in the existing literature, 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 WTP to
4-58
-------
reduce the risk. The appropriateness of a distribution of WTP based on the
current VSL literature 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
the extent to which the risks being valued are similar and the extent to which the
subjects in the studies are similar to the population affected by changes in
pollution concentrations.
Level of risk reduction: The transferability of estimates of the VSL from the
wage-risk studies to the context of the CAIR analysis rests on the assumption that,
within a reasonable range, WTP for reductions in mortality risk is linear in risk
reduction. 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 reduction is 1/10,000. If WTP for
reductions in mortality risk is linear in risk reduction, then a WTP of $50 for a
reduction of 1/100,000 implies a WTP of $500 for a risk reduction of 1/10,000
(which is 10 times the risk reduction valued in the study). Under the assumption
of linearity, the estimate of the VSL does not depend on the particular amount of
risk reduction being valued. This assumption has been shown to be reasonable
provided the change in the risk being valued is within the range of risks evaluated
in the underlying studies (Rowlatt et al., 1998).
Voluntariness of risks evaluated: Although job-related mortality risks may differ
in several ways from air pollution-related mortality risks, the most important
difference may be that job-related risks are incurred voluntarily, or generally
assumed to be, whereas air pollution-related risks are incurred involuntarily.
Some evidence suggests 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 understate WTP to reduce involuntarily incurred
air pollution-related mortality risks.
Sudden versus protracted death: A final important difference related to the nature
of the risk may be that some workplace mortality 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 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.
Self-selection and skill in avoiding risk: Recent research (Shogren and Stamland,
2002) suggests that VSL estimates based on hedonic wage studies may overstate
4-59
-------
the average value of a risk reduction. This is based on the fact that the risk-wage
trade-off revealed in hedonic studies reflects the preferences of the marginal
worker (i.e., that worker who demands the highest compensation for his risk
reduction). This worker must have either higher risk, lower risk tolerance, or
both. However, the risk estimate used in hedonic studies is generally based on
average risk, so the VSL may be upwardly biased because the wage differential
and risk measures do not match.
4.1.5.4.2 Valuing Reductions in the Riskof Chronic Bronchitis. The best available
estimate of WTP to avoid a case of CB comes from Viscusi et al. (1991). The Viscusi et al.
study, however, describes a severe case of CB to the survey respondents. We therefore
employ an estimate of WTP to avoid a pollution-related case of CB, based on adjusting the
Viscusi et al. (1991) estimate of the WTP to avoid a severe case. This is done to account for
the likelihood that an average case of pollution-related CB is not as severe. The adjustment
is made by applying the elasticity of WTP with respect to severity reported in the Krupnick
and Cropper (1992) study. Details of this adjustment procedure are provided in the Benefits
TSD for the Nonroad Diesel rulemaking (Abt Associates, 2003).
We use the mean of a distribution of WTP estimates as the central tendency estimate
of WTP to avoid a pollution-related case of CB in this analysis. The distribution
incorporates uncertainty from three sources: the WTP to avoid a case of severe CB, as
described by Viscusi et al.; the severity level of an average pollution-related case of CB
(relative to that of the case described by Viscusi et al.); and the elasticity of WTP with
respect to severity of the illness. 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. The expected value (i.e.,
mean) of this distribution, which is about $331,000 (2000$), is taken as the central tendency
estimate of WTP to avoid a PM-related case of CB.
4.1.5.4.3 Valuing Reductions in Nonfatal Myocardial Infarctions (Heart A ttacks).
The Agency has recently incorporated into its analyses the impact of air pollution on the
expected number of nonfatal heart attacks, although it has examined the impact of reductions
in other related cardiovascular endpoints. We were not able to identify a suitable WTP value
for reductions in the risk of nonfatal heart attacks. Instead, we use a COI unit value with two
components: the direct medical costs and the opportunity cost (lost earnings) associated with
the illness event. Because the costs associated with a myocardial infarction extend beyond
the initial event itself, we consider costs incurred over several years. Using age-specific
annual lost earnings estimated by Cropper and Krupnick (1990) and a 3 percent discount
4-60
-------
rate, we estimated a present discounted value in lost earnings (in 2000$) over 5 years due to
a myocardial infarction of $8,774 for someone between the ages of 25 and 44, $12,932 for
someone between the ages of 45 and 54, and $74,746 for someone between the ages of 55
and 65. The corresponding age-specific estimates of lost earnings (in 2000$) using a 7
percent discount rate are $7,855, $11,578, and $66,920, respectively. Cropper and Krupnick
(1990) do not provide lost earnings estimates for populations under 25 or over 65. As such,
we do not include lost earnings in the cost estimates for these age groups.
We found three possible sources in the literature of estimates of the direct medical
costs of myocardial infarction:
• Wittels et al. (1990) estimated expected total medical costs of myocardial
infarction over 5 years to be $51,211 (in 1986$) for people who were admitted to
the hospital and survived hospitalization. (There does not appear to be any
discounting used.) Wittels et al. was used to value coronary heart disease in the
812 Retrospective Analysis of the Clean Air Act. Using the CPI-U for medical
care, the Wittels estimate is $109,474 in year 2000$. This estimated cost is based
on a medical cost model, which incorporated therapeutic options, projected
outcomes, and prices (using "knowledgeable cardiologists" as consultants). The
model used medical data and medical decision algorithms to estimate the
probabilities of certain events and/or medical procedures being used. The authors
note that the average length of hospitalization for acute myocardial infarction has
decreased over time (from an average of 12.9 days in 1980 to an average of 11
days in 1983). Wittels et al. used 10 days as the average in their study. It is
unclear how much further the length of stay for myocardial infarction may have
decreased from 1983 to the present. The average length of stay for ICD code 410
(myocardial infarction) in the year-2000 Agency for Healthcare Research and
Quality (AHRQ) HCUP database is 5.5 days. However, this may include patients
who died in the hospital (not included among our nonfatal myocardial infarction
cases), whose length of stay was therefore substantially shorter than it would be if
they had not died.
• Eisenstein et al. (2001) estimated 10-year costs of $44,663 in 1997$, or $49,651
in 2000$ for myocardial infarction patients, using statistical prediction
(regression) models to estimate inpatient costs. Only inpatient costs (physician
fees and hospital costs) were included.
• Russell et al. (1998) estimated first-year direct medical costs of treating nonfatal
myocardial infarction of $15,540 (in 1995$) and $1,051 annually thereafter.
Converting to year 2000$, that would be $23,353 for a 5-year period (without
discounting) or $29,568 for a 10-year period.
4-61
-------
In summary, the three different studies provided significantly different values (see
Table 4-13).
Table 4-13. Alternative Direct Medical Cost of Illness Estimates for Nonfatal Heart
Attacks
Study
Wittelsetal. (1990)
Russell etal. (1998)
Eisenstein et al. (2001)
Russell etal. (1998)
Direct Medical Costs (2000$)
$109,474a
$22,33 lb
$49,65 lb
$27,242b
Over an x-Year Period, for x =
5
5
10
10
a Wittels et al. did not appear to discount costs incurred in future years.
b Using a 3 percent discount rate.
As noted above, the estimates from these three studies are substantially different, and
we have not adequately resolved the sources of differences in the estimates. Because the
wage-related opportunity cost estimates from Cropper and Krupnick (1990) cover a 5-year
period, we used estimates for medical costs that similarly cover a 5-year period (i.e.,
estimates from Wittels et al. (1990) and Russell et al. (1998). We used a simple average of
the two 5-year estimates, or $65,902, and added it to the 5-year opportunity cost estimate.
The resulting estimates are given in Table 4-14.
Table 4-14. Estimated Costs Over a 5-Year Period (in 2000$) of a Nonfatal Myocardial
Infarction
Age Group
0-24
25^4
45-54
55-65
>65
Opportunity Cost
$0
$8,774b
$12,253b
$70,6 19b
$0
Medical Cost3
$65,902
$65,902
$65,902
$65,902
$65,902
Total Cost
$65,902
$74,676
$78,834
$140,649
$65,902
a An average of the 5-year costs estimated by Wittels et al. (1990) and Russell et al. (1998).
b From Cropper and Krupnick (1990), using a 3 percent discount rate.
4-62
-------
4.1.5.4.4 Valuing Reductions in School Absence Days. School absences associated
with exposure to ozone are likely to be due to respiratory-related symptoms and illnesses.
Because the respiratory symptom and illness endpoints we are including are all PM-related
rather than ozone-related, we do not have to be concerned about double-counting of benefits
if we aggregate the benefits of avoiding ozone-related school absences with the benefits of
avoiding PM-related respiratory symptoms and illnesses.
One possible approach to valuing a school absence is using a parental opportunity
cost approach. This method requires two steps: estimate the probability that, if a school-
aged child stays home from school, a parent will have to stay home from work to care for the
child, and value the lost productivity at the person's wage. Using this method, we would
estimate the proportion of families with school-age children in which both parents work and
value a school loss day as the probability of a work loss day resulting from a school loss day
(i.e., the proportion of households with school-age children in which both parents work)
times some measure of lost wages (whatever measure we use to value work loss days).
There are two significant problems with this method, however. First, it omits WTP to avoid
the symptoms/illness that resulted in the school absence. Second, it effectively gives zero
value to school absences that do not result in a work loss day (unless we derive an alternative
estimate of the value of the parent's time for those cases in which the parent is not in the
labor force). We are investigating approaches using WTP for avoid the symptoms/illnesses
causing the absence. In the interim, we use the parental opportunity cost approach.
For the opportunity-cost approach, we make an explicit, conservative assumption that
in married households with two working parents, the female parent will stay home with a
sick child. From the U.S. Census Bureau, Statistical Abstract of the United States: 2001, we
obtained 1) the numbers of single, married, and "other" (i.e., widowed, divorced, or
separated) women with children in the workforce and 2) the rates of participation in the
workforce of single, married, and "other" women with children. From these two sets of
statistics, we inferred the numbers of single, married, and "other" women with children and
the corresponding percentages. These percentages were used to calculate a weighted-average
participation rate, as shown in Table 4-15.
Our estimated daily lost wage (if a mother must stay at home with a sick child) is
based on the median weekly wage among women age 25 and older in 2000 (U.S. Census
Bureau, Statistical Abstract of the United States: 2001, Section 12: Labor Force,
Employment, and Earnings, Table No. 621). This median wage is $551. Dividing by 5 gives
an estimated median daily wage of $103.
4-63
-------
Table 4-15. Women with Children: Number and Percentage in the Labor Force, 2000,
and Weighted Average Participation Ratea
Single
Married
Otherb
Total:
Number (in
millions) in
Labor Force
(1)
3.1
18.2
4.5
Participation
Rate
(2)
73.9%
70.6%
82.7%
Implied Total
Number in
Population (in
millions)
(3) = (l)/(2)
4.19
25.78
5.44
35.42
Implied
Percent in
Population
(4)
11.84%
72.79%
15.36%
Weighted
Average
Participation
Rate [=sum
(2)*(4) over
rows]
72.85%
Data in columns (1) and (2) are from U.S. Census Bureau, Statistical Abstract of the United States: 2001,
Section 12: Labor Force, Employment, and Earnings, Table No. 577.
Widowed, divorced, or separated.
The expected loss in wages due to a day of school absence in which the mother would
have to stay home with her child is estimated as the probability that the mother is in the
workforce times the daily wage she would lose if she missed a day = 72.85 percent of $103,
or $75.19 Of course, non-working parent time also has value. Determining that value is not
straightforward. In a world with perfect labor markets, economic theory suggests that a
non-working parent's time is worth at least (and generally exactly) that of a working parent.
Otherwise, they would choose to work. Imperfect labor markets could imply a lower value
of time, but still one above zero. Compounding the uncertainty about the value of time is
uncertainty about the amount of time sacrificed by a non-working parent to care for a sick
child. Since the value of a non-working parent's time is a function of the usual activities
they perform, and caring for a child is presumably one of those activities, the crucial question
is how much additional time is spent caring for a sick child during a school absence and what
other activities are foregone as a result. Assuming a negligible reallocation of these activities
19In a recent article, Hall et al. (2003) use a similar methodology to derive a mid-estimate value per school
absence day for California of between $70 and $81, depending on differences in incomes between three
counties in California. Our national average estimate of $75 per absence is consistent with these published
values.
4-64
-------
and thus zero costs of a school absence to a household with a non-working parent provides a
lower-bound on the benefits estimate for averting these impacts.
4.1.6 Human Welfare Impact Assessment
PM and ozone have numerous documented effects on environmental quality that
affect human welfare. These welfare effects include direct damages to property, either
through impacts on material structures or by soiling of surfaces, direct economic damages in
the form of lost productivity of crops and trees, indirect damages through alteration of
ecosystem functions, and indirect economic damages through the loss in value of recreational
experiences or the existence value of important resources. EPA's Criteria Documents for
PM and ozone list numerous physical and ecological effects known to be linked to ambient
concentrations of these pollutants (EPA, 1996a; 1996b). This section describes individual
effects and how we quantify and monetize them. These effects include changes in
commercial crop and forest yields, visibility, and nitrogen deposition to estuaries.
4.1.6.1 Visibility Benefits
Changes in the level of ambient PM caused by the reduction in emissions from CAIR
will change the level of visibility in much of the Eastern United States. Visibility directly
affects people's enjoyment of a variety of daily activities. Individuals value visibility both in
the places they live and work, in the places they travel to for recreational purposes, and at
sites of unique public value, such as the Great Smokey Mountains National Park. This
section discusses the measurement of the economic benefits of improved visibility.
It is difficult to quantitatively define a visibility endpoint that can be used for
valuation. Increases in PM concentrations cause increases in light extinction, a measure of
how much the components of the atmosphere absorb light. More light absorption means that
the clarity of visual images and visual range is reduced, ceteris paribus. Light absorption is
a variable that can be accurately measured. Sisler (1996) created a unitless measure of
visibility, the deciview, based directly on the degree of measured light absorption.
Deciviews are standardized for a reference distance in such a way that one deciview
corresponds to a change of about 10 percent in available light. Sisler characterized a change
in light extinction of one deciview as "a small but perceptible scenic change under many
4-65
-------
circumstances." Air quality models were used to predict the change in visibility, measured
in deciviews, of the areas affected by the control options.20
EPA considers benefits from two categories of visibility changes: residential
visibility and recreational visibility. In both cases economic benefits are believed to consist
of use values and nonuse values. Use values include the aesthetic benefits of better visibility,
improved road and air safety, and enhanced recreation in activities like hunting and
birdwatching. Nonuse values are based on people's beliefs that the environment ought to
exist free of human-induced haze. Nonuse values may be more important for recreational
areas, particularly national parks and monuments.
Residential visibility benefits are those that occur from visibility changes in urban,
suburban, and rural areas and also in recreational areas not listed as federal Class I areas.21
For the purposes of this analysis, recreational visibility improvements are defined as those
that occur specifically in federal Class I areas. A key distinction between recreational and
residential benefits is that only those people living in residential areas are assumed to receive
benefits from residential visibility, while all households in the United States are assumed to
derive some benefit from improvements in Class I areas. Values are assumed to be higher if
the Class I area is located close to their home.22
Only two existing studies provide defensible monetary estimates of the value of
visibility changes. One is a study on residential visibility conducted in 1990 (McClelland et
al., 1993) and the other is a 1988 survey on recreational visibility value (Chestnut and Rowe,
1990a; 1990b). Although there are a number of other studies in the literature, they were
conducted in the early 1980s and did not use methods that are considered defensible by
current standards. Both the Chestnut and Rowe and McClelland et al. studies use the CV
method. There has been a great deal of controversy and significant development of both
20A change of less than 10 percent in the light extinction budget represents a measurable improvement in
visibility but may not be perceptible to the eye in many cases. Some of the average regional changes in
visibility are less than one deciview (i.e., less than 10 percent of the light extinction budget) and thus less
than perceptible. However, this does not mean that these changes are not real or significant. Our
assumption is then that individuals can place values on changes in visibility that may not be perceptible.
This is quite plausible if individuals are aware that many regulations lead to small improvements in visibility
that, when considered together, amount to perceptible changes invisibility.
21The Clean Air Act designates 156 national parks and wilderness areas as Class I areas for visibility protection.
22For details of the visibility estimates discussed in this chapter, please refer to the Benefits TSD for the
Nonroad Diesel rulemaking (Abt Associates, 2003).
4-66
-------
theoretical and empirical knowledge about how to conduct CV surveys in the past decade. In
EPA's judgment, the Chestnut and Rowe study contains many of the elements of a valid CV
study and is sufficiently reliable to serve as the basis for monetary estimates of the benefits
of visibility changes in recreational areas.23 This study serves as an essential input to our
estimates of the benefits of recreational visibility improvements in the primary benefits
estimates. Consistent with SAB advice, EPA has designated the McClelland et al. study as
significantly less reliable for regulatory benefit-cost analysis, although it does provide useful
estimates on the order of magnitude of residential visibility benefits (EPA-SAB-COUNCIL-
ADV-00-002, 1999). Residential visibility benefits are not calculated for this analysis.
The Chestnut and Rowe study measured the demand for visibility in Class I areas
managed by the National Park Service (NFS) in three broad regions of the country:
California, the Southwest, and the Southeast. Respondents in five states were asked about
their WTP to protect national parks or NPS-managed wilderness areas within a particular
region. The survey used photographs reflecting different visibility levels in the specified
recreational areas. The visibility levels in these photographs were later converted to
deciviews for the current analysis. The survey data collected were used to estimate a WTP
equation for improved visibility. In addition to the visibility change variable, the estimating
equation also included household income as an explanatory variable.
The Chestnut and Rowe study did not measure values for visibility improvement in
Class I areas outside the three regions. Their study covered 86 of the 156 Class I areas in the
United States. We can infer the value of visibility changes in the other Class I areas by
transferring values of visibility changes at Class I areas in the study regions. A complete
description of the benefits transfer method used to infer values for visibility changes in Class
I areas outside the study regions is provided in the Benefits TSD for the Nonroad Diesel
rulemaking (Abt Associates, 2003).
The Chestnut and Rowe study (Chestnut and Rowe, 1990a; 1990b), although
representing the best available estimates, has a number of limitations. These include the
following:
23An SAB advisory letter indicates that "many members of the Council believe that the Chestnut and Rowe
study is the best available" (EPA-SAB-COUNCIL-ADV-00-002, 1999, p. 13). However, the committee did
not formally approve use of these estimates because of concerns about the peer-reviewed status of the study.
EPA believes the study has received adequate review and has been cited in numerous peer-reviewed
publications (Chestnut and Dennis, 1997).
4-67
-------
• The age of the study (late 1980s) will increase the uncertainty about the
correspondence of the estimated values to those that might be provided by current
or future populations.
• The survey focused only on populations in five states, so the application of the
estimated values to populations outside those states requires that preferences of
populations in the five surveyed states be similar to those of nonsurveyed states.
• There is an inherent difficulty in separating values expressed for visibility
improvements from an overall value for improved air quality. The Chestnut and
Rowe study attempted to control for this by informing respondents that "other
households are being asked about visibility, human health, and vegetation
protections in urban areas and at national parks in other regions." However, most
of the respondents did not feel that they were able to segregate visibility at
national parks entirely from residential visibility and health effects.
• It is not clear exactly what visibility improvements the respondents to the
Chestnut and Rowe survey were valuing. For the purpose of the benefits analysis
for this rule, EPA assumed that respondents provided values for changes in
annual average visibility. Because most policies will result in a shift in the
distribution of visibility (usually affecting the worst days more than the best
days), the annual average may not be the most relevant metric for policy analysis.
• The WTP question asked about changes in average visibility. However, the
survey respondents were shown photographs of only summertime conditions,
when visibility is generally at its worst. It is possible that the respondents
believed those visibility conditions held year-round, in which case they would
have been valuing much larger overall improvements in visibility than what
otherwise would be the case.
• The survey did not include reminders of possible substitutes (e.g., visibility at
other parks) or budget constraints. These reminders are considered to be best
practice for stated preference surveys.
• The Chestnut and Rowe survey focused on visibility improvements in and around
national parks and wilderness areas. The survey also focused on visibility
improvements of national parks in the southwest United States. Given that
national parks and wilderness areas exhibit unique characteristics, it is not clear
whether the WTP estimate obtained from Chestnut and Rowe can be transferred
to other national parks and wilderness areas, without introducing additional
uncertainty.
In general, the survey design and implementation reflect the period in which the survey was
conducted. Since that time, many improvements to the stated preference methodology have
4-68
-------
been developed. As future survey efforts are completed, EPA will incorporate values for
visibility improvements reflecting the improved survey designs.
The estimated relationship from the Chestnut and Rowe study is only directly
applicable to the populations represented by survey respondents. EPA used benefits transfer
methodology to extrapolate these results to the population affected by the final CAIR. A
general WTP equation for improved visibility (measured in deciviews) was developed as a
function of the baseline level of visibility, the magnitude of the visibility improvement, and
household income. The behavioral parameters of this equation were taken from analysis of
the Chestnut and Rowe data. These parameters were used to calibrate WTP for the visibility
changes resulting from CAIR. The method for developing calibrated WTP functions is based
on the approach developed by Smith et al. (2002). Available evidence indicates that
households are willing to pay more for a given visibility improvement as their income
increases (Chestnut, 1997). The benefits estimates here incorporate Chestnut's estimate that
a 1 percent increase in income is associated with a 0.9 percent increase in WTP for a given
change in visibility.
Using the methodology outlined above, EPA estimates that the total WTP for the
visibility improvements in Southeastern Class I areas brought about by CAIR is $1.14 billion
in 2010 and $1.78 billion in 2015. This value includes the value to households living in the
same state as the Class I area as well as values for all households in the United States living
outside the state containing the Class I area, and the value accounts for growth in real
income.
The benefits resulting from visibility improvements in Southeastern Class I areas
under the final CAIR are presented in Figure 4-2. This figure presents these benefits both in
terms of the total benefits modeled for each of the Class I areas (i.e., the "Park Benefits"
map) and the benefits realized by the populations in each of the 48 contiguous states (i.e., the
"State Benefits" map). The latter results reflect the WTP of state residents for visibility
improvements occurring in Class I areas in the Southeastern United States.
One major source of uncertainty for the visibility benefits estimate is the benefits
transfer process used. Judgments used to choose the functional form and key parameters of
the estimating equation for WTP for the affected population could have significant effects on
the size of the estimates. Assumptions about how individuals respond to changes in visibility
that are either very small or outside the range covered in the Chestnut and Rowe study could
also affect the results.
4-69
-------
\
/•• Liiiville Gorge
J { ~ <
j Joyce Kilmer- S'lickrock
: Dolb- Sods \'
Benefits in 2015 U.S. Dollars
« 0 to 1
® 1 to 10 million
HI 10 to 100 million
HP 100 million to 1 billion!
11 to 1.3 billion
Qkefenokee
: Ctassihowitzbi j
...I Evetglades
Figure 4-2. CAIR Final Rule Visibility Improvements in Class I Areas in the Southeast
4.1.6.2 Agricultural, Forestry, and Other Vegetation-Related Benefits
The Ozone Criteria Document notes that ozone affects vegetation throughout the
United States, impairing crops, native vegetation, and ecosystems more than any other air
pollutant (EPA, 1996a, page 5-11). Changes in ground-level ozone resulting from the
control options are expected to affect crop and forest yields throughout the affected area.
Well-developed techniques exist to provide monetary estimates of these benefits to
agricultural producers and to consumers. These techniques use models of planting decisions,
yield response functions, and the supply of and demand for agricultural products. The
resulting welfare measures are based on predicted changes in market prices and production
costs. Models also exist to measure benefits to silvicultural producers and consumers.
However, these models have not been adapted for use in analyzing ozone-related forest
impacts. Because of resource limitations, we are unable to provide agricultural or benefits
estimates for the final CAIR rule.
4.1.6.2.1 Agricultural Benefits. Laboratory and field experiments have shown
reductions in yields for agronomic crops exposed to ozone, including vegetables (e.g.,
4-70
-------
lettuce) and field crops (e.g., cotton and wheat). The most extensive field experiments,
conducted under the National Crop Loss Assessment Network (NCLAN), examined 15
species and numerous cultivars. The NCLAN results show that several economically
important crop species are sensitive to ozone levels typical of those found in the United
States (EPA, 1996a). In addition, economic studies have shown a relationship between
observed ozone levels and crop yields (Garcia et al., 1986).
4.1.6.2.2 Forestry Benefits. Ozone also has been shown conclusively to cause
discernible injury to forest trees (EPA, 1996a; Fox and Mickler, 1996). In our previous
analysis of the HD Engine/Diesel Fuel rule, we were able to quantify the effects of changes
in ozone concentrations on tree growth for a limited set of species. Because of resource
limitations, we were not able to quantify such impacts for this analysis.
4.1.6.2.3 Other Vegetation Effects. An additional welfare benefit expected to accrue
as a result of reductions in ambient ozone concentrations in the United States is the economic
value the public receives from reduced aesthetic injury to forests. There is sufficient
scientific information available to reliably establish that ambient ozone levels cause visible
injury to foliage and impair the growth of some sensitive plant species (EPA, 1996a).
However, present analytic tools and resources preclude EPA from quantifying the benefits of
improved forest aesthetics.
Urban ornamentals (floriculture and nursery crops) represent an additional vegetation
category likely to experience some degree of negative effects associated with exposure to
ambient ozone levels and likely to affect large economic sectors. In the absence of adequate
exposure-response functions and economic damage functions for the potential range of
effects relevant to these types of vegetation, no direct quantitative economic benefits analysis
has been conducted. The farm production value of ornamental crops was estimated at over
$14 billion in 2003 (USDA, 2004). This is therefore a potentially important welfare effects
category. However, information and valuation methods are not available to allow for
plausible estimates of the percentage of these expenditures that may be related to impacts
associated with ozone exposure.
The EGU program, by reducing NOX emissions, will also reduce nitrogen deposition
on agricultural land and forests. There is some evidence that nitrogen deposition may have
positive effects on agricultural output through passive fertilization. Holding all other factors
constant, farmers' use of purchased fertilizers or manure may increase as deposited nitrogen
is reduced. Estimates of the potential value of this possible increase in the use of purchased
fertilizers are not available, but it is likely that the overall value is very small relative to other
4-71
-------
health and welfare effects. The share of nitrogen requirements provided by this deposition is
small, and the marginal cost of providing this nitrogen from alternative sources is quite low.
In some areas, agricultural lands suffer from nitrogen over saturation due to an abundance of
on-farm nitrogen production, primarily from animal manure. In these areas, reductions in
atmospheric deposition of nitrogen represent additional agricultural benefits.
Information on the effects of changes in passive nitrogen deposition on forests and
other terrestrial ecosystems is very limited. The multiplicity of factors affecting forests,
including other potential stressors such as ozone, and limiting factors such as moisture and
other nutrients, confound assessments of marginal changes in any one stressor or nutrient in
forest ecosystems. However, reductions in the deposition of nitrogen could have negative
effects on forest and vegetation growth in ecosystems where nitrogen is a limiting factor
(EPA, 1993).
On the other hand, there is evidence that forest ecosystems in some areas of the
United States are nitrogen saturated (EPA, 1993). Once saturation is reached, adverse effects
of additional nitrogen begin to occur such as soil acidification, which can lead to leaching of
nutrients needed for plant growth and mobilization of harmful elements such as aluminum.
Increased soil acidification is also linked to higher amounts of acidic runoff to streams and
lakes and leaching of harmful elements into aquatic ecosystems.
4.1.6.3 Benefits from Reductions in Materials Damage
The control options that we modeled are expected to produce economic benefits in
the form of reduced materials damage. There are two important categories of these benefits.
Household soiling refers to the accumulation of dirt, dust, and ash on exposed surfaces.
Criteria pollutants also have corrosive effects on commercial/industrial buildings and
structures of cultural and historical significance. The effects on historic buildings and
outdoor works of art are of particular concern because of the uniqueness and irreplaceability
of many of these objects.
Previous EPA benefits analyses have been able to provide quantitative estimates of
household soiling damage. Consistent with SAB advice, we determined that the existing
data (based on consumer expenditures from the early 1970s) are too out of date to provide a
reliable estimate of current household soiling damages (EPA-SAB-COUNCIL-ADV-98-003,
1998).
EPA is unable to estimate any benefits to commercial and industrial entities from
reduced materials damage. Nor is EPA able to estimate the benefits of reductions in PM-
4-72
-------
related damage to historic buildings and outdoor works of art. Existing studies of damage to
this latter category in Sweden (Grosclaude and Soguel, 1994) indicate that these benefits
could be an order of magnitude larger than household soiling benefits.
4.1.6.4 Benefits from Reduced Ecosystem Damage
The effects of air pollution on the health and stability of ecosystems are potentially
very important but are at present poorly understood and difficult to measure. The reductions
in NOX caused by the final rule could produce significant benefits. Excess nutrient loads,
especially of nitrogen, cause a variety of adverse consequences to the health of estuarine and
coastal waters. These effects include toxic and/or noxious algal blooms such as brown and
red tides, low (hypoxic) or zero (anoxic) concentrations of dissolved oxygen in bottom
waters, the loss of submerged aquatic vegetation due to the light-filtering effect of thick algal
mats, and fundamental shifts in phytoplankton community structure (Bricker et al., 1999).
Direct functions relating changes in nitrogen loadings to changes in estuarine benefits
are not available. The preferred WTP-based measure of benefits depends on the availability
of these functions and on estimates of the value of environmental responses. Because neither
appropriate functions nor sufficient information to estimate the marginal value of changes in
water quality exist at present, calculation of a WTP measure is not possible.
If better models of ecological effects can be defined, EPA believes that progress can
be made in estimating WTP measures for ecosystem functions. These estimates would be
superior to avoided cost estimates in placing economic values on the welfare changes
associated with air pollution damage to ecosystem health. For example, if nitrogen or sulfate
loadings can be linked to measurable and definable changes in fish populations or definable
indexes of biodiversity, then CV studies can be designed to elicit individuals' WTP for
changes in these effects. This is an important area for further research and analysis and will
require close collaboration among air quality modelers, natural scientists, and economists.
4.2 Benefits Analysis—Results
Applying the impact and valuation functions described previously in this chapter to
the estimated changes in ozone and PM yields estimates of the changes in physical damages
(e.g., premature mortalities, cases, admissions, change in light extinction) and the associated
monetary values for those changes. Estimates of physical health impacts are presented in
Table 4-16. Monetized values for both health and welfare endpoints are presented in
Table 4-17, along with total aggregate monetized benefits. All of the monetary benefits are
in constant-year 1999 dollars.
4-73
-------
Table 4-16. Clean Air Interstate Rule: Estimated Reduction in Incidence of Adverse
Health Effects3
2010 2015
Health Effect Incidence Reduction
PM-Related Endpoints
Premature Mortality15'0
Adult, age 30 and over
Infant, age <1 year
Chronic bronchitis (adult, age 26 and over)
Nonfatal myocardial infarction (adults, age 18 and older)
Hospital admissions — respiratory (all ages)d
Hospital admissions — cardiovascular (adults, age >18)e
Emergency room visits for asthma (age 18 years and younger)
Acute bronchitis (children, age 8-12)
Lower respiratory symptoms (children, age 7-14)
Upper respiratory symptoms (asthmatic children, age 9-18)
Asthma exacerbation (asthmatic children, age 6-18)
Work loss days (adults, age 18-65)
Minor restricted-activity days (adults, age 18-65)
13,000
29
6,900
17,000
4,300
3,800
10,000
16,000
190,000
150,000
240,000
1,400,000
8,100,000
17,000
36
8,700
22,000
5,500
5,000
13,000
19,000
230,000
180,000
290,000
1,700,000
9,900,000
Ozone-Related Endpoints
Hospital admissions — respiratory causes (adult, 65 and older/
Hospital admissions — respiratory causes (children, under 2)
Emergency room visit for asthma (all ages)
Minor restricted-activity days (adults, age 18-65)
School absence days
610
380
100
280,000
180,000
1,700
1,100
280
690,000
510,000
a Incidences are rounded to two significant digits. These estimates represent benefits for CAIR Nationwide
for the final CAIR program inclusive of the proposal to include SO2 and annual NOX controls for New Jersey
and Delaware. Note these estimates may be slightly overstated due to the inclusion of SO2 and annual NOX
controls for Arkansas.
b Premature mortality benefits associated with ozone are not addressed in the primary analysis.
0 PM premature mortality impacts for adults are based on application of the effect estimate derived from the
Pope et al (2002) cohort study. Infant premature mortality based upon studies by Woodruff, et al 1997.
d Respiratory hospital admissions for PM include admissions for COPD, pneumonia, and asthma.
e Cardiovascular hospital admissions for PM include total cardiovascular and subcategories for ischemic heart
disease, dysrhythmias, and heart failure.
f Respiratory hospital admissions for ozone include admissions for total respiratory and subcategories for
COPD and pneumonia.
4-74
-------
Table 4-17. Estimated Monetary Value in Reductions in Incidence of Health and
Welfare Effects (in millions of 1999$)a b
Health Effect
Premature mortalityc'd
Adult, age 30 and over
3% discount rate
7% discount rate
Infant, < 1 year
Chronic bronchitis (adults, 26 and over)
Non-fatal acute myocardial infarctions
3% discount rate
7% discount rate
Hospital admissions for respiratory causes
Hospital admissions for cardiovascular causes
Emergency room visits for asthma
Acute bronchitis (children, age 8-12)
Lower respiratory symptoms (children, 7-14)
Upper respiratory symptoms (asthma, 9-11)
Asthma exacerbations
Work loss days
Minor restricted-activity days (MRADs)
School absence days
Worker productivity (outdoor workers, 18-65)
Recreational visibility, 81 Class I areas
Monetized Total6
Base Estimate:
3% discount rate
7% discount rate
Pollutant
PM25
PM25
PM25
PM2 5, 03
PM25
PM25,O3
PM25
PM25
PM25
PM25
PM25,
PM25,03
03
03
PM25
PM25,03
2010
Estimated
$67,300
$56,600
$168
$2,520
$1,420
$1,370
$45.2
$80.7
$2.84
$5.63
$2.98
$3.80
$10.3
$180
$422
$12.9
$7.66
$1,140
$73,300 H
$62.600 H
2015
Value of Reductions
$92,800
$78,100
$222
$3,340
$1,850
$1,790
$78.9
$105
$3.56
$7.06
$3.74
$4.77
$12.7
$219
$543
$36.4
$19.9
$1,780
-B $101,000 + B
- B $86.300 + B
Monetary benefits are rounded to three significant digits for ease of presentation and computation. Benefits in
this table are nationwide (with the exception of visibility and ozone) and are associated with NOX and SO2
reductions at EGU sources. Ozone benefits relate to the eastern United States. Visibility benefits relate to Class
I areas in the southeastern United States. These estimates relate to the final CAIR program inclusive of the
proposal to include SO2 and annual NOX controls for New Jersey and Delaware. Note that these estimates may
be slightly overstated due to the inclusion of SO2 and annual NOX controls for Arkansas in the modeling.
Monetary benefits adjusted to account for growth in real GDP per capita between 1990and the analysis year
(2010 or 2015)
Valuation assumes discounting over the SAB recommended 20 year segmented lag structure described earlier.
Results reflect the use of 3 percent and 7 percent discount rates consistent with EPA and OMB guidelines for
preparing economic analyses (EPA, 2000b; OMB, 2003).
Adult premature mortality estimates based upon studies by Pope, et al 2002. Infant premature mortality based
upon Woodruff et al 1997.
B represents the monetary value of health and welfare benefits and disbenefits not monetized. A detailed listing
is provided in Table 4-2.
4-75
-------
Not all known PM- and ozone-related health and welfare effects could be quantified or
monetized. The monetized value of these unquantified effects is represented by adding an
unknown "B" to the aggregate total. The estimate of total monetized health benefits is thus
equal to the subset of monetized PM- and ozone-related health and welfare benefits plus B,
the sum of the nonmonetized health and welfare benefits.
Total monetized benefits are dominated by benefits of mortality risk reductions. The
primary analysis estimate projects that the final rule will result in 13,000 avoided premature
deaths annually in 2010 and 17,000 avoided premature deaths annually in 2015. The
increase in annual benefits from 2010 to 2015 reflects additional emission reductions from
the standards, as well as increases in total population and the average age (and thus baseline
mortality risk) of the population. Note that unaccounted for changes in baseline mortality
rates over time may lead to reductions in the estimated number of avoided premature
mortalities.
Our estimate of total monetized benefits in 2010 for the final rule is $73.3 billion
using a 3 percent discount rate and $62.6 billion using a 7 percent discount rate. In 2015, the
monetized benefits are estimated at $101 billion using a 3 percent discount rate and $86.3
billion using a 7 percent discount rate. Health benefits account for 98 percent of total
benefits, in part because we are unable to quantify most of the nonhealth benefits. These
unquantified benefits may be substantial and could exceed the costs of the rule, although the
magnitude of these benefits is highly uncertain. The monetized benefit associated with
reductions in the risk of premature mortality, which accounts for $67.3 billion in 2010 and
$92.8 billion in 2015, is over 90 percent of total monetized health benefits. The next largest
benefit is for reductions in chronic illness (CB and nonfatal heart attacks), although this
value is more than an order of magnitude lower than for premature mortality. Hospital
admissions for respiratory and cardiovascular causes, visibility, MRADs, work loss days,
school absence days, and worker productivity account for the majority of the remaining
benefits. The remaining categories each account for a small percentage of total benefit;
however, they represent a large number of avoided incidences affecting many individuals. A
comparison of the incidence table to the monetary benefits table reveals that there is not
always a close correspondence between the number of incidences avoided for a given
endpoint and the monetary value associated with that endpoint. For example, there are
almost 100 times more work loss days than premature mortalities, yet work loss days account
for only a very small fraction of total monetized benefits. This reflects the fact that many of
the less severe health effects, while more common, are valued at a lower level than the more
severe health effects. Also, some effects, such as hospital admissions, are valued using a
4-76
-------
proxy measure of WTP. As such, the true value of these effects may be higher than that
reported in Table 4-16.
Ozone benefits are in aggregate positive for the nation. However, because ozone
increases occur during certain hours of the day in some urban areas, there is a dampening of
overall ozone benefits in both 2010 and 2015, although the net incidence and benefits
estimates for all health effects categories are net positive. Overall, ozone benefits are low
relative to PM benefits for similar endpoint categories because of the increases in ozone
concentrations during some hours of some days in certain urban areas.
4.2.1 Potential Benefits of the New Jersey and Delaware Proposal
The benefit estimate results presented reflect benefits for the final CAIR program
inclusive of the New Jersey and Delaware proposal. Air quality modeling was not conducted
for the New Jersey and Delaware proposal. For this reason, an analysis of the potential
benefits for the New Jersey and Delaware proposal could not be completed with any degree
of specificity. However based on the air quality modeling results for the CAIR, we make
rough estimates of the benefits that might occur with this proposal. Including New Jersey
and Delaware in the CAIR program would result in additional reductions of SO2 and NOX
emissions. We estimate that approximately $630 million of the total annual CAIR program
benefits previously discussed could be attributable to annual SO2 and NOX controls for New
Jersey and Delaware in 2010 and approximately $1.1 billion could be attributable to annual
SO2 and NOX controls for New Jersey and Delaware in 2015.
4.3 Probabilistic Analysis of Uncertainty in the Benefits Estimates
The recent NRC report on estimating public health benefits of air pollution regulations
recommended that EPA begin to move the assessment of uncertainties from its ancillary
analyses into its primary analyses by conducting probabilistic, multiple-source uncertainty
analyses (NRC, 2002). The probability distributions required for these analyses should be
based on available data and expert judgment. The NRC also recommended that EPA use
both internal and external experts as needed, in each case identifying those experts whose
judgments are used and the rationales and empirical bases for their judgments.
As part of an overall program to improve the Agency's characterization of
uncertainties in health benefits analyses, this section describes EPA's initial efforts to
address uncertainties associated with the PM mortality C-R relationship and valuation.
Similar to our approach in the Nonroad Diesel RIA, for the CAIR, we present two types of
probabilistic approaches. The first approach generates a distribution of benefits based on the
4-77
-------
sampling error and variability in the underlying health and economic valuation studies used
in the benefits modeling framework. The second approach uses the results from a pilot
expert elicitation project designed to characterize key aspects of uncertainty in the ambient
PM2 5/mortality relationship. Both approaches provide insights into the likelihood of
different outcomes and about the state of knowledge regarding the benefits estimates. Both
approaches have different strengths and weaknesses, that are summarized below.
We provide likelihood distributions both for the total dollar benefits estimate and for
the incidence of premature mortality to show the uncertainty described by each expert's
judgment relative to the range of uncertainty associated with the standard error in the Pope et
al. (2002) study. The uncertainty about the total dollar benefit associated with any single
endpoint combines the uncertainties from two sources—the C-R relationship and the
valuation—and is estimated with a Monte Carlo method.24 Our estimates of the likelihood
distributions for total benefits should be viewed within the context of the wide range of
sources of uncertainty that we have not incorporated, including uncertainty in emissions, air
quality, and baseline health effect incidence rates.
In benefit analyses of air pollution regulations conducted to date, the estimated impact
of reductions in premature mortality has accounted for 85 to 95 percent of total benefits.
Therefore, in characterizing the uncertainty related to the estimates of total benefits it is
particularly important to attempt to characterize the uncertainties associated with this
endpoint. We conducted two different Monte Carlo analyses, one based on the distribution
of reductions in premature mortality characterized by the mean effect estimate and standard
error from the Pope et al. (2002) study (our primary estimate), and one based on the results
from a pilot expert elicitation project (lEc, 2004). The Pope et al. study is described earlier
in this chapter. A detailed discussion of the pilot expert elicitation project is provided in
Appendix B. We summarize several key points about the project below.
As a first step in addressing the NRC recommendations regarding expert elicitation,
EPA, in collaboration with OMB, conducted a pilot expert elicitation to characterize
uncertainties in the relationship between ambient PM2 5 and mortality. This pilot was
designed to provide EPA with an opportunity to improve its understanding of the design and
24In each iteration of the Monte Carlo procedure, a value is randomly drawn from the incidence distribution, and
a value is randomly drawn from the unit dollar value distribution. The total dollar benefit for that iteration is
the product of the two. If this is repeated for many (e.g., thousands of) iterations, the distribution of total
dollar benefits associated with the endpoint is generated. For details on the specific Monte Carlo approach
we used, see Appendix B.
4-78
-------
application of expert elicitation methods to economic benefits analysis, to lay the
groundwork for a more comprehensive elicitation, and to provide more information about the
uncertainty in the PM2 5-mortality relationship in the context of the Nonroad Diesel RIA and
similar analyses conducted in the near term.
The pilot project elicited the judgments of five experts in the PM health sciences, all
members of at least one of two recent National Academy of Sciences scientific committees
focused on particulate matter. The specific process used to select experts is summarized in
Appendix B and detailed in the technical report describing the elicitation (lEc, 2004) along
with additional information about the experts' affiliations and fields of expertise. The
responses of each expert to questions enumerated in the elicitation protocol provide the
inputs for developing distributions of mortality benefits.
The elicitation approach tested in the pilot was peer reviewed by four experts
(Mansfield, 2004) and generally received favorable comments, including
• the pilot project followed "best practices" for expert elicitation and was well
documented,
• the elicitation was well conducted and is an appropriate technique for
characterizing uncertainty,
• experts chosen are well known and respected and represent a range of views, and
• the expert selection process/protocol was very good.
The peer-review report also provides recommendations on how to improve the process
for a more comprehensive expert elicitation addressing the PM mortality C-R relationship.
Specifically, the peer-review recommendations included
• holding a pre-elicitation workshop to ensure that all experts are properly
motivated and conditioned (i.e., on the same footing) before the interviews and to
allow for information sharing prior to the elicitation;
• allowing for some form of communication following the individual interviews to
allow review of the experts' responses and allow them to adjust their estimates if
necessary—one way to accomplish this is through a post-elicitation workshop;
and
• changing the encoding process to ensure that extreme values (upper and lower
ranges) are collected prior to judgments on central tendency. This sequencing
would avoid anchoring or adjustment heuristics associated with biased estimates
of uncertainty.
4-79
-------
In addition, although not listed explicitly as a criticism, some of the peer reviewers
noted the small number of experts participating in the pilot and suggested a larger number of
experts should be used in EPA's next elicitation.
The peer reviewers also offered varying comments on the methods for combining the
results of the pilot elicitation. Several of the reviewers preferred that the expert opinions not
be combined or stated that they knew of no agreed-upon method for combining results from
expert elicitations. They stated that presenting each expert's response independently allows
for differences in the individual distributions to be recognized. Two of the reviewers
indicated that they were reasonably comfortable with the method used in this study to
combine the results, while the other two reviewers offered comments on the combined result
of the elicitation. One reviewer stated that the combined distributions do not adequately
capture the opinions of individual experts but rather average them out. He states that it is
possible in such cases that the combined judgments may generate results that none of the
experts would agree on. Another reviewer stated that expert elicitation studies typically do
not combine judgments, but if one were to combine them, he recommended that the response
of each be maintained independently from the other experts and run through the benefits
model completely prior to combining the results.
For more details regarding the peer-review comments, see Appendix B. The full peer-
review report is also available at www.epa.gov/ttn/ecas/benefits.html.
The distributions of all other, nonmortality health endpoints are characterized by the
reported mean and standard deviations from the epidemiology literature. Details on the
distributions used for individual health endpoints are provided in Appendix B. We are
unable at this time to characterize the uncertainty in the estimate of benefits of improvements
in visibility at Class I areas. As such, we treat the visibility benefits as fixed and add them to
all percentiles of the health benefits distribution. Given this unequal treatment of endpoints,
it is likely that these distributions do not capture the full range of benefits, and in fact are
likely to understate the uncertainty, especially on the high end of the range due to omission
of potentially significant benefit categories. We include them here primarily as an
illustration of the impacts of using probabilistic (expert elicitation and statistical error-based)
distributions for premature mortality associated with PM25 compared with EPA's traditional
approach.
Figure 4-3 presents box plots of the distributions of the reduction in PM2 5-related
premature mortality based on the C-R distributions provided by each expert, as well as that
for our primary estimate, based on the statistical error associated with Pope et al. (2002).
4-80
-------
60,000
50,000
Distribution o
Primary
Distributions Based on Application of Pilot Expert Elicitation Results
o
CM
c
I"
s
o
ro
£
c
O
&
40,000
Estimate
30,000
20,000
10,000
16.70C
13,700
7,100
2,800
11,600
22,800
Pope et al (2002)
Statistical Error
Expert A
Expert B
Expert C
Expert D
Expert E
Note: Distributions labeled Expert A - Expert E are based on individual expert responses. The distribution labeled Pope et al. (2002) Statistical Error is based on the mean
and standard error of the C-R function from the study.
Figure 4-3. Results of Illustrative Application of Pilot Expert Elicitation: Annual Reductions in Premature Mortality in
2015 Associated with the Clean Air Interstate Rule
4-81
-------
The distributions are depicted as box plots with the diamond symbol (+) showing the mean,
the dash (-) showing the median (50th percentile), the box defining the interquartile range
(bounded by the 25th and 75th percentiles), and the whiskers defining the 90 percent
confidence interval (bounded by the 5th and 95th percentiles of the distribution). Our
primary estimate based on the Pope et al. (2002) study shows that the average number of
premature deaths avoided in 2015 is 16,700. This is higher than four of the experts and
lower than one expert and falls within the uncertainty bounds of all but one expert. The
figure shows that the average annual number of premature deaths avoided in 2015 ranges
from approximately 2,800 (based on the judgments of Expert C) to 22,800 (based on the
judgments of Expert E). The medians span zero to 20,000, with the zero value due to the
high threshold associated with one of the expert's distributions. The statistical uncertainty
bounds of all of the estimates, including the Pope et al.-based distribution, overlap.
Although the uncertainty bounds for each expert include zero, and some distributions have
significant percentiles at zero, all of the distributions have a positive mean estimate.
Figure 4-4 presents box plots of the distributions of monetized benefits of reductions
in premature mortality associated with use of the Pope et al. (2002) and expert-based
mortality incidence distributions. Our primary estimate based on the Pope et al. (2002) study
shows that the mean annual benefit is roughly $93 billion. Mean annual benefits for each
expert elicited during the pilot expert elicitation range from approximately $16 billion (based
on judgments of Expert C) to $130 billion (based on the judgments of Expert E). Impacts on
the distribution of total benefits (including visibility and non-mortality health benefits) are
discussed in Appendix B.
The uncertainty estimates based on statistical error have the strength of presenting a
statistical measure of the uncertainty in the underlying studies serving as the basis for the
estimates used in the analysis. However, this approach captures only a limited portion of the
uncertainty about the parameters. The 5th and 95th percentile points of the distributions are
based on statistical error and cross-study variability and provide some insight into how
uncertain our estimate is with regard to those sources of uncertainty. However, it does not
capture other sources of uncertainty regarding the model specification and other inputs to the
model, including emissions, air quality, and aspects of the health science not captured in the
studies, such as the likelihood that PM is causally related to premature mortality and other
serious health effects.
4-82
-------
$400,000
$300,000
Distribution of
Primary
Estimate
Distributions Based on Application of Pilot Expert Elicitation
Results
0 $200,000
$100,000
$0
$76,000
• $40,000
$16,000
$64,000
$130,000
Pope et al (2002) Statistical
Error
Expert A
Expert B
Expert C
Expert D
Expert E
Note: Mortality distributions labeled Expert A - Expert E are based on individual expert responses. The mortality distribution labeled Pope et al. (2002) statistical error is based on the mean
and standard error of the C-R function from the study. Mortality valuation is based on a normally distributed VSL with a mean of $5.5 million and a 95% Cl between $1 and $10 million. The
VSL distribution has then been adjsuted for income growth out to 2015 using an adjustment factor of 1.15.
Figure 4-4. Results of Illustrative Application of Pilot Expert Elicitation: Dollar Value of Annual Reductions in Premature
Mortality in 2015 Associated with the Clean Air Interstate Rule
4-83
-------
4.4 Discussion
This analysis has estimated the health and welfare benefits of reductions in ambient
concentrations of particulate matter and ozone resulting from reduced emissions of NOX and
SO2 from affected EGUs. The result suggests there will be significant health and welfare
benefits arising from regulating emissions from EGUs in the United States. Our estimate
that 17,000 premature mortalities would be avoided when the emissions reductions from the
regulation are fully realized provides additional evidence of the important role that pollution
from the EGU sector plays in the public health impacts of air pollution.
Other uncertainties that we could not quantify include the importance of unquantified
effects and uncertainties in the modeling of ambient air quality. Inherent in any analysis of
future regulatory programs are uncertainties in projecting atmospheric conditions and source-
level emissions, as well as population, health baselines, incomes, technology, and other
factors. The assumptions used to capture these elements are reasonable based on the
available evidence. However, data limitations prevent an overall quantitative estimate of the
uncertainty associated with estimates of total economic benefits. If one is mindful of these
limitations, the magnitude of the benefits estimates presented here can be useful information
in expanding the understanding of the public health impacts of reducing air pollution from
EGUs.
EPA will continue to evaluate new methods and models and select those most
appropriate for estimating the health benefits of reductions in air pollution. It is important to
continue improving benefits transfer methods in terms of transferring economic values and
transferring estimated impact functions. The development of both better models of current
health outcomes and new models for additional health effects such as asthma, high blood
pressure, and adverse birth outcomes (such as low birth weight) will be essential to future
improvements in the accuracy and reliability of benefits analyses (Guo et al., 1999; Ibald-
Mulli et al., 2001). Enhanced collaboration between air quality modelers, epidemiologists,
toxicologists, and economists should result in a more tightly integrated analytical framework
for measuring health benefits of air pollution policies.
4-84
-------
4.5 References
Abbey, D.E., S.D. Colome, P.K. Mills, R. Burchette, W.L. Beeson, and Y. Tian. 1993b.
"Chronic Disease Associated With Long-Term Concentrations of Nitrogen Dioxide."
Journal of Exposure Analysis and Environmental Epidemiology 3 (2): 181 -202.
Abbey, D.E., B.L. Hwang, RJ. Burchette, T. Vancuren, and P.K. Mills. 1995. "Estimated
Long-Term Ambient Concentrations of PM(10) and Development of Respiratory
Symptoms in a Nonsmoking Population." Archives of Environmental Health 50(2):
139-152.
Abbey, D.E., N. Nishino, W.F. McDonnell, RJ. Burchette, S.F. Knutsen, W. Lawrence
Beeson, and J.X. Yang. 1999. "Long-term inhalable particles and other air pollutants
related to mortality in nonsmokers [see comments]." American Journal of
Respiratory and Critical Care Medicine 159(2):373-382.
Abt Associates, Inc. 1995. Urban Ornamental Plants: Sensitivity to Ozone and Potential
Economic Losses. Prepared for the U.S. Environmental Protection Agency, Office of
Air Quality Planning and Standards; Research Triangle Park, NC.
Abt Associates, Inc. April 2003. ProposedNonroad Land-based Diesel Engine Rule: Air
Quality Estimation, Selected Health and Welfare Benefits Methods, and Benefit
Analysis Results. Prepared for Office of Air Quality Planning and Standards, U.S.
EPA.
Adams, P.P., G.E. Hendershot, and M.A. Marano. 1999. "Current Estimates from the
National Health Interview Survey, 1996." Vital Health Statistics 10(200): 1-212.
Agency for Healthcare Research and Quality (AHRQ). 2000. HCUPnet, Healthcare Cost
and Utilization Project.
American Lung Association. 1999. "Chronic Bronchitis." Available at
http://www.lungusa.org/diseases/lungchronic.html.
American Lung Association. 2002a. Trends in Asthma Morbidity andMortality. American
Lung Association, Best Practices and Program Services, Epidemiology and Statistics
Unit.
American Lung Association. 2002b. Trends in Chronic Bronchitis and Emphysema:
Morbidity and Mortality. American Lung Association, Best Practices and Program
Services, Epidemiology and Statistics Unit.
4-85
-------
American Lung Association. 2002c. Trends in Morbidity and Mortality: Pneumonia,
Influenza, and Acute Respiratory Conditions. American Lung Association, Best
Practices and Program Services, Epidemiology and Statistics Unit.
Ayyub, B.M. 2002. Elicitationof Expert Opinions for Uncertainty and Risk. CRC Press,
Florida.
Banzhaf, S., D. Burtraw, D. Evans, and A. Krupnick. September 2004. "Valuation of
Natural Resource Improvements in the Adirondacks," Resources for the Future.
Banzhaf, S., D. Burtraw, and K. Palmer. October 2002. "Efficient Emission Fees in the U.S.
Electricity Sector." Resources for the Future Discussion Paper 02-45.
Belanger, K., W. Beckett, E. Triche, M.B. Bracken, T. Holford, P. Ren, I.E. McSharry, D.R.
Gold, T.A. Platts-Mills, and B.P. Leaderer. 2003. "Symptoms of Wheeze and
Persistent Cough in the First Year of Life: Associations with Indoor Allergens, Air
Contaminants, and Maternal History of Asthma." American Journal of Epidemiology
158:195-202.
Bell, M.L., A. McDermott, S.L. Zeger, J.M. Samet, and F. Dominici. 2004. "Ozone and
Short-term Mortality in 95 U.S. Urban Communities, 1987-2000." Journal of the
American Medical Association 292:23 72-23 78.
Bell, M., A. McDermott, S. Zeger, J. Samet, and F. Dominici. 2005. "A Meta-Analysis of
Time-Series Studies of Ozone and Mortality with Comparison to the National
Morbidity, Mortality, and Air Pollution Study." Epidemiology (In Press).
Berger, M.C., G.C. Blomquist, D. Kenkel, and G.S. Tolley. 1987. "Valuing Changes in
Health Risks: A Comparison of Alternative Measures." The Southern Economic
Journal 53:977-984.
Bricker, S.B., C.G. Clement, D.E. Pirhalla, S.P. Orlando, and D.R.G. Farrow. 1999.
National Estuarine Eutrophication Assessment: Effects of Nutrient Enrichment in the
Nation's Estuaries. National Oceanic and Atmospheric Administration, National
Ocean Service, Special Projects Office and the National Centers for Coastal Ocean
Science. Silver Spring, Maryland.
4-86
-------
Burnett, R.T., M. Smith-Doiron, D. Stieb, M.E. Raizenne, J.R. Brook, R.E Dales, J.A. Leech,
S. Cakmak, D. Krewski. 2001. "Association between Ozone and Hospitalization for
Acute Respiratory Diseases in Children less than 2 Years of Age." American Journal
of Epidemiology 153:444-52.
CDC Wonder, Wide-ranging OnLine Data for Epidemiologic Research (Wonder) (data from
years 1996-1998), Centers for Disease Control and Prevention (CDC), U.S.
Department of Health and Human Services, http://wonder.cdc.gov.
Carnethon, M.R., D. Liao, G.W. Evans, W.E. Cascio, L.E. Chambless, W.D. Rosamond, and
G. Heiss. 2002. "Does the Cardiac Autonomic Response to Postural Change Predict
Incident Coronary Heart Disease and Mortality? The Atherosclerosis Risk in
Communities Study." American Journal of Epidemiology 155(l):48-56.
Chay, K.Y., and M. Greenstone. 2003. "The Impact of Air Pollution on Infant Mortality:
Evidence from Geographic Variation in Pollution Shocks Induced by a Recession."
Quarterly Journal of Economics 118(3).
Chen, L., B.L. Jennison, W. Yang, and S.T. Omaye. 2000. "Elementary School
Absenteeism and Air Pollution." Inhalation Toxicology 12(11):997-1016.
Chestnut, L.G. April 15, 1997. Draft Memorandum: Methodology for Estimating Values
for Changes in Visibility at National Parks.
Chestnut, L.G. and R.L. Dennis. 1997. "Economic Benefits of Improvements in Visibility:
Acid Rain Provisions of the 1990 Clean Air Act Amendments." Journal of the Air
and Waste Management Association 47:395-402.
Chestnut, L.G., and R.D. Rowe. 1990a. "A New National Park Visibility Value Estimates."
In Visibility and Fine Particles, Transactions of an AWMA/EPA International
Specialty Conference, C. V. Mathai, ed. Air and Waste Management Association,
Pittsburgh.
Chestnut, L.G., and R.D. Rowe. 1990b. Preservation Values for Visibility Protection at the
National Parks: Draft Final Report. Prepared for Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC, and
Air Quality Management Division, National Park Service, Denver, CO.
4-87
-------
Cody, R.P., C.P. Weisel, G. Birnbaum, and P.J. Lioy. 1992. "The Effect of Ozone
Associated with Summertime Photochemical Smog on the Frequency of Asthma
Visits to Hospital Emergency Departments." Environmental Research 58(2): 184-94.
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.
Cropper, M.L., and AJ. Krupnick. 1990. "The Social Costs of Chronic Heart and Lung
Disease." Resources for the Future. Washington, DC. Discussion Paper QE 89-16-
REV.
Daniels, M.J., F. Dominici, J.M. Samet, and S.L. Zeger. 2000. "Estimating Particulate
Matter-Mortality Dose-Response Curves and Threshold Levels: An Analysis of
Daily Time-Series for the 20 Largest U.S. Cities." American Journal of
Epidemiology 152(5):397-406.
Dekker, J.M., R.S. Crow, A.R. Folsom, P.J. Hannan, D. Liao, C.A. Swenne, and E.G.
Schouten. 2000. "Low Heart Rate Variability in a 2-Minute Rhythm Strip Predicts
Risk of Coronary Heart Disease and Mortality From Several Causes: The ARIC
Study." Circulation 2000 102:1239-1244.
Dockery, D.W., C.A. Pope, X.P. Xu, J.D. Spengler, J.H. Ware, M.E. Fay, E.G. Ferris, and
F.E. Speizer. 1993. "An Association between Air Pollution and Mortality in Six
U.S. Cities." New England Journal of Medicine 329(24): 1753-1759.
Dockery, D.W., J. Cunningham, A.I. Damokosh, L.M. Neas, J.D. Spengler, P. Koutrakis,
J.H. Ware, M. Raizenne, and F.E. Speizer. 1996. "Health Effects of Acid Aerosols
On North American Children-Respiratory Symptoms." Environmental Health
Perspectives 104(5): 500-505.
Eisenstein, E.L., L.K. Shaw, K.J. Anstrom, C.L. Nelson, Z. Hakim, V. Hasselblad and D.B.
Mark. 2001. "Assessing the Clinical and Economic Burden of Coronary Artery
Disease: 1986-1998." Medical Care 39(8):824-35.
EPA-SAB-COUNCIL-ADV-98-003. 1998. "Advisory Council on Clean Air Compliance
Analysis Advisory on the Clean Air Act Amendments (CAAA) of 1990 Section 812
Prospective Study: Overview of Air Quality and Emissions Estimates: Modeling,
Health and Ecological Valuation Issues Initial Studies."
4-88
-------
EPA-SAB-COUNCIL-ADV-99-012. July 1999. "The Clean Air Act Amendments (CAAA)
Section 812 Prospective Study of Costs and Benefits (1999): Advisory by the Health
and Ecological Effects Subcommittee on Initial Assessments of Health and
Ecological Effects: Part 1."
EPA-SAB-COUNCIL-ADV-99-005. February 1999. "An SAB Advisory on the Health and
Ecological Effects Initial Studies of the Section 812 Prospective Study: Report to
Congress: Advisory by the Health and Ecological Effects Subcommittee."
EPA-SAB-COUNCIL-ADV-01-004. 2001. "Review of the Draft Analytical Plan for EPA's
Second Prospective Analysis—Benefits and Costs of the Clean Air Act 1990-2020:
An Advisory by a Special Panel of the Advisory Council on Clean Air Compliance
Analysis." September.
EPA-SAB-EEAC-00-013. July 2000. "An SAB Report on EPA's White Paper Valuing the
Benefits of Fatal Cancer Risk Reduction."
Fox, S., and R.A. Mickler. 1996. "Impact of Air Pollutants on Southern Pine Forests."
Ecological Studies 118. New York: Springer Verlag.
Freeman, A.M. III. 1993. The Measurement of Environmental and Resource Values: Theory
and Methods. Washington, DC: Resources for the Future.
Garcia, P., B. Dixon, and J. Mjelde. 1986. "Measuring the Benefits of Environmental
Change Using a Duality Approach: The Case of Ozone and Illinois Cash Grain
Farms." Journal of Environmental Economics and Management 13:69-80.
Gilliland, F.D., K. Berhane, E.B. Rappaport, D.C. Thomas, E. Avol, W.J. Gauderman, S.J.
London, H.G. Margolis, R. McConnell, K.T. Islam, and J.M. Peters. 2001. "The
Effects of Ambient Air Pollution on School Absenteeism due to Respiratory
Illnesses." Epidemiology 12(l):43-54.
Gold, D.R., A. Litonjua, J. Schwartz, E. Lovett, A. Larson, B. Nearing, G. Allen, M. Verrier,
R. Cherry., and R. Verrier. 2000. "Ambient Pollution and Heart Rate Variability."
Circulation 101(11): 1267-73.
4-89
-------
Greenbaum, D. 2002. Letter to colleagues dated May 30, 2002. [Available at
www.healtheffects.org]. Letter from L.D. Grant, Ph.D. to Dr. P. Hopke re: external
review of EPA's Air Quality Criteria for Particulate Matter, with copy of 05/30/02
letter from Health Effects Institute re: re-analysis of National Morbidity, Mortality
and Air Pollution Study data attached. Docket No. A-2000-01. Document No. IV-A-
145.
Grosclaude, P., and N.C. Soguel. 1994. "Valuing Damage to Historic Buildings Using a
Contingent Market: A Case Study of Road Traffic Externalities." Journal of
Environmental Planning and Management 37: 279-287.
Guo, Y.L., Y.C. Lin, F.C. Sung, S.L. Huang, Y.C. Ko, J.S. Lai, H.J. Su, C.K. Shaw, R.S. Lin,
and D.W. Dockery. 1999. "Climate, Traffic-Related Air Pollutants, and Asthma
Prevalence in Middle-School Children in Taiwan." Environmental Health
Perspectives 107:1001-1006.
Hall, J.V., V. Brajer, and F.W. Lurmann. 2003. "Economic Valuation of Ozone-related
School Absences in the South Coast Air Basin of California." Contemporary
Economic Policy 21 (4): 407-417.
Harrington, W., and P.R. Portney. 1987. "Valuing the Benefits of Health and Safety
Regulation." Journal of Urban Economics 22:101-112.
Health Effects Institute (HEI). 2003. Revised Analyses of Time-Series Studies of Air
Pollution and Health; Revised Analyses of the National Morbidity, Mortality and Air
Pollution Study, Part II; Revised Analyses of Selected Time-Series Studies. Health
Effects Institute, Boston, MA.
Hollman, F.W., T.J. Mulder, and I.E. Kalian. January 2000. "Methodology and
Assumptions for the Population Projections of the United States: 1999 to 2100."
Population Division Working Paper No. 38, Population Projections Branch,
Population Division, U.S. Census Bureau, Department of Commerce.
Ibald-Mulli, A., J. Stieber, H.-E. Wichmann, W. Koenig, and A. Peters. 2001. "Effects of
Air Pollution on Blood Pressure: A Population-Based Approach." American Journal
of Public Health 91:571-577.
Industrial Economics, Incorporated (TEc). March 31., 1994. Memorandum to Jim
DeMocker, Office of Air and Radiation, Office of Policy Analysis and Review, U.S.
Environmental Protection Agency.
4-90
-------
Industrial Economics, Incorporated (ffic). April 2004. "Expert Judgment Assessment of the
Relationship Between PM2 5 Exposure and Mortality." Available at
.
Ito, K. 2003. "Associations of Parti culate Matter Components with Daily Mortality and
Morbidity in Detroit, Michigan." In Revised Analyses of Time-Series Studies of Air
Pollution and Health. Special Report. Health Effects Institute, Boston, MA.
Ito, K., S. De Leon, and M. Lippman. 2005. Associations between ozone and daily
mortality: a review and an additional analysis. Epidemiology (In Press).
Ito, K., and G.D. Thurston. 1996. "Daily PMlO/Mortality Associations: An Investigations
of At-Risk Subpopulations." Journal of Exposure Analysis and Environmental
Epidemiology 6(l):79-95.
Ito, K., and G.D. Thurston. 2001. "Epidemiological studies of acute ozone exposures and
mortality." Journal of Exposure Analysis and Environmental Epidemiology
ll(4):286-294.
Kleckner, N., and J. Neumann. June 3, 1999. "Recommended Approach to Adjusting WTP
Estimates to Reflect Changes in Real Income." Memorandum to Jim Democker, U.S.
EPA/OPAR.
Krewski D., R.T. Burnett, M.S. Goldbert, K. Hoover, J. Siemiatycki, M. Jerrett, M.
Abrahamowicz, and W.H. White. July 2000. Reanalysis of the Harvard Six Cities
Study and the American Cancer Society Study of Paniculate Air Pollution and
Mortality. Special Report to the Health Effects Institute, Cambridge MA.
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.
Kunzli, N., S. Medina, R. Kaiser, P. Quenel, F. Horak Jr, and M. Studnicka. 2001.
"Assessment of Deaths Attributable to Air Pollution: Should We Use Risk Estimates
Based on Time Series or on Cohort Studies?" American Journal of Epidemiology
153(ll):1050-55.
Kunzli, N., R. Kaiser, S. Medina, M. Studnicka, O. Chanel, P. Filliger, M. Kerry, F. Horak Jr.,
V. Puybonnieux-Texier, P. Quenel, J. Schneider, R. Seethaler, J-C Vergnaud, and H.
Sommer. 2000. "Public-Health Impact of Outdoor and Traffic-Related Air Pollution:
A European Assessment." The Lancet 356:795-801.
4-91
-------
Lave, L.B., and E.P. Seskin. 1977. Air Pollution and Human Health. Baltimore: Johns
Hopkins University Press for Resources for the Future.
Levy, J.I., J.K. Hammitt, Y. Yanagisawa, and J.D. Spengler. 1999. "Development of a New
Damage Function Model for Power Plants: Methodology and Applications."
Environmental Science and Technology 33:4364-4372.
Levy, J.I., TJ. Carrothers, J.T. Tuomisto, J.K. Hammitt, and J.S. Evans. 2001. "Assessing
the Public Health Benefits of Reduced Ozone Concentrations." Environmental Health
Perspectives 109:1215-1226.
Levy, J.I., S. M. Chemerynski, and J.A. Sarnat. 2005. "Ozone Exposure and Mortality Risk:
An Empirical Bayes Meta-Regression Analysis." Epidemiology (In Press).
Liao, D., J. Cai, W.D. Rosamond, R.W. Barnes, R.G. Hutchinson, E.A. Whitsel, P.
Rautaharju, and G. Heiss. 1997. "Cardiac Autonomic Function and Incident
Coronary Heart Disease: A Population-Based Case-Cohort Study. The ARIC Study.
Atherosclerosis Risk in Communities Study." American Journal of Epidemiology
145(8):696-706.
Liao, D., J. Creason, C. Shy, R. Williams, R. Watts, and R. Zweidinger. 1999. "Daily
Variation of Particulate Air Pollution and Poor Cardiac Autonomic Control in the
Elderly." Environ Health Perspect 107:521-5.
Lipfert, F.W., S.C. Morris, and R.E. Wyzga. 1989. "Acid Aerosols—the Next Criteria Air
Pollutant." Environmental Science & Technology 23(11):1316-1322.
Lipfert, F.W., H. Mitchell Perry Jr., J. Philip Miller, Jack D. Baty, Ronald E. Wyzg, and
Sharon E. Carmody. 2000. "The Washington University-EPRI Veterans' Cohort
Mortality Study: Preliminary Results." Inhalation Toxicology 12:41-74.
Magari, S.R., R. Hauser, J. Schwartz, P.L. Williams, T.J. Smith, and D.C. Christiani. 2001.
"Association of Heart rate Variability with Occupational and Environmental Exposure
to Paniculate Air Pollution." Circulation 104(9):986-91.
Mansfield, C. 2004. "Peer Review of Expert Elicitation" Research Triangle Park, NC. RTI
International. Available at: .
4-92
-------
McClelland, G., W. Schulze, D. Waldman, J. Irwin, D. Schenk, T. Stewart, L. Deck, and M.
Thayer. September 1993. Valuing Eastern Visibility: A Field Test of the Contingent
Valuation Method. Prepared for Office of Policy, Planning and Evaluation, U.S.
Environmental Protection Agency.
McConnell, R., K. Berhane, F. Gilliland, SJ. London, H. Vora, E. Avol, WJ. Gauderman,
H.G. Margolis, F. Lurmann, D.C. Thomas, and J.M. Peters. 1999. "Air Pollution and
Bronchitic Symptoms in Southern California Children with Asthma." Environmental
Health Perspectives 107(9):757-760.
McConnell, R., K. Berhane, F. Gilliland, SJ. London, T. Islam, WJ. Gauderman, E. Avol,
H.G. Margolis, and J.M. Peters. 2002. "Asthma in Exercising Children Exposed to
Ozone: A Cohort Study." Lancet 359(9309):896.
Moolgavkar, S.H. 2000. "Air Pollution and Hospital Admissions for Diseases of the
Circulatory System in Three U.S. Metropolitan Areas." Journal of the Air and Waste
Management Association 50:1199-1206.
Moolgavkar, S.H. 2003. "Air Pollution and Daily Deaths and Hospital Admissions in Los
Angeles and Cook Counties." In Revised Analyses of Time-Series Studies of Air
Pollution and Health. Special Report. Boston, MA: Health Effects Institute.
Moolgavkar, S.H., E.G. Luebeck, and E.L. Anderson. 1997. "Air Pollution and Hospital
Admissions for Respiratory Causes in Minneapolis-St. Paul and Birmingham."
Epidemiology 8:364-370.
Mrozek, J.R., and L.O. Taylor. 2002. "What Determines the Value of Life? AMeta-
Analysis." Journal of Policy Analysis and Management 21(2):253-270.
National Hospital Ambulatory Medical Care Survey (NHAMCS). 2000. National Center for
Health Statistics, Centers for Disease Control and Prevention (CDC), US Department
of Health and Human Services. Available at: .
NHDS Public Use Data Files. 1999. Centers for Disease Control and Prevention (CDC), U.S.
Department of Health and Human Services. Available at: ftp://ftp.cdc.gov/pub/
Health Statistics/NCHS/datasets/NHDS.
4-93
-------
National Health Interview Survey (NHIS). 2000. National Center for Health Statistics,
Centers for Disease Control and Prevention (CDC), U.S. Department of Health and
Human Services, Web site: ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/
Datasets/NHIS/2000/.
National Center for Education Statistics. 1996. "The Condition of Education 1996, Indicator
42: Student Absenteeism and Tardiness." Washington, DC: U.S. Department of
Education National Center for Education Statistics.
National Research Council (NRC). 1998. Research Priorities for Airborne Paniculate
Matter: I. Immediate Priorities and a Long-Range Research Portfolio. Washington,
DC: The National Academies Press.
National Research Council (NRC). 2002. Estimating the Public Health Benefits of Proposed
Air Pollution Regulations. Washington, DC: The National Academies Press.
Neumann, I.E., M.T. Dickie, and R.E. Unsworth. March 31, 1994. "Linkage Between Health
Effects Estimation and Morbidity Valuation in the Section 812 Analysis—Draft
Valuation Document." Industrial Economics Incorporated (lEc) Memorandum to Jim
DeMocker, U.S. Environmental Protection Agency, Office of Air and Radiation,
Office of Policy Analysis and Review.
Norris, G., S.N. YoungPong, J.Q. Koenig, T.V. Larson, L. Sheppard, and J.W. Stout. 1999.
"An Association between Fine Particles and Asthma Emergency Department Visits for
Children in Seattle." Environmental Health Perspectives 107(6):489-493.
Ostro, B.D. 1987. "Air Pollution and Morbidity Revisited: A Specification Test." Journal of
Environmental Economics Management 14:87-98.
Ostro, B.D. and S. Rothschild. 1989. "Air Pollution and Acute Respiratory Morbidity: An
Observational Study of Multiple Pollutants." Environmental Research 50:238-247.
Ostro, B.D., MJ. Lipsett, M.B. Wiener, and J.C. Seiner. 1991. "Asthmatic Responses to
Airborne Acid Aerosols." American Journal of Public Health 81(6):694-702.
Ostro, B., and L. Chestnut. 1998. "Assessing the Health Benefits of Reducing Particulate
Matter Air Pollution in the United States." Environmental Research, Section A,
76:94-106.
4-94
-------
Ostro, B., M. Lipsett, J. Mann, H. Braxton-Owens, and M. White. 2001. "Air Pollution and
Exacerbation of Asthma in African-American Children in Los Angeles."
Epidemiology 12(2):200-208.
Ozkaynak, H., and G.D. Thurston. 1987. "Associations between 1980 U.S. Mortality Rates
and Alternative Measures of Airborne Particle Concentration." Risk Analysis
7(4):449-461.
Peters, A., D.W. Dockery, I.E. Muller, and M.A. Mittleman. 2001. "Increased Particulate
Air Pollution and the Triggering of Myocardial Infarction." Circulation
103:2810-2815.
Poloniecki, J.D., R.W. Atkinson., A.P. de Leon., and H.R. Anderson. 1997. "Daily Time
Series for Cardiovascular Hospital Admissions and Previous Day's Air Pollution in
London, UK." Occupational and Environmental Medicine 54(8):535-540.
Pope, C.A. 2000. "Invited Commentary: Particulate Matter-Mortality Exposure-Response
Relations and Thresholds." American Journal of Epidemiology 152:407-412.
Pope, C.A., III, MJ. 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." American Journal of Respiratory Critical Care
Medicine 151:669-674.
Pope, C.A., III, R.T. Burnett, MJ. Thun, E.E. Calle, D. Krewski, K. Ito, and G.D. Thurston.
2002. "Lung Cancer, Cardiopulmonary Mortality, and Long-term Exposure to Fine
Particulate Air Pollution." Journal of the American Medical Association
287:1132-1141.
Pope, C.A., III, D.W. Dockery, J.D. Spengler, and M.E. Raizenne. 1991. "Respiratory
Health and PM10 Pollution: A Daily Time Series Analysis." American Review of
Respiratory Diseases 144:668-674.
Ransom, M.R., and C.A. Pope. 1992. "Elementary School Absences and PM(10) Pollution in
Utah Valley." Environmental Research 58(2):204-219.
Rosamond, W., G. Broda, E. Kawalec, S. Rywik, A. Pajak, L. Cooper, and L. Chambless.
1999. "Comparison of Medical Care and Survival of Hospitalized Patients with Acute
Myocardial Infarction in Poland and the United States." American Journal of
Cardiology 83:1180-1185.
4-95
-------
Rossi, G., M.A. Vigotti, A. Zanobetti, F. Repetto, V. Gianelle, and J. Schwartz. 1999. "Air
Pollution and Cause-Specific Mortality in Milan, Italy, 1980-1989." Archives of
Environmental Health 54(3): 158-164
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. Environmental Protection Agency, Office of Policy Analysis. EPA-
230-09-86-018. Washington, DC.
Rowlatt, et al. 1998. "Valuation of Deaths from Air Pollution." NERA and CASPAR for
DETR.
Russell, M.W., D.M. Huse, S. Drowns, E.C. Hamel, and S.C. Hartz. 1998. "Direct Medical
Costs of Coronary Artery Disease in the United States." American Journal of
Cardiology 81(9): 1110-1115.
Samet, J.M., S.L. Zeger, I.E. Kelsall, J. Xu, andL.S. Kalkstein. March 1997. "Air Pollution,
Weather, and Mortality in Philadelphia 1973-1988." Cambridge, MA: Health Effects
Institute.
Samet, J.M., S.L. Zeger, F. Dominici, F. Curriero, I. Coursac, D.W. Dockery, J. Schwartz, and
A. Zanobetti. June 2000. The National Morbidity, Mortality and Air Pollution Study:
Part II: Morbidity, Mortality and Air Pollution in the United States. Research Report
No. 94, Part II. Health Effects Institute, Cambridge MA.
Schwartz, J. 1993. "Particulate Air Pollution and Chronic Respiratory Disease."
Environmental Research 62:7-13.
Schwartz, J. 1994a. "PM(10) Ozone, and Hospital Admissions for the Elderly in
Minneapolis-St Paul, Minnesota." Archives of Environmental Health 49(5):366-374.
Schwartz, J. 1994b. "Air Pollution and Hospital Admissions for the Elderly in Detroit,
Michigan." American Journal of Respiratory and Critical Care Medicine
150(3):648-655.
Schwartz, J. 1995. "Short Term Fluctuations in Air Pollution and Hospital Admissions of the
Elderly for Respiratory Disease." Thorax 50(5):531-538
4-96
-------
Schwartz, I, D.W. Dockery, L.M. Neas, D. Wypij, J.H. Ware, J.D. Spengler, P. Koutrakis,
F.E. Speizer, and E.G. Ferris, Jr. 1994. "Acute Effects of Summer Air Pollution on
Respiratory Symptom Reporting in Children." American Journal of Respiratory
Critical Care Medicine 15 0:123 4-1242.
Schwartz, J., and A. Zanobetti. 2000. "Using Meta-Smoothing to Estimate Dose-Response
Trends across Multiple Studies, with Application to Air Pollution and Daily Death."
Epidemiology 11:666-672.
Schwartz, J., and L.M. Neas. 2000. "Fine Particles are More Strongly Associated than
Coarse Particles with Acute Respiratory Health Effects in Schoolchildren."
Epidemiology 11:6-10.
Schwartz, J., F. Laden, and A. Zanobetti. 2002. "The Concentration-Response Relation
between PM(2.5) and Daily Deaths." Environmental Health Perspectives
110:1025-1029.
Sheppard, L. 2003. "Ambient Air Pollution and Nonelderly Asthma Hospital Admissions in
Seattle, Washington, 1987-1994." In Revised Analyses of Time-Series Studies of Air
Pollution and Health. Special Report. Boston, MA: Health Effects Institute.
Sheppard, L., D. Levy, G. Norris, T.V. Larson, and J.Q. Koenig. 1999. "Effects of Ambient
Air Pollution on Nonelderly Asthma Hospital Admissions in Seattle, Washington,
1987-1994." Epidemiology 10:23-30.
Shogren, J., and T. Stamland. 2002. "Skill and the Value of Life." Journal of Political
Economy 110:1168-1197.
Sisler, J.F. 1996. Spatial and Seasonal Patterns and Long Term Variability of the
Composition of the Haze in the United States: An Analysis of Data from the IMPROVE
Network. Colorado State University, Cooperative Institute for Research in the
Atmosphere. Fort Collins, CO. July. See EPA Air Docket A-96-56, Document No.
VI-B-09-(ee).
Smith, A., T. Kim, M. Fuentes, and D. Spitzner. 2000. "Threshold Dependence of Mortality
Effects for Fine and Coarse Particles in Phoenix, Arizona." Journal of the Air and
Waste Management Association 5:1367-1379.
4-97
-------
Smith, D.H., D.C. Malone, K.A. Lawson, LJ. Okamoto, C. Battista, and W.B. Saunders.
1997. "A National Estimate of the Economic Costs of Asthma." American Journal of
Respiratory and Critical Care Medicine 156(3 Pt l):787-793.
Smith, V.K., G. Van Houtven, and S.K. Pattanayak. 2002. "Benefit Transfer via Preference
Calibration." Land Economics 78:132-152.
Standard and Poor's. 2000. "The U.S. Economy: The 25 Year Focus." Winter. Website:
http://www2 standardandpoor s. com.
Stanford, R., T. McLaughlin, and LJ. Okamoto. 1999. "The Cost of Asthma in the
Emergency Department and Hospital." American Journal of Respiratory and Critical
Care Medicine 160(1):211-215.
Stieb, D.M., R.T. Burnett, R.C. Beveridge, and J.R. Brook. 1996. "Association between
Ozone and Asthma Emergency Department Visits in Saint John, New Brunswick,
Canada." Environmental Health Perspectives 104(12): 1354-1360.
Sweeney, Jeff. "EPA's Chesapeake Bay Program Air Strategy." October 26, 2004.
Thurston, G.D., and K. Ito. 2001. "Epidemiological Studies of Acute Ozone Exposures and
Mortality." Journal of Exposure Analysis and Environmental Epidemiology
ll(4):286-294.
Tolley, G.S. et al. January 1986. Valuation of Reductions in Human Health Symptoms and
Risks. University of Chicago. Final Report for the U.S. Environmental Protection
Agency.
Tsuji, H., M.G. Larson, F.J. Venditti, Jr., E.S. Manders, J.C. Evans, C.L. Feldman, D. Levy.
1996. "Impact of Reduced Heart Rate Variability on Risk for Cardiac Events. The
Framingham Heart Study." Circulation 94(11):2850-2855.
U.S. Bureau of Census. Annual Projections of the Total Resident Population, Middle Series,
1999-2100. Available at: < http://www.census.gov/population/www/
project! ons/natsum-Tl .html>.
U.S. Bureau of Census. 2000. Population Projections of the United States by Age, Sex, Race,
Hispanic Origin and Nativity: 1999 to 2100. Population Projections Program,
Population Division, U.S. Census Bureau, Washington, DC. Available at:
4-98
-------
U.S. Bureau of the Census. 2002. Statistical Abstract of the United States: 2001.
Washington, DC.
Floriculture and Nursery Crops Situation and Outlook Yearbook. 2004. Market and Trade
Economics Division, Economic Research Service, U.S. Department of Agriculture,
FLO-2004.
NPS/NERCHAL/NRTR-03/090. Philadelphia, PA: National Park Service, Northeast Region.
Available at http://www.nps.gov/shen/air_quality.htm.
U.S. Environmental Protection Agency (EPA). 1993. External Draft, Air Quality Criteria for
Ozone and Related Photochemical Oxidants. Volume II. U.S. EPA, Office of Health
and Environmental Assessment. Research Triangle Park, NC, EP A/600/AP-
93/004b.3v.
U.S. Environmental Protection Agency (EPA). 1996a. Review of the National Ambient Air
Quality Standards for Ozone: Assessment of Scientific and Technical Information.
Office of Air Quality Planning and Standards, Research Triangle Park, NC, EPA
report no. EPA/4521R-96-007.
U.S. Environmental Protection Agency (EPA). 1996b. Review of the National Ambient Air
Quality Standards for P articulate Matter: Assessment of Scientific and Technical
Information. Office of Air Quality Planning and Standards, Research Triangle Park,
NC; EPA report no. EPA/4521R-96-013.
U.S. Environmental Protection Agency (EPA). 1997. The Benefits and Costs of the Clean
Air Act, 1970 to 1990. Prepared for U.S. Congress by U.S. EPA, Office of Air and
Radiation/Office of Policy Analysis and Review, Washington, DC.
U.S. Environmental Protection Agency (EPA). 1999a. The Benefits and Costs of the Clean
Air Act, 1990-2010. Prepared for U.S. Congress by U.S. EPA, Office of Air and
Radiation/Office of Policy Analysis and Review, Washington, DC, November; EPA
report no. EPA-410-R-99-001.
U.S. Environmental Protect!on Agency (EPA). 1999b. "An SAB Advisory: The Clean Air
Act Section 812 Prospective Study Health and Ecological Initial Studies." Prepared
by the Health and Ecological Effects Subcommittee (HEES) of the Advisory Council
on the Clean Air Compliance Analysis, Science Advisory Board, U.S. Environmental
Protection Agency. Washington DC. EPA-SAB-COUNCIL-ADV-99-005.
4-99
-------
U.S. Environmental Protection Agency (EPA). September 2000b. Guidelines for Preparing
Economic Analyses. EPA 240-R-00-003.
U.S. Environmental Protection Agency (EPA). 2000c. Integrated Risk Information System;
website access available at www.epa.gov/ngispgm3/iris. Data as of December 2000.
U.S. Environmental Protection Agency (EPA). 2000d. Regulatory Impact Analysis: Heavy-
Duty Engine and Vehicle Standards and Highway Diesel Fuel Sulfur Control
Requirements. Prepared by: Office of Air and Radiation. Available at
http://www.epa.gov/otaq/diesel.htm. Accessed March 20, 2003.
U.S. Environmental Protection Agency. 2003c. Clear Skies Act—Technical Report:
Section B.
U.S. Environmental Protection Agency. January 2004a. Air Quality Data Analysis Technical
Support Document for the Proposed Interstate Air Quality Rule.
U.S. Nuclear Regulatory Commission (NRC). 1996. "Branch Technical Position on the Use
of Expert Elicitation in the High-Level Radioactive Waste Program." November,
1996.
U.S. Office of Management and Budget. October 1992. "Guidelines and Discount Rates for
Benefit-Cost Analysis of Federal Programs." Circular No. A-94.
U.S. Office of Management and Budget (OMB). 2003. Circular A-4 Guidance for Federal
Agencies Preparing Regulatory Analyses, Available at:
.
Vedal, S., J. Petkau, R. White, and J. Blair. 1998. "Acute Effects of Ambient Inhalable
Particles in Asthmatic and Nonasthmatic Children." American Journal of Respiratory
and Critical Care Medicine 15 7(4): 103 4-1043.
Viscusi, W.K., W.A. Magat, and J. Huber. 1991. "Pricing Environmental Health Risks:
Survey Assessments of Risk-Risk and Risk-Dollar Trade-Offs for Chronic
Bronchitis." Journal of Environmental Economics and Management 21:32-51.
Viscusi, V.K., and IE. Aldy. 2003. "The Value of a Statistical Life: A Critical Review of
Market Estimates Throughout the World." Journal of Risk and Uncertainty
27(l):5-76.
4-100
-------
Weisel, C.P., R.P. Cody, and PJ. Lioy. 1995. "Relationship between Summertime Ambient
Ozone Levels and Emergency Department Visits for Asthma in Central New Jersey."
Environmental Health Perspectives 103 Suppl 2:97-102.
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.
Wittels, E.H., J.W. Hay, and A.M. Gotto, Jr. 1990. "Medical Costs of Coronary Artery
Disease in the United States." American Journal of Cardiology 65(7):432-440.
Woodruff, T.J., J. Grille, and K.C. Schoendorf. 1997. "The Relationship Between Selected
Causes of Postneonatal Infant Mortality and Paniculate Air Pollution in the United
States." Environmental Health Perspectives 105(6):608-612.
Woods & Poole Economics, Inc. 2001. "Population by Single Year of Age CD." Woods &
Poole Economics, Inc.
World Health Organization (WHO). 2002. "Global Burden of Disease Study." World Health
Organization.
Yu, O., L. Sheppard, T. Lumley, J.Q. Koenig, and G.G. Shapiro. 2000. "Effects of Ambient
Air Pollution on Symptoms of Asthma in Seattle-Area Children Enrolled in the CAMP
Study." Environmental Health Perspectives 108(12): 1209-1214.
4-101
-------
CHAPTER 5
QUALITATIVE ASSESSMENT OF NONMONETIZED BENEFITS
5.1 Introduction
In addition to the enumerated human health and welfare benefits resulting from
reductions in ambient levels of PM and ozone, CAIR will result in benefits that we are not
able to monetize. This chapter discusses welfare benefits associated with reduced sulfur and
nitrogen deposition that affects acidification of ecosystems and eutrophication in water
bodies. Other welfare benefits including potential visibility improvements, agricultural yield
increases, forestry production increases, reductions in soiling and materials damage, mercury
health and welfare benefits, and other welfare categories are discussed in Chapter 4 of this
report.
5.2 Atmospheric Deposition of Sulfur and Nitrogen—Quantification of Impacts for
the Rule
The benefits associated with reduced sulfur and nitrogen deposition to aquatic, forest,
and coastal ecosystems are qualitatively discussed in this section. We also quantify the
reductions in acid deposition for the nation's lakes in the northeast, including the
Adirondacks, and in southeast streams that are expected to occur as a result of this rule.
Figures 5-1 and 5-2 present the reductions in sulfur and nitrogen deposition anticipated to
occur in 2010 and 2015 in the Eastern United States as a result of emission reductions
anticipated under the final CAIR. As shown in Figure 5-1, CAIR is predicted to result in
sulfur deposition reductions ranging from slightly less than 1 to 59 percent in specific areas
of the CAIR region and other eastern states with reductions of 16.6 percent on average for
the areas east of the 100th parallel in 2015. As Figure 5-2 depicts, changes in all forms of
nitrogen deposition in the region are expected to range from small areas of slight increases
that do not exceed 1.5 percent to reductions up to 19 percent and average reductions of 4.6
percent for areas east of the 100th parallel in 2015.
Figure 5-3 reports average nitrogen and sulfur deposition reductions anticipated for
the Adirondacks, New England, the Blue Ridge in 2015 under CAIR. As shown in
Figure 5-3, sulfur depositions reductions from baseline conditions are predicted in 2015 of
5-1
-------
Projected changes In suifur deposition,
compared to the base case In 2010
Projected changes in suifur deposition,
compared to the base case in 2015
y
r-TL-^^7rA
ri
\ L--A P w« ent Reduction
PT#
5 to 10
!0to15
15 to 20
20 to 25
* Base case assume*
implementation of existing
Clean Air Act programs.
Figure 5-1. Percentage Reduction of All Forms of Sulfur Deposition for the Years 2010
and 2015
Prejteted changes in nitrogen deposition
compared to the base case in 2010
100th
parallel
\
Percent Reduction
-1.3to5
5 to 10
10 to 15
i 15to20
Projected changes in nitrogen deposition
compared to the base case in 2016
Figure 5-2. Percentage Reduction of All Forms of Nitrogen Deposition for the Years
2010 and 2015
5-2
-------
,. Adirondacks
"New England
je
2015 Average Nitrogen
and Sulfur Deposition
Percent Reductions
UlHler CAIR (compared with a
1>1 ojectetl2015lose oset
Adirondack*
J0*o Snlfiir Reduction
8 ° o Nitrogen Reduction
New England
24°« Sulfur Redmtiou
6*0 Nitrogen Reduction
Blue Ridge
5-1*0 Sulfui Redncdvn
8 "o Nitrogen Redaction
Figure 5-3. CAIR Nitrogen and Sulfur Deposition Reductions in the Adirondacks, New
England, and the Blue Ridge
approximately 30 percent in the Adirondacks, 24 percent in New England, and 34 percent in
the Blue Ridge. In 2015, these same regions are expected to experience nitrogen deposition
reductions ranging from 6 to 8 percent from baseline conditions.
5.3 Atmospheric Deposition of Sulfur and Nitrogen—Impacts on Aquatic, Forest,
and Coastal Ecosystems
Atmospheric deposition of sulfur and nitrogen, more commonly known as acid rain,
occurs when emissions of SO2 and NOX react in the atmosphere (with water, oxygen, and
oxidants) to form various acidic compounds. These acidic compounds fall to Earth in either
a wet form (rain, snow, and fog) or a dry form (gases and particles). Prevailing winds
transport the acidic compounds hundreds of miles, often across state and national borders.
Acidic compounds (including small particles such as sulfates and nitrates) cause many
negative environmental effects. These pollutants
5-3
-------
• acidify lakes and streams,
• harm sensitive forests, and
• harm sensitive coastal ecosystems.
The effect of atmospheric deposition of acids on freshwater and forest ecosystems depends
largely on the ecosystem's ability to neutralize the acid (Driscoll et al., 2001). This is
referred to as an ecosystem's acid neutralizing capacity (ANC). Acid neutralization occurs
when positively charged ions such as calcium, potassium, sodium, and magnesium,
collectively known as base cations, are released. As water moves through a watershed, two
important chemical processes act to neutralize acids. The first involves cation exchange in
soils, a process by which hydrogen ions from the acid deposition displace other cations from
the surface of soil particles, releasing these cations to soil and surface water. The second
process is mineral weathering, where base cations bound in the mineral structure of rocks are
released as the minerals gradually break down over long time periods. As the base cations
are released by weathering, they neutralize acidity and increase the pH level in soil water and
surface waters. Acid deposition, because it consists of acid anions (e.g., sulfate, nitrate),
leaches some of the accumulated base cation reserves from the soils into drainage waters.
The leaching rate of these base cations may accelerate to the point where it significantly
exceeds the resupply via weathering (Driscoll et al., 2001). CAIR is expected to reduce
atmospheric deposition of nitrogen and sulfur and to reduce the total nitrogen and sulfur load
in areas of the East.
Soils, forests, surface waters and aquatic biota (fish, algae, and the rest), and coastal
ecosystems share water, nutrients, and other essential ecosystem components and are
inextricably linked by the chemical processes described above. For example, the same base
cations that help to neutralize acidity in lakes and streams are also essential nutrients in
forest soils, meaning that cation depletion both increases freshwater acidification and
decreases forest productivity. Similarly, the same nitrogen atom that contributes to stream
acidification can ultimately contribute to coastal eutrophication as it travels downstream to
an estuarine environment. Therefore, to understand the full effects of atmospheric
deposition, it is necessary to recognize the interactions between all of these systems.
5.3.1 Freshwater A cidiftcation
Acid deposition causes acidification of surface waters. In the 1980s, acid rain was
found to be the dominant cause of acidification in 75 percent of acidic lakes and 50 percent
of acidic streams. Areas especially sensitive to acidification include portions of the
5-4
-------
Northeast (particularly the Adirondack and Catskill Mountains, portions of New England,
and streams in the mid-Appalachian highlands) and Southeastern streams. Some high-
elevation Western lakes, particularly in the Rocky Mountains, have become acidic,
especially during snowmelt. However, although many Western lakes and streams are
sensitive to acidification, they are not subject to continuously high levels of acid deposition
and so have not become chronically acidified (NAPAP, 1990).
ANC, a key indicator of the ability of the water and watershed soil to neutralize the
acid deposition it receives, depends largely on the watershed's physical characteristics:
geology, soils, and size. Waters that are sensitive to acidification tend to be located in small
watersheds that have few alkaline minerals and shallow soils. Conversely, watersheds that
contain alkaline minerals, such as limestone, tend to have waters with a high ANC.
As acidity increases, aluminum leached from the soil flows into lakes and streams
and can be toxic to aquatic species. The lower pH levels and higher aluminum levels that
result from acidification make it difficult for some fish and other aquatic species to survive,
grow, and reproduce. In some waters, the number of species offish able to survive has been
directly correlated to water acidity. Acidification can also decrease fish population density
and individual fish size (U.S. Department of the Interior, 2003).
Recent watershed mass balance studies in the Northeast reveal that loss of sulfate
from the watershed exceeds atmospheric sulfur deposition (Driscoll et al., 2001). This
suggests that these soils have become saturated with sulfur, meaning that the supply of sulfur
from deposition exceeds the sulfur demands of the ecosystem. As a result, sulfur is gradually
being released or leached from the watershed into the surface waters as sulfate. Scientists
now expect that the release of sulfate that previously accumulated in watersheds will delay
the recovery of surface waters in the Northeast that is anticipated in response to the recent
SO2 emission controls (Driscoll et al., 2001).
A major study of the ecological response to acidification is taking place in the Bear
Brook Watershed in Maine. Established in 1986 as part of the EPA's Watershed
Manipulation Project, the project has found that experimental additions of sulfur and
nitrogen to the watershed increased the concentrations of both sulfate and nitrate in the West
Bear Brook stream. Stream water concentrations of several other ions, including base
cations, aluminum, and ANC, changed substantially as well (Norton et al., 1999). During the
first year of treatment, 94 percent of the nitrogen added experimentally to the Bear Brook
watershed was retained, while the remainder leached into streams as nitrate. Nitrogen
retention decreased to about 82 percent in subsequent years (Kahl et al., 1993; 1999).
5-5
-------
Although the forest ecosystem continued to accumulate nitrogen, nitrate leaching into the
stream continued at elevated levels throughout the length of the experiment. This nitrate
contributed to both episodic and chronic acidification of the stream. This and other similar
studies have allowed scientists to quantify acidification and recovery relationships in eastern
watersheds in much more detail than was possible in 1990.
The Appalachian Mountain region receives some of the highest rates of acid
deposition in the United States (Herlihy et al., 1993). The acid-base status of stream waters
in forested upland watersheds in the Appalachian Mountains was extensively investigated in
the early 1990s (e.g., Church et al. [1992]; Herlihy et al. [1993]; Webb et al. [1994]; van
Sickle and Church [1995]). A more recent assessment of the southern Appalachian region
from West Virginia to Alabama identified watersheds that are sensitive to acid deposition
using geologic bedrock and the associated buffering capacity of soils to neutralize acid. The
assessment found that approximately 59 percent of all trout stream length in the region is in
areas that are highly vulnerable to acidification and that 27 percent is in areas that are
moderately vulnerable (SAMAB, 1996). Another study estimated that 18 percent of
potential brook trout streams in the mid-Appalachian Mountains are too acidic for brook
trout survival (Herlihy et al., 1996). Perhaps the most important study of acid-base
chemistry of streams in the Appalachian region in recent years has been the Virginia Trout
Stream Sensitivity Study (Webb et al., 1994). Trend analyses of these streams indicate that
few long-term sampling sites are recovering from acidification, most are continuing to
acidify, and the continuing acidification is at levels that are biologically significant for brook
trout populations (Webb et al., 2000).
5.3.1.1 Water/Watershed Modeling
Researchers have used models to help them understand and predict atmospheric,
environmental, and human health responses to acid deposition for well over 20 years. Since
1990, watershed modeling capabilities have also improved as researchers are continuing to
refine and expand models that project acidification of waterbodies. Unlike the response of
air quality and deposition to changes in emissions, lakes and streams take years to decades to
fully reflect reductions in acid deposition. In some cases, soil chemistry has been
significantly altered, and ions must either build up or be leached out before the chemistry can
return to its pre-acidification status. Therefore, lake and stream conditions are presented for
2030. These results may still not reflect the full scope of ecosystem response to CAIR
emissions reductions.
5-6
-------
5.3.1.2 Description of the MAGIC Model and Methods
A number of mathematical models of soil and surface water acidification in response
to atmospheric deposition were developed in the early 1980s (e.g., Christopherson and
Wright [1981]; Christopherson et al. [1982]; Schnoor et al. [1984]; Booty and Kramer
[1984]; Goldstein et al. [1984]; Cosby et al. [1985a,b,c]). These models were based on
process-level information about the acidification process and were built for a variety of
purposes ranging from estimating transient water quality responses for individual storm
events to estimating chronic acidification of soils and base flow surface water. One of these
models (MAGIC—the Model of Acidification of Groundwater In Catchments; Cosby et al.
[1985a,b,c]) has been in use now for more than 15 years. MAGIC has been applied
extensively in North America and Europe to both individual sites and regional networks of
sites and has also been used in Asia, Africa, and South America. The utility of MAGIC for
simulating a variety of water and soil acidification responses at the laboratory, plot, hillslope,
and catchment scales has been tested using long-term monitoring data and experimental
manipulation data. MAGIC has been widely used in policy and assessment activities in the
United States and in several countries in Europe.
5.3.1.3 Model Structure
MAGIC is a lumped-parameter model of intermediate complexity, developed to
predict the long-term effects of acidic deposition on surface water chemistry. The model
simulates soil solution chemistry and surface water chemistry to predict the monthly and
annual average concentrations of the major ions in these waters. MAGIC consists of the
following: (1) a section in which the concentrations of major ions are assumed to be
governed by simultaneous reactions involving sulfate adsorption, cation exchange,
dissolution-precipitation-speciation of aluminum, and dissolution-speciation of inorganic
carbon; and (2) a mass balance section in which the flux of major ions to and from the soil is
assumed to be controlled by atmospheric inputs, chemical weathering, net uptake, and loss in
biomass and losses to runoff. At the heart of MAGIC is the size of the pool of exchangeable
base cations in the soil. As the fluxes to and from this pool change over time owing to
changes in atmospheric deposition, the chemical equilibria between soil and soil solution
shift to give changes in surface water chemistry. The degree and rate of change of surface
water acidity thus depend both on flux factors and the inherent characteristics of the affected
soils.
Cation exchange is modeled using equilibrium (Gaines-Thomas) equations with
selectivity coefficients for each base cation and aluminum. Sulfate adsorption is represented
5-7
-------
by a Langmuir isotherm. Aluminum dissolution and precipitation are assumed to be
controlled by equilibrium with a solid phase of aluminum trihydroxide. Aluminum
speciation is calculated by considering hydrolysis reactions as well as complexation with
sulfate, fluoride, and dissolved organic compounds. Effects of carbon dioxide on pH and on
the speciation of inorganic carbon are computed from equilibrium equations. Organic acids
are represented in the model as tri-protic analogues. Weathering rates are assumed to be
constant. Two alternative mechanisms are offered for simulating nitrate and ammonium in
soils: either (1) first order equations representing net uptake and retention or (2) a set of
equations and compartments describing process-based nitrogen dynamics in soils controlled
by soil nitrogen pools. Input-output mass balance equations are provided for base cations
and strong acid anions, and charge balance is required for all ions in each compartment.
Given a description of the historical, current, and expected future deposition at a site, the
model equations are solved numerically to give long-term reconstructions of surface water
chemistry (for complete details of the model see Cosby et al. [1985 a,b,c]; [2001]).
MAGIC has been used to reconstruct the history of acidification, to examine current
patterns of recovery, and to simulate the future trends in stream water acidity in both
individual catchment and regional applications at a large number of sites across North
America and Europe (e.g., Beier et al. [1995]; Cosby et al. [1985b,1990, 1995, 1996]; Cosby
and Wright, 1998; Ferrier et al. [2001]; Hornberger et al. [1989]; Jenkins et al. [1990];
Moldan et al. [1998]; Norton et al. [1992]; Whitehead et al. [1988, 1997]; Wright et al.
[1990, 1994, 1998]).
5.3.1.4 Model Implementation
Atmospheric deposition and net uptake-release fluxes for the base cations and strong
acid anions are required as inputs to the model. These inputs are generally assumed to be
uniform over the catchment. Atmospheric fluxes are calculated from concentrations of the
ions in precipitation and the rainfall volume into the catchment. The atmospheric fluxes of
the ions must be corrected for dry deposition of gas, particulates, and aerosols and for inputs
in cloud/fog water. The volume discharge for the catchment must also be provided to the
model. In general, the model is implemented using average hydrologic conditions and
meteorological conditions in annual or seasonal simulations (i.e., mean annual or mean
monthly deposition); precipitation and lake discharge are used to drive the model. Values for
soil and surface water temperature, partial pressure of carbon dioxide, and organic acid
concentrations must also be provided at the appropriate temporal resolution.
5-8
-------
In this application, deposition input to MAGIC was estimated from the policy
scenario emissions data based on previous known relationships between emissions of acid
deposition precursors and acid deposition. As implemented in this project, the model is a
two-compartment representation of a catchment. Atmospheric deposition enters the soil
compartment, and the equilibrium equations are used to calculate soil water chemistry. The
water is then routed to the stream compartment, and the appropriate equilibrium equations
are reapplied to calculate runoff chemistry.
Once initial conditions (initial values of variables in the equilibrium equations) have
been established, the equilibrium equations are solved for soil water and surface water
concentrations of the remaining variables. These concentrations are used to calculate the
lake discharge output fluxes of the model for the first time step. The mass balance equations
are (numerically) integrated over the time step, providing new values for the total amounts of
base cations and strong acid anions in the system. These in turn are used to calculate new
values of the remaining variables, new lake discharge fluxes, and so forth. The output from
MAGIC is thus a time trace for all major chemical constituents for the period of time chosen
for the integration.
5.3.1.5 Calibration Procedure
The aggregated nature of the model requires that it be calibrated to observed data
from a system before it can be used to examine potential system response. Calibration is
achieved by setting the values of certain parameters within the model that can be directly
measured or observed in the system of interest (called "fixed" parameters). The model is
then run (using observed atmospheric and hydrologic inputs) and the simulated values of
surface water and soil chemical variables (called "criterion" variables) are compared to
observed values of these variables. If the observed and simulated values differ, the values of
another set of parameters in the model (called "optimized" parameters) are adjusted to
improve the fit. After a number of iterations, the simulated-minus-observed values of the
criterion variables usually converge to zero (within some specified tolerance). The model is
then considered calibrated. If new assumptions (or values) for any of the fixed variables or
inputs to the model are subsequently adopted, the model must be recalibrated by readjusting
the optimized parameters until the simulated-minus-observed values of the criterion variables
again fall within the specified tolerance.
Calibrations are based on volume-weighted mean annual or seasonal fluxes for a
given period of observation. The length of the period of observation used for calibration is
not arbitrary. Model output will be more reliable if the annual flux estimates used in
5-9
-------
calibration are based on a number of years rather than just 1 year. There is a lot of year-to-
year variability in atmospheric deposition and catchment runoff. Averaging over a number
of years reduces the likelihood that an "outlier" year (very dry, etc.) is the primary data on
which model forecasts are based. On the other hand, averaging over too long a period may
remove important trends in the data that the model needs to simulate.
The calibration procedure requires that stream water quality, soil chemical and
physical characteristics, and atmospheric deposition data be available for each catchment.
The water quality data needed for calibration are the concentrations of the individual base
cations (Ca, Mg, Na, and K) and acid anions (Cl, SO4, and NO3) and the pH. The soil data
used in the model include soil depth and bulk density, soil pH, soil cation-exchange capacity,
and exchangeable bases on the soil (Ca, Mg, Na, and K). The atmospheric deposition inputs
to the model must be estimates of total deposition, not just wet deposition. In some
instances, direct measurements of either atmospheric deposition or soil properties may not be
available for a given site with stream water data. In these cases, the required data can often
be estimated by assigning soil properties based on some landscape classification of the
catchment and assigning deposition using model extrapolations from some national or
regional atmospheric deposition monitoring network.
Soil Physical and Chemical Properties. Soils data for model calibration are usually
derived as averaged values of soil parameters within a catchment. If soils data for a given
location are vertically stratified, the soils data for the individual soil horizons at that
sampling site can be aggregated based on horizon, depth, and bulk density to obtain single
vertically aggregated values for the site, or the stratified data can be used directly in the
model.
Total Atmospheric Deposition. Total atmospheric deposition consists of three
components: wet deposition, the flux of ions occurring in precipitation; dry deposition,
resulting from gaseous and particulate fluxes; and cloud/fog deposition (which can be
particularly important in mountainous inland areas or moderate highlands in areas adjacent to
oceans or seas). Estimates of precipitation volume and ionic concentrations in precipitation
can be used to calculate wet deposition for a site. Observations of dry deposition or
cloud/fog deposition are very infrequent. The approach usually used to quantify these
components relies on some estimate of the ratio of estimated total deposition to the observed
wet deposition for important ions (e.g., sulphate, nitrate, and ammonium ions). These ratios
(called dry deposition factors) are then used to calculate total deposition from the observed
wet deposition data.
5-10
-------
Historical Loading. Calibration of the model (and estimation of the historical
changes at the sites) requires a temporal sequence of historical anthropogenic deposition.
Our current understanding of ecosystem responses to acidic deposition suggests that future
ecosystem responses can be strongly conditioned by historical acidic loadings. Thus, as part
of the model calibration process, the model should be constrained by some measure of
historical deposition to the site. However, such long-term, continuous historical deposition
data do not exist. The usual approach is to use historical emissions data as a surrogate for
deposition. The emissions for each year in the historical period can be normalized to
emissions in a reference year (a year for which observed deposition data are available).
Using this scaled sequence of emissions, historical deposition can be estimated by
multiplying the total deposition estimated for each site in reference year by the emissions
scale factor for any year in the past to obtain deposition for that year.
5.3.1.6 MAGIC Modeling Results
Watershed modeling was undertaken for the CAIR proposal. It was determined that
the watershed modeling conducted for the proposal is essentially unchanged for the final
rule. This modeling projects that 1 percent of northeastern lakes would be chronically acidic
in 2030 as a result of CAIR. In contrast, the same model used to analyze existing control
programs projects 6 percent of northeastern lakes would be chronically acidic in 2030. The
modeling projects that, under CAIR, 28 percent of northeastern lakes would be episodically
acidic in 2030, compared to 25 percent in 2030 under existing control programs. For
Adirondack lakes, a subset of northeastern lakes, the signals of surface water chemical
recovery are much stronger. Under CAIR, no Adirondack lakes would be chronically acidic,
and 64 percent would be episodically acidic in 2030, as opposed to 12 percent chronically
acidic and 52 percent episodically acidic in 2030 under current control programs.
Because of the age and types of soils in many high-elevation areas of the southeast,
streams in that region are more frequently characterized by a delayed response to changes in
deposition. For the ecosystems modeled in this region, 17 percent of streams are currently
chronically acidic, and this level stays the same under CAIR in 2030; the proportion of
episodically acidic streams increases from 19 percent under current conditions to 23 percent
under CAIR, which reflects a decrease in the proportion of nonacidic streams from 64
percent under current conditions to 60 percent under CAIR in 2030. It is important to note
that, under the base case, the proportion of nonacidic streams decreases even further,
dropping from 64 percent under current conditions to 58 percent in 2030. Thus, in the
southeast, CAIR would slow the deterioration of stream health (episodically acidic) expected
5-11
-------
under the base case and would prevent additional streams from becoming chronically acidic.
Results of the MAGIC modeling are summarized in Table 5-1.
Table 5-1. Acidification Changes in Water Bodies as a Result of CAIR
Base Case Transport
Current (2030) Rule (2030)
Northeastern Lakes
chronically acidic 10% 6% 1%
episodically acidic 21% 25% 28%
nonacidic 69% 69% 71%
Adirondack Lakes
chronically acidic 21% 12% 0%
episodically acidic 43% 52% 64%
nonacidic 36% 36% 36%
Southeastern Streams
chronically acidic 17% 17% 17%
episodically acidic 19% 25% 23%
nonacidic 64% 58% 60%
5.3.7.7 Study of Benefits of Natural Resource Improvements in the Adirondacks
The EPA conducted MAGIC modeling to quantify acidification improvements likely
to occur as a result of CAIR in lakes in the northeastern U.S., the Adirondack Lakes, and
streams in the southeastern U.S. However, we were unable to estimate the monetary benefits
associated with the improvements predicted by the MAGIC modeling. A study conducted by
RFF estimates the monetary benefits associated with natural resource improvements due to
acidification reductions in the Adirondacks (Banzhaf, 2004). Since CAIR results in
acidification improvements in the Adirondack Lakes, the RFF study may be relevant to this
rulemaking. The RFF study estimates the total benefits (i.e., the sum of use and nonuse
values) of natural resource improvements for the Adirondacks resulting from a program that
5-12
-------
would reduce acidification in 40 percent of the lakes in the Adirondacks of concern for
acidification. RFF estimates that there are significant benefits for the hypothesized
improvements in lake health for New York State households. The study is undergoing peer
review with an intent to publish in the near future. Based on the peer review, the Agency
will determine whether and how it should be used to quantify the benefits associated with
reductions in acidification. As the MAGIC modeling indicates CAIR is also expected to
result in significant acid deposition reductions for the Adirondacks, leading to reduced
acidification and improved lake health. The benefits of these improvements in lake health
relative to the RFF estimates will depend upon the exact magnitude of the acid deposition
reductions, the number of lakes with improvements, and the magnitude of acidification
improvements. The RFF study suggests that the benefits of acid deposition reductions for
CAIR could be substantial in terms of the total monetized value for ecological endpoints.
5.3.2 Forest Ecosystems
Reductions in sulfur and nitrogen deposition under CAIR are expected to reduce the
effects of acid deposition on forests. Our current understanding of the effects of acid
deposition on forest ecosystems has come to focus increasingly on the effects of
biogeochemical processes that affect plant uptake, retention, and cycling of nutrients within
forested ecosystems. Research results from the 1990s indicate that documented decreases in
base cations (calcium, magnesium, potassium, and others) from soils in the northeastern and
southeastern United States are at least partially attributable to acid deposition (Lawrence et
al., 1997; Huntington et al., 2000). Base cation depletion is a cause for concern because of
the role these ions play in acid neutralization and, in the case of calcium, magnesium, and
potassium, their importance as essential nutrients for tree growth. It has been known for
some time that depletion of base cations from the soil interferes with the uptake of calcium
by roots in forest soils (Shortle and Smith, 1988). Recent research indicates it also leads to
aluminum mobilization (Lawrence et al., 1995), which can have harmful effects on fish
(Dept. of Interior, 2003).
The plant physiological processes affected by reduced calcium availability include
cell wall structure and growth, carbohydrate metabolism, stomatal regulation, resistance to
plant pathogens, and tolerance of low temperatures (DeHayes et al., 1999). Soil structure,
macro and micro fauna, decomposition rates, and nitrogen metabolism are also important
processes that are significantly influenced by calcium levels in soils. The importance of
calcium as an indicator of forest ecosystem function is due to its diverse physiological roles,
coupled with the fact that calcium mobility in plants is very limited and can be further
5-13
-------
reduced by tree age, competition, and reduced soil water supply (McLaughlin and Wimmer,
1999).
A clear link has now been established in red spruce stands between acid deposition,
calcium supply, and sensitivity to abiotic stress. Red spruce uptake and retention of calcium
is affected by acid deposition in two main ways: leaching of important stores of calcium
from needles (DeHayes et al., 1999) and decreased root uptake of calcium due to calcium
depletion from the soil and aluminum mobilization (Smith and Shortle, 2001; Shortle et al.,
1997; Lawrence et al., 1997). Acid deposition leaches calcium from mesophyll cells of
1-year-old red spruce needles (Schaberg et al., 2000), which in turn reduces freezing
tolerance (DeHayes et al., 1999). These changes increase the sensitivity of red spruce to
winter injuries under normal winter conditions in the Northeast, result in the loss of needles,
and impair the overall health of forest ecosystems (DeHayes et al., 1999). Red spruce must
also expend more metabolic energy to acquire calcium from soils in areas with low
calcium/aluminum ratios, resulting in slower tree growth (Smith and Shortle, 2001).
Losses of calcium from forest soils and forested watersheds have now been
documented as a sensitive early indicator of the soil response to acid deposition for a wide
range of forest soils in the United States (Lawrence et al., 1999; Huntington et al., 2000).
There is a strong relationship between acid deposition and leaching of base cations from
hardwood forest (e.g., maple, oak) soils, as indicated by long-term data on watershed mass
balances (Likens et al., 1996; Mitchell et al., 1996), plot- and watershed-scale acidification
experiments in the Adirondacks (Mitchell et al., 1994) and in Maine (Norton et al., 1994;
Rustad et al., 1996), and studies of soil solution chemistry along an acid deposition gradient
from Minnesota to Ohio (MacDonald et al., 1992).
Although sulfate is the primary cause of base cation leaching, nitrate is a significant
contributor in watersheds that are nearly nitrogen saturated (Adams et al., 1997). Recent
studies of the decline of sugar maples in the Northeast demonstrate a link between low base
cation availability, high levels of aluminum and manganese in the soil, and increased levels
of tree mortality due to native defoliating insects (Horsley et al., 2000). The chemical
composition of leaves and needles may also be altered by acid deposition, resulting in
changes in organic matter turnover and nutrient cycling.
5-14
-------
5.3.3 Coastal Ecosystems
Since 1990, a large amount of research has been conducted on the impact of nitrogen
deposition to coastal waters. It is now known that nitrogen deposition is a significant source
of nitrogen to many estuaries (Valigura et al., 2001; Howarth, 1998). The amount of
nitrogen entering estuaries due to atmospheric deposition varies widely, depending on the
size and location of the estuarine watershed and other sources of nitrogen in the watershed.
For a handful of estuaries, atmospheric deposition of nitrogen contributes well over 40
percent of the total nitrogen load; however, in most estuaries for which estimates exist, the
contribution from atmospheric deposition ranges from 15 to 30 percent. The area with the
highest deposition rates (30 percent deposition rates) stretches from Massachusetts to the
Chesapeake Bay and along the central Gulf of Mexico coast.
Nitrogen is often the limiting nutrient in coastal ecosystems. Increasing the levels of
nitrogen in coastal waters can cause significant changes to those ecosystems. Approximately
60 percent of estuaries in the United States (65 percent of the estuarine surface area) suffer
from overenrichment of nitrogen, a condition known as eutrophication (Bricker et al., 1999).
Symptoms of eutrophication include changes in the dominant species of plankton (the
primary food source for many kinds of marine life) that can cause algal blooms, low levels of
oxygen in the water column, fish and shellfish kills, and cascading population changes up the
food chain. Many of the most highly eutrophic estuaries are along the Gulf and mid-Atlantic
coasts, overlapping many of the areas with the highest nitrogen deposition, but there are
eutrophic estuaries in every region of the coterminous U.S. coastline.
Figure 5-4 shows the percent reductions in nitrogen deposition to coastal hydrologic
regions expected to result from CAIR within the CAIR region. As Figure 5-4 depicts,
changes in all forms of nitrogen deposition in the hydrologic regions within the CAIR region
are expected to range from small areas of slight increases that do not exceed 1.5 percent to
reductions up to 19 percent and average reductions of 4.6 percent for areas east of the 100th
parallel in 2015.
The Chesapeake Bay Program estimated the reduced mass of delivered nitrogen loads
likely to result from CAIR, based upon the CAIR proposal deposition estimates published in
January 2004 (Sweeney, 2004). Atmospheric deposition of nitrogen accounts for a
significant portion of the nitrogen loads to the Chesapeake with 28 percent of the nitrogen
loads from the watershed coming from air deposition. Based upon the CAIR proposal
nitrogen deposition rates published in the January 2004 proposal, the Chesapeake Bay
5-15
-------
2015 Percent Reductions of
Nitrogen To Coastal Hvdrologic
^J> 4 ^^
Regions Under CAIR c
with a projected 2015 b^se case)
Percent Reduction
-1 3 to 5
5 to 10
10 ro 15
• - 15 to 20
Coastal HydrologK Regions
*Hydrologic Regions based on USG!
Hydrologic Unit Classification
System. More information can be
found at:
http ://water.usgs. go v/GIS/huc.html
Figure 5-4. CAIR Nitrogen Deposition Reductions in Hydrologic Regions
5-16
-------
Program finds that CAIR will likely reduce the nitrogen loads to the Bay by 10 million
pounds per year by 2010. Nitrogen deposition reductions for the final CAIR are anticipated
to be comparable to the proposed rule in this region. These substantial nitrogen load
reductions more than fulfill the EPA's commitment to reduce atmospheric deposition
delivered to the Chesapeake Bay by 8 million pounds annually. The monetized value of such
reductions for the Bay are likely to be substantial, but we are unable to estimate the monetary
value of these reductions at this time.
5.3.4 Potential Other Impacts
This rule is expected to result in many categories of benefits that we are currently
unable to quantify or monetize. It is possible that reductions in nitrogen deposition resulting
from this rule may lessen the benefits of passive fertilization for forests and terrestrial
ecosystems where nutrients are a limiting factor and for some croplands.
The effects of ozone and particulate matter on radiative transfer in the atmosphere
can also lead to effects of uncertain magnitude and direction on the penetration of ultraviolet
light and climate. Ground level ozone makes up a small percentage of total atmospheric
ozone (including the stratospheric layer) that attenuates penetration of UVb radiation to the
ground. EPA's past evaluation of the information indicates that potential disbenefits would
be small, variable, and with too many uncertainties to attempt quantification of relatively
small changes in average ozone levels over the course of a year (EPA, 2005a). EPA's most
recent provisional assessment of the currently available information indicates that potential
but unquantifiable benefits may also arise from reducing ozone-related attenuation of UVb
radiation (EPA, 2005b). Sulfate and nitrate particles also scatter UVb, which can decrease
exposure of horizontal surfaces to UVb, but increase exposure of vertical surfaces. In this
case as well, both the magnitude and direction of the effect of reductions in sulfate and
nitrate particles are too uncertain to quantify (EPA, 2004). Ozone is a greenhouse gas, and
sulfates and nitrates can reduce the amount of solar radiation reaching the earth, but EPA
believes that we are unable to quantify any net climate-related disbenefit or benefit
associated with the combined ozone and PM reductions in this rule.
5.4 References
Adams, M.B., T.R. Angradi, and J.N. Kochenderfer. 1997. "Stream Water and Soil Solution
Responses to 5 Years of Nitrogen and Sulfur Additions at the Fernow Experimental
Forest, West Virginia." Forest Ecology and Management 95:79-91.
5-17
-------
Banzhaf, Spencer, Dallas Burtraw, David Evans, and Alan Krupnick. September 2004.
"Valuation of Natural Resource Improvements in the Adirondacks." Resources for
the Future (RFF).
Beier, C., H. Hultberg, F. Moldan, and Wright. 1995. "MAGIC Applied to Roof
Experiments (Risdalsheia, N.; Gardsjon, S.; Klosterhede, D.K.) to Evaluate the Rate
of Reversibility of Acidification Following Experimentally Reduced Acid
Deposition." Water Air Soil Pollution 85:1745-1751.
Booty, W.G. and J.R. Kramer. 1984. "Sensitivity Analysis of a Watershed Acidification
Model." Philos. Trans. R. Soc. London, Ser. B 305:441-449.
Bricker, S.B., C.G. Clement, D.E. Pirhalla, S.P. Orlando, and D.R.G. Farrow. 1999.
National Estuarine Eutrophication Assessment: Effects of Nutrient Enrichment in the
Nation's Estuaries. National Oceanic and Atmospheric Administration, National
Ocean Service, Special Projects Office and the National Centers for Coastal Ocean
Science. Silver Spring, Maryland.
Christopherson, N., and R.F. Wright. 1981. "Sulfate Budget and a Model for Sulfate
Concentrations in Streamwater at Birkenes, a Small Forested Catchment in
Southernmost Norway." Water Resources Research 17:377-389.
Christopherson, N., H.M. Seip, and R.F. Wright. 1982. "A Model for Streamwater
Chemistry at Birkenes, Norway." Water Resources Research 18:977-996.
Church, M.R., P.W. Shaffer, K.W. Thornton, D.L. Cassell, C.I. Liff, M.G Johnson, D.A.
Lammers, JJ. Lee, G.R. Holdren, J.S. Kern, L.H. Liegel, S.M. Pierson, D.L. Stevens,
B.P. Rochelle, and R.S. Turner. 1992. Direct/DelayedResponse Project: Future
Effects of Long-Term Sulfur Deposition on Stream Chemistry in the Mid-Appalachian
Region of the Eastern United States. U.S. Environmental Agency, EPA/600/R-
92/186, Washington, DC. 384pp.
Cosby, B.J., G.M. Hornberger, R.F. Wright, E.B. Rastetter, and J.N. Galloway. 1995.
"Estimating Catchment Water Quality Response to Acid Deposition Using
Mathematical Models of Soil Ion Exchange Processes." Geoderma 38:77-95.
Cosby, BJ. and R.F. Wright. 1998. "Modelling Regional Response of Lakewater Chemistry
to Changes in Acid Deposition: The MAGIC Model Applied to Lake Surveys in
Southernmost Norway 1974-1986-1995." Hydrology and Earth Systems Sciences
2:563-576.
5-18
-------
Cosby, B.J., R.F. Wright, G.M. Hornberger, and J.N. Galloway. 1985a. Modelling the
Effects of Acid Deposition: Assessment of a Lumped Parameter Model of Soil Water
and Streamwater Chemistry." Water Resources Research 21:51 -63.
Cosby, B.J., R.F. Wright, G.M. Hornberger, and J.N. Galloway. 1985b. "Modelling the
Effects of Acid Deposition: Estimation of Long-Term Water Quality Responses in a
Small Forested Catchment." Water Resources Research 21:1591-1601.
Cosby, B.J., G.M. Hornberger, J.N. Galloway, and R.F. Wright. 1985c. "Time Scales of
Catchment Acidification: A Quantitative Model for Estimating Freshwater
Acidification." Environmental Science Technology 19:1144-1149.
Cosby, B.J., A. Jenkins, J.D. Miller, R.C. Ferrier, and T.A.B. Walker. 1990. "Modelling
Stream Acidification in Afforested Catchments: Long-Term Reconstructions at Two
Sites in Central Scotland." Journal of Hydrology 120:143-162.
Cosby, B.J., R.F. Wright, and E. Gjessing. 1995. "An Acidification Model (MAGIC) with
Organic Acids Evaluated Using Whole-Catchment Manipulations in Norway."
Journal of Hydrology 170:101-122.
Cosby, B.J., R.C. Ferrier, A. Jenkins, and R.F. Wright. 2001. "Modeling the Effects of Acid
Deposition: Refinements, Adjustments and Inclusion of Nitrogen Dynamics in the
MAGIC Model." Hydrology and Earth Systems Sciences 5:499-517.
DeHayes, D.H., P.G. Schaberg, G.J. Hawley, and G.R. Strimbeck. 1999. "Acid Rain
Impacts Calcium Nutrition and Forest Health: Alteration of Membrane-Associated
Calcium Leads to Membrane Destabilization and Foliar Injury in Red Spruce."
Bioscience 49:789-800.
Driscoll, C.T., G. Lawrence, A. Bulger, T. Butler, C. Cronan, C. Eagar, K.F. Lambert, G.E.
Likens, J. Stoddard, and K. Weathers. 2001. "Acid Deposition in the Northeastern
U.S.: Sources and Inputs, Ecosystem Effects, and Management Strategies."
Bioscience 51:180-198.
Ferrier, R.C., R.C. Helliwell, B.J. Cosby, A. Jenkins, and R.F. Wright. 2001. "Recovery
from Acidification of Lochs in Galloway, South-west Scotland, UK: 1979-1998."
Hydrology and Earth System Sciences 5:421 -431.
Goldstein, R.A., S.A. Gherini, C.W. Chen, L. Mak, and R.J.M. Hudson. 1984. "Integrated
Acidification Study (ILWAS): A Mechanistic Ecosystem Analysis. " Phil. Trans. R.
Soc. London, Ser. B, 305:409-425.
5-19
-------
Herlihy, A.T., P.R. Kaufmann, M.R. Church, P.J. Wigington, Jr., J.R. Webb, and MJ. Sale.
1993. "The Effects of Acid Deposition on Streams in the Appalachian Mountain and
Piedmont Region of the Mid-Atlantic United States." Water Resources Research
29:2687-2703.
Herlihy, A.T., P.R. Kaufmann, J.L. Stoddard, K.N. Eshleman, and AJ. Bulger. 1996.
Effects of Acidic Deposition on Aquatic Resources in the Southern Appalachians with
a Special Focus on Class I Wilderness Areas. The Southern Appalachian Mountain
Initiative (SAMI).
Hornberger, G.M., BJ. Cosby, and R.F. Wright. 1989. "Historical Reconstructions and
Future Forecasts of Regional Surface Water Acidification in Southernmost Norway."
Water Resources Research 25:2009-2018.
Horsley, S.B., R.P. Long, S.W. Bailey, R.A. Hallett, and T.J. Hall. 2000. "Factors
Associated with the Decline Disease of Sugar Maple on the Allegheny Plateau."
Canadian Journal of Forest Research 30:1365-1378.
Howarth, Robert. 1998. "An Assessment of Human Influences on Fluxes of Nitrogen from
the Terrestrial Landscape to the Estuaries and Continental Shelves of the North
Atlantic Ocean." Nutrient Cycling in Agroe cosy stems 52(2/3):213-223.
Huntington, T.G., R.P. Hooper, C.E. Johnson, B.T. Aulenbach, R. Cappellato, and A.E.
Blum. 2000. "Calcium Depletion in a Southeastern United States Forest
Ecosystem." Soil Science Society of America Journal 64:1845-1858.
Jenkins, A., P.O. Whitehead, B.J. Cosby, and H.J.B. Birks. 1990. "Modelling Long-Term
Acidification: A Comparison with Diatom Reconstructions and the Implications for
Reversibility." Phil. Trans. R. Soc. London B 327:435-440.
Kahl, J., S. Norton, I. Fernandez, L. Rustad, and M. Handley. 1999. "Nitrogen and Sulfur
Input-Output Budgets in the Experimental and Reference Watersheds, Bear Brook
Watershed, Maine (BBWM)." Environmental Monitoring and Assessment
55:113-131.
Kahl, J.S., S.A. Norton, I.J. Fernandez, K.J. Nadelhoffer, C.T. Driscoll, and J.D. Aber. 1993.
"Experimental Inducement of Nitrogen Saturation at the Watershed Scale."
Environmental Science and Technology 27:565-568.
Lawrence, G.B., M.B. David, and W.C. Shortle. 1995. "A New Mechanism for Calcium
Loss in Forest-Floor Soils." Nature 378:162-165.
5-20
-------
Lawrence, G.W., M.B. David, S.W. Bailey, and W.C. Shortle. 1997. "Assessment of
Calcium Status in Soils of Red Spruce Forests in the Northeastern United States."
Biogeochemistry 38:19-39.
Lawrence, G.B., M.B. David, G.M. Lovett, P.S. Murdoch, D.A. Burns, J.L. Stoddard, B.P.
Baldigo, J.H. Porter, and A.W. Thompson. 1999. "Soil Calcium Status and the
Response of Stream Chemistry to Changing Acidic Deposition Rates in the Catskill
Mountains of New York." Ecological Applications 9:1059-1072.
Likens, G.E., C.T. Driscoll, and D.C. Buso. 1996. "Long-Term Effects of Acid Rain:
Responses and Recovery of a Forest Ecosystem." Science 272:244-246.
MacDonald, N.W., AJ. Burton, H.O. Liechty, J.A. Whitter, K.S. Pregitzer, G.D. Mroz, and
D.D. Richter. 1992. "Ion Leaching in Forest Ecosystems along a Great Lakes Air
Pollution Gradient." Journal of Environmental Quality 21:614-623.
McLaughlin S.B. andR. Wimmer. 1999. "Tansley Review No. 104, Calcium Physiology
and Terrestrial Ecosystem Processes." New Phytologist 142:373-417.
Mitchell, M.J., M.B. David, IJ. Fernandez, R.D. Fuller, K. Nadelhoffer, L.E. Rustad, and
A.C. Stam. 1994. "Response of Buried Mineral Soil Bags to Experimental
Acidification of Forest Ecosystem." Soil Science Society of America Journal
58:556-563.
Mitchell, M.J., C.T. Driscoll, J.S. Kahl, GE. Likens, P.S. Murdoch, and L.H. Pardo. 1996.
"Climate Control on Nitrate Loss from Forested Watersheds in the Northeast United
States." Environmental Science and Technology 30:2609-2612.
Moldan, F., R.F. Wright, R.C. Ferrier, B.I., Andersson, and H. Hultberg. 1998. "Simulating
the Gardsjon Covered Catchment Experiment with the MAGIC Model." In
Experimental Reversal of Acid Rain Effects. The Gardsjon Roof Project, Hultberg,
H. and Skeffmgton, R.A. (eds.), p. 351-362, Chichester, UK: Wiley and Sons,
466 pp.
Murdoch, P.S., D.S. Burns, and G.B. Lawrence. 1998. "Relation of Climate Change to the
Acidification of Surface Waters by Nitrogen Deposition." Environmental Science
and Technology 32:1642-1647.
National Acid Precipitation Assessment Program (NAPAP). 1991. 1990 Integrated
Assessment Report. Washington, DC: National Acid Precipitation Assessment
Program Office of the Director.
5-21
-------
Norton, S.A. R.F. Wright, J.S. Kahl, and J.P. Scofield. 1992. "The MAGIC Simulation of
Surface Water Acidification at, and Preliminary Results from, the Bear Brook
Watershed Manipulation, Maine." Environmental Pollution 77:279-286.
Norton, S.A., J.S. Kahl, I.J. Fernandez, L.E. Rustad, J.P. Schofield, and T.A. Haines. 1994.
"Response of the West Bear Brook Watershed, Maine, USA, to the addition of
(NH4)2SO4: 3-Year Results." Forest and Ecology Management 68:61-73.
Norton, S.A., R.F. Wright, J.S. Kahl, and J.P. Scofield. 1998. "The MAGIC Simulation of
Surface Water Acidification at, and First Year Results from, the Bear Brook
Watershed Manipulation, Maine, USA." Environmental Pollution 77:279-286.
Norton, S.A., J.S. Kahl, I.J. Fernandez, T.A. Haines, L.E. Rustad, S. Nodvin, J.P. Scofield, T.
Strickland, H. Erickson, P. Wiggington, and J. Lee. 1999. "The Bear Brook
Watershed, Maine, (BBWP) USA." Environmental Monitoring and Assessment
55:7-51.
Rustad, L.E., I.J. Fernandez, M.B. David, M.J. Mitchell, K.J. Nadelhoffer, and R.D. Fuller.
1996. "Experimental Soil Acidification and Recovery at the Bear Brook Watershed
in Maine." Soil Science of 'America Journal 60:1933-1943.
Schaberg, P.G., D.H. DeHayes, G.J. Hawley, G.R. Strimbeck, J.R. Cumming, P.F.
Murakami, and C.H. Borer. 2000. "Acid Mist, Soil Ca and Al Alter the Mineral
Nutrition and Physiology of Red Spruce." Tree Physiology 20:73-85.
Schnoor, J.L., W.D. Palmer, Jr., and G.E. Glass. 1984. "Modeling Impacts of Acid
Precipitation for Northeastern Minnesota." In Modeling of Total Acid Precipitation
Impact Schnoor, J.L. (ed.), pp. 155-173. Boston: Butterworth.
Smith, K.T. and W.C. Shortle. 2001. "Conservation of Element Concentration in Xylem
Sap of Red Spruce." Trees 15:148-153.
Shortle, W.C. and K.T. Smith. 1988. "Aluminum-Induced Calcium Deficiency Syndrome in
Declining Red Spruce Trees." Science 240:1017-1018.
Shortle, W.C., K.T. Smith, R. Minocha, G.B. Lawrence, and M.B. David. 1997. "Acidic
Deposition, Cation Mobilization, and Biochemical Indicators of Stress in Healthy
Red Spruce." Journal of Environmental Quality 26:871-876.
5-22
-------
Southern Appalachian Man and the Biosphere (SAMAB). 1996. The Southern Appalachian
Assessment: Summary Report. Atlanta, GA: U.S. Department of Agriculture, Forest
Service, Southern Region.
Sweeney, Jeff. "EPA's Chesapeake Bay Program Air Strategy." October 26, 2004.
U.S. Department of the Interior, National Park Service. 2003. Assessment of Air Quality
and Related Values in Shenandoah National Park. Technical Report
NPS/NERCHAL/NRTR-03/090. Philadelphia, PA: National Park Service, Northeast
Region. http://www.nps.gov/shen/air_quality.htm.
U.S. Environmental Protection Agency (EPA). 2004. Air Quality Criteria for Particulate
Matter (October 2004). .
U.S. Environmental Protection Agency (EPA). 2005a. Air Quality Criteria for Ozone and
Related Photochemical Oxidants (First External Review Draft).
.
U.S. Environmental Protection Agency (EPA). 2005b. Draft Air Quality Criteria for Ozone
and Related Photochemical Oxidants E-Docket No. ORD-2004-0015 [Federal
Register: January 31, 2005 (Volume 70, Number 19)]
.
Valigura, R.A., R.B. Alexander, M.S. Castro, T.P. Meyers, H.W. Paerl, P.E. Stacy, and R.E.
Turner. 2001. Nitrogen Loading in Coastal Water Bodies: An Atmospheric
Perspective. Washington, DC: American Geophysical Union.
Van Sickle, J. and M.R. Church. 1995. Methods for Estimating the Relative Effects of Sulfur
and Nitrogen Deposition on Surface Water Chemistry. EPA/600/R-95/172.
Washington, DC: U.S. Environmental Protection Agency.
Webb, J.R., F.A. Deviney, Jr., B.J. Cosby, A.J. Bulger, and J.N. Galloway. 2000. Change in
Acid- Base Status in Streams in the Shenandoah National Park and the Mountains
of Virginia. American Geophysical Union, Biochemical Studies of the
Shenandoah National Park, http://www.nps.gov/shen/air_quality.htm.
5-23
-------
Webb, J.R., F.A. Deviney, J.N. Galloway, C.A. Rinehart, P. A Thompson, and S. Wilson.
1994. The Acid-Base Status of Native Brook Trout Streams in the Mountains of
Virginia. A Regional Assessment Based on the Virginia Trout Stream Sensitivity
Study. Charlottesville, VA: Univ. of Virginia, http://www.nps.gov/shen/
air_quality.htm.
Whitehead, P.O., B. Reynolds, G.M. Hornberger, C. Neal, BJ. Cosby, and P. Paricos. 1988.
"Modelling Long-Term Stream Acidification Trends in Upland Wales at
Plynlimon." HydrologicalProcesses 2:357-368.
Whitehead, P.O., J. Barlow, E.Y. Haworth, and J.K. Adamson. 1997. "Acidification in
Three Lake District Tarns: Historical Long Term Trends and Modelled Future
Behaviour under Changing Sulphate and Nitrate Deposition." Hydrology and
Earth System Sciences 1:197-204.
Wright, R.F., BJ. Cosby, M.B. Flaten, and J.O. Reuss. 1990. "Evaluation of an
Acidification Model with Data from Manipulated Catchments in Norway." Nature
343:53-55.
Wright, R.F., BJ. Cosby, R.C. Ferrier, A Jenkins, AJ. Bulger, and R. Harriman. 1994.
"Changes in the Acidification of Lochs in Galloway, Southwestern Scotland, 1979-
1988: The MAGIC Model Used to Evaluate the Role of Afforestation, Calculate
Critical Loads, and Predict Fish Status." Journal of Hydrology 161:257-285.
Wright, R.F., B.A. Emmett, and A. Jenkins. 1998. "Acid Deposition, Land-Use Change and
Global Change: MAGIC7 Model Applied to Risdalsheia, Norway (RAIN and
CLEVIEX projects) and Aber, UK (NITREX project)." Hydrology and Earth
System Sciences 2:385-397.
5-24
-------
CHAPTER 6
ELECTRIC POWER SECTOR PROFILE
This chapter discusses important aspects of the power sector as they relate to CAIR,
including the types of power-sector sources affected by CAIR, and provides background on
the power sector and EGUs. In addition, this chapter provides some historical background
on EPA regulation of and future projections for the power sector.
6.1 Power-Sector Overview
The functions of the power sector can be separated into three distinct operating
activities: generation, transmission, and distribution.
6.1.1 Generation
Electricity generation is the first process in the delivery of electricity to consumers.
The process of generating electricity, in most cases, involves creating heat to rotate turbines
which, in turn, create electricity. The power sector consists of over 16,000 generating units,
consisting of fossil-fuel fired units, nuclear units, and hydroelectric and renewable sources
dispersed throughout the country (see Table 6-1).
Table 6-1. Existing Electricity Generating Capacity by Energy Source, 2002
Energy Source
Coal
Petroleum
Natural Gas
Dual Fired
Other Gases
Nuclear
Hydroelectric
Other Renewables
Other
Total
Number of Generators
1,566
3,076
2,890
2,974
104
104
4,157
1,501
41
16T413
Generator Nameplate Capacity (MW)
338,199
43,206
194,968
180,174
2,210
104,933
96,343
18,797
756
979T585
Source: El A
6-1
-------
These electric-generating sources provide electricity for commercial, industrial, and
residential uses, each of which consumes roughly one-third of the total electricity produced
(see Table 6-2).
Table 6-2. Total U.S. Electric Power Industry Retail Sales in 2003 (Billion kWh)
N %
Residential
Commercial
Industrial
Other
All Sectors
1,280
1,119
991
109
3,500
37%
32%
28%
3%
100%
Source: El A
In 2003, electric-generating sources produced 3,848 billion kWh to meet electricity
demand. Roughly 70 percent of this electricity was produced through the combustion of
fossil fuels, primarily coal and natural gas, with coal accounting for more than half of the
total (see Table 6-3).
Table 6-3. Electricity Net Generation in 2003 (Billion kWh)
N %
Coal
Petroleum
Natural Gas
Other Gases
Nuclear
Hydroelectric
Other
Total
1,970
118
629
11
764
275
81
3,848
51%
3%
16%
0.3%
20%
7%
2%
100%
Source: EIA
Note: Retail sales and net generation may not correspond exactly because net generation data may include
net exported electricity and loss of electricity.
6-2
-------
Coal-fired generating units typically supply "base-load" electricity, which means
these units operate continuously throughout the day. Coal-fired generation, along with
nuclear generation, meet the part of demand that is relatively constant. Gas-fired generation,
however, typically supplies "peak" power, when there is increased demand for electricity
(e.g., when businesses operate throughout the day or when people return home from work
and run appliances and heating/air-conditioning, versus late at night or very early morning
when demand for electricity is reduced).
6.1.2 Transmission
Transmission is the term used to describe the movement of electricity, through use of
high voltage lines, from electric generators to substations where power is stepped down for
local distribution. Transmission systems have been traditionally characterized as a collection
of independently operated networks or grids interconnected by bulk transmission interfaces.
Within a well-defined service territory, the regulated utility has historically had
responsibility for all aspects of developing, maintaining, and operating transmission of
electricity. These responsibilities typically included system planning and expanding,
maintaining power quality and stability, and responding to failures.
6.1.3 Distribution
Distribution of electricity involves networks of smaller wires and substations that
take the higher voltage from the transmission system and step it down to lower levels to
match the needs of customers. The transmission and distribution system is the classic
example of a natural monopoly because it is not practical to have more than one set of lines
running from the electricity-generating sources to neighborhoods or from the curb to the
house.
Transmission and distribution have been considered differently than generation in
current efforts to restructure the industry. Transmission has generally been developed by the
larger vertically integrated utilities that typically operate generation and distribution
networks. Distribution is handled by a large number of utilities that often only sell
electricity. Electricity restructuring has focused primarily on converting the industry to fully
compete the sale of electricity production or generation and not the transmission or
distribution of electricity. The restructuring of the industry is, in large part, the separating of
generation assets from the transmission and distribution assets into separate economic
6-3
-------
entities in many state efforts. Transmissions and distribution remain price regulated
throughout the country based on the cost of service.
6.2 Deregulation and Restructuring
The ongoing process of deregulation of wholesale and retail electric markets is
changing the structure of the electric power industry. In addition to reorganizing asset
management between companies, deregulation is aimed at the functional unbundling of
generation, transmission, distribution, and ancillary services the power sector has historically
provided to competition in the generation segment of the industry.
Beginning in the 1970s, government policy shifted against traditional regulatory
approaches and in favor of deregulation for many important industries, including
transportation, communications, and energy, which were all thought to be natural
monopolies (prior to 1970) that warranted governmental control of pricing. Some of the
primary drivers for deregulation of electric power included the desire for more efficient
investment choices, the possibility of lower electric rates, reduced costs of combustion
turbine technology that opened the door for more companies to sell power, and complexity of
monitoring utilities' cost of service and establishing cost-based rates for various customer
classes (see Figure 6-1). The pace of restructuring in the electric power industry slowed
significantly in response to market volatility and financial turmoil associated with
bankruptcy filings of key energy companies in California. By the end of 2001, restructuring
had either been delayed or suspended in eight states that previously enacted legislation or
issued regulatory orders for its implementation. Another 18 other states that had seriously
explored the possibility of deregulation in 2000 reported no legislative or regulatory activity
in 2001 (DOE, EIA, 2003a). Currently, there are 17 states where price deregulation of
generation (restructuring) has occurred. Fifteen of these states are in the CAIR region. The
effort is more or less at a standstill; however, at the federal level, there are efforts in the form
of proposed legislation and proposed Federal Energy Regulatory Commission (FERC)
actions aimed at reviving restructuring. For states that have not begun restructuring efforts,
it is unclear when and at what pace these efforts will proceed.
6-4
-------
H Restructuring Active
| Restructuring Delayed
_J R eslr uctur ing Suspended
_J Restruduring NotActive
Figure 6-1. Status of State Electricity Industry Restructuring Activities (as of February
2003)
6.3 Pollution and EPA Regulation of Emissions
The burning of fossil fuels, which generates about 70 percent of our electricity
nationwide, results in air emissions of SO2 and NOX, important precursors in the formation of
fine particles and ozone (NOX only). The power sector is a major contributor of both these
pollutants, and reductions of SO2 and NOX emissions are critical to EPA's efforts to bring
about attainment with the fine particle and ozone NAAQS through programs like CAIR. In
2003, the power sector accounted for 67 percent of total nationwide SO2 emissions and 22
percent of total nationwide NOX emissions (see Figure 6-2).
Different types of fossil fuel-fired units vary widely in their air emissions levels for
SO2 and NOX, particularly when uncontrolled. For coal-fired units, NOX emission rates can
vary from under 0.05 Ibs/mmBtu (for a unit with selective catalytic reduction for NOX
removal) to over 1 Ib/mmBtu for an uncontrolled cyclone boiler. NOX emissions from
coal-fired power plants are formed during combustion and are a result of both nitrogen in
coal and nitrogen in the air. SO2 emission rates can vary from under 0.1 Ibs/mmBtu (for
some units with flue gas desulfurization for SO2 removal) to over 5 Ibs/mmBtu for units
6-5
-------
Sulfur Dioxide
Nitrogen Oxides
Figure 6-2. Emissions of SO2 and NOX from the Power Sector (2003)
burning higher sulfur coal. For an uncontrolled coal plant, SO2 emissions are directly related
to the amount of sulfur in the coal.
Oil and gas-fired units also have a wide range of NOX emissions depending on both
the plant type and the controls installed. Units with selective catalytic reduction (SCR) can
have emission rates under 0.01 Ibs/mmBtu, while completely uncontrolled units can have
emission rates in excess of 0.5 Ibs/mmBtu. Gas-fired units have very little SO2 emissions.
NOX emission rates on oil-fired units can range from under 0.1 Ibs/mmBtu (for units with
new combustion controls) to over 0.6 Ibs/mmBtu for units without combustion controls. SO2
emissions for oil-fired units can range from under 0.1 Ibs/mmBtu for units burning low sulfur
distillate oil to over 2 Ibs/mmBtu for units burning high sulfur residual oil.
6.4 Pollution Control Technologies
There are two primary options for reducing SO2 emissions from coal-burning power
plants. Units may switch from higher to lower sulfur coal, or they may use flue gas
desulfurization (FGD, commonly referred to as scrubbers). According to data submitted to
EPA for compliance with the Title IV Acid Rain Program, the SO2 emission rates for
coal-fired units varied from under 0.4 Ibs/mmBtu to over 5 Ibs/mmBtu depending on the type
of coal combusted.
6-6
-------
It is generally easier to switch to a coal within the same rank (e.g., bituminous or
sub-bituminous) because these coals will have similar heat contents and other characteristics.
Switching completely to sub-bituminous coal (which typically has a lower sulfur content)
from bituminous coal is likely to require some modifications to the unit. Limited blending of
sub-bituminous coal with bituminous coal can often be done with much more limited
modifications.
The two most commonly used scrubber types include wet scrubbers and spray dryers.
Wet scrubbers can use a variety of sorbents to capture SO2 including limestone and
magnesium enhanced lime. The choice of sorbent can affect the performance, size, and
capital and operating costs of the scrubber. New wet scrubbers typically achieve at least
95 percent SO2 removal. Spray dryers can achieve over 90 percent removal.
One method of reducing NOX emissions is through the use of combustion controls
(such as low NOX burners and over-fired air). Combustion controls reduce NOX, by ensuring
that the combustion of coal occurs under conditions under which less formation of NOX
occurs. Post-combustion controls reduce NOX by removing the NOX after it has been formed.
The most common post-combustion control is SCR. SCR systems inject ammonia (NH3),
which combines with the NOX in the flue gas, to form nitrogen and water and uses a catalyst
to enhance the reaction. These systems can reduce NOX by 90 percent and achieve emission
rates of around 0.06 Ibs/mmBtu. Selective noncatalytic reduction also removes NOX by
injecting ammonia, but no catalyst is used. These systems can reduce NOX by up to 40
percent.
For more detail on the cost and performance assumptions of pollution controls, see
the documentation for the Integrated Planning Model (IPM), a dynamic linear programming
model that EPA uses to examine air pollution control policies for SO2 and NOX throughout
the contiguous United States for the entire power system. Documentation for IPM can be
found at www.epa.gov/airmarkets/epa-ipm.
6.5 Regulation of the Power Sector
At the federal level, efforts to reduce emissions of SO2 and NOX have been occurring
since 1970. Policy makers have recognized the need to address these harmful emissions, and
incremental steps have been taken to ensure that the country meets air quality standards.
CAIR is the next step towards realizing attainment of the standards.
6-7
-------
Federal regulation of SO2 and NOX emissions at power plants began with the 1970
Clean Air Act. The Act required the Agency to develop performance standards for a number
of source categories including coal-fired power plants. The first New Source Performance
Standards (NSPS) for power plants (subpart D) required new units to limit SO2 emissions
either by using scrubbers or by using low sulfur coal. NOX was required to be limited
through the use of low NOX burners. A new NSPS (subpart Da), promulgated in 1978,
tightened the standards for SO2 requiring scrubbers on all new units.
The 1990 Clean Air Act Amendments (CAAA) placed a number of new requirements
on power plants. The Acid Rain Program, established under Title IV of the 1990 CAAA,
requires major reductions of SO2 and NOX emissions. The SO2 program sets a permanent cap
on the total amount of SO2 that can be emitted by electric power plants in the contiguous
United States at about one-half of the amount of SO2 these sources emitted in 1980. Using a
market-based cap-and-trade mechanism allows flexibility for individual combustion units to
select their own methods of compliance. The program uses a more traditional approach to
NOX emission limitations for certain coal-fired electric utility boilers, with the objective of
achieving a 2 million ton reduction from projected NOX emission levels that would have been
emitted in 2000 without implementation of Title IV.
The Acid Rain Program comprises two phases for SO2 and NOX. Phase I applied
primarily to the largest coal-fired electric generation sources from 1995 through 1999 for
SO2 and from 1996 through 1999 for NOX. Phase II for both pollutants began in 2000. For
SO2, it applies to thousands of combustion units generating electricity nationwide; for NOX it
generally applies to affected units that burned coal during 1990 through 1995. The Acid
Rain Program has led to the installation of a number of scrubbers on existing coal-fired units
as well as significant fuel switching to lower sulfur coals. Under the NOX provisions of Title
IV, most existing coal-fired units were required to install low NOX burners.
The CAAA also placed much greater emphasis on control of NOX to reduce ozone
nonattainment. This has led to the formation of several regional NOX trading programs as
well as an intrastate NOX trading program in Texas. The Ozone Transport Commission (a
group of northeast states) created an interstate NOX trading program that began in 1999. In
1998, EPA promulgated regulations (the NOX SIP Call) that required 21 states in the eastern
United States and the District of Columbia to reduce NOX emissions that contributed to
nonattainment in downwind states using the cap-and-trade approach. This program began in
6-8
-------
the summer of 2004 and has resulted in the installation of significant amounts of selective
catalytic reduction.
In addition to federal programs to reduce emissions of SO2 and NOX, several states
have also taken action. Several states, like North Carolina, New York, Connecticut, and
Massachusetts, have moved to control these emissions to address nonattainment.
6.6 Cap and Trade
The cap-and-trade system under CAIR, which is largely based on the Acid Rain
Trading Program and the NOX SIP Call, provides the power sector with considerable
flexibility in meeting the emission reduction requirements. Cap-and-trade regulation is an
extremely efficient tool that allows for environmental goals to be met in the most cost-
effective manner, because firms have economic incentives to achieve emissions reductions
where they are cheapest. The system allows for various compliance options, with each firm
determining what option works best given certain costs, such as fuel costs or costs of
pollution controls.
In addition to the pollution control options discussed above, companies can comply
with cap-and-trade programs through more efficient use of the generating fleet to take
advantage of generating sources that emit less and run more efficiently, commonly referred
to as dispatch changes. By shifting generation to these more efficient units, the power sector
is reducing the cost of compliance because there is a cost to pollute under a cap. Another
option is purchasing additional allowances to cover emissions.
6.7 Clean Air Interstate Rule
To address air quality problems and improve public health and the environment, EPA
is finalizing CAIR. The final CAIR requires annual SO2 and NOX reductions in 23 States and
the District of Columbia, and also requires ozone season NOX reductions in 25 States and the
District of Columbia. Many of the CAIR States are affected by both the annual SO2 and NOX
reduction requirements and the ozone season NOX requirements. CAIR allows affected states
to adopt a two-phased cap-and-trade program to meet emissions reduction requirements of
roughly 73 percent for SO2 and 61 percent for NOX from 2003 levels.
The rule would affect roughly 3,000 fossil fuel-fired units with a nameplate capacity
greater than 25 MW. These sources accounted for roughly 89 percent of nationwide SO2
emissions and 79 percent of nationwide NOX emissions in 2003 (see Table 6-4).
6-9
-------
Table 6-4. Emissions of SO2 and NOX in 2003 and Percentage of Emissions in the CAIR
Affected Region (tons)
CAIR Region
Nationwide
CAIR Emissions
as % of Nationwide Emissions
SO2
9,407,406
10,595,069
89%
NOX
3,222,636
4,165,026
79%
Source: EPA.
Note: Region includes states covered for the annual SO2 and NOX trading programs (Alabama, District of
Columbia, Florida, Georgia, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maryland, Michigan, Minnesota,
Mississippi, Missouri, New York, North Carolina, Ohio, Pennsylvania, South Carolina, Tennessee, Texas,
Virginia, West Virginia, and Wisconsin).
EPA modeling1 shows that coal-fired and oil/gas-fired generation will continue to
play an important part of the electricity generating portfolio in the United States. Electricity
demand is anticipated to grow by 1.6 percent a year, and total electricity demand is projected
to be 4,198 billion kWh by 2010. Table 6-5 shows current electricity generation and
projected levels in 2010 and 2015 using EPA modeling.
Table 6-5. Current Electricity Net Generation and EPA Projections for 2010 and 2015
(Billion kWh)
2003 2010 2015
Coal 1,970 2,198 2,242
Oil/Gas 758 777 1,026
Other 1,119 1,223 1,235
Total 3,848 4,198 4,503
Source: 2003 data is from EIA. Projections are from the Integrated Planning Model run by EPA.
'EPA uses the IPM to make power-sector forecasts about emissions, costs, and other key factors of the power
sector. Industry projections presented here are from EPA's base case scenario. For more information about
IPM, see http://www.epa.gov/airmarkets/epa-ipm/index.html.
6-10
-------
6.8 Price Elasticity of Electricity
Electricity performs a vital and high-value function in the economy; as a result,
electricity consumers are generally unable or unwilling to alter consumption as the price
increases. Demand for electricity, especially in the short run, is not very sensitive to changes
in prices and is considered relatively price inelastic although some demand reduction does
occur. With that in mind, EPA modeling does not incorporate a "demand response" to any
increases in electricity prices because of the reasons mentioned. Electricity demand is
considered to be constant in EPA modeling applications and the reduction in production
costs that would result from lower demand is not considered. This leads to some
overstatement in the private compliance costs that EPA estimates.
6-11
-------
CHAPTER 7
COST, ECONOMIC, AND ENERGY IMPACTS
This chapter reports the cost, economic, and energy impact analysis performed for CAIR.
EPA used the IPM, developed by ICF Consulting, to conduct its analysis. IPM is a dynamic
linear programming model that can be used to examine air pollution control policies for SO2 and
NOX throughout the contiguous United States for the entire power system. Documentation for
IPM can be found at www.epa.gov/airmarkets/epa-ipm.
7.1 Modeling Background
The analysis presented here covers the electric power sector, a major source of SO2 and
NOX emissions nationwide and the industry that EPA assumes that States will control in setting
State emissions reduction requirements. EPA has also assumed that States implement those
reductions through a cap-and-trade program. For SO2 and NOX, EPA modeled an annual, two-
phased control strategy for 26 eastern States and the District of Columbia (see Figure 7-1). For
NOX, separate ozone season caps were applied to Connecticut and Massachusetts. See Table 7-1
for total annual emissions caps under CAIR.
• States controlled for both SO2 and NOx
• States controlled for ozone season NOx only
D States not covered by CAIR
Figure 7-1. CAIR Modeled Region
7-1
-------
Table 7-1. CAIR Annual Emissions Caps (Million Tons)
S02
NO,
2010-2014 ('09-' 14 for NOX)
3.6
1.5
2015-Thereafter
2.5
1.3
The final CAIR requires annual SO2 and NOX reductions in 23 States and the District of
Columbia, and also requires ozone season NOX reductions in 25 States and the District of
Columbia. Many of the CAIR States are affected by both the annual SO2 and NOX reduction
requirements and the ozone season (May-September) NOX requirements. Using IPM, EPA
modeled the cost and emissions impacts of CAIR. For further discussion about the scope and
requirements of CAIR, see the Final CAIR preamble.
EPA initially conducted IPM modeling for today's final action using a control strategy
that is similar, but not identical to, the final CAIR requirements. The control strategy that EPA
initially modeled included three additional States (Arkansas, Delaware and New Jersey) within
the region and required these States to make annual SO2 and NOX reductions. However, these
three States are not required to make annual reductions under the final CAIR. In the "Proposed
Rules" section of today's Federal Register publication, EPA is publishing a proposal to include
Delaware and New Jersey in the CAIR region for annual SO2 and NOX reductions. This RIA is
to serve as EPA's analytical assessment of both today's Final CAIR and the proposed rule for
incorporating Delaware and New Jersey into the annual SO2 and NOX requirements of CAIR.
The addition of Arkansas, Delaware, and New Jersey brought the total number of affected States
for annual SO2 and NOX to 26 plus the District of Columbia for the initial model run. Arkansas
will not be included in the annual SO2 and NOX requirements either as part of today's Final
CAIR or the "Proposed Rule," but is included for the ozone season CAIR requirement as part of
today's Final Rule. The initial model run also included individual State ozone season NOX caps
for Connecticut and Massachusetts, and did not include ozone season NOX caps for any other
CAIR States.
The Agency conducted revised final IPM modeling that reflects the final CAIR control
strategy. The final IPM modeling includes regionwide annual SO2 and NOX caps on the 23
States and the District of Columbia for States required to make annual reductions, and includes a
regionwide ozone season NOX cap on the 25 States and the District of Columbia required to
make ozone season reductions. EPA modeled the final CAIR NOX strategy as an annual NOX cap
with a nested, separate ozone season NOX cap.
7-2
-------
In this chapter of the RIA, the projected CAIR costs and emissions are derived from the
IPM run reflecting CAIR with Arkansas, Delaware, and New Jersey included for the annual
requirements, and without a separate ozone season NOX cap for ozone season States. However,
where IPM results differ significantly between the scenario with Arkansas, Delaware, and New
Jersey, and the scenario with these three States included for NOX summer season only, EPA has
highlighted these differences and included them in this chapter as well. The air quality and
benefits analyses done in support of CAIR are based on emission projections from the initial
IPM run with Arkansas, Delaware, and New Jersey included for annual SO2 and NOX.
EPA believes that the differences between the initial IPM run and the final IPM run have
very little impact on projected control costs, emissions, and other impacts. Modeling the CAIR
region without Arkansas, Delaware, and New Jersey do not change the results presented here in
any significant way, and in any event, this chapter generally reflects the geographic scope of the
CAIR program as EPA intends it to be ultimately. IPM output files for the model runs used in
CAIR analyses are available in the CAIR docket.
CAIR was designed to achieve significant emissions reductions in a highly cost-effective
manner to reduce the transport of fine particles that have been found to contribute to
nonattainment. EPA analysis has found that the most efficient method to achieve the emissions
reduction targets is through a cap-and-trade system on the power sector that States have the
option of adopting. The power sector accounted for 67 percent of nationwide SO2 emissions and
22 percent of nationwide NOX emissions in 2002. States, in fact, can choose not to participate in
the optional cap-and-trade program and can choose to obtain equivalent emissions reductions
from other sectors. However, EPA believes that a region-wide cap-and-trade system for the
power sector is the best approach for reducing emissions. The modeling done with IPM assumes
a region-wide cap and trade system on the power sector for the States covered. However, EPA
recognizes that States may choose to cover other sources and may use a different approach for
reducing emissions.
IPM has been used for evaluating the economic and emission impacts of environmental
policies for over a decade. The model's base case incorporates title IV of the Clean Air Act (the
Acid Rain Program), the NOX SIP Call, various New Source Review (NSR) settlements, and
several State rules affecting emissions of SO2 and NOX that were finalized prior to April of 2004.
The NSR settlements include agreements between EPA and Southern Indiana Gas and Electric
Company (Vectren), Public Service Enterprise Group, Tampa Electric Company, We Energies
(WEPCO), Virginia Electric & Power Company (Dominion), and Santee Cooper. IPM also
includes various current and future State programs in Connecticut, Illinois, Maine,
Massachusetts, Minnesota, New Hampshire, North Carolina, New York, Oregon, Texas, and
-------
Wisconsin. IPM includes State rules that have been finalized and/or approved by a State's
legislature or environmental agency. The base case is used to provide a reference point to
compare environmental policies and assess their impacts and does not reflect a future scenario
that EPA predicts will occur.
The economic modeling presented in this chapter has been developed for specific
analyses of the power sector. Thus, the model has been designed to reflect the industry as
accurately as possible. As a result, EPA has used discount rates in IPM that are appropriate for
the various types of investments and other costs that the power sector incurs. The discount rates
used in IPM may differ from discount rates used in other EPA analyses done for CAIR,
particularly the discount rates used in the benefits analysis that are assumed to be social discount
rates. EPA uses the best available information from utilities, financial institutions, debt rating
agencies, and government statistics as the basis for the discount rates used for power sector
modeling. These discount rates have undergone review by the power sector and the Energy
Information Administration. EPA's discount rate approach has not been challenged in court.
EPA's modeling is based on its best judgment for various input assumptions that are
uncertain, particularly assumptions for future fuel prices and electricity demand growth. To
some degree, EPA addresses the uncertainty surrounding these two assumptions through its
sensitivity analysis, which is discussed in Appendix D.
More detail on IPM can be found in the model documentation, which provides additional
information on the assumptions discussed here as well as all other assumptions and inputs to the
model (www.epa.gov/airmarkets/epa-ipm).
7.2 Projected SO2 and NOX Emissions and Reductions
Because of the existence of a bank of allowances under the title IV Acid Rain Program
that sources will be allowed to use under the requirements of CAIR, emissions of SO2 in 2010
and 2015 will be higher than the caps that are required for CAIR. Table 7-2 provides projected
emissions levels.
As shown in Figure 7-2, the results of EPA modeling of CAIR show that substantial SO2
emissions reductions occur in the Midwest and Mid-Atlantic regions of the country. Significant
NOX emissions reductions occur across the CAIR region (see Figure 7-3), and with CAIR, ozone
season NOX emissions reductions are lower than they would have been with the NOX SIP Call
(base case) (see Figure 7-4). For NOX, the annual CAIR cap achieves greater emission
reductions during the ozone season than the NOX SIP Call summer requirement.
7-4
-------
Table 7-2. Projected Emissions of SO2 and NOX with the Base Case" (No Further Controls)
and with CAIR (Million Tons)
Coverage
S02
(annual)
NOX
(annual)
NOX
(summer)
Nationwide
CAIR Region
Nationwide
CAIR Region
Nationwide
CAIR Region
2010
C3
u
m
9.7
8.8
3.6
2.8
1.2
0.8
d
U
6.1
5.2
2.4
1.5
1.0
0.7
fl
'&
1
3.6
3.6
1.2
1.3
0.2
0.2
2015
C3
o
m
8.9
8.0
3.7
2.8
1.2
0.8
d
U
4.9
4.1
2.1
1.3
0.9
0.6
fl
c
i
1
4.0
4.0
1.5
1.5
0.3
0.3
2020
C3
u
m
8.6
7.9
3.7
2.9
1.2
0.8
d
U
4.2
3.4
2.1
1.3
1.0
0.6
fl
fl
1
1
4.5
4.5
1.6
1.6
0.3
0.3
Note: Numbers may not add due to rounding. The emissions data presented here are EPA modeling results and
the CAIR region includes States modeled for the annual SO2 and NOX requirements. "Summer" is from
May 1-September 30, which is the ozone season.
a Base case includes title IV Acid Rain Program, NOX SIP Call, and State rules finalized before March 2004.
Source: Integrated Planning Model run by EPA.
7.3 Projected Costs
For the modeled region, EPA projects that the annual incremental costs of CAIR are $2.4
billion in 2010 and $3.6 billion in 2015 (see Table 7-3). In 2020, the annual costs are $4.4
billion. The cost of electricity generation represents roughly one-third to one-half of total
electricity costs, with transmission and distribution costs representing the remaining portion. A
better impact measure is the impact on electricity pricing, which is shown in a later table. The
marginal costs of CAIR for SO2 and NOX can also be found in Table 7-3.
7.4 Projected Control Technology Retrofits
CAIR is projected to result in the installation of an additional 64 GW of flue gas
desulfurization (scrubbers) on existing coal-fired generation capacity for SO2 control and an
additional 34 GW of selective catalytic reduction technology (SCR) on existing coal-fired
generation capacity for NOX control by 2015 (see Table 7-4). The first phase of CAIR will result
in 37 GW of additional scrubbers and 14 GW of SCR by 2010. Much of the NOX reductions
7-5
-------
Base Case 2010
CAIR2010
Base Case 2015
CAIR2015
Scale: I 1,373,038 tons in Ohio, Base Case 2010
Figure 7-2. SO2 Emissions from the Power Sector in 2010 and 2015 with and without
CAIR
Source: Integrated Planning Model run by EPA.
Note: Arkansas, Delaware, and New Jersey are not included in the annual SO2 and NOX requirements of the
Final CAIR. These States are included in the ozone season requirement only. Modeling presented here did not
include an ozone season requirement and covered a region that is inclusive of these three States for the annual
requirements. Please see earlier modeling background discussion for more detail.
7-6
-------
^^1 Base Case 2010
I I CAIR2010
I I Base Case 2015
CAIR2015
Scale: [I 274,372 tons in Ohio, Base Case 2015
Figure 7-3. NOX Emissions from the Power Sector in 2010 and 2015 With and Without
CAIR.
Source: Integrated Planning Model run by EPA.
Note: Arkansas, Delaware, and New Jersey are not included in the annual SO2 and NOX requirement of the Final
CAIR. These States are included in the ozone season requirement only. Modeling presented here did not include an
ozone season requirement and covered a region that is inclusive of these three States for the annual requirements.
Please see earlier modeling background discussion for more detail.
7-7
-------
Base Case 2010
CAIR2010
Base Case 2015
CAIR2015
I
Scale: I 93,685 tons in Texas, Base Case 2010
Figure 7-4. Ozone Season NOX Emissions from the Power Sector in 2010 and 2015 with and
without CAIR
Source: Integrated Planning Model run by EPA.
Note: Arkansas, Delaware, and New Jersey are not included in the annual SO2 and NOX requirements of the Final CAIR.
These States are included in the ozone season requirements only. Modeling presented here did not included an ozone season
requirement and covered a region that is inclusive of these three States for the annual requirements. Please see earlier
modeling background discussion for more detail.
-------
Table 7-3. Annualized Regional Cost of CAIR and Marginal Cost of SO2 and NOX
Reductions with CAIR ($1999)
Annualized Cost (billions)
Marginal Cost ($/ton)
SO2
NOX
2010
$2.4
$700
$1,300
2015
$3.6
$1,000
$1,600
2020
$4.4
$1,400
$1,600
Note: Numbers rounded to the nearest hundred million for annualized cost and nearest hundred for marginal cost.
Source: Integrated Planning Model run by EPA.
Table 7-4. Pollution Controls by Technology with the Base Case (No Further Controls)
and with CAIR (GW)
Technology
Scrubbers
SCR
Base Case Total
(Cumulative)
2010 2015 2020
110 116 117
111 119 121
Incremental with CAIR
2010
37
14
2015 2020
64 82
34 33
Total with CAIR
(Cumulative)
2010 2015 2020
147 180 199
126 152 154
Note: Numbers may not add due to rounding. Base case retrofits include existing scrubbers and SCR as well as
additional retrofits for the Title IV Acid Rain Program, the NOX SIP call, NSR settlements, and various
State rules.
Source: Integrated Planning Model run by EPA.
achieved in the first phase of the rule can be attributed to the large pool of existing SCR that are
used during the ozone season in the NOX SIP call region that, for relatively little cost, run the
SCRs year-round. A small number of coal-fired units also install selective noncatalytic
reduction technology (SNCR) for NOX control under CAIR. Emission reductions are achieved
through a combination of compliance options, such as additional pollution control installations,
generation shifts towards more efficient electricity producing units, and fuel and coal switching.
7.5 Projected Generation Mix
Table 7-5 and Figure 7-5 show the generation mix with CAIR. Coal-fired generation and
natural gas-fired generation are projected to remain relatively unchanged because of the phased-
in nature of CAIR, which allows industry the appropriate amount of time to install the necessary
pollution controls.
7-9
-------
Table 7-5. Generation Mix with the Base Case (No Further Controls) and with CAIR
(Thousand GWhs)
Generating
Fuel Use
Coal
Oil/Natural Gas
Other
Total
2003
1,970
758
1,120
3,848
Base
Case
2,198
111
1,223
4,198
2010
CAIR
2,163
809
1,218
4,190
Change
from
Base
Case
-1.6%
4.1%
-0.4%
-0.2%
2015
Base
Case
2,242
1,026
1,235
4,503
CAIR
2,195
1,072
1,233
4,499
Change
from
Base
Case
-2.1%
4.5%
-0.2%
-0.1%
2020
Base
Case
2,410
1,221
1,218
4,850
CAIR
Percent
Change
2,381
1,250
1,217
4,847
Change
from
Base
Case
-1.2%
2.3%
-0.1%
-0.1%
Note: Numbers may not add due to rounding.
Source: 2003 data are from EIA and projections are from the Integrated Planning Model run by EPA.
$000
Other
Coal
Base
Case
CAIR
Base
Case
CAIR
Base
Case
CAIR
Figure 7-5. Generation Mix with and without CAIR
Source: Integrated Planning Model run by EPA.
7-10
-------
Relative to the base case, about 5.3 GW of coal-fired capacity is projected to be
uneconomic to maintain (about 1.7 percent of all coal-fired capacity and 0.5 percent of all
generating capacity), and about 1 GW of coal-fired capacity is projected to repower to natural
gas or Integrated Gasification Combined Cycle (IGCC). Uneconomic units, for the most part,
are small and infrequently used generating units that are dispersed throughout the CAIR region.
In practice, units projected to be uneconomic to maintain may be "mothballed," retired, or kept
in service to ensure transmission reliability in certain parts of the grid. EPA modeling is unable
to distinguish between these potential outcomes. IPM can only predict that specific generating
units are uneconomic to maintain, based on their fuel, operating and fixed costs, and whether
they are needed to meet both demand and reliability reserve requirements. "Repowering"
converts units to combined-cycle natural gas or IGCC.
7.6 Projected Capacity Additions
In addition, EPA projects that future growth in electric demand will be met with a
combination of new natural gas- and coal-fired capacity (see Table 7-6).
Table 7-6. Total Coal and Natural Oil/Gas-Fired Capacity by 2020 (GW)
Pulverized Coal
IGCC
Oil/Gas
Current
305
0.6
395
Base Case
318
8
467
CAIR
314
9
469
Source: Current data are from EPA's NEEDS 2004. Projections are from the Integrated Planning Model run by
EPA.
7.7 Projected Coal Production for the Electric Power Sector
Coal production for electricity generation is expected to increase relative to current
levels, with or without CAIR (see Table 7-7 and Figure 7-6). The reductions in emissions from
the power sector will be met through the installation of pollution controls for SO2 and NOX
removal. The pollution controls can achieve up to a 95 percent SO2 removal rate, which allows
industry to rely more heavily on local bituminous coal in the eastern and central parts of the
country that has a higher sulfur content and is less expensive to transport than western
subbituminous coal.
7.8 Projected Retail Electricity Prices
Retail electricity prices for the CAIR region are projected to increase a small amount
with CAIR (see Table 7-8 and Figure 7-7). The cap-and-trade approach allows industry to meet
7-11
-------
Table 7-7. Coal Production for the Electric Power Sector with the Base Case (No Further
Controls) and with CAIR (Million Tons)
Supply Area
Appalachia
Interior
West
National
2000
299
131
475
905
2003
275
135
526
936
Base Case
2010
325
161
603
1,089
2015
315
162
631
1,109
2020
301
173
714
1,188
CAIR
2010
306
164
607
1,077
2015
310
193
579
1,082
2020
331
219
607
1,156
Source: 2000 and 2003 data are derived from EIA data. All projections are from the Integrated Planning Model run
by EPA.
Northwest
owder River
Basin North Dakota
Lignite
Appalachia
lAf__ L
West
* Other"
Western V
^Interior A ~o n o mt
Interior si ii|
Scale: Appalachia 2000 = 299 million tons
Figure 7-6. Current Coal Production Levels and Projected Production with
CAIR
Source: Integrated Planning Model run by EPA.
7-12
-------
Table 7-8. Projected Regional Retail Electricity Prices with the Base Case (No Further
Controls) and with CAIR (Mills/kWh)
Year
Base Case
CAIR
Percent Change
2010
2015
2020
58
61
61
59
62
62
2.0%
2.7%
1.8%
Source: EPA's Retail Electricity Price Model.
.n
i
ou.u
7n n
60.0
en n
Af\ e\
OA A
on n
m n
n n
^
2000 2010 2015
Base Case CAI R
Figure 7-7. Regional Electricity Prices with and without CAIR
Source: Integrated Planning Model run by EPA.
the requirements of CAIR in the most cost-effective manner, thereby minimizing the costs
passed on to consumers. Regional retail electricity prices are projected to be 2 to 3 percent
higher with CAIR. Retail electricity prices by NERC region are provided in Table 7-9
(Figure 7-8). These results show small increases in retail prices for the NERC regions in the
eastern part of the country. By 2020, CAIR region retail electricity prices are projected to be
roughly 1.8 percent higher with CAIR (Table 7-8 and Figure 7-7).
7-13
-------
Table 7-9. Retail Electricity Prices by NERC Region with the Base Case (No Further
Controls) and with CAIR (Mills/kWh)
Power
Region Primary States Included
ECAR (1) OH, MI, IN, KY, WV, PA
ERCOT(2) TX
MAAC (3) PA, NJ, MD, DC, DE
MAIN (4) IL, MO, WI
MAPP (5) MN, IA, SD, ND, NE
NY (6) NY
NE (7) VT, NH, ME, MA, CT, RI
FRCC (8) FL
STV (9) VA, NC, SC, GA, AL, MS,
TN, AR, LA
SPP (10) KS, OK, MO
Regionwide
2000
57.4
65.1
80.4
61.2
57.4
104.3
89.9
67.9
59.3
59.3
66.0
Base Case
2010 2015 2020
51.7 55.2 56.1
57.9 64.4 62.6
59.3 69.4 72.2
52.6 57.8 61.0
52.8 49.3 47.6
82.8 87.9 88.1
77.4 83.9 82.8
71.2 71.3 69.5
56.2 55.1 55.3
54.2 57.0 56.7
58.0 60.8 61.0
CAIR
2010 2015 2020
53.8 58.5 58.0
59.3 64.6 63.3
61.2 71.7 72.8
54.0 60.3 62.0
52.9 49.6 48.0
83.3 88.8 88.4
77.5 84.7 83.0
71.7 72.3 70.5
57.0 56.2 56.6
54.6 57.5 57.0
59.2 62.4 62.1
Percentage Change
2010 2015 2020
4.0% 5.9% 3.4%
2.5% 0.2% 1.2%
3.2% 3.4% 0.8%
2.6% 4.3% 1.7%
0.2% 0.7% 0.8%
0.5% 1.0% 0.3%
0.1% 1.0% 0.2%
0.8% 1.3% 1.5%
1.5% 2.1% 2.3%
0.7% 0.9% 0.6%
2.0% 2.7% 1.8%
Source: EPA's Retail Electricity Price Model. 2000 prices are from EIA's AEO 2003.
Figure 7-8. NERC Power Regions
7-14
-------
7.9 Projected Fuel Price Impacts
The impacts of CAIR on coal prices and natural gas prices before shipment are shown
below (Table 7-10).
Table 7-10. Henry Hub Natural Gas Prices and Average Delivered Coal Prices with the
Base Case (No Further Controls) and with CAIR ($1999)
Fuel
Natural Gas
Coal
2000
4.15
1.25
Base Case
2010
3.20
1.05
2015
3.25
1.01
2020
3.16
0.96
CAIR
2010
3.25
1.05
2015
3.30
0.98
2020
3.20
0.93
Percentage Change
2010 2015 2020
1.6% 1.5% 1.3%
0.0% -3.0% -3.1%
Source: Integrated Planning Model run by EPA. 2000 natural gas data are from Platts GASdat. 2000 coal prices
are from EIA.
Note: Coal price changes largely result from changes in the mix of coal types used. Delivered coal prices vary
widely, but large changes in the cost of each type of coal are not projected.
7.10 Key Differences in EPA Model Runs for Final CAIR Modeling
As previously stated, the emissions, cost, air quality, and benefits analyses done for the
final CAIR are from a modeling scenario that requires annual SO2 and NOX reductions in 26
States and the District of Columbia and ozone season NOX requirements in Connecticut and
Massachusetts (See Figure 7-1). This modeling differs from today's final CAIR, in that
Arkansas, Delaware, and New Jersey are not included in the annual SO2 and NOX requirements,
and various States are required to meet an ozone season NOX requirement (See Figure 7-9).
Modeling was done based upon the Final CAIR region, and in large part, results under this
scenario are not significantly different in scope or magnitude.
Coal production, minemouth coal and wellhead natural gas prices, generation and
generating capacity, and electricity prices do not differ significantly between the scenarios (all
less than a 1 percent difference in modeling results). Nationwide emissions of SO2 and NOX are
roughly 1 to 3 percent higher when Arkansas, Delaware, and New Jersey are not included in the
annual programs. The only other notable difference in the modeling results is the difference in
retrofits; retrofits of SCR and FGD are roughly 2-4 percent lower without these three States. All
IPM runs done in support of CAIR and used as part of the final CAIR package are in the final
CAIR Docket and can be found on EPA's website: (http://www.epa.gov/airmarkets/
epa-ipm/iaqr.html). A complete list of IPM runs can be found in Appendix D of this RIA.
7-15
-------
• States controlled for both SO2 and NOx
S States controlled for Ozone Season NOx
D States not covered by CAIR
Figure 7-9. Final CAIR Region
Note: Delaware and New Jersey are not included in the Final CAIR for the annual SO2 and NOX
requirements. However, EPA intends on incorporating these two States in the annual CAIR program
through a separate rulemaking. See earlier discussion for more detail.
In addition, EPA did an analysis to determine the incremental impact of including
Delaware and New Jersey in the annual SO2 and NOX requirements from inclusion in the ozone
season only NOX requirement. Inclusion of these two States leads to roughly an additional fifty
thousand tons of SO2 reduction and ten thousand tons of NOX reduction in 2015. The
incremental cost of including these two States is estimated to be roughly $40 million in 2015.
Please see "Proposed Rules" section of today's Federal Register publication for more detail on
the inclusion of Delaware and New Jersey for the annual CAIR requirements.
7.11 Projected Primary PM Emissions from Power Plants
IPM does not project primary PM emissions from power plants. These emissions are
projected using a combination of IPM outputs and emission factors. Separate methodologies are
used to project filterable PM emissions and condensible PM emissions. The sum is the total
projected primary PM emissions.
For filterable PM emissions, emission factors were developed for each unit based on
historical information regarding installed emissions controls and types of fuel burned. This
methodology tends to underpredict reductions in filterable PM emissions between the base case
and the control case because no changes are assumed in the emission factors even if a unit is
projected to install a control such as an FGD, which could lead to a decrease in filterable PM
emissions.
For condensible PM emissions, emission factors were changed between the base case and
the control case to reflect SO2 controls projected to be installed in the control case. Although
7-16
-------
EPA used the best emission factors available for its analysis, these emission factors did not
account for the potential changes in condensible PM emissions due to the installations of SCRs.
The formation of additional condensible PM (in the form of SO3 and H2SO4) in units with SCRs
depends on a number of factors, including coal sulfur content, combustion conditions and
characteristics of the catalyst used in the SCR, and is likely to vary widely from unit to unit.
SCRs are generally designed and operated so that they minimize increases in condensible PM.
This limitation leads to an overprediction of reductions in condensible PM emissions for units
with SCRs. For a more complete description of the methodologies used to project PM emissions
see "Document for the 2001 Electrical Generating Unit (EGU) Trends Procedures Report,"
Section 4.2, September 25, 2003.
7.12 Limitations of Analysis
EPA's modeling is based on its best judgment for various input assumptions that are
uncertain. Assumptions for future fuel prices and electricity demand growth deserve particular
attention because of the importance of these two key model inputs to the power sector. To some
degree, EPA addresses the uncertainty surrounding these two assumptions through its sensitivity
analysis, which is discussed in Appendix D. As a general matter, the Agency selects the best
available information from available engineering studies of air pollution controls and has set up
what it believes is the most reasonable modeling framework for analyzing the cost, emission
changes, and other impacts of regulatory controls.
The annualized cost estimates of the private compliance costs that are provided in this
analysis are meant to show the increase in production (engineering) costs of CAIR to the power
sector. In simple terms, the private compliance costs that are presented are the annual increase
in revenues required for the industry to be as well off after CAIR is implemented as before. To
estimate these annualized costs, EPA uses a conventional and widely-accepted approach that is
commonplace in economic analysis of power sector costs for estimating engineering costs in
annual terms. For estimating annualized costs, EPA has applied a capital recovery factor (CRF)
multiplier to capital investments and added that to the annual incremental operating expenses.
The CRF is derived from estimates of the cost of capital (private discount rate), the amount of
insurance coverage required, local property taxes, and the life of capital. The private compliance
costs presented earlier are EPA's best estimate of the direct private compliance costs of CAIR.
The annualization factor used for pure social cost calculations (for annualized costs)
normally includes the life of capital and the social discount rate. For purposes of benefit-cost
analysis of this rule, EPA has calculated the annualized social costs using the discount rates from
the benefits analysis for CAIR (3 percent and 7 percent and a 30 year life of capital. The cost of
added insurance necessary because of CAIR was included in the calculations, but local taxes
7-17
-------
were not included because they are considered to be transfer payments, and not a social cost).
Using these discount rates, the social costs of CAIR are $1.91 billion in 2010 and $2.56 billion in
2015 using a discount rate of 3 percent, and $2.14 billion in 2010 and $3.07 billion in 2015 using
a discount rate of 7 percent.
The annualized regional cost of CAIR, as quantified here, is EPA's best assessment of
the cost of implementing CAIR, assuming that States adopt the model cap and trade program.
These costs are generated from rigorous economic modeling of changes in the power sector due
to CAIR. This type of analysis using IPM has undergone peer review and federal courts have
upheld regulations covering the power sector that have relied on IPM's cost analysis.
The direct private compliance cost includes, but is not limited to, capital investments in
pollution controls, operating expenses of the pollution controls, investments in new generating
sources, and additional fuel expenditures. EPA believes that the cost assumptions used for CAIR
reflect, as closely as possible, the best information available to the Agency today. The relatively
small cost associated with monitoring emissions, reporting, and record keeping for affected
sources is not included in these annualized cost estimates, but EPA has done a separate analysis
and estimated the cost to be less than $42 million (see Section X. B. Paperwork Reduction Act).
Furthermore, there are some unquantified costs that EPA wants to identify as limits to its
analysis. These costs include the costs of federal and State administration of the program, which
we believe are modest given our experience with the Acid Rain Program and the NOX Budget
Trading Program and likely to be less than the alternative of States developing approvable SIPs,
securing EPA approval of those SIPs, and Federal/State enforcement to deal with the air
pollution transport problem that CAIR addresses. There also may be unquantified costs of
transitioning to CAIR, such as the costs associated with the retirement of smaller or less efficient
electricity generating units, and employment shifts as workers are retrained at the same company
or re-employed elsewhere in the economy. There are certain relatively small permitting costs
associated with Title IV that new program entrants face (we believe there are far less than 1,000
new entrants who may require one day of additional work for trading permits). In a separate
analysis explained later in this RIA, the EPA estimates the indirect costs and impacts of higher
electricity prices on the entire economy (see Regulatory Impact Analysis for the Final Clean Air
Interstate Rule, Appendix E [March 2005]).
Cost estimates for CAIR are based on results from ICF's Integrated Planning Model.
The model minimizes the costs of producing electricity (including abatement costs) while
meeting load demand and other constraints (full documentation for IPM can be found at
www.epa.gov/airmarkets/epa-ipm). The structure of the model assumes that the electric utility
industry will be able to meet the environmental emission caps at least cost. Montgomery (1972)
7-18
-------
has shown that this least cost solution corresponds to the equilibrium of an emission permit
system. See also Atkinson and Tietenburg (1982), Krupnick et al. (1980), and McGartland and
Gates (1985). However, to the extent that transaction and/or search costs, combined with
institutional barriers, restrict the ability of utilities to exhaust all the gains from emissions
trading, costs are underestimated by the model. Utilities in the IPM model also have "perfect
foresight." To the extent that utilities misjudge future conditions affecting the economics of
pollution control, costs may be understated as well.
From another vantage point, this modeling analysis does not take into account the
potential for advancements in the capabilities of pollution control technologies for SO2 and NOX
removal as well as reductions in their costs over time. Market-based cap and trade regulation
serves to promote innovation and the development of new and cheaper technologies. As an
example, recent cost estimates of the Acid Rain SO2 trading program by Resources for the
Future (RFF) and MIT's Center for Energy and Environmental Policy Research (CEEPR) have
been as much as 83 percent lower than originally projected by the EPA (see Carlson et al., 2000;
Ellerman, 2003). It is important to note that the original analysis for the Acid Rain Program
done by EPA also relied on an optimization model like IPM. Ex ante, EPA cost estimates of
roughly $2.7 to $6.2 billion1 in 1989 were an overestimate of the costs of the program in part
because of the limitation of economic modeling to predict technological improvement of
pollution controls and other compliance options such as fuel switching. Ex post estimates of the
annual cost of the Acid Rain SO2 trading program range from $1.0 to $1.4 billion. Harrington et
al. have examined cost analyses of EPA programs and found a tendency for predicted costs to
overstate actual implementation costs in market-based programs (Harrington, Morgenstern, and
Nelson, 2000). In recognition of this, EPA's mobile source program uses adjusted engineering
cost estimates of pollution control equipment and installation costs to account for this fact, which
EPA has not done in this case.2
It is also important to note that the capital cost assumptions for scrubbers used in EPA
modeling applications are highly conservative. These are a substantial part of the compliance
costs. Data available from recent published sources show the reported FGD costs from recent
installations to be below the levels projected by IPM.3 In addition, EPA also conducted a survey
1 2010 Phase II cost estimate in $1995.
2 See recent regulatory impact analysis for the Tier 2 Regulations for passenger vehicles (1999) and Heavy-Duty
Diesel Vehicle Rules (2000). There is also evidence that scrubber costs will decrease in the future because of the
learning-by-doing phenomenon, as more scrubbers are installed (see Manson, Nelson, and Neumann, 2002).
3 There is also evidence that scrubber costs will decrease in the future because of the learning-by-doing phenomenon,
as more scrubbers are installed (see Manson, Nelson, and Neumann, 2002).
7-19
-------
of recent FGD installations and compared the costs of these installations to the costs used in
IPM. This survey included small, mid-size, and large units. Examples of the comparison of
these referenced published data with the FGD capital cost estimates obtained from IPM are
provided in the Final CAIR docket.
EPA's latest update of IPM incorporates State rules or regulations adopted before March
2004 and various NSR settlements. Documentation for IPM can be found at
www.epa.gov/airmarkets/epa-ipm. Any State or settlement action since that time has not been
accounted for in our analysis in this chapter.
As configured in this application, the IPM model does not take into account demand
response (i.e., consumer reaction to electricity prices). The increased retail electricity prices
shown in Tables 7-8 and 7-9 would prompt end users to curtail (to some extent) their use of
electricity and encourage them to use substitutes.4 The response would lessen the demand for
electricity, resulting in electricity price increases slightly lower than IPM predicts, which would
also reduce generation and emissions. Because of demand response, certain unquantified
negative costs (i.e., savings) result from the reduced resource costs of producing less electricity
because of the lower quantity demanded. To some degree, these saved resource costs will offset
the additional costs of pollution controls and fuel switching that we would anticipate with CAIR.
Although the reduction in electricity use is likely to be small, the cost savings from such a large
industry ($250 billion in revenues in 2003) is likely to be substantial. EIA analysis examining
multi-pollutant legislation under consideration in 2003 indicates that the annualized costs of
CAIR may be overstated substantially by not considering demand response, depending on the
magnitude and coverage of the price increases.5
Recent research suggests that the total social costs of a new regulation may be affected
by interactions between the new regulation and pre-existing distortions in the economy, such as
taxes. In particular, if cost increases due to a regulation are reflected in a general increase in the
price level, the real wage received by workers may be reduced, leading to a small fall in the total
amount of labor supplied. This "tax interaction effect" may result in an increase in deadweight
loss in the labor market and an increase in total social costs. Although there is a good case for
the existence of the tax interaction effect, recent research also argues for caution in making prior
assumptions about its magnitude. However, there are currently no government-wide economic
4 The degree of substitution/curtailment depends on the price elasticity of demand for electricity.
5 See "Analysis of S. 485, the Clear Skies Act of 2003, and S. 843, the Clean Air Planning Act of 2003." Energy
Information Administration. September, 2003. EIA modeling indicated that the Clear Skies Act of 2003 (a
nationwide cap and trade program for SO2, NOX, and mercury), demand response could lower present value costs
by as much as 47% below what it would have been without an emission constraint similar to CAIR.
7-20
-------
analytical guidelines which discuss the tax interaction effect and its potential relevance for
estimation of federal program costs and benefits. The limited empirical data available to support
quantification of any such effect leads to this qualitative identification of the costs.
On balance, after consideration of various unquantified costs (and savings that are
possible), EPA believes that the annual private compliance costs that we have estimated are
more likely to overstate the future annual compliance costs that industry will incur, rather than
understate those costs.
7.13 Significant Energy Impact
According to E. O. 13211: Actions that Significantly Affect Energy Supply, Distribution,
or Use, this rule is significant because it has a greater than 1 percent impact on the cost of
natural gas and electricity production and it results in the retirement of greater than 500 MW of
coal-fired generation.
Several aspects of CAIR are designed to minimize the impact on energy production.
First, EPA recommends a trading program rather than the use of command-and-control
regulations. Second, compliance deadlines are set cognizant of the impact that those deadlines
have on electricity production. Both of these aspects of CAIR reduce the impact of the proposal
on the electricity sector.
7.14 Industry-Sector Impacts
EPA estimates the direct costs of implementing CAIR at $3.6 billion in 2015 in the CAIR
region. Given the impact of this rule on electricity generators, we believe it is important to
gauge the extent to which the rule might affect other industry sectors. To do so, we conducted a
limited analysis of the economy-wide effects of implementing CAIR.
EPA was particularly interested in learning how anticipated changes in electricity prices
might affect industry sectors that are large electricity users. The models we employed indicated
those impacts would be small, even without incorporating the beneficial economic effects of
CAIR-related air quality improvements such as improved worker health and productivity.
Rather, our analyses continue to show that the value of even the limited subset of CAIR benefits
we were able to quantify substantially outweigh implementation costs.
By focusing only on cost-side spillover effects on the economy, the industry-sector
impacts projected by our macroeconomic models are likely overstated, primarily because the
positive market impacts of CAIR on labor availability and productivity are excluded. In this
regard, an independent panel of experts has encouraged EPA to work toward incorporating both
7-21
-------
beneficial and costly effects when modeling the economy-wide consequences of regulation.
EPA is working to develop this capability.
Although neither model has yet been configured to include the indirect economic benefits
of air quality improvements, EPA employed two distinct computable general equilibrium models
to gauge the potential magnitude of the economy-wide effects of C AIR implementation costs.
The first model, known as IGEM, has a long track record and was used by the Agency for the
first of the two Clean Air Act Section 812 studies. The other model, called EMPAX-CGE, is
currently in peer review and has the advantage of disaggregating the United States into multiple
regions. As with all models, these tools have their respective strengths and weaknesses, and
differences in data and choice of functional form imply that the models are likely to show
slightly different results. Despite the differences between the models, the results of the
respective analyses show similarly small impacts of CAIR on energy-intensive industries. For
example, production changes for the chemical manufacturing industry are estimated at -0.01
percent to -0.04 percent in 2010. Furthermore, if labor productivity improvements and other
benefits of improved air quality were included, the small production output decreases projected
by both models might be partially or entirely offset. Please see Appendix E for more details.
7.15 References
Atkinson, S., and T. Tietenberg. 1982. "The Empirical Properties of Two Classes of Design for
Transferable Discharge Permit Markets." Journal of Environmental Economics and
Management 9:101-121
Carlson, Curtis, Dallas R. Burtraw, Maureen, Cropper, and Karen L. Palmer. 2000. "Sulfur
Dioxide Control by Electric Utilities: What Are the Gains from Trade?" Journal of
Political Economy 108(#6): 1292-1326.
Ellerman, Denny. January 2003. Ex Post Evaluation of Tradable Permits: The U.S. SO2
Cap-and-Trade Program. Massachusetts Institute of Technology Center for Energy and
Environmental Policy Research.
Harrington, W., R.D. Morgenstern, and P. Nelson. 2000. "On the Accuracy of Regulatory Cost
Estimates." Journal of Policy Analysis and Management 19(2):297-322.
Krupnick, A., W. Gates, and E. Van De Verg. 1980. "On Marketable Air Pollution Permits:
The Case for a System of Pollution Offsets." Journal of Environmental Economics and
Management 10:233-47.
Manson, Nelson, and Neumann. 2002. "Assessing the Impact of Progress and Learning Curves
on Clean Air Act Compliance Costs." Industrial Economics Incorporated.
7-22
-------
McGartland, A., and W. Gates. 1985. "Marketable Permits for the Prevention of Environmental
Deterioration." Journal of Environmental Economics and Management 12:207-228.
Montgomery, W. David. 1972. "Markets in Licenses and Efficient Pollution Control
Programs." Journal of Economic Theory 5(3):395-418.
7-23
-------
CHAPTER 8
STATUTORY AND EXECUTIVE ORDER IMPACT ANALYSES
This chapter presents discussion and analyses relating to relevant Executive Orders
and statutory requirements relevant for CAIR. We discuss potential impacts to affected
small entities as required by the Regulatory Flexibility Act (RFA), as amended by the Small
Business Regulatory Enforcement Fairness Act (SBREFA). We also describe the analysis
conducted to meet the requirements of the Unfunded Mandates Reform Act of 1995
(UMRA) that assess the impact of CAIR for state, local and Tribal governments and the
private sector. Analyses conducted to comply with the Paperwork Reduction Act (PRA) are
also discussed. In addition, we address the requirements of Executive Order (EO) 13045:
Protection of Children from Environmental Health and Safety Risks; EO 13175:
Consultation and Coordination with Indian Tribal Governments; and EO 12898: Federal
Actions to Address Environmental Justice in Minority Populations and Low-Income
Populations. Discussion of Executive Order 13211: Actions that Significantly Affect
Energy Supply, Distribution or Use is provided in Chapter 7 of this report.
8.1 Small Entity Impacts
The Regulatory Flexibility Act (5 U.S.C. § 601 et seq.), as amended by the Small
Business Regulatory Enforcement Fairness Act (Public Law No. 104-121), provides that
whenever an agency is required to publish a general notice of proposed rulemaking, it must
prepare and make available an initial regulatory flexibility analysis, unless it certifies that the
proposed rule, if promulgated, will not have "a significant economic impact on a substantial
number of small entities" (5 U.S.C. § 605[b]). Small entities include small businesses, small
organizations, and small governmental jurisdictions.
For the purposes of assessing the impacts of CAIR on small entities, a small entity is
defined as:
(1) A small business according to the Small Business Administration size
standards by the North American Industry Classification System (NAICS)
category of the owning entity. The range of small business size standards for
electric utilities is 4 billion kilowatt-hours of production or less;
8-1
-------
(2) a small government jurisdiction that is a government of a city, county, town,
district, or special district with a population of less than 50,000; and
(3) a small organization that is any not-for-profit enterprise that is independently
owned and operated and is not dominant in its field.
Table 8-1 lists entities potentially affected by this proposed rule with applicable
NAICS code. It is important to note that the proposed rule leaves states to decide which
sources to control, such that states may choose to regulate source categories in addition to
those listed in Table 8-1.
Table 8-1. Potentially Regulated Categories and Entities"
Category
Industry
Federal
Government
NAICS
Code"
221112
221112C
Examples of Potentially Regulated Entities
Fossil fuel-fired electric utility steam generating units.
Fossil fuel-fired electric utility steam generating units owned by
the federal government.
221112° Fossil fuel-fired electric utility steam generating units owned by
State/Local/ municipalities.
Tribal
Government 921150 Fossil fuel-fired electric utility steam generating units in Indian
Country.
a Include NAICS categories for source categories that own and operate electric generating units only.
b North American Industry Classification System.
0 Federal, state, or local government-owned and operated establishments are classified according to the
activity in which they are engaged.
Courts have interpreted the RFA to require a regulatory flexibility analysis only when
small entities will be subject to the requirements of the rule.1 This rule would not establish
specific requirements applicable to small entities. Instead, it would require states to develop,
adopt, and submit SIP revisions that would achieve the necessary SO2 and NOX reductions,
leaving to states the task of determining how and by which entities these reductions will be
obtained. Because affected states would decide which sources to control and the extent of
emissions reductions each selected source would have to achieve, EPA cannot definitively
1 See Michigan v. EPA. 213 F.3d 663, 668-69 (D.C. Cir. 2000), cert, den. 121 S.Ct. 225, 149 L.Ed.2d 135
(2001). An agency's certification need consider the rule's impact only on entities subject to the rule.
8-2
-------
project the effect of the proposed rule on small entities. Although not required, EPA
conducts a general analysis of the potential impact of CAIR on small entities.
EPA examined the potential economic impacts to small entities associated with this
rulemaking based on assumptions of how the affected states will implement control measures
to meet their NOX and SO2 budgets. This analysis assumes that all affected states choose to
meet their budgets by controlling EGUs only. This analysis does not examine potential
indirect economic impacts associated with CAIR, such as employment effects in industries
providing fuel and pollution control equipment, or the potential effects of electricity price
increases on industries and households.
As is noted in Chapter 7, EPA initially conducted IPM modeling for today's final
action using a control strategy that is similar, but not identical to, the final CAIR
requirements. The control strategy that EPA initially modeled included three additional
States (Arkansas, Delaware and New Jersey) within the region and required these States to
make annual SO2 and NOX reductions. While these three States are included in this analysis,
they are not required to make annual reductions under the final CAIR (in the "Proposed
Rules" section of today's Federal Register publication, EPA is publishing a proposal to
include Delaware and New Jersey in the CAIR region for annual SO2 and NOX reductions).
The implication of this is that total impacts on small entities are somewhat overstated in this
analysis.
8.1.1 Identification of Small Entities
EPA used EGRID data as a basis for compiling the list of potentially affected small
entities. EGRID is EPA's Emissions & Generation Resource Integrated Database, which
contains emissions and resource mix data for virtually every power plant and company that
generates electricity in the United States.2 The data set contains detailed ownership and
corporate affiliation information. For plants burning fossil fuel as the primary fuel,
plant-level boiler and generator capacity, heat input, generation, and emissions data were
aggregated by owner and then parent company. Entities with more than 4 billion kWh of
annual electricity generation were removed from the list, as were municipal-owned entities
serving a population greater than 50,000. Finally, for cooperatives, investor-owned utilities,
and subdivisions that generate less than 4 billion kWh of electricity annually but may be part
of a large entity, additional research on power sales, operating revenues, and other business
activities was performed to make a final determination regarding size. Because the rule does
2 eGRID is available at http://www.epa.gov/cleanenergy/egrid/download.htm.
8-3
-------
not affect units with a generating capacity of 25 MW or less, small entities that do not own at
least one generating unit with a capacity greater than 25 MW were dropped from the data set.
According to EPA's analysis, approximately 185 small entities were exempted by this
provision. Finally, small entities for which IPM does not project generation in 2010 or 2015
in the base case were omitted from the analysis because they are not projected to be
operating and thus will not face the costs of compliance with C AIR. Four small municipal
entities are omitted for this reason. After omitting entities for the reasons above, EPA
identified a total of 75 potentially affected small entities, out of a possible 264. The number
of potentially affected small entities by ownership type is listed in Table 8-2.
Table 8-2. Projected Impact of CAIR on Small Entities
ECU
Ownership
Type
Cooperative
Investor-
Owned Utility
Municipal
Subdivision
Other
Total
Number of
Potentially
Affected
Entities
17
2
49
5
2
75
Total Net
Compliance Cost
($1999 millions)
2010 2015
13.2 22.3
3.1 2.8
-28.2 - 9.1
-15.2 -1.7
0.1 0.1
-27.0 14.4
Number of Small
Entities with
Compliance Costs
>1% of Generation
Revenues
2010 2015
8 11
1 2
15 29
3 3
1 1
28 46
Number of Small
Entities with
Compliance Costs
>3% of Generation
Revenues
2010 2015
5 9
0 0
15 19
0 1
0 1
20 30
Note: The total number of potentially affected entities in this table excludes the 189 entities that have been
dropped because they will not be affected by CAIR. Also, the total number of entities with costs
greater than 1 percent or 3 percent of revenues includes only entities experiencing positive costs. A
negative cost value implies that the group of entities experiences a net savings under CAIR.
Source: IPM and TRUM analysis
8.1.2 Overview of Analysis and Results
This section presents the methodology and results for estimating the impact on CAIR
to small entities in 2010 and 2015 based on the following endpoints:
• annual economic impacts of CAIR on small entities and
• ratio of small entity impacts to revenues from electricity generation.
8-4
-------
8.1.2.1 Methodology for Estimating Impacts ofCAIR on Small Entities
An entity can comply with CAIR through some combination of the following:
installing retrofit technologies, purchasing allowances, switching to a cleaner fuel, or
reducing emissions through a reduction in generation. Additionally, units with more
allowances than needed can sell these allowances on the market. The chosen compliance
strategy will be primarily a function of the unit's marginal control costs and its position
relative to the marginal control costs of other units.
To attempt to account for each potential control strategy, EPA estimates compliance
costs as follows:
^-Compliance ~ ^ ^ Operating^Retrofit ^ ^Fuel ^ ^Allowances ^ ^Transaction ~ A K (o. 1)
where C represents a component of cost as labeled, and A R represents the retail value of
foregone electricity generation.
In reality, compliance choices and market conditions can combine such that an entity
may actually experience a savings in any of the individual components of cost. Under CAIR,
for example, EPA projects that the price of low-sulfur coal will fall as many units install
scrubbers and switch away from low-sulfur coal to cheaper bituminous coal, such that many
entities actually experience a reduction in fuel costs as a result of lower prices due to the
demand shift. Similarly, although some units will forgo some level of electricity generation
(and thus revenues) to comply, this impact will be lessened on these entities by the projected
increase in electricity prices under CAIR as well as reductions in fuel costs, and those not
reducing generation levels will see an increase in electricity revenues. Elsewhere, units
burning high or medium sulfur coal might decide to pay relatively more for low-sulfur coal
under CAIR and sell allowances on the market, in the hopes of negating some or all of their
compliance cost. Because this analysis evaluates the total costs along each of the four
compliance strategies laid out above for each entity, it inevitably captures savings or gains
such as those described. As a result, what we describe as cost is really more of a measure of
the net economic impact of the rule on small entities.
For this analysis, EPA used IPM-parsed output to estimate costs based on the
parameters above, at the unit level. These impacts were then summed for each small entity,
adjusting for ownership share. Net impact estimates were based on the following: operating
and retrofit costs, sale or purchase of allowances, and the change in fuel costs or electricity
generation revenues under CAIR relative to the base case. These individual components of
compliance cost were estimated as follows:
8-5
-------
(1) Operating and retrofit costs: Using the IPM-parsed output for the base case
and CAIR (http://www.epa.gov/airmarkets/epa-ipm/iaqr.html), EPA identified
units that install control technology under CAIR and the technology installed.
The equations for calculating retrofit costs were adopted from EPA's
Technology Retrofit and Updating Model (TRUM). The model calculates the
capital cost (in $/MW); the fixed operation and maintenance (O&M) cost (in
$/MW-year); the variable O&M cost (in $/MWh); and the total annualized
retrofit cost for units projected to install FGD, SCR, or SNCR.
(2) Sale or purchase of allowances: EPA estimated the value of initial SO2 and
NOX allowance holdings. For SO2, units were assumed to retain their Phase II
allowance allocations as determined under EPA's 1998 reallocation of Acid
Rain allowances, adjusted to reflect the 50 percent reduction in 2010 and
65 percent reduction in 2015 under CAIR. Because of the resources involved
in compiling allowance-holding data, the value of banked SO2 allowances was
not considered in this analysis. The implication of this is that the annual net
purchase of allowances may be overstated for some units. For NOX, the state
emission budgets were assumed to be apportioned to units on a heat-input
basis. Each unit was assumed to receive a share of the state NOX emission
budget equal to its share of the total state heat input for that year in the base
case. This is a simplification of what is included in the model rule, which
proposes allocating NOX allowances based on heat input from 1999-2002.3
However, states can ultimately decide how to allocate NOX allowances.
To estimate the value of allowances holdings, allocated allowances were
subtracted from projected emissions, and the difference was then multiplied
by the allowance prices projected by IPM for 2010 and 2015. Units were
assumed to purchase or sell allowances to exactly cover their projected
emissions under CAIR.
(3) Fuel costs: Fuel costs were estimated by multiplying fuel input (MMBtu) by
region and fuel-type-adjusted fuel prices ($/MMBtu) from TRUM. The
change in fuel expenditures under CAIR was then estimated by taking the
difference in fuel costs between CAIR and the base case.
3 A similar approach was used in regulatory impact analyses for the 126 FIP and NOX SIP Call.
8-6
-------
(4) Value of electricity generated: EPA estimated electricity generation by first
estimating unit capacity factor and maximum fuel capacity. Unit capacity
factor is estimated by dividing fuel input (MMBtu) by maximum fuel capacity
(MMBtu). The maximum fuel capacity was estimated by multiplying
capacity (MW) * 8,760 operating hours * heat rate (MMBtu/MWh). The
value of electricity generated is then estimated by multiplying capacity
(MW)*capacity factor*8,760*regional-adjusted retail electricity price
($/MWh).
As discussed later in this analysis, the small entities projected to be affected
by CAIR do not have to operate in a competitive market environment and thus
should be able to pass compliance costs on to consumers. To somewhat
account for this, we incorporated the projected regional-adjusted retail
electricity price calculated under CAIR in our estimation of generation
revenue under CAIR.
(5) Administrative costs: Because most affected units are already monitored as
a result of other regulatory requirements, EPA considered the primary
administrative cost to be transaction costs related to purchasing or selling
allowances. EPA assumed that transaction costs were equal to 1.5 percent of
the total absolute value of a unit's allowances. This assumption is based on
market research by ICF Consulting.
8.1.2.2 Results
The potential impacts of CAIR on small entities are summarized in Table 8-2. All
costs are presented in $1999. EPA estimated the annualized net compliance cost to small
entities to be approximately -$27.0 million in 2010 and $14.4 million in 2015.4 As discussed
below, these negative net compliance costs in 2010 are largely due to higher electricity prices
(and therefore higher generation revenue) and reduced low-sulfur coal prices under CAIR.
Based on EPA analysis, small entities experiencing the greatest impact under CAIR in terms
of cost as a share of revenue are those projected to both reduce output and purchase
allowances.
4 Neither the costs nor the revenues of units that retire under CAIR are included in the impact estimates.
Because these units are better off retiring under CAIR than continuing operation, the true cost of the rule on
these units is not represented by our modeling. The true cost of CAIR for these units is the differential
between their costs in the base case and the costs of meeting their customers' demand under the rule.
8-7
-------
Furthermore, approximately 180 MW of small cooperative capacity (5 units of 61 in
this analysis) are projected by IPM to be uneconomic to maintain under CAIR relative to the
base case, as are approximately 265 MW of small municipality-owned capacity (6 units of
101 in this analysis). One plant owned by a cooperative is projected to be uneconomic, as is
one plant owned by a municipality. Overall, about 445 MW of total small entity capacity, or
1.0 percent of total small entity capacity in the CAIR region, is projected to be uneconomic
to maintain under CAIR relative to the base case. To put these numbers in context, of all
affected capacity under CAIR, about 5.3 GW (1.7 percent) of coal-fired capacity is projected
to be uneconomic to maintain relative to the base case. This comparison suggests that small
entities should not be disproportionately affected by CAIR. In practice, units projected to be
uneconomic to maintain may be "mothballed," retired, or kept in service to ensure
transmission reliability in certain parts of the grid. Our IPM modeling is unable to
distinguish between these potential outcomes. Notably, none of the units affected are likely
to be in a competitive market environment and thus should be able to pass compliance costs
on to consumers.
EPA further assessed the economic and financial impacts of the rule using the ratio of
compliance costs to the value of revenues from electricity generation, focusing in particular
on entities for which this measure is greater than 1 percent. Although this metric is
commonly used in EPA impact analyses, it makes the most sense when as a general matter
an analysis is looking at small businesses that operate in competitive environments.
However, small businesses in the electric power industry often operate in a price-regulated
environment where they are able to recover expenses through rate increases. Given this,
EPA considers the 1 percent measure in this case a crude measure of the price increases these
small entities will be asking of rate commissions or making at publicly owned companies.
Of the 75 small entities considered in this analysis, and 264 total small entities in the
CAIR region, 28 entities may experience compliance costs greater than 1 percent of
generation revenues in 2010, while 46 may in 2015. Entities that experience negative net
costs under CAIR are excluded from these totals. These results do not fully account for the
reality that none of these entities operate in a competitive market and thus should be able to
recover all of their costs of complying with CAIR. It should also be emphasized that under
CAIR, states, through their choice of NOX allowance allocation methodologies, can
potentially mitigate adverse affects of CAIR on small entities. The number of entities with
compliance costs exceeding 3 percent of generation revenues is also included in Table 8-2.
-------
The distribution across entities of economic impacts as a share of base case revenue
is summarized in Table 8-3. Although the distributions of economic impacts on each
ownership type are in general fairly tight, there are a few outliers for which the percentage of
economic impacts as a share of revenue is either very low or very high relative to the
capacity-weighted average. In the cases where entities are projected to experience negative
net impacts that are a high percentage of revenues, these entities have units that are able to
switch to a cheaper, lower-sulfur coal to comply with CAIR and are able to maintain or
increase generation levels, thus increasing revenues. Additionally, entities in regions for
which we project large electricity price increases relative to other regions tend to be among
those at the lower end of the distribution. In the cases where entities are projected to
experience positive net impacts that are a high percentage of revenues, these entities do not
find it economic to retrofit and are unable to switch to a lower sulfur coal. Thus, these
entities comply primarily by purchasing allowances and reducing generation.
Table 8-3. Summary of Distribution of Economic Impacts of CAIR on Small Entities
ECU Ownership
Type
Cooperative
Investor-owned utility
Municipal
Subdivision
Other
All
C ap acity- Weighted
Average Economic
Impacts as a % of
Generation Revenues
2010 2015
1.0% 1.8%
1.6% 1.5%
-3.8% -1.3%
-0.1% 0.0%
-0.3% -0.2%
-0.2% 0.0%
Min
2010
-20.9%
0.4%
-13.8%
- 80.0%
-0.7%
-80.0%
2015
-13.1%
1.5%
-20.4%
- 27.6%
-0.8%
-27.6%
Max
2010
11.5%
2.0%
17.2%
1.9%
1.9%
17.2%
2015
8.4%
1.5%
43.4%
3.1%
3.1%
43.4%
Source: IPM and TRUM analysis
8-9
-------
The separate components of annualized costs to small entities under CAIR are
summarized in Table 8-4. The most significant components of incremental cost to these
entities under CAIR are retrofit and operating cost and allowance purchases. Fuel costs fall
over all ownership groups, because of the combination of switching to bituminous coal,
reduced overall fuel use, and lower prices for low-sulfur coal. Additionally, increases in
electricity generation revenue are experienced over municipal, subdivision, and other, in
2010, and over all ownership types in 2015. This is due largely to the projected increase in
electricity prices under CAIR.
Table 8-4. Incremental Annualized Costs under CAIR Summarized by Ownership
Group and Cost Category ($1999 millions)
ECU Ownership
Type
Cooperative
Investor-Owned
Utility
Municipal
Subdivision
Other
Retrofit +
Operating
Cost
2010 2015
4.1 15.7
-0.1 0.0
2.2 5.6
10.4 9.4
0.374 0.375
Net Purchase of
Allowances
2010 2015
35.2 54.0
6.5 11.1
17.7 30.7
-1.6 1.7
-0.146 -0.106
Fuel Cost
2010 2015
-34.0 -34.8
-6.0 -6.9
-18.7 -27.7
-3.2 -8.4
-0.138 -0.117
Lost Electricity
Revenue
2010 2015
7.5 -13.1
2.6 -1.5
-29.6 -18.0
-20.9 -4.5
-0.007 -0.018
Administrative
Cost
2010 2015
0.3 0.5
0.0 0.1
0.2 0.3
0.05 0.06
0.001 0.001
Note: Numbers may not add to totals in Table 8-2 due to rounding.
Source: IPM and TRUM analysis.
8.1.3 Summary of Small Entity Impacts
EPA examined the potential economic impacts to small entities associated with this
rulemaking based on assumptions of how the affected states will implement control measures
to meet their emissions. While EPA concludes that the RFA as amended by SBREFA does
not apply to CAIR, these impacts have been calculated to provide additional understanding
of the nature of potential impacts, and additional information to the states as they revise SIPs
to meet the emissions budgets set by this rulemaking.
Overall, about 445 MW of total small entity capacity, or 1.0 percent of total small
entity capacity in the CAIR region, is projected to be uneconomic to maintain under CAIR
relative to the base case. In practice, units projected to be uneconomic to maintain may be
8-10
-------
"mothballed," retired, or kept in service to ensure transmission reliability in certain parts of
the grid. Our IPM modeling is unable to distinguish between these potential outcomes.
Furthermore, of the 75 small entities potentially affected, and the 264 small entities in
the CAIR region that are included in EPA's modeling, 28 may experience compliance costs
in excess of 1 percent of revenues in 2010, and 46 may in 2015, based on our assumptions of
how the affected states implement control measures to meet their emissions budgets as set
forth in this rulemaking. Potentially affected small entities experiencing compliance costs in
excess of 1 percent of revenues have some potential for significant impact resulting from
implementation of CAIR. However, as noted above, it is EPA's position that because none
of the affected entities currently operate in a competitive market environment, they should be
able to pass the costs of complying with CAIR on to rate-payers. Furthermore, the decision
to include only units greater than 25 MW in size exempts 185 small entities that would
otherwise be potentially affected by CAIR.
Two other points should be considered when evaluating the impact of CAIR,
specifically, and cap-and-trade programs more generally, on small entities. First, under
CAIR, the cap-and-trade program is designed such that states determine how NOX allowances
are to be allocated across units. A state that wishes to mitigate the impact of the rule on
small entities might choose to allocate NOX allowances in a manner that is favorable to small
entities. .
8.2 Unfunded Mandates Reform Act (UMRA) Analysis
Title II of the UMRA of 1995 (Public Law 104-4)(UMRA) establishes requirements
for federal agencies to assess the effects of their regulatory actions on state, local, and Tribal
governments and the private sector. Under Section 202 of the UMRA, 2 U.S.C. 1532, EPA
generally must prepare a written statement, including a cost-benefit analysis, for any
proposed or final rule that "includes any Federal mandate that may result in the expenditure
by State, local, and Tribal governments, in the aggregate, or by the private sector, of
$100,000,000 or more ... in any one year." A "Federal mandate" is defined under Section
421(6), 2 U.S.C. 658(6), to include a "Federal intergovernmental mandate" and a "Federal
private sector mandate." A "Federal intergovernmental mandate," in turn, is defined to
include a regulation that "would impose an enforceable duty upon State, Local, or Tribal
governments," Section 421(5)(A)(i), 2 U.S.C. 658(5)(A)(i), except for, among other things, a
duty that is "a condition of Federal assistance," Section 421(5)(A)(i)(I). A "Federal private
sector mandate" includes a regulation that "would impose an enforceable duty upon the
private sector," with certain exceptions, Section 421(7)(A), 2 U.S.C. 658(7)(A).
8-11
-------
Before promulgating an EPA rule for which a written statement is needed under
Section 202 of the UMRA, Section 205, 2 U.S.C. 1535, of the UMRA generally requires
EPA to identify and consider a reasonable number of regulatory alternatives and adopt the
least costly, most cost-effective, or least burdensome alternative that achieves the objectives
of the rule.
EPA prepared a written statement for the Supplemental Notice of Proposed
Rulemaking (SNPR) consistent with the requirements of Section 202 of the UMRA.
Furthermore, as EPA stated in the proposal, EPA is not directly establishing any regulatory
requirements that may significantly or uniquely affect small governments, including Tribal
governments. Thus, under CAIR, EPA is not obligated to develop under Section 203 of the
UMRA a small government agency plan. Furthermore, in a manner consistent with the
intergovernmental consultation provisions of Section 204 of the UMRA, EPA carried out
consultations with the governmental entities affected by this rule.
For several reasons, EPA is not concluding that the requirements of the UMRA apply
to CAIR. First, it is questionable whether a requirement to submit a SIP revision would
constitute a federal mandate in any case. The obligation for a state to revise its SIP that
arises out of Section 110(a) of the CAA is not legally enforceable by a court of law and at
most is a condition for continued receipt of highway funds. Therefore, it is possible to view
an action requiring such a submittal as not creating any enforceable duty within the meaning
of Section 421(5)(9a)(I) of UMRA (2 U.S.C. 658 (a)(I)). Even if it did, the duty could be
viewed as falling within the exception for a condition of federal assistance under Section
421(5)(a)(i)(I) of UMRA (2 U.S.C. 658(5)(a)(i)(I)).
As noted earlier, however, notwithstanding these issues, EPA prepared for the SNPR
the statement that would be required by the UMRA if its statutory provisions applied, and
EPA has consulted with governmental entities as would be required by the UMRA. While
not required for CAIR, EPA analyzed the economic impacts of CAIR on government entities
for informational purposes. This analysis does not examine potential indirect economic
impacts associated with CAIR, such as employment effects in industries providing fuel and
pollution control equipment, or the potential effects of electricity price increases on
industries and households.
As is noted in Chapter 7, EPA initially conducted IPM modeling for today's final
action using a control strategy that is similar, but not identical to, the final CAIR
requirements. The control strategy that EPA initially modeled included three additional
States (Arkansas, Delaware and New Jersey) within the region and required these States to
8-12
-------
make annual SO2 and NOX reductions. While these three States are included in this analysis,
they are not required to make annual reductions under the final CAIR (in the "Proposed
Rules" section of today's Federal Register publication, EPA is publishing a proposal to
include Delaware and New Jersey in the CAIR region for annual SO2 and NOX reductions).
The implication of this is that total impacts on government-owned entities are somewhat
overstated in this analysis.
8.2.1 Identification of Government-Owned Entities
Using eGRID data, EPA identified state- and municipality-owned utilities and
subdivisions in the CAIR region. EPA then used IPM-parsed output to associate these plants
with individual generating units. Entities that did not own at least one unit with a generating
capacity of greater than 25 MW were omitted from the analysis because of their exemption
from the rule. This exempts 179 entities owned by state or local governments. Additionally,
government-owned entities for which IPM does not project generation in either 2010 or 2015
under the base case or CAIR were exempted from this analysis, because they are not
projected to be operating and thus will not face the costs of compliance with CAIR. Five
municipal entities were dropped from the analysis for this reason. Thus, EPA identified 81
state and municipality-owned utilities that are potentially affected by CAIR, out of a possible
265, which are summarized in Table 8-5.
Table 8-5. Summary of Potential Impacts on Government Entities under CAIR
ECU
Ownership
Type
Subdivision
State
Municipal
Total
Potentially
Affected
Entities
5
7
69
81
Projected
Annualized Costs
($1,000,000)
2010 2015
-$15.2 -$1.7
-$134.2 -$110.5
-$162.5 -$97.9
-$311.9 -$210.2
Number of
Government Entities
with Compliance
Costs >1% of
Generation Revenues
2010 2015
3 3
0 2
17 34
20 39
Number of
Government Entities
with Compliance
Costs >3% of
Generation Revenues
2010 2015
0 1
0 0
17 22
17 23
Note: The total number of potentially affected entities in this table excludes the 184 entities that have been
dropped because they will not be affected by CAIR. Also, the total number of entities with costs
greater than 1 percent or 3 percent of revenues includes only entities experiencing positive costs. A
negative cost value implies that the group of entities experiences a net savings under CAIR.
Source: IPM and TRUM analysis
8-13
-------
8.2.2 Overview of Analysis and Results
After identifying potentially affected government entities, EPA estimated the impact
of CAIR in 2010 and 2015 based on the following:
• total impacts of compliance on government entities and
• ratio of small entity impacts to revenues from electricity generation.
The financial burden to owners of EGUs under CAIR is composed of compliance and
administrative costs. This section outlines the compliance and administrative costs for the 81
potentially affected government-owned units in the CAIR region.
8.2.2.1 Methodology for Estimating Impacts of CAIR on Government Entities
The primary burden on state and municipal governments that operate utilities under
CAIR is the cost of installing control technology on units to meet SO2 and NOX emission
limits or the cost of purchasing allowances. However, an entity can comply with CAIR
through any combination of the following: installing retrofit technologies, purchasing
allowances, switching to a cleaner fuel, or reducing emissions through a reduction in
generation. Additionally, units with more allowances than needed can sell these allowances
on the market. The chosen compliance strategy will be primarily a function of the unit's
marginal control costs and its position relative to the marginal control costs of other units.
To attempt to account for each potential control strategy, EPA estimates compliance
costs as follows:
^Compliance ~ ^ ^Operating+Retrofit ^ ^Fuel ^ ^Allowances ^ ^Transaction ~ A K \°-2-)
where C represents a component of cost as labeled, and A R represents the retail value of
foregone electricity generation.
In reality, compliance choices and market conditions can combine such that an entity
may actually experience a savings in any of the individual components of cost. Under CAIR,
for example, EPA projects that the price of low-sulfur coal will fall as many units install
scrubbers and switch away from low-sulfur coal to cheaper bituminous coal, such that many
entities burning low-sulfur coal actually experience a reduction in fuel costs as a result of the
demand shift. Similarly, although some units will forgo some level of electricity generation
(and thus revenues) to comply, this impact will be lessened on these entities by the projected
increase in electricity prices under CAIR as well as reductions in fuel costs, while those not
reducing generation levels will see an increase in electricity revenues. Elsewhere, units
8-14
-------
burning high- or medium-sulfur coal might decide to pay relatively more for low-sulfur coal
under CAIR and sell allowances on the market, in the hopes of negating some or all of their
compliance cost. Because this analysis evaluates the total costs along each of the four
compliance strategies laid out above for each entity, it inevitably captures savings or gains
such as those described. As a result, what we describe as cost is really more of a measure of
the net economic impact of the rule on small entities.
In this analysis, EPA used IPM-parsed output for the base case and CAIR
(http://www.epa.gov/airmarkets/epa-ipm/iaqr.html) to estimate compliance cost at the unit
level. These costs were then summed for each small entity, adjusting for ownership share.
Compliance cost estimates were based on the following: operating and retrofit costs, sale or
purchase of allowances, and the change in fuel costs or electricity generation revenues under
CAIR relative to the base case. These components of compliance cost were estimated as
follows:
(1) Retrofit and operating costs: Using the IPM-parsed output for the base case
and CAIR, EPA identified units that install control technology under CAIR
and the technology installed. The equations for calculating retrofit costs for
SCR, SNCR, and FGD were adopted from EPA's TRUM. The model
calculates the capital cost (in $/MW), the fixed O&M cost (in $/MW-year),
the variable O&M cost (in $/MWh), and the total annualized retrofit and
operating cost by unit.
(2) Sale or purchase of allowances: EPA estimated the value of initial SO2 and
NOX allowance holdings. For SO2, units were assumed to retain their Phase II
allowance allocations as determined under EPA's 1998 reallocation of Acid
Rain allowances, adjusted to reflect the 50 percent reduction in 2010 and
65 percent reduction in 2015 under CAIR. The value of banked SO2
allowances was not considered in this analysis. Because the use of banked
allowances is expected to be a significant compliance strategy, this analysis
most likely overstates annualized compliance costs. For NOX, the state
emission budgets were assumed to be apportioned to units on a heat-input
basis. Each unit was assumed to receive a share of the state NOX emission
budget equal to its share of the total state heat input for that year in the base
case. This is a simplification of what is included in the model rule, which
proposes allocating NOX allowances based on heat input from 1999 through
8-15
-------
1992.5 However, states can ultimately decide how to allocate NOX
allowances.
To estimate the value of allowances holdings, allocated allowances were
subtracted from projected emissions, and the difference was then multiplied
by the allowance price projected by IPM. Units were assumed to purchase or
sell allowances to exactly cover their projected emissions under CAIR.
(3) Fuel costs: Fuel costs were estimated by multiplying fuel input (MMBtu) by
region and fuel type-adjusted fuel prices ($/MMBtu) from TRUM. The
change in fuel expenditures under CAIR was then estimated by taking the
difference in fuel costs between CAIR and the base case.
(4) Value of electricity generated: EPA estimated electricity generation by first
estimating the unit capacity factor and maximum fuel capacity. The unit
capacity factor is estimated by dividing fuel input (MMBtu) by maximum fuel
capacity (MMBtu). The maximum fuel capacity was estimated by
multiplying capacity (MW) * 8,760 operating hours * heat rate
(MMBtu/MWh). The value of electricity generated was then estimated by
multiplying capacity (MW)*capacity factor*8,760*regional-adjusted retail
electricity price ($/MWh).
(5) Administrative costs: Because most affected units are already monitored as
a result of other regulatory requirements, EPA considered the primary
administrative cost to be transaction costs related to purchasing or selling
allowances. EPA assumed that transaction costs were equal to 1.5 percent of
the total absolute value of a unit's allowances. This assumption is based on
market research by ICF Consulting.
8.2.2.2 Results
A summary of economic impacts on government-owned entities is presented in
Table 8-5. According to EPA's analysis, the total net economic impact on each category of
government-owned entity (state- and municipality-owned utilities and subdivisions) is
expected to be negative in both 2010 and 2015.6 IPM modeling of CAIR projects that
approximately 340 MW (8 units of 219 in this analysis) of municipality-owned capacity
5A similar approach was used in impact analyses for the 126 FIP and NOX SIP Call.
6 All costs are reported in 1999 dollars.
8-16
-------
would be uneconomic to maintain under CAIR, beyond what is projected in the base case.
This represents about 0.4 percent of all subdivision, state, and municipality capacity in the
CAIR region. For comparison, overall affected capacity under CAIR, about 5.3GW, or
1.7 percent of all coal-fired capacity is projected to be uneconomic to maintain relative to the
base case. This comparison suggests that government entities should not face a
disproportionate burden under CAIR. In practice, units projected to be uneconomic to
maintain may be "mothballed," retired, or kept in service to ensure transmission reliability in
certain parts of the grid. Our IPM modeling is unable to distinguish between these potential
outcomes.
As was done for the small entities analysis, EPA further assessed the economic and
financial impacts of the rule using the ratio of compliance costs to the value of revenues from
electricity generation in the base case, also focusing specifically on entities for which this
measure is greater than 1 percent.7 EPA projects that 20 government entities will have
compliance costs greater than 1 percent of revenues from electricity generation in 2010, and
39 will in 2015. Entities that are projected to experience negative compliance costs under
CAIR are not included in those totals. This approach is more indicative of a significant
impact when an analysis is looking at entities operating in a competitive market
environment. Government-owned entities do not operate in a competitive market
environment and therefore will be able to recover expenses under CAIR through rate
increases. Given this, EPA considers the 1 percent measure in this case a crude measure of
the extent to which rate increases will be made at publicly owned companies.
The distribution across entities of economic impacts as a share of base case revenue
is summarized in Table 8-6. For state-owned entities and subdivisions, the maximum
economic impact as a share of base case revenues is approximately 3 percent. A few
municipality-owned entities experience economic impacts that are significantly higher than
the capacity-weighted average for this group. In the cases where entities are projected to
experience positive net costs that are a high percentage of revenues, these entities do not find
it economic to retrofit and are unable to switch to a lower-sulfur coal. Thus, these entities
comply primarily through the purchase of allowances and reductions in generation.
7Neither the costs nor the revenues of units that retire under CAIR are included in this portion of the analysis.
Because these units are better off retiring under CAIR than continuing operation, the true cost of the rule on
these units is not represented by our modeling. The true cost of CAIR for these units is the differential
between their costs in the base case and the costs of meeting their customers' demand under the rule.
8-17
-------
Table 8-6. Distribution of Economic Impacts on Government Entities under CAIR
ECU Ownership Type
Sub-division
State
Municipal
All
Capacity-Weighted
Average Economic
Impacts as a % of
Generation Revenues
2010 2015
-3.6% -2.0%
-5.2% -3.9%
-5.9% -0.3%
-4.2% -2.3%
Min
2010
-80.0%
-11.4%
-13.8%
-80.0%
2015
-27.6%
-10.2%
-20.4%
-27.6%
Max
2010
1.9%
0.2%
17.2%
17.2%
2015
3.1%
2.8%
43.5%
43.5%
Source: IPM and TRUM analysis
Additionally, a few entities are projected to experience negative net costs that are a
high percentage of base case revenues. These entities have units that are able to switch to a
cheaper, lower-sulfur coal to comply with CAIR and are able to maintain or increase
generation levels, thus increasing revenues. Additionally, entities in regions for which we
project large electricity price increases relative to other regions tend to be among those at the
lower end of the distribution.
The various components of annualized incremental cost under CAIR to each group of
government entities are summarized in Table 8-7. Overall, with the exceptions of
subdivisions in 2010, each group is a net purchaser of allowances. Additionally, each group
experiences both a reduction in fuel expenditures and an increase in electricity revenue under
CAIR. Incremental fuel costs are negative because of the combination of a reduction in total
coal use, switching to bituminous coal, and reduced low-sulfur coal prices under CAIR.
Additionally, although total electricity generation by government entities falls slightly under
CAIR, the total loss in revenues is more than exceeded by the revenue gains projected as a
result of retail electricity prices rising under CAIR.
8.2.3 Summary of Government Entity Impacts
EPA examined the potential economic impacts on state and municipality-owned
entities associated with this rulemaking based on assumptions of how the affected states will
implement control measures to meet their emissions. Although EPA does not conclude that
the requirements of the UMRA apply to CAIR, these impacts have been calculated to
8-18
-------
Table 8-7. Incremental Annualized Costs under CAIR Summarized by Ownership
Group and Cost Category ($1,000,000)
ECU Ownership
Type
Subdivision
State
Municipal
Retrofit +
Operating Cost
2010 2015
10.4 9.4
20.1 25.9
21.3 26.7
Net Purchase of
Allowances
2010 2015
-1.6 1.7
29.2 52.5
39.3 94.8
Fuel Cost
2010 2015
-3.2 -8.4
-116.8 -143.1
-120.0 -156.8
Lost Electricity
Revenue
2010 2015
-20.9 -4.5
-67.0 -46.3
-103.7 -63.7
Administrative
Cost
2010 2015
0.0 0.1
0.3 0.5
0.7 1.0
Source: IPM and TRUM analysis
provide additional understanding of the nature of potential impacts and additional
information to the states as they revise SIPs to meet the emissions budgets set by this
rulemaking.
According to EPA's analysis, the total net economic impact on government-owned
entities is expected to be negative in both 2010 and 2015. However, IPM modeling projects
that about 340 MW of municipality-owned capacity (about 0.4 percent of all subdivision,
state, and municipality capacity in the CAIR region) would be uneconomic to maintain under
CAIR, beyond what is projected in the base case. In practice, units projected to be
uneconomic to maintain may be "mothballed," retired, or kept in service to ensure
transmission reliability in certain parts of the grid. Our IPM modeling is unable to
distinguish between these potential outcomes.
Of the 81 government entities considered in this analysis and the 265 government
entities in the CAIR region that are included in EPA's modeling, 20 may experience
compliance costs in excess of 1 percent of revenues in 2010, and 39 may in 2015, based on
our assumptions of how the affected states implement control measures to meet their
emissions budgets as set forth in this rulemaking.
Government entities projected to experience compliance costs in excess of 1 percent
of revenues have some potential for significant impact resulting from implementation of
CAIR. However, as noted above, it is EPA's position that because these government entities
can pass on their costs of compliance to rate-payers, they will not be significantly affected.
Furthermore, the decision to include only units greater than 25 MW in size exempts 179
government entities that would otherwise be potentially affected by CAIR.
8-19
-------
The above points aside, potential adverse impacts of CAIR on state- and
municipality-owned entities could be limited by the fact that the cap-and-trade program is
designed such that states determine how NOX allowances are to be allocated across units. A
state that wishes to mitigate the impact of the rule on state- or municipality-owned entities
might choose to allocate NOX allowances in a manner that is favorable to these entities.
Finally, in general, the use of cap-and-trade programs in general will limit impacts on entities
owned by small governments relative to a less flexible command-and-control program.
8.3 Paperwork Reduction Act
In compliance with the Paperwork Reduction Act (44 U.S.C. 3501 et seq.X EPA
submitted a proposed Information Collection Request (ICR) (EPA ICR number 2512.01) to
the Office of Management and Budget (OMB) for review and approval on July 19, 2004
(FR 42720-42722). The ICR describes the nature of the information collection and its
estimated burden and cost associated with the final rule. In cases where information is
already collected by a related program, the ICR takes into account only the additional
burden. This situation arises in states that are also subject to requirements of the
Consolidated Emissions Reporting Rule (EPA ICR number 0916.10; OMB control number
2060-0088) or for sources that are subject to the Acid Rain Program (EPA ICR number
1633.13; OMB control number 2060-0258) or NOX SIP Call (EPA ICR number 1857.03;
OMB number 2060-0445) requirements.
EPA solicited comments on specific aspects of the information collection. The
purpose of the ICR is to estimate the anticipated monitoring, reporting, and record-keeping
burden estimates and associated costs for states, local governments, and sources that are
expected to result from CAIR.
The record-keeping and reporting burden to sources resulting from states choosing to
participate in a regional cap-and-trade program is approximately $42 million annually. This
estimate includes the annualized cost of installing and operating appropriate SO2 and NOX
emissions monitoring equipment to measure and report the total emissions of these pollutants
from affected EGUs (serving generators greater than 25 megawatt electrical). The burden to
state and local air agencies includes any necessary SIP revisions, performance of monitoring
certification, and fulfilling of audit responsibilities. More information on the ICR analysis is
included in the official CAIR docket.
In accordance with the Paperwork Reduction Act on July 19, 2004, an ICR was made
available to the public for comment. The 60-day comment period expired September 19,
2004, with no public comments received specific to the ICR.
8-20
-------
8.4 Children's Health
E.O. 13045, "Protection of Children from Environmental Health Risks and Safety
Risks" (62 FR 19885, April 23, 1997), applies to any rule that (1) is determined to be
"economically significant" as defined under E.O. 12866 and (2) concerns an environmental
health or safety risk that EPA has reason to believe may have a disproportionate effect on
children. If the regulatory action meets both criteria, Section 5-501 of the Order directs the
Agency to evaluate the environmental health or safety effects of the planned rule on children
and explain why the planned regulation is preferable to other potentially effective and
reasonably feasible alternatives considered by the Agency.
This final rule is not subject to this E.O., because it does not involve decisions on
environmental health or safety risks that may disproportionately affect children. EPA
believes that the emissions reductions from the strategies in this rule will further improve air
quality and will further improve children's health. Chapter 4 of the RIA outlines benefits
such as reduced incidences of respiratory illness, acute bronchitis, and asthma attacks for
children that are anticipated to occur as a result of this rule.
8.5 Tribal Impacts
E.O. 13175, entitled "Consultation and Coordination with Indian Tribal
Governments" (65 FR 67249, November 9, 2000), requires EPA to develop an accountable
process to ensure "meaningful and timely input by Tribal officials in the development of
regulatory policies that have Tribal implications." This rule does not have "Tribal
implications" as specified in E.O. 13175.
This rule addresses transport of pollutants that are precursor for ozone and PM25.
The CAA provides for states and Tribes to develop plans to regulate emissions of air
pollutants within their jurisdictions. The regulations clarify the statutory obligations of states
and Tribes that develop plans to implement this rule. The Tribal Authority Rule (TAR) gives
Tribes the opportunity to develop and implement CAA programs, but it leaves to the
discretion of the Tribe whether to develop these programs and which programs, or
appropriate elements of a program, the Tribe will adopt.
This rule does not have Tribal implications as defined by E.O. 13175. It does not
have a substantial direct effect on one or more Indian Tribes, because no Tribe has
implemented an air quality management program at this time. Furthermore, this rule does
not affect the relationship or distribution of power and responsibilities between the federal
government and Indian Tribes. The CAA and the TAR establish the relationship of the
8-21
-------
federal government and Tribes in developing plans to attain the NAAQS, and this rule does
nothing to modify that relationship. Because this rule does not have Tribal implications,
E.O. 13175 does not apply.
If one assumes a Tribe is implementing a Tribal Implementation Plan, this final rule
would have implications for that Tribe, but it would neither impose substantial direct costs
on the Tribe nor preempt Tribal law. As provided above, EPA has estimated that the total
annual private costs for the rule for the CAIR region as implemented by state, local, and
Tribal governments is approximately $2.4 billion in 2010 and $3.6 billion in 2015 (1999$).
There are currently very few emissions sources in Indian country that could be affected by
this rule and the percentage of Tribal land that will be impacted is very small. For Tribes
that choose to regulate sources in Indian country, the costs would be attributed to inspecting
regulated facilities and enforcing adopted regulations.
Although E.O. 13175 does not apply to this rule, EPA consulted with Tribal officials
in developing this rule. EPA has encouraged Tribal input at an early stage. Also, EPA held
periodic meetings with the states and the Tribes during the technical development of this
rule. Three meetings were held with the Crow Tribe, where the Tribe expressed concerns
about potential impacts of the rule on the coal mine operations. In addition, EPA held three
calls with Tribal environmental professionals to address concerns specific to the Tribes.
These discussions have given EPA valuable information about Tribal concerns regarding the
development of this rule. EPA has provided briefings for Tribal representatives, and the
newly formed National Tribal Air Association (NTAA), and other national Tribal forums.
Input from Tribal representatives has been taken into consideration in developing this rule.
8.6 Environmental Justice
E.O. 12898, "Federal Actions to Address Environmental Justice in Minority
Populations and Low-Income Populations," requires federal agencies to consider the impact
of programs, policies, and activities on minority populations and low-income populations.
According to EPA guidance, agencies are to assess whether minority or low-income
populations face risks or a rate of exposure to hazards that are significant and that
"appreciably exceed or is likely to appreciably exceed the risk or rate to the general
population or to the appropriate comparison group" (EPA, 1998).
In accordance with E.O. 12898, the Agency has considered whether this rule may
have disproportionate negative impacts on minority or low income populations. The Agency
expects this rule to lead to reductions in air pollution and exposures generally. For this
8-22
-------
reason, negative impacts to these subpopulations that appreciably exceed similar impacts to
the general population are not expected.
8.7 Reference
U.S. Environmental Protection Agency (EPA). April 1998. Guidance for Incorporating
EnvironmentalJustice Concerns in EPA 's NEPA Compliance Analyses. Washington,
DC: Office of Federal Activities.
8-23
-------
SECTION 9
COMPARISON OF BENEFITS AND COSTS
The estimated social costs to implement the final CAIR program, as described in this
document, are approximately $1.91 or $2.14 billion annually for 2010 and $2.56 or $3.07
billion annually for 2015 (1999 dollars, 3 percent and 7 percent discount rate, respectively).
Thus, the net benefits (social benefits minus social costs) of the program in 2010 are
approximately $71.4 + B billion or $60.4 + B billion annually and in 2015 are $98.5 + B
billion or $83.2 + B billion annually (1999 dollars, based on a discount rate of 3 percent and
7 percent, respectively). (B represents the sum of all unquantified benefits and disbenefits of
the regulation.) Therefore, implementation of this rule is expected, based purely on
economic efficiency criteria, to provide society with a significant net gain in social welfare,
even given the limited set of health and environmental effects we were able to quantify.
Addition of ozone-, directly emitted PM2 5-, mercury-, acidification-, and eutrophication-
related impacts would likely increase the net benefits of the rule. Table 9-1 presents a
summary of the benefits, costs, and net benefits of the final rule.
The benefits and costs reported represent estimates for a complete CAIR program
that includes the CAIR promulgated rule and the proposal to include annual SO2 and NOX
controls for New Jersey and Delaware. Annual SO2 and NOX controls for Arkansas are
included in the modeling used to develop these estimates resulting in a minimal
overstatement of the reported benefits and costs for the complete CAIR program.
Air quality modeling was not conducted for the New Jersey and Delaware proposal.
For this reason, an analysis of the potential benefits for the New Jersey and Delaware
proposal could not be completed with any degree of specificity. However based on the air
quality modeling results for the CAIR, we make rough estimates of the benefits and net
benefits that might occur with this proposal. Including New Jersey and Delaware in the
CAIR program would result in additional reductions of SO2 and NOX emissions. We
estimate that approximately $630 million of the total annual CAIR program benefits
previously discussed may be attributable to annual SO2 and NOX controls for New Jersey and
Delaware in 2010. This estimate increases to approximately $1.1 billion annually in 2015.
The full CAIR analysis including New Jersey and Delaware showed a benefit-cost ratio of
9-1
-------
Table 9-1. Summary of Annual Benefits, Costs, and Net Benefits of the Clean Air
Interstate Rule3 (billions of 1999 dollars)
Description
Social costsb
3 percent discount rate
7 percent discount rate
Social benefits0'"'6
3 percent discount rate
7 percent discount rate
Health-related benefits:
3 percent discount rate
7 percent discount rate
Visibility benefits
Net benefits (benefits-costs)6 f
3 percent discount rate
7 percent discount rate
2010
$1.91
$2.14
73.3 +B
62.6 + B
72.1
61.4
1.14
$71.4 +B
$60.4 + B
2015
$2.56
$3.07
101+B
86.3 +B
99.3
84.5
1.78
$98.5 +B
$83.2 + B
a All estimates are rounded to three significant digits for ease of presentation and computation. Estimates
represent a complete CAIR program that includes the CAIR promulgated rule and the proposal to include
annual SO2 and NOX controls for New Jersey and Delaware. Annual SO2 and NOX controls for Arkansas are
included in the modeling used to develop these estimates resulting in a minimal overstatement of the benefits
and costs for the complete CAIR program.
b Note that costs are the annualized total costs of reducing pollutants including NOX and SO2 for the EGU
source category in the CAIR region in the years 2010 and 2015.
0 As this table indicates, total benefits are driven primarily by PM-related health benefits. The reduction in
premature fatalities each year accounts for over 90 percent of total monetized benefits. Benefits in this table
are nationwide (with the exception of ozone and visibility) and are associated with NOX and SO2 reductions.
Ozone benefits represent benefits in the eastern United States. Visibility benefits represent benefits in Class I
areas in the southeastern United States.
d Not all possible benefits or disbenefits are quantified and monetized in this analysis. B is the sum of all
unquantified benefits and disbenefits. Potential benefit categories that have not been quantified and
monetized are listed in Table 1-4.
e Valuation assumes discounting over the SAB-recommended 20-year segmented lag structure described in
Chapter 4. Results reflect the use of 3 percent and 7 percent discount rates consistent with EPA and OMB
guidelines for preparing economic analyses (EPA, 2000; OMB, 2003).
f Net benefits are rounded to the nearest $100 million. Columnar totals may not sum due to rounding.
9-2
-------
around 39:1 in 2015. Based on the relatively low estimated private costs of including New
Jersey and Delaware of $30 million in 2010 and $40 million in 2015, it is highly unlikely
that costs of including New Jersey and Delaware would exceed benefits even if benefits of
controlling SO2 and NOX for New Jersey and Delaware were substantially lower than the
average benefit we used to estimate the benefits. It is highly unlikely that benefits are much
lower than average given the urban nature of much of New Jersey, and the proximity of New
Jersey and Delaware to many heavily populated urban areas.
As with any complex analysis of this scope, there are several uncertainties inherent in
the final estimate of benefits and costs, that are described fully in Chapters 4 and 7. In
addition to the uncertainty characterization provided in these chapters, we also present two
types of probabilistic approaches to characterize uncertainty in the benefit estimate of the
CAIR program. The first approach generates a distribution of benefits based on the classical
statistical error expressed in the underlying health and economic valuation studies used in the
benefits modeling framework. The second approach uses the results from a pilot expert
elicitation project designed to characterize key aspects of uncertainty in the ambient
PM2 5/mortality relationship, and augments the uncertainties in the mortality estimate with
the statistical error reported for other endpoints in the benefit analysis.
9.1 References
Pope, C.A., III, R.T. Burnett, M.J. Thun, E.E. Calle, D. Krewski, K. Ito, and G.D. Thurston.
2002. "Lung Cancer, Cardiopulmonary Mortality, and Long-term Exposure to Fine
Particulate Air Pollution." Journal of the American Medical Association
287:1132-1141.
U.S. Environmental Protection Agency (EPA). September 2000. Guidelines for Preparing
Economic Analyses. EPA 240-R-00-003.
U.S. Office of Management and Budget (OMB). 2003. Circular A-4 Guidance to Federal
Agencies on Preparation of Regulatory Analysis.
Woodruff, T.J., J. Grillo, and K.C. Schoendorf. 1997. "The Relationship Between Selected
Causes of Postneonatal Infant Mortality and Particulate Infant Mortality and
Particulate Air Pollution in the United States." Environmental Health Perspectives
105(6):608-612.
9-3
-------
APPENDIX A
BENEFITS AND COSTS OF THE CLEAN AIR INTERSTATE RULE, THE CLEAN
AIR VISIBILITY RULE, AND THE CLEAN AIR INTERSTATE RULE PLUS THE
CLEAN AIR VISIBILITY RULE
This appendix presents the benefits and costs for the CAIR program (CAIR final rule
plus the New Jersey and Delaware proposal),1 the EGU requirements for Best Available
Retrofit Technology (BART) Guidelines for the Regional Haze Rule, and the CAIR program
in the CAIR region plus BART control for EGUs elsewhere in the country (CAIR Plus
BART in the Non-CAIR Region). It is important to note that the CAIR, CAIR Plus BART in
the Non-CAIR Region, and BART Nationwide benefit and costs estimates reflect controls
for the EGU source category only, while the BART regulation will potentially affect 26
source categories. The analysis presented in this Appendix was conducted to show the
benefits and costs of the alternative programs addressing the EGU sector.2
As Table A-l shows, annual net benefits for the CAIR program are $97.4 billion in
2015. This estimate compares to annual net benefits of $44.3 billion for BART nationwide
program and $100 billion for CAIR Plus BART in the Non-CAIR Region (assuming a 3
percent discount rate). These estimates become $82.7 billion for CAIR, $84.9 billion for
CAIR Plus BART in the Non-CAIR Region, and $37.0 billion for the BART nationwide
program assuming a 7 percent discount rate. The analysis shows that if one assumes the
CAIR program exists, the incremental benefits of requiring BART controls for the EGU
source category only in areas outside the CAIR region are approximately $3.2 to $4 billion
(7 percent and 3 percent discount rate, respectively). Related incremental costs are
approximately $1 billion. All estimates are shown in 1999 dollars. Table A-2 lists the
1 The modeling for the rule includes annual SO2 and NOX controls for Arkansas and results in a minimal
overstatement of the benefits and costs of the CAIR program (CAIR final plus the New Jersey and Delaware
proposal).
2Note that the net benefits reported in this appendix are estimated using the private costs of the respective rules
rather than social costs. Thus the net benefits shown for the CAIR program in this appendix differ
somewhat from the estimates presented in the body of this report.
A-l
-------
Table A-l. Summary of Annual Benefits, Costs, and Net Benefits of the Clean Air
Interstate Rule, 2015 (billions of 1999 dollars)3
CAIR Plus BART
in the Non-CAIR
Description CAIR Program8 Region8 BART Nationwide
Private costsb $3.57 $4.55 $5.19
Social benefitscAe
3 percent discount rate $101 + B $105 + B $49.5+ B
7 percent discount rate $86.3 + B $89.5 + B $42.2 + B
Health-related benefits:
3 percent discount rate 99.3 103 48.8
7 percent discount rate 84.5 87.6 41.5
Visibility benefits 1.78 1.89 0.699
Net benefits (benefits-costs)e>f
3 percent discount rate $97.4 + B $100 + B $44.3 + B
7 percent discount rate $82.7+ B $84.9+B $37.0+B
a All estimates are rounded to three significant digits for ease of presentation and computation. These annual
estimates represent the benefits and costs of these regulatory programs expected to occur in 2015. BART
estimates reflect benefits and costs for controls for the EGU source category only.
b Note that costs presented are the annual total private costs to the power sector of reducing pollutants
including NOX and SO2. The costs are estimated using the IPM and assume affected firms face cost of
capital ranging from 5.34 to 6.74 percent. CAIR costs reflect costs for the CAIR region. Costs for CAIR
Plus BART in the Non-CAIR Region and BART nationwide are national cost estimates.
0 Total benefits are driven primarily by PM-related health benefits. The reduction in premature fatalities each
year accounts for over 90 percent of total monetized benefits in 2015. Benefits in this table are nationwide
(with the exception of ozone and visibility) and are associated with NOX and SO2 reductions. Ozone benefits
relate to the eastern United States. Visibility benefits relate to Class I areas in the southeastern United
States. While ozone benefits are expected for each of these programs, ozone benefits are included in the
CAIR program benefits estimates only. The benefit estimates for CAIR Plus in the Non-CAIR Region
BART and BART nationwide do not include ozone benefits. Inclusion of ozone benefits would not likely
alter the conclusions reached on the magnitude of the difference between the options.
d Not all possible benefits or disbenefits are quantified and monetized in this analysis. B is the sum of all
unqualified benefits and disbenefits. Potential effects categories that have not been quantified and
monetized are listed in Table 1-4 and Table 1-5.
e Valuation assumes discounting over the SAB-recommended 20-year segmented lag structure described in
Chapter 4. Results reflect 3 percent and 7 percent discount rates consistent with EPA and OMB guidelines
for preparing economic analyses (EPA, 2000; OMB, 2003).
f Net benefits are rounded to the nearest $100 million. Columnar totals may not sum due to rounding.
g CAIR costs and benefits are the estimates for the CAIR program that includes the promulgated CAIR and the
proposal to include annual SO2 and NOX controls for New Jersey and Delaware. Modeling for CAIR
assumes annual SO2 and NOX controls for Arkansas that is not a part of the CAIR program. Thus the
benefits and costs reported are slightly overstated.
A-2
-------
Table A-2. Clean Air Interstate Rule: Estimated Reduction in Health Effects
(Incidence)—2015
Health Effect
CAIR
Program*
CAIR Plus
BART in the
Non-CAIR
Region*
BART
Nation-
wide
PM-Related Endpoints:
Premature mortality15
Adult, age 30 and over
Infant, age <1 year
Chronic bronchitis (adult, age 26 and over)
Nonfatal myocardial infarction (adults, age 18 and older)
Hospital admissions — respiratory (all ages)0
Hospital admissions — cardiovascular (adults, age >18)d
Emergency room visits for asthma (age 18 years and
younger)
Acute bronchitis (children, aged 8-12)
Lower respiratory symptoms (children, aged 7-14)
Upper respiratory symptoms (asthmatic children, aged 9-18)
Asthma exacerbation (asthmatic children, aged 6-18)
Work loss days (adults, aged 18-65)
Minor restricted-activity days (adults, aged 18-65)
17,000
36
8,700
22,000
5,500
5,000
13,000
19,000
230,000
180,000
290,000
1,700,000
9,900,000
17,000
38
9,100
23,000
5,700
5,200
13,000
20,000
240,000
190,000
310,000
1,700,000
10,300,000
8,200
19
4,400
11,000
2,700
2,500
6,500
10,000
120,000
92,000
150,000
830,000
5,000,000
Ozone-Related Endpoints "
Hospital admissions — respiratory causes (adult, 65 and older)
Hospital admissions — respiratory causes (children, under 2)
Emergency room visit for asthma (all ages)
Minor restricted-activity days (adults, aged 18-65)
School absence days
1,700
1,100
280
690,000
510,000
NE
NE
NE
NE
NE
NE
NE
NE
NE
NE
a Incidences are rounded to two significant digits. BART estimates reflect incidences for controls for the EGU
source category only.
b Premature mortality benefits associated with ozone are not quantified in the primary analysis. Adult
premature mortality estimates are based upon studies by Pope et al., 2002. Infant premature mortality
estimates are based upon studies by Woodruff, Grille, and Schoendorf, 1997.
0 Respiratory hospital admissions for PM include admissions for COPD, pneumonia, and asthma.
d Cardiovascular hospital admissions for PM include total cardiovascular and subcategories for ischemic heart
disease, dysrhythmias, and heart failure.
e Although ozone benefits are expected to occur for CAIR Plus BART in the Non-CAIR Region and BART
nationwide, ozone benefits are estimated for the CAIR program only.
f These health effects incidences reflect estimates for the CAIR program (the promulgated CAIR and the
proposal to include annual SO2 and NOX controls for New Jersey and Delaware in CAIR). Modeling for
CAIR assumes annual SO2 and NOX controls for Arkansas that is not a part of the CAIR program. Thus the
incidence estimates reported for CAIR are slightly overstated.
NE = Not estimated
A-3
-------
reduction in health incidence resulting from the CAIR program, CAIR Plus BART in the
Non-CAIR Region, and BART nationwide. Table A-3 depicts the monetary value of the
benefit categories listed on Table A-2. We were unable to estimate all of the benefits and
disbenefits associated with these rulemakings as summarized in Table 1-4. These
unquantified effects are represented by the letter B.
Table A-3. Estimated Monetary Value in Reductions in Incidence of Health and
Welfare Effects (in millions of 1999$)—2015a bc
Health Effect
Premature mortality"1
Adult >30 years
3% discount rate
7% discount rate
Child <1 year
Chronic bronchitis (adults, 26 and over)
Nonfatal acute myocardial infarctions
3% discount rate
7% discount rate
Hospital admissions for respiratory causes
Hospital admissions for cardiovascular causes
Emergency room visits for asthma
Acute bronchitis (children, aged 8-12)
Lower respiratory symptoms (children, 7-14)
Upper respiratory symptoms (asthma, 9-11)
Asthma exacerbations
Work loss days
Minor restricted-activity days (MRADs)
School absence days
Worker productivity (outdoor workers, 18-65)
Recreational visibility, 81 Class I areas
Monetized Total e
Base estimate:
3% discount rate
7% discount rate
Pollutant
PM25
PM25
PM25
PM2 5, 03
PM25
PM2 5, 03
PM25
PM25
PM25
PM25
PM25
PM25,03
03
03
PM25
CAIR
Program*
$92,800
78,100
222
3,340
1,850
1,790
78.9
105
3.56
7.06
3.74
4.77
12.7
219
543
36.4
19.9
1,780
$101+B
S86.3+B
CAIR Plus
BART in the
Non-CAIR
Region*
$96,300
81,100
232
3,510
1,920
1,860
43.6
82.6
3.62
7.41
3.87
4.69
13.4
209
528
NE
NE
1,890
$105+B
S89.5+B
BART
Nation-
wide
$45,700
38,400
116
1,690
905
876
20.6
39.3
1.79
3.68
1.91
2.30
6.56
101
256
NE
NE
699
$49.5+
B
$42.2+
B
Monetary benefits are rounded to three significant digits for ease of presentation and computation. Benefit
estimates relate to emissions reductions for the EGU source category only. Estimates represent nationwide
benefits (with the exception of ozone and visibility) and are associated with NOX and SO2 reductions. Ozone
benefits represent benefits for the eastern United States. Visibility estimates relate to Class I areas in the
southeastern United States. BART estimates reflect benefits for controls for the EGU source category only.
(continued)
A-4
-------
Table A-3. Estimated Monetary Value in Reductions in Incidence of Health and
Welfare Effects (in millions of 1999$)—2015a bc (continued)
b Monetary benefits adjusted to account for growth in real GDP per capita between 1990 and 2015.
0 Ozone benefits are estimated for the final CAIR. While ozone benefits are anticipated for CAIR plus BART
in the Non-CAIR Region and BART nationwide, these ozone benefits were not estimated.
d Valuation assumes discounting over the SAB-recommended 20-year segmented lag structure described
earlier. Results show 3 percent and 7 percent discount rates consistent with EPA and OMB guidelines for
preparing economic analyses (EPA, 2000; OMB, 2003). Adult premature mortality estimates are based upon
studies by Pope et al., 2002. Infant premature mortality estimates are based upon studies by Woodruff,
Grillo, and Schoendorf, 1997.
e B represents the monetary value of health and welfare benefits not monetized. A detailed listing of
unqualified benefits is provided in Table 1-4. Columnar totals may not add due to rounding.
f These benefits reflect estimates for the CAIR program (the promulgated CAIR and the proposal to include
annual SO2 and NOX controls for New Jersey and Delaware in CAIR). Modeling for CAIR assumes annual
SO2 and NOX controls for Arkansas that is not a part of the CAIR program. Thus the benefit estimates
reported for CAIR are slightly overstated.
NE = Not estimated
A.I References
Pope, C.A., III, R.T. Burnett, M.J. Thun, E.E. Calle, D. Krewski, K. Ito, and G.D. Thurston.
2002. "Lung Cancer, Cardiopulmonary Mortality, and Long-term Exposure to Fine
Particulate Air Pollution." Journal of the American Medical Association
287:1132-1141.
U.S. Environmental Protection Agency (EPA). September 2000. Guidelines for Preparing
Economic Analyses. EPA 240-R-00-003.
U.S. Office of Management and Budget (OMB). 2003. Circular A-4 Guidance to Federal
Agencies on Preparation of Regulatory Analysis.
Woodruff, T.J., J. Grillo, and K.C. Schoendorf. 1997. "The Relationship Between Selected
Causes of Postneonatal Infant Mortality and Particulate Infant Mortality and
Particulate Air Pollution in the United States." Environmental Health Perspectives
105(6):608-612.
A-5
-------
APPENDIX B
SUPPLEMENTAL ANALYSES ADDRESSING UNCERTAINTIES IN THE
BENEFITS ANALYSES
B.I Introduction
The recent NAS report on estimating public health benefits of air pollution
regulations recommended that EPA begin to move the assessment of uncertainties from its
ancillary analyses into its primary analyses by conducting probabilistic, multiple-source
uncertainty analyses. In this appendix, we describe our progress toward improving our
approach of characterizing the uncertainties in our economic benefits estimates, with
particular emphasis on the C-R function relating premature mortality to exposures to ambient
PM2 5. We present two approaches to generating probabilistic distributions designed to
illustrate the potential influence of some aspects of the uncertainty in the C-R function in a
PM benefits analysis. The first approach generates a probabilistic estimate of statistical
uncertainty based on standard errors reported in the underlying studies used in the benefit
modeling framework. The second approach uses the results from a pilot expert elicitation
designed to characterize certain aspects of uncertainty in the ambient PM2 5/mortality
relationship.
In recent benefit analyses of air pollution regulations, estimation of the reduction in
premature mortality from the control of particles accounts for 85 to 95 percent of total
benefits. Therefore, it is an endpoint that will be an important focus for characterizing the
uncertainty related to the estimates of total benefits. As part of a collaboration between
EPA's Office of Air and Radiation (OAR) and the Office of Management and Budget
(OMB) on the Nonroad Diesel Rule, we conducted a pilot expert elicitation intended to more
fully characterize uncertainty in the estimate of mortality resulting from exposure to PM.
It should be recognized that in addition to uncertainty, the annual benefits estimates
for the final CAIR also are inherently variable, due to the truly random processes that govern
pollutant emissions and ambient air quality in a given year. Factors such as hourly rate of
emissions and daily weather display constant variability regardless of our ability to
accurately measure them.
B-l
-------
B.1.1 General Approach
In addition to the two approaches to characterize uncertainty for PM mortality, we
also wanted to incorporate information on uncertainties of other endpoints in the benefits
model. For this rule we did not attempt to assign probabilities to all of the uncertain
parameters in the model because of a lack of resources and reliable methods. At this time,
we simply generate estimates of the distributions of dollar benefits for PM health effects and
for total dollar benefits including visibility. For non-mortality endpoints, we provide a
likelihood distribution for the total benefits estimate, based solely on the statistical
uncertainty surrounding the estimated C-R functions and the assumed distributions around
the unit values. Visibility benefits are also included in the estimate of total benefits, but
because of data limitations, are characterized as a constant value rather than a distribution.
This effectively shifts the distribution of total benefits upwards at all percentiles by the same
amount.
Our estimate of the likelihood distribution for total benefits should be viewed as
incomplete because of the wide range of sources of uncertainty that we have not
incorporated. The 5th and 95th percentile points of our estimate are based on statistical error,
and cross-study variability provides some insight into how uncertain our estimate is with
regard to those sources of uncertainty. However, it does not capture other sources of
uncertainty regarding other inputs to the model, including emissions, air quality, baseline
population incidence, projected exposures, or the model itself, including aspects of the health
science not captured in the studies, such as the likelihood that PM is causally related to
premature mortality and other serious health effects and the likelihood that ozone has an
independent effect on mortality. Thus, a likelihood description based on the standard error
would provide a misleading picture about the overall uncertainty in the estimates.
Both the uncertainty about the incidence changes1 and uncertainty about unit dollar
values can be characterized by distributions. Each "likelihood distribution" characterizes our
beliefs about what the true value of an unknown variable (e.g., the true change in incidence
of a given health effect in relation to PM exposure) is likely to be, based on the available
1 Because this is a national analysis in which, for each endpoint, a single C-R function is applied everywhere,
there are two sources of uncertainty about incidence: statistical uncertainty (due to sampling error) about the
true value of the pollutant coefficient in the location where the C-R function was estimated and uncertainty
about how well any given pollutant coefficient approximates (3*.
B-2
-------
information from relevant studies.2 Unlike a sampling distribution (which describes the
possible values that an estimator of an unknown variable might take on), this likelihood
distribution describes our beliefs about what values the unknown variable itself might be.
Such likelihood distributions can be constructed for each underlying unknown variable (such
as a particular pollutant coefficient for a particular location) or for a function of several
underlying unknown variables (such as the total dollar benefit of a regulation). In either
case, a likelihood distribution is a characterization of our beliefs about what the unknown
variable (or the function of unknown variables) is likely to be, based on all the available
relevant information. A likelihood description based on such distributions is typically
expressed as the interval from the 5th percentile point of the likelihood distribution to the 95th
percentile point. If all uncertainty had been included, this range would be the "credible
range" within which we believe the true value is likely to lie with 90 percent probability.
B.2 Monte-Carlo Based Uncertainty Analysis
The uncertainty about the total dollar benefit associated with any single endpoint
combines the uncertainties from these two sources (the C-R relationship and the valuation)
and is estimated with a Monte Carlo method. In each iteration of the Monte Carlo procedure,
a value is randomly drawn from the incidence distribution, another value is randomly drawn
from the unit dollar value distribution; the total dollar benefit for that iteration is the product
of the two.3 When this is repeated for many (e.g., thousands of) iterations, the distribution of
total dollar benefits associated with the endpoint is generated.
Using this Monte Carlo procedure, a distribution of dollar benefits can be generated
for each endpoint. As the number of Monte Carlo draws gets larger and larger, the Monte
Carlo-generated distribution becomes a better and better approximation of a joint likelihood
distribution (for the considered parameters) making up the total monetary benefits for the
endpoint.
After endpoint-specific distributions are generated, the same Monte Carlo procedure
can then be used to combine the dollar benefits from different (nonoverlapping) endpoints to
generate a distribution of total dollar benefits.
2 Although such a "likelihood distribution" is not formally a Bayesian posterior distribution, it is very similar in
concept and function (see, for example, the discussion of the Bayesian approach in Kennedy (1990),
pp. 168-172).
3 This method assumes that the incidence change and the unit dollar value for an endpoint are stochastically
independent.
B-3
-------
The estimate of total benefits may be thought of as the end result of a sequential
process in which, at each step, the estimate of benefits from an additional source is added.
Each time an estimate of dollar benefits from a new source (e.g., a new health endpoint) is
added to the previous estimate of total dollar benefits, the estimated total dollar benefits
increases. However, our bounding or likelihood description of where the true total value lies
also increases as we add more sources.
As an example, consider the benefits from reductions in PM-related hospital
admissions for cardiovascular disease. Because the actual dollar value is unknown, it may be
described using a variable, with a distribution describing the possible values it might have. If
this variable is denoted as Xl3 then the mean of the distribution, E(Xj) and the variance of Xl3
denoted Var(Xj), and the 5th and 95th percentile points of the distribution (related to Va^Xj)),
are ways to describe the likelihood for the true but unknown value for the benefits reduction.
Now suppose the benefits from reductions in PM-related hospital admissions for
respiratory diseases are added. Like the benefits from reductions in PM-related hospital
admissions for cardiovascular disease, the likelihood distribution for where we expect the
true value to be may be considered a variable, with a distribution. Denoting this variable as
X2, the benefits from reductions in the incidence of both types of hospital admissions is Xt +
X2. This variable has a distribution with mean E(Xl + X2) = E(Xj) + E(X2), and a variance of
Var(Xj + X2) = Var(Xj) + Var(X2) + 2Cov(X1,X2); if X1 and X2 are stochastically
independent, then it has a variance of Var(Xx + X2) = Var(Xj) + Var(X2), and the covariance
term is zero.
The benefits from reductions in all nonoverlapping PM-related health and welfare
endpoints are (Xm+1, ..., XJ is X = Xl + ... + X,,. The mean of the distribution of total
benefits, X, is
E(X) = E(Xl) + E(X2} + ...+E(Xn) (B.I)
and the variance of the distribution of total benefits — assuming that the components are
stochastically independent of each other (i.e., no covariance between variables) — is
Var(X) = Var(X,} + Var(X2} + ... + Var(Xn) (B .2)
If all the means are positive, then each additional source of benefits increases the
point estimate (mean) of total benefits. However, with the addition of each new source of
benefits, the variance of the estimate of total benefits also increases. That is,
E(Xl)
-------
Var(X1) < Var(X, + X2) < Var(X, + X2+X3)<...< Var(X, + ...+Xa)= Var(X). (B .4)
That is, the addition of each new source of benefits results in a larger mean estimate of total
benefits (as more and more sources of benefits are included in the total) about which there is
less certainty. This phenomenon occurs whenever estimates of benefits are added.
Calculated with a Monte Carlo procedure, the distribution of X is composed of
random draws from the components of X. In the first draw, a value is drawn from each of
the distributions, Xl3 X2, through X,,; these values are summed; and the procedure is repeated
again, with the number of repetitions set at a high enough value (e.g., 5,000) to reasonably
trace out the distribution of X. The 5th percentile point of the distribution of X will be
composed of points pulled from all points along the distributions of the individual
components and not simply from the 5th percentile. Although the sum of the 5th percentiles
of the components would be represented in the distribution of X generated by the Monte
Carlo, it is likely that this value would occur at a significantly lower percentile. For a similar
reason, the 95th percentile of X will be less than the sum of the 95th percentiles of the
components, and instead the 95th percentile of X will be composed of component values that
are significantly lower than the 95th percentiles.
The physical effects estimated in this analysis are assumed to occur independently. It
is possible that, for any given pollution level, there is some correlation between the
occurrence of physical effects, due to say avoidance behavior or common causal pathways
and treatments (e.g., stroke, some kidney disease, and heart attack are related to treatable
blood pressure). Estimating accurately any such correlation, however, is beyond the scope of
this analysis, and instead it is simply assumed that the physical effects occur independently.
We conducted two different Monte Carlo analyses, one based on the distribution of
reductions in premature mortality characterized by the mean effect estimate and standard
error from the epidemiology study of PM-associated mortality associated with the primary
estimate in Chapter 4 (Pope et al., 2002), and one based on the results from a pilot expert
elicitation project (lEc, 2004). In both analyses, the distributions of all other health
endpoints are characterized by the reported mean and standard deviations from the
epidemiology literature. Distributions for unit dollar values are based on reported ranges or
distributions of values in the economics literature and are summarized in Table B-l. We are
unable at this time to characterize the uncertainty in the estimate of benefits of improvements
in visibility at Class I areas. As such, we treat the visibility benefits as fixed and add them to
B-5
-------
Table B-l. Distributions for Unit Values of Health Endpoints
Health Endpoint
Mean Value,
Adjusted for
Income Growth to
2030
Derivation of Distribution
Premature Mortality
(Value of a Statistical
Life)
$5,500,000
Normal distribution with mean of $5.5 million and standard
deviation of $2.3 million, anchored at 2.5th and 97.5th percentiles of
$1 and $10 million, respectively. Confidence interval is based on
two meta-analyses of the wage-risk VSL literature. $1 million
represents the lower end of the interquartile range from the Mrozek
and Taylor (2000) meta-analysis. $10 million represents the upper
end of the interquartile range from the Viscusi and Aldy (2003)
meta-analysis. The VSL represents the value of a small change in
mortality risk aggregated over the affected population. Normal
distribution chosen through best professional judgment.
Chronic Bronchitis (CB)
$430,000
The WTP to avoid a case of pollution-related CB is calculated as
WTP; = WTP13 * e-f "<"-''', where x is the severity of an average CB
case, WTP13 is the WTP for a severe case of CB, and p is the
parameter relating WTP to severity, based on the regression results
reported in Krupnick and Cropper (1992). The distribution of
WTP for an average severity-level case of 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, with the most
likely value at severity level 6.5 and endpoints at 1.0 and 12.0; 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]). This process and the
rationale for choosing it is described in detail in the Costs and
Benefits of the Clean Air Act, 1990 to 2010 (EPA, 1999).
(continued)
-------
Table B-l. Distributions for Unit Values of Health Endpoints (continued)
Health Endnoint
Nonfatal Myocardial
Infarction (heart attack)
3% discount rate
Age 0-24
Age 25^4
Age 45-54
Age 55-65
Age 66 and over
7% discount rate
Age 0-24
Age 25^4
Age 45-54
Age 55-65
Age 66 and over
Mean Value,
Adjusted for
Income Growth to
2030
$66,902
$74,676
$78,834
$140,649
$66,902
$65,293
$73,149
$76,871
$132,214
$65,293
Derivation of Distribution
No distribution available. Age-specific COI values reflecting lost
earnings and direct medical costs over a 5-year period following
nonfatal MI. Lost earnings estimates based on Cropper and
a
Krupnick (1990). Direct medical costs based on simple average of
estimates from Russell et al. (1998) and Wittels et al. (1990).
Lost earnings:
Cropper and Krupnick (1 990). Present discounted value of 5 years
of lost earnings:
age of onset at 3% at 7%
25^4 $8,774 $7,855
45-54 $12,932 $11,578
55-65 $74,746 $66,920
Direct medical expenses: An average of:
1. Wittels et al. (1990) ($102,658— no discounting)
2. Russell et al. (1998), 5-year period. ($22,331 at 3% discount
rate; $21,1 13 at 7% discount rate)
Hospital Admissions
All Respiratory
(ICD codes 480-487,
490-492, 494-496)
$12,378
No distributions available. The COI point estimates (lost earnings
plus direct medical costs) are based on ICD-9 code-level
information (e.g., average hospital care costs, average length of
hospital stay, and weighted share of total COPD category illnesses)
reported in Agency for Healthcare Research and Quality (2000)
(www.ahrq.gov).
All Cardiovascular
(ICD codes 390-429)
$18,387
No distribution available. The COI point estimates (lost earnings
plus direct medical costs) are based on ICD-9 code-level
information (e.g., average hospital care costs, average length of
hospital stay, and weighted share of total cardiovascular category
illnesses) reported in Agency for Healthcare Research and Quality
(2000) (www.ahrq.gov).
(continued)
B-7
-------
Table B-l. Distributions for Unit Values of Health Endpoints (continued)
Health Endpoint
Emergency Room Visits
for Asthma
Mean Value,
Adjusted for
Income Growth to
2030
$286
Derivation of Distribution
No distribution available. The COI point estimate is the simple
average of two unit COI values:
(1) $31 1.55, from Smith et al. (1997) and
(2) $260.67, from Stanford et al. (1999).
Respiratory Ailments Not Requiring Hospitalization
Upper Respiratory
Symptoms (URS)
$27
Combinations of the three symptoms for which WTP estimates are
available that closely match those listed by Pope et al. result in
seven different "symptom clusters," each describing a "type" of
URS. A dollar value was derived for each type of URS, using
mid-range estimates of WTP (lEc, 1994) to avoid each symptom in
the cluster and assuming additivity of WTPs. In the absence of
information surrounding the frequency with which each of the
seven types of URS occurs within the URS symptom complex, we
assumed a uniform distribution between $10 and $45.
Lower Respiratory
Symptoms (LRS)
$17
Combinations of the four 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 dollar value was derived for each type of LRS, using mid-range
estimates of WTP (lEc, 1994) to avoid each symptom in the cluster
and assuming additivity of WTPs. The dollar value for LRS is the
average of the dollar values for the 11 different types of LRS. In
the absence of information surrounding the frequency with which
each of the 11 types of LRS occurs within the LRS symptom
complex, we assumed a uniform distribution between $8 and $25.
Asthma Exacerbations
$45
Asthma exacerbations are valued at $45 per incidence, based on
the mean of average WTP estimates for the four severity
definitions of a "bad asthma day," described in Rowe and Chestnut
(1986). This study surveyed asthmatics to estimate WTP for
avoidance of a "bad asthma day," as defined by the subjects. For
purposes of valuation, an asthma exacerbation is assumed to be
equivalent to a day in which asthma is moderate or worse as
reported in the Rowe and Chestnut (1986) study. The value is
assumed have a uniform distribution between $17 and $73.
(continued)
-------
Table B-l. Distributions for Unit Values of Health Endpoints (continued)
Health Endpoint
Mean Value,
Adjusted for
Income Growth to
2030
Derivation of Distribution
Acute Bronchitis
$390
Assumes a 6-day episode, with the distribution of the daily value
specified as uniform with the low and high values based on those
recommended for related respiratory symptoms in Neumann et al.
(1994). The low daily estimate of $ 10 is the sum of the mid-range
values recommended by lEc (1994) for two symptoms believed to
be associated with acute bronchitis: coughing and chest tightness.
The high daily estimate was taken to be twice the value of a minor
respiratory restricted-activity day, or $110.
Restricted Activity, Work Loss, and School Absence Days
School Absences
Work Loss Days
(WLDs)
Minor Restricted-
Activity Days (MRADs)
$75
Variable
$55
No distribution available. Point estimate is based on (1 ) the
probability that if a school child stays home from school, a
working mother will have to stay home from work to care for the
child (0.73); and (2) lost productivity at the female parent's wage
(average of $103).
No distribution available. Point estimate is based on county-
specific median annual wages divided by 50 (assuming 2 weeks of
vacation) and then by 5 — to get median daily wage. U.S. Year
2000 Census, compiled by Geolytics, Inc.
Median WTP estimate to avoid one MRAD from Tolley et al.
(1986). Distribution is assumed to be triangular with a minimum
of $22 and a maximum of $83, with a most likely value of $55.
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.
all percentiles of the PM health benefits distribution. Results of the Monte Carlo analysis
based on the Pope et al. (2002) distribution are presented in the next section. Results of the
Monte Carlo analysis based on the pilot expert elicitation are presented in Section B.3.
B. 2.1 Monte Carlo Analysis Using Classical Statistical Sources of Uncertainty
Based on the Monte Carlo techniques described earlier, we generated likelihood
distributions for the dollar value of total annual benefits including PM health, ozone health,
and visibility benefits for the final CAIR. For this analysis, the likelihood descriptions for
the true value for each of the health endpoint incidence measures, including premature
B-9
-------
mortality, were based on classical statistical uncertainty measures, including the mean and
standard deviation for the C-R relationships in the epidemiological literature, and assumption
of particular likelihood distribution shapes for the valuation for each health endpoint value
based on reported values in the economic literature. Table B-l summarizes the chosen
parameters for likelihood distributions for unit values for each of the PM health effects
included in the Monte Carlo simulation. The distributions for the value used to represent
incidence of a health effect in the total benefits valuation represent both the simple statistical
uncertainty surrounding individual effect estimates and, for those health endpoints with
multiple effects from different epidemiology studies, interstudy variability. Visibility
benefits are also included in the distribution of total benefits; however, we were unable to
characterize a distribution for visibility benefits. As such, they are simply added to each
percentile of the distribution of health benefits.
Results of the Monte Carlo simulations are presented in Table B-2. The table
provides the estimated means of the distributions and the estimated 5th and 95th percentiles of
the distributions. The contribution of mortality to the mean benefits and to both the 5th and
95th percentiles of total benefits is substantial, with mortality accounting for 93 percent of the
mean estimate, and even the 5th percentile of mortality benefits dominating the 95th percentile
of all other benefit categories. Thus, the choice of value and the shape for likelihood
distribution for VSL should be examined closely and is key information to provide to
decision makers for any decision involving this variable. The 95th percentile of total benefits
is approximately twice the mean, while the 5 th percentile is approximately one-fourth of the
mean. The overall range from 5th to 95th represents about one order of magnitude.
B.3 Pilot Expert Elicitation of PM Mortality Uncertainty
Expert elicitation is a formal, highly structured and well-documented process
whereby expert judgments, usually of multiple experts, are obtained. Formal expert
elicitation usually involves experts with training and expertise in statistics, decision analysis,
and probability encoding who work with subject matter experts to structure questions about
uncertain relationships or parameters and who design and implement the process used to
obtain probability and other judgments from subject matter experts. Several academic
traditions—judgment and decision-making, human factors, cognitive sciences, expert
systems, management science, to name a few—have sought to understand how to
successfully elicit probabilistic judgments from both lay people and experts (Morgan and
Henri on, 1990; Cooke, 1991; Wright and Ayton, 1994; Ayyub, 2002). Over the past 2
decades, an increasing number of studies have used expert judgment techniques to
characterize uncertainty in quantities of interest to environmental risk analysis and decision
B-10
-------
Table B-2. Distribution of Value of Annual Human Health and Welfare Benefits in
2015 for the Final CAIR Rule3
Endpoint
Premature mortality0
Long-term exposure, (adults, >30yrs)
Long-term exposure (child
-------
making. North and Merkhofer (1976) considered using expert judgment in evaluating
emission control strategies. As referred to by the NAS, EPA's Office of Air Quality
Planning and Standards (OAQPS) successfully used expert judgment to characterize
uncertainty in the health effects of exposure to lead (McCurdy and Richmond, 1983;
Whitfield and Wallsten, 1989) and to ozone (Whitfield et al., 1991; Winkler et al., 1995).
Amaral (1983) and Morgan et al. (1984) used expert judgment to evaluate the transport and
impacts of sulfur air pollution. Several studies have been done in the area of climate change
(Manne and Richels, 1994; Nordhaus, 1994; Morgan and Keith, 1995; Reilly et al., 2001).
Hawkins and Evans (1989) used industrial hygienists to predict toluene exposures to workers
involved in a batch chemical process. In a more recent use of expert judgment in exposure
analysis, Walker et al. (2001, 2003) asked experts to estimate ambient, indoor and personal
air concentrations of benzene. A few studies have used expert judgment to characterize
uncertainty in chemical dose-response: Hawkins and Graham (1988) and Evans et al.
(1994a) for formaldehyde and Evans et al. (1994b) for risk of exposure to chloroform in
drinking water. Expert judgment has also been used to characterize residential radon risks
(Krewski et al., 1999). The Non-Road Diesel Rule (Mansfield, 2004) was the first
illustration of an application of the results of an expert elicitation study to a regulatory policy
analysis.
In its 2002 report, the NAS provides a number of recommendations for how EPA
might improve the characterization of uncertainty in its benefits analyses. One
recommendation was that:
"EPA should begin to move the assessment of uncertainties from its ancillary
analyses into its primary analyses by conducting probabilistic, multiple-source
uncertainty analyses. This shift will require specification of probability distributions
for major sources of uncertainty. These distributions should be based on available
data and expert judgment" (NAS, 2002, p. 14). The NAS elaborated on this
recommendation by saying "although the specific methods for selection and
elicitation of experts may need to be modified somewhat, the protocols that have
been developed and tested by OAQPS [in prior EPA projects—see below] provide a
solid foundation for future work in the area. EPA may also consider having its
approaches reviewed by decision analysts, biostatisticians, and psychologists from
other fields where expert judgment is applied" (NAS, 2002, p. 140).
They recommended the use of formally elicited expert judgments but noted that a number of
issues must be addressed and that sensitivity analyses would be needed for distributions that
are based on expert judgment. They also recommended that EPA clearly distinguish between
B-12
-------
data-derived components of an uncertainty assessment and those based on expert opinions.
As a first step in addressing the NAS recommendations regarding expert elicitation, EPA, in
collaboration with OMB, conducted a pilot expert elicitation to characterize uncertainties in
the relationship between ambient PM2 5 and mortality.
B. 3.1 Elicitation Method
This pilot was designed to provide EPA with an opportunity to improve its
understanding of the design and application of expert elicitation methods to economic
benefits analysis and lay the groundwork for a more comprehensive elicitation. The scope of
the pilot was limited to a one-year period to allow for inclusion in the Non-Road Diesel Rule.
As such, we focused the elicitation on the C-R function of PM mass rather than on individual
issues surrounding an estimate of the change in mortality due to PM exposure. We selected
experts for participation from two previously established expert panels of the NAS. Due to
time constraints, we chose not to conduct a workshop with the experts prior to the elicitation,
which is often included in the protocol of elicitations to help condition and prepare the
experts for the elicitation. A full description of the pilot is contained in a report titled "An
Expert Judgment Assessment of the Concentration-Response Relationship between PM2 5
Exposure and Mortality" (lEc, 2004) available in the public docket of the Non-Road Diesel
Rule.
The analytic plan for the pilot was developed based on established elicitation
methods as suggested by the NAS and published in the peer-reviewed literature. The plan
and protocol were reviewed in three separate contexts. It was internally reviewed by EPA
and OMB scientists with experience using expert elicitation methods. The project team that
implemented the pilot consisted of individuals with experience in expert elicitation and
individuals with expertise in PM health effects and health benefits. Second, as part of a
review of the analytical blueprint of EPA's Second Prospective Analysis of the Costs and
Benefits of the Clean Air Act under Section 812 of the Act, a panel of outside experts—the
Health Effects Subcommittee (HES) of the Advisory Council on Clean Air Compliance
Analysis (Council)4—provided a limited5 and preliminary review of the methodology and
4 The Council is an advisory committee with an independent statutory charter that is organized and supported
under EPA's SAB.
5 Council/HES report: "...in view of the fact that the pilot project is well-underway, the experts have already
been selected, and many (if not all) of the interviews have been conducted, the HES sees little potential
benefit in providing detailed suggestions about the design or conduct of the pilot study" (EPA-SAB-
COUNCIL-ADV-04-002, March 2004, page 34).
B-13
-------
design of the expert elicitation (EPA-SAB-COUNCIL-ADV-04-002, March 2004), and
provided the following comments:
• "We applaud the Agency's interest in exploring the use of formal expert judgment
as a tool for improving uncertainty analysis and believe that the proposed pilot
study has great potential to yield important insights. The pilot is well designed to
inform subsequent and more comprehensive expert elicitation projects, but relies
on the opinions of a relatively small group of experts. It may provide preliminary
information about the general magnitude of the mortality effects, and may yield a
sense of both the uncertainty inherent in these estimates and the factors largely
responsible for such uncertainty. However, until the pilot study methods and
results have been subjected to peer review, it may be unwise for the Agency to
rely directly on these preliminary results in key policy decisions."
• In presenting results of the pilot elicitation, "the HES advises EPA to present the
entire collection of individual judgments; to carefully examine the collection of
individual judgments noting the extent of agreement or disagreement; to
thoughtfully assess the reasons for any disagreement; and to consider formal
combinations of judgments only after such deliberation and with full awareness of
the context..."
• "The HES recognizes that in order to make the pilot tractable it was necessary to
limit participation, and is aware of the many factors which must be balanced in
the selection of expert panels (Hawkins and Graham, 1988), but is concerned
about whether the judgments of such a limited group can reasonably be
interpreted as representing a fair and balanced view of the current state of
knowledge."
The Council, however, did not provide a peer review of the final report, or the interpretation
and application of results. The protocol was then tested on PM scientists from within EPA
and external to the Agency, who would not be part of the final elicitation process.
Upon completion of the final report, the pilot was reviewed in a third context. The
EPA commissioned a peer review of the pilot from a panel of four experts on the topics of
expert elicitation, decision analysis, and uncertainty characterization6 (Mansfield, 2004).
The review was generally positive. Overall, the reviewers indicated approval of the
procedures EPA followed to conduct the expert elicitation, commenting that the procedures
"were well documented and followed the standard elicitation protocol." Reviewers cited the
approach for selecting experts as a strength of the assessment. They also provided a variety
of specific suggestions, including:
1 A full report of the peer review is available at www.epa.gov/ttn/ecas^enefits.html.
B-14
-------
Some commented that the number of experts included in the elicitation was
small and could possibly be expanded in the future; however, there is not an
established method to determine the appropriate number of experts.
Several reviewers discussed was the lack of a pre- and post-elicitation
workshop with the participating experts, which is typically conducted in other
elicitations. They urged EPA to allow sufficient time to include these steps in
future elicitations but recognized the need to exclude these steps in the pilot.
Specifically, the reviewers stated that the experts should have communicated
with each other before and/or after the individual interviews. Group
communication prior to the individual interviews would have aided in the
motivation and conditioning steps of the elicitation, while communication,
either in person or through a summary document, would have allowed an
expert to adjust his response based on the responses of the other experts.
The encoding process of elicitation could be improved. In the elicitation
process, the reviewers interpreted that some of the experts provided
judgments based on a central tendency before providing judgments on
extreme values (upper and lower ranges). This type of sequencing may
introduce anchoring or adjustment heuristics, which are associated with
biased estimates of uncertainty. Although the authors of the elicitation report
introduced the topic of heuristics for the experts, reviewers felt that a more
substantive discussion on how the study addressed known sources and any
other potential forms of bias was necessary.
Reviewers also provided considerable comment on whether results of the pilot
should be combined into a single estimate. Several of the reviewers preferred
that the expert opinions not be combined or stated that they knew of no
agreed-upon method for combining results from expert elicitations. This
allows for the differences in the individual distributions to be recognized.
Two of the reviewers indicated that they were reasonably comfortable with
the method used in this study to combine the results, while the other two
reviewers offered comments on the combined result of the elicitation. One
reviewer stated that the combined distributions do not adequately capture the
opinions of individual experts but rather average them out. It is possible in
such cases that the combined judgments may generate results that none of the
experts could agree on. Another reviewer stated that expert elicitation studies
typically do not combine judgments, but if one were to combine them, he
B-15
-------
recommended that the response of each be maintained independently from the
other experts and run through the benefits model completely prior to
combining the results.
B. 3.2 Elicitation Results
Figure B-l displays the responses of the experts to the quantitative elicitation
question for the mortality effects of changes in long-term PM25 exposures. The distributions
provided by each expert, identified by the letters A through E, are depicted as box plots with
the diamond symbol showing the median (50th percentile), a circle symbol showing the mean
estimate, the box defining the interquartile range (bounded by the 25th and 75th percentiles),
and the whiskers defining each expert's 90 percent confidence interval (bounded by the 5th
and 95th percentiles of the distribution).
As illustrated by the figures, the experts exhibited considerable variation in both the
median values they reported and in the spread of uncertainty about the median. In response
to the question concerning the effects of changes in long-term exposures to PM2 5, the median
value ranged from values at or near zero to a 0.7 percent increase in annual nonaccidental
mortality per 1 |ig/m3 increase in annual mean PM25 concentration (within a range of PM25
concentrations from 8 to 20 |ig/m3). The variation in the responses for the effects of long-
term exposures largely reflects differences of opinion among the experts on a number of
factors such as the interpretation of key epidemiological results from long-term cohort
studies, the likelihood of a causal relationship, and the shape of the C-R function. Some
observations concerning the outcome of the individual expert judgments are provided below:
Key Cohort Studies. The experts' nonzero responses for the percentage change in
annual mortality were mostly influenced by the Krewski et al. (2000) reanalysis of the
original American Cancer Society (ACS) cohort study and by the later Pope et al. (2002)
update of the ACS study that included additional years of follow-up. In the characterization
of uncertainty upper bounds, none of the experts placed substantial weight on the mortality
estimates from the Six-Cities study (Dockery et al., 1993) in composing their quantitative
responses, despite citing numerous strengths of that analysis. Concern about sample size and
representativeness of the Six-Cities study for the entire United States appeared to be major
reasons for de-emphasizing those results. In addition, all of the experts gave a value of zero
at the 5th percentile, and thus the C-R functions are bounded by zero. We use a normal
distribution to characterize the pilot results, but the distribution could potentially be skewed
because of the bounding at zero.
B-16
-------
Q.
CO
E
2 --
1.5 --
1 --
0.5 --
C-8 ug/m3 C-10 ug/m3 C-15ug/m3 C-20 ug/m3
Expert
Figure B-l. Summary of Experts' Judgments About the Percentage Increase in Annual
Average Nonaccidental Mortality Associated with a 1 ng/m3 Increase in Annual
Average Exposures to PM2 5
* Expert B specified this distribution for the PM/mortality coefficient above an uncertain threshold that he
characterized as ranging between 4 and 15 with a modal value of 12 ng/m3. As illustrated here, considerable
variation exists in both the median values and the spread of uncertainty provided by the experts. The median
value of the percentage change in annual nonaccidental mortality per unit change in annual PM2 5
concentration (within a range of PM25 concentrations from 8 to 20 ng/m3) ranged from values at or near zero
to a value of 0.7 percent. The variation in the responses largely reflects differences in the amount of
uncertainty each expert considered inherent in the key epidemiological results from long-term cohort studies,
the likelihood of a causal relationship, and the shape of the C-R function. The technical report (lEc, 2004)
provides detailed descriptions of the experts' judgments about these factors, but we present a few brief
observations relative to their responses below.
** Expert C specified a nonlinear model and provided distributions for the slope of the curve at four discrete
concentrations within the range.
B-17
-------
Causality for Long-Term Effects. Three of the five experts gave distributions more
heavily weighted towards zero. Those experts were also the ones who expressed the lowest
probability of a causal effect of long-term exposure to PM2 5 in response to the preliminary
questions. All of the experts placed at least a 5 percent probability on the possibility that
there is no causal relationship between fine PM exposure and mortality; as a result, all
experts gave a fifth percentile value for the C-R coefficient of zero. For most of the experts,
this was based primarily on residual concerns about the strength of the mechanistic link
between the exposures and mortality.
Shape of the C-R Function for Long-Term Effects. The other key determinant of
each expert's responses for long-term effects was his assumption about the nature of the C-R
function across the range of baseline annual average PM25 concentrations assumed in the
pilot (8 to 20 jig/m3). Three experts (A, D, and E) assumed that the function relating
mortality with PM concentrations would be log-linear with constant slope over the specified
range. They therefore gave a single estimate of the distribution of the slope describing that
log-linear function. The other two experts provided more complex responses.
Expert B assumed a population threshold in his model, below which there would be
no effect of increased PM25 exposure and above which the relationship would be log-linear.
He characterized his estimate of a possible threshold as uncertain, ranging between 4 |ig/m3
and 15 i-ig/m3, with a modal value of 12 |ig/m3. He then described a distribution for the slope
for the log-linear function that might exist above the threshold; this distribution is depicted in
Figure B-2. The effect of incorporating the uncertain threshold is essentially to shift his
entire distribution downward.
Expert C believed that the increased relative risks for mortality observed in the cohort
studies were likely to be the result of exposures at the higher end of the exposure range, and
he expected there to be a declining effect on mortality with decreasing levels of PM25. He
also argued that some practical concentration threshold was likely to exist below which we
would not observe any increase in mortality. He reflected these beliefs by developing a
nonlinear model within the range from 8 to 20 |ig/m3; he described the model by providing
distributions for the slope of the curve at four discrete concentrations within the range.
B-18
-------
IO
0.
fO
E
"01
3
a>
Q.
c
-------
B.3.3 Experts' Views of Sources of Uncertainty
The experts were asked at several points during the interview to discuss the key
sources of potential bias and uncertainty in current evidence on which they relied for their
judgments. In the context of the quantitative discussion they were asked to list the top five
issues. They were encouraged to think about how these issues would affect the uncertainty
surrounding their best estimate of the potential impact on total mortality of a small change in
long-term exposure to PM2 5. The tables summarizing the factors identified by each expert
may be found in Appendix E of the technical report (TEc, 2004).
Many of the same factors appeared in the list of the five experts. However, the
experts often differed on whether a particular factor was a source of potential bias or
uncertainty. Some of the common concerns raised as either sources of bias or uncertainty
were
• residual confounding by smoking,
• residual confounding by "lifestyle" or other personal factors or "stressors,"
• exposure errors/misclassification,
• the role of co-pollutants as confounders or effect modifiers,
• impact of the relative toxicity of PM components,
• representativeness of the cohort populations with respect to the general U.S.
population, and
• investigator/publication biases.
Despite the many qualitative discussions about sources of uncertainty, because the
pilot study did not elicit quantitative judgments about the size and nature of impacts of each
source of uncertainty and bias, we were unable to systematically evaluate the nature of the
influence of these factors on the quantitative results provided by each expert unless an expert
explicitly adjusted his estimates by a particular factor.
B. 3.5 Limitations in Pilot Elicitation Design
The pilot elicitation has afforded many opportunities for learning about expert
elicitation in the context of economic benefits analysis. During the process of developing the
protocol, the project team noted limitations inherent in the design of the pilot, some of which
were also commented on by the SAB/HES in their review of the protocol prior to conducting
B-20
-------
the elicitation. Additional detail on the strengths and weaknesses of the pilot are provided in
the technical report (lEc, 2004).
The limited scope of the pilot meant that a full expert elicitation process was
truncated, and many aspects of the uncertainty surrounding the PM2 5/mortality relationship
could not be characterized quantitatively. Recognizing this, the results of the pilot are only
used in this benefits estimation for illustrative purposes.
• Small panel of experts—Because of resource constraints we limited the pilot to a
panel of five experts. As noted above, the SAB-HES (in Section B.3.1) brings
into question whether the judgments of such a limited group can reasonably be
interpreted as representing a fair and balanced view of the current state of
knowledge.
Little analytical research has been conducted on the more difficult question of
how to determine the ideal number of experts for a particular application. We
have not found any analyses of the effect of expert panel size based on
comparisons of empirical results of expert judgment studies. A theoretical
analysis by Clemen and Winkler (1985) suggests that where data sources are
moderately positively dependent there are diminishing marginal returns to the
value of information associated with each additional data source. In the context
of expert judgment studies, such a result implies that when dealing with experts
of similar backgrounds who rely on the same models and studies, a larger expert
panel may not provide significantly higher quality results than a smaller one.
However, the addition of an expert expected to provide a more independent
assessment, such as an expert from a different but pertinent field, would be
expected to exhibit a much greater value of information. Clemen and Winkler
(1999) note that "heterogeneity among experts is highly desirable." These
findings would appear to support addressing complex issues using a panel
comprising relatively small subgroups (perhaps three to five experts each) from
multiple disciplines. Although the decision analysis field tends to use relatively
small sample sizes (i.e., typically 5 to 10 experts), some are not comfortable with
obtaining a combined distribution from such small numbers in the absence of an
a priori assessment of the degree to which the expert panel is likely to be
representative of the overall population of relevant experts on the question of
interest.
• Use of an aggregate elicitation question—The expert judgment literature
discusses two broad approaches to elicitation of judgments: an aggregated and a
disaggregated approach. As the term implies, an aggregated approach asks the
expert to estimate the quantity of interest directly, for example, the numbers of
newspapers sold in the United States in a particular year. In a disaggregated
approach, the expert (or group of experts) is asked to construct a model for
estimating the quantity of interest and is asked directly about the inputs to that
B-21
-------
model (e.g., population in each state, percentage of the population that reads
newspapers). The intuition is that it is easier for experts to answer questions
about the intermediate quantities than about the total quantity.
The project team carefully considered the relative advantages and disadvantages
of the two approaches, however, the time and resources necessary to develop a
disaggregated model structure drove the decision to undertake an aggregate
approach to elicit the C-R coefficient for the PM2 5/mortality relationship. Based
on advice from the SAB-HES, the project team felt that separate questions to
address effects of short- and long-term exposures, though still at a high level of
aggregation, would prove to be easier for experts to address than a question that
"rolled up" all the effects into a single estimate. This level of disaggregation also
enabled the elicitation team to explore with experts possible overlap in reported
mortality effects detected using short-term and long-term epidemiological studies.
Nonetheless, a major goal of the preliminary and follow-up questions in the
protocol was to identify critical issues that could be addressed by developing a
more disaggregated approach in a future assessment.
The aggregated design limits our ability to determine the influence of any one key
factor over others in a large list of issues that the experts were to consider prior to
answering the quantitative question. It also limited the ability of the experts to
express their views about the difference in the C-R function based on the location
in the United States (i.e., the demographics of the exposed population, the air
concentration of PM and/or PM mixture).
No workshop was conducted—It is customary to conduct a workshop prior to the
elicitation interview with the experts. This allows the experts to become familiar
with the protocol and to discuss methods to limit bias during the interview.
Because of time constraints for the pilot, we did not conduct a pre-elicitation
workshop.
No calibration of experts—In some elicitation studies, the authors use a
calibration measure to weight the experts appropriately for the purposes of
combining the results of the elicitation. We do not have calibration measures that
could be used to assess the results of this pilot. At this point, we can only assess
the process—did the pilot assessment employ a structure, supporting materials,
and a process that enabled experts to make judgments that would be likely to be
well calibrated? Without calibration measures, we cannot weight experts based
on their performance on calibration and thus, we present only an equal weighting
of the responses. The peer reviewers agree that if a combination is provided, then
the equal weighting of judgments is preferred.
B-22
-------
B. 3.6 Combining the Expert Judgments for Application to Economic Benefit Analyses
Many methods are available to combine the responses from the experts. Each method
has advantages and disadvantages from a statistical viewpoint. The project team is not aware
of any rule of thumb in statistics that would provide guidance for combining linear and
nonlinear functions. Therefore, we considered four alternative methods for combining the
results as an illustration of potential combinations of the results. The peer review provides
extensive comments on whether a combination is necessary and on the methods employed
here.
Analysts must give careful thought to whether and how to combine the results of
individual expert judgments into a single distribution. When dealing with a limited number
of experts, the analyst must be particularly careful to identify the influence of each expert's
response on the combined distribution. Therefore, we considered four alternative methods
for combining the pilot results, each of which had limitations. The peer review also provides
a lengthy discussion of issues in combining expert judgments. In this section, we discuss the
issues we considered in combining the results of the pilot and how we came to the conclusion
that for the illustrative benefits analysis presented in Section B.5 below, we would present
both the individual quantitative distributions of the C-R coefficient elicited from the five
experts interviewed and results based on a probabilistic estimate that represents the
combined results of the pilot based on an equal weighting of the calculated change in
mortality incidence based on the individual judgments.
B.3.6.1 Background
Some investigators (e.g., Hawkins and Graham [1990]; Winkler and Wallsten [1995];
and Morgan et al. [1984]) have preferred to keep expert opinions separate to preserve the
diversity of opinion on the issues of interest. In such situations, presenting a range of values
expressed by the experts can help decision makers to understand the sensitivity of their
analyses to the analytical model chosen, thereby bounding possible outcomes. Individual
judgments also can illustrate varying opinions arising from different disciplinary
perspectives or from the rational selection of alternative theoretical models or data sets
(Morgan and Henri on, 1990). Nonetheless, analysts are often interested in developing a
single distribution of values that reflects a synthesis of the judgments elicited from a group of
experts.
On the other hand, there are advantages to combining the results across experts. An
extensive literature describes methods for combining expert judgments. These methods can
be broadly classified as either mathematical or behavioral (Clemen and Winkler, 1999).
B-23
-------
Mathematical approaches range from simple averaging of responses to much more complex
models incorporating information about the quality of expert responses, potential dependence
among expert judgments, or (in the case of Bayesian methods) prior probability distributions
about the variable of interest. Behavioral approaches require the interaction of experts in an
effort to encourage them to achieve consensus, either through face-to-face meetings or
through the exchange of information about judgments among experts. As noted in the
technical report (lEc, 2004),analysts have raised both methodological and practical concerns
with respect to the behavioral approach. Therefore, we used a mathematical combination
process to derive a single distribution.
One advantage of mathematical combination over behavioral approaches is the ability
to be completely transparent about how weights have been assigned to the judgments of
specific experts and about what assumptions have been made concerning the degree of
correlation between experts. Several approaches can be used to assign weights to individual
experts. Weights can be assigned based on the analyst's opinion of the relative expertise of
each expert; on a quantitative assessment of the calibration and informativeness (i.e.,
precision) of each expert as determined from their responses to a set of calibration questions
(as described in Cooke [1991]); or on weights assigned by each expert, either to him or
herself or to the other experts on the panel (see Evans et al. [1994b] for an example of this
approach). Ideally, such a weighting system would address problems of uneven calibration
and informativeness across experts, as well as potential motivational biases (Cooke, 1991).7
At the design stages of the pilot, we decided that the resulting expert judgments
would be combined using equal weights, essentially calculating the arithmetic mean of the
expert responses, for simplicity and transparency. Some empirical evidence suggests that the
simple combination rules, like equal weighting, perform equally well when compared to
more complex methods in terms of calibration scores for the combined results (Clemen and
Winkler, 1999).
B.3.6.2 Alternative Combination Methods
Although a combination method using equal weights for the results of each expert is
straightforward in principle, applying it in this context of the results of the pilot was
complicated by the fact that the elicitation protocol gave the experts freedom to specify
different forms for the C-R function. If all the experts had chosen the same form of the C-R
7 "Motivational bias" refers to the willful distortion of an expert's true judgments. The origins of this bias can
vary but could include, for example, a reluctance to contradict views expressed by one's employer or a
deliberate attempt to skew the outcome of the study for political gain.
B-24
-------
function (e.g., if each expert had specified a log-linear C-R function with a constant, but
uncertain, C-R coefficient [i.e., slope] over the PM range specified in the protocol), the
combination of their distributions for the C-R coefficient would require a simple averaging
across experts at each elicited percentile. However, in this assessment, three experts
specified log-linear functions with constant C-R coefficients over the specified range of
PM2 5 concentrations, and two of the experts specified the C-R coefficient was likely to vary
over the range of specified PM2 5 concentrations (as discussed in Section B.4.2 above).
These more complex C-R functions necessitated some additional steps in calculating the
combined results.
As discussed in the technical report for the pilot (lEc, 2004), individual responses
either can be combined before application of the benefits model or during the application of
the model, allowing each expert's C-R function to be estimated in the benefits model
independently. We considered three alternative approaches to combining the expert
judgments before application to the benefits model each of which differs in how the
combined estimate accounts for the underlying particulate air pollution levels. We
considered the use of (1) a uniform distribution and equal weighting, which involves taking a
simple average of the responses across experts for each percentile, (2) a normal distribution
describing the population-weighted annual average PM2 5 concentration data, and (3) a
combined distribution specified at four intervals of PM25 concentrations that coincide with
those specified by Expert B.
Overall, the combination methods considered result in fairly similar results at the
median and mean relative risk estimate. However, slight differences occur in the tails of the
distribution in their characterization of uncertainty. In particular, when combining the upper
bound is averaged out to a lower value that may skew the results (in comparison to the views
expressed by the experts). Thus, the resulting estimates from the combined distribution may
not be estimates to which the experts would agree (i.e, Expert E may not agree with a
lowering of the upper bound estimate due to averaging across experts). In Figure B-3, the
C-R function for the population-weighted combination method was compared to the existing
cohort epidemiological studies of the long-term PM2 5/mortality relationship. We observe
that the results of the pilot elicitation are generally within the range of findings from these
epidemiological studies. However, as expected, the elicitation results in a larger spread of
uncertainty than is given by the standard errors of the individual studies.
B-25
-------
CM
Q_
CO
E
I1
I
3
5
ro
-------
provided by Experts B and C (although some adjustments must still be made).8 Details of the
illustration are provided in Section B.6.
B.4 Illustrative Application of Pilot Expert Elicitation Results
In this section, we illustrate how expert judgments can be applied in a benefits
analysis. We estimated avoided incidence of mortality associated with the CAIR rule for
each expert and also applied the pooled approach for combining results across participants to
develop a combined distribution of avoided incidence of premature mortality. We then used
Monte Carlo simulations to combine the expert judgment-based distributions with the
distribution for VSL and with the values for other health and welfare endpoints. We present
the resulting distributions of total dollar benefits to demonstrate how using expert elicitation
can improve the characterization of the overall uncertainty in benefits estimates. The values
generated below are not intended to replace the primary estimate of benefits of CAIR. They
are included solely as an illustration of expert elicitation-based distributions that
characterize the uncertainty of the estimate of premature mortality associated with long-term
exposure to PM2 5 rather than a data-derived distribution.
B.4.1 Method
B. 4.1.1 C-R Distribution Based on Combined Results Across Experts
As discussed in Section B.4.5, we converted each expert's percentile responses about
mortality associated with long-term exposure into a custom distribution such that each
percentile is correctly represented and percentiles in between are represented as continuous
functions (custom distributions were generated using Crystal Ball and are represented as
15,000 equally probable points).
For Experts A, D, and E, we used a standard log-linear functional form:
! Expert B specified a distribution for the C-R coefficient for PM2 5 concentrations above a threshold and
assigned the coefficient a value of zero for all PM concentrations below the threshold. He then specified a
probability distribution to describe the uncertainty about the threshold value. Expert C specified separate
distributions for the C-R coefficient at four discrete points within the concentration ranges defined in the
protocol to represent a continuous C-R function whose slope varied with the PM2 5 concentration. Expert C
indicated that the coefficient value between these points was best modeled as a continuous function, rather
than a step function. Both experts assumed the same functional forms in responding to elicitation question.
B-27
-------
where we set p equal to ln(l+B/100), where B is the percentage change in all-cause mortality
associated with a 1 |ig reduction in PM25. BenMAP then represents the distribution of Ay
based on the custom distribution of p.
The conditional functions of Experts B and C required us to estimate some values on
the C-R function between the points that were elicited, which requires an extrapolation from
the response provided in the pilot to create continuous distributions. This can alter the true
response given by these experts to the elicitation. This component of the combination was
not included in the peer-review material.
Specifically, Expert C provided a set of conditional C-R functions for different
baseline levels of PM25. Expert C provided four conditional responses, one for 8 i-ig/m3, one
for 10 i-ig/m3, one for 15 i-ig/m3, and one for 20 |ig/m3. To "fill in" the C-R function for
intermediate baseline PM2 5 values, we linearly interpolated between the responses for each
pair of points (e.g., 10 to 15 or 15 to 20). We calculated interpolated values for 13 points,
ranging from 8 |ig to 20 |ig. For baseline values less than 8 |ig, we assigned a value of zero
(essentially assuming a threshold at 8 |ig). For baseline values greater than 20 |ig, we
assigned the values provided by Expert C for 20 |ig. This may result in an underestimate of
the incidence of mortality for Expert C. For each of the conditional functions, we used a log-
linear specification, similar to A, D, and E. Total incidence of mortality for Expert C is the
sum of the conditional estimates over the range of baseline air concentrations.
Expert B provided a log-linear C-R function, conditional on an unknown threshold
characterized by a triangular distribution bounded by 4 |ig and 15 |ig, with a mode at 12 |ig.
We discretized the triangular distribution into 12 ranges of unit length (e.g., 4 to 5, 5 to 6)
and calculated the expected value of the response at each population grid cell based on the
observed baseline PM2 5 and the probability of that baseline value exceeding the potential
threshold. We assumed that if a grid cell has a baseline value above the threshold, then the
full value of the reduction in PM2 5 at that grid cell is associated with a reduction in mortality.
This may result in an overestimate of the mortality impact for Expert B because for grid cells
where the baseline level is only marginally above the threshold, a benefit might only accrue
to the change in PM2 5 down to the threshold. The rest of the change would not result in any
mortality reduction. Because most of the changes in air quality are relatively small
(population-weighted change in annual mean PM2 5 is around 1 |ig), and larger changes tend
to occur where there are higher baseline PM2 5 levels, this should not be a large issue.
To provide context for the estimates based on the experts providing conditional and
threshold specifications, it is useful to summarize the CAIR baseline PM2 5 levels in 2015.
Table B-5 lists the population distribution of baseline concentrations of PM25 in 2015.
B-28
-------
Table B-5. Population Distribution of Baseline Ambient PM2
-2.5
Baseline PM2 5 (|j,g/m3)
PM25<5
5
-------
60,000
50,000
40,000
Note: Distributions labeled Expert A - Expert E are based on individual expert
responses. The distribution labeled Combined Experts is based on Monte Carlo
sampling from the individual expert distributions assuming equal weights. The
distribution labeled Pope et al. (2002) Statistical Error is based on the mean and
standard error of the C-R function from the study.
15
r
o
£
3 30,000
re
o 20,000
'o
3
•D
10,000
13,700
7,100
2,800
11,600
22,800
11,500
16,700
Expert A
Expert B
Expert C
Expert D
Expert E Combined Experts Pope et al (2002)
Statistical Error
Figure B-4. Results of Illustrative Application of Pilot Expert Elicitation: Annual Reductions in Premature Mortality in
2015 Associated with the Clean Air Interstate Rule
B-30
-------
pooled estimate assuming equal weight for each expert's distribution and the Pope et al.
(2002) results for comparison. Corresponding distributions of the dollar value of the
reductions in premature mortality are presented in Figure B-6. The distributions are depicted
as box plots with the diamond symbol (+) showing the mean, the dash (-) showing the
median (50th percentile), the box defining the interquartile range (bounded by the 25th and
75th percentiles), and the whiskers defining the 90 percent confidence interval (bounded by
the 5th and 95th percentiles of the distribution). For comparison, the figure also displays the
distribution derived from the statistical error associated with Pope et al. (2002).
The figure shows that the average annual number of premature deaths avoided
because of CAIR in 2015 ranges from approximately 3,000 to 23,000, depending on the C-R
function used. The medians span zero to 20,000, with the zero value due to the low threshold
associated with one of the expert's distributions. At the means of the distributions,
45 percent more premature deaths are predicted to be avoided by the estimate based on Pope
et al. (2002) than based on the estimate pooled across the five experts. Specifically, because
less than a quarter of the population is expected to live in areas with PM25 levels above the
threshold specified by Expert C, and much of the decrease in PM25 associated with CAIR
occurs below that threshold, a much smaller decrease in premature morality is predicted for
Expert C than those experts who provided continuous C-R functions down to zero (PM2 5) as
well as for Expert B who provided an uncertain threshold. Furthermore, note that above the
50th percentile, the C-R functions provided by all of the experts predict positive benefits from
the modeled control option.
The boxplots displayed in Figure B-4 are derived by applying the C-R distributions
specified by each expert (as presented in Figure B-l) to the change in air quality predicted by
the final CAIR. Although the Figures B-4 and B-l show similar patterns, there are important
differences. Specifically, the ratio of 75th percentiles of the C-R functions specified by
Experts A and B (as denoted in Figure B-l) is 0.4, whereas the ratio of the predicted change
in incidence of premature mortality associated with the modeled preliminary control option
is 0.5. This 25 percent increase (from 0.4 to 0.5) in the ratio highlights the impact of the
extent of the predicted air quality change on the choice of C-R function used in the benefits
analysis.
The combined expert distribution depicted in Figure B-4 provides additional insights.
The combined (average) distribution has a 90 percent credible interval between zero and
37,000. When compared with results derived from the Pope et al. (2002) study, it is clear
that the combined expert distribution reflects greater uncertainty about the estimated
reduction in premature mortality, which is expected given that the elicitation exercise was
designed to encompass more sources of uncertainty than were addressed in the Pope et al.
study, including fundamental model uncertainty. In addition, the expert judgments place
more weight on the lower end of the distribution than the Pope et al. study. The mean
estimate from the combined expert distribution is over 30 percent lower than the mean
B-31
-------
derived from the Pope et al. (2002) distribution. However, the 90 percent confidence interval
based on the standard error from Pope et al. (2002) is completely contained within the 90
percent credible interval of the combined expert distribution.
Figure B-5 shows the same data using cumulative distribution functions (CDFs).
This figure is valuable for demonstrating differences in degree of certainty in achieving
specific reductions in premature mortality. For instance, the Pope et al. (2002) C-R
distribution predicts a 90 percent chance that there will be more than 10,000 fewer premature
deaths, whereas the pooled distribution from the pilot predicts less than a 50 percent chance
of more than 10,000 fewer premature deaths resulting from CAIR. The probabilities
associated with the individual experts for avoiding 10,000 or more premature deaths range
from about 5 percent to 80 percent, demonstrating once again the sensitivity of the estimate
to assumptions regarding the C-R function. The CDFs of the estimated reductions in
premature mortality show that for several experts there is a small probability of a
substantially more reduction in pre-mature mortality. For example, the 75th percentile of the
distribution based on Expert B's responses is at 11,000 fewer deaths, while the 98th percentile
for that distribution is over three times higher, at 35,000. The CDF also shows that, although
most of the experts provided fairly wide distributions reflecting incorporation of information
beyond what is demonstrated in any one empirical study, the CDF based on Expert C's
responses is much narrower, reflecting the high degree of confidence he placed on the
existence of a threshold below 15 |ig.
Figures B-6 and B-7 use box plots and CDFs to display the estimated dollar value of
these annual reductions in premature mortality. Whereas the average based on the Pope et al.
(2002) distribution is $93 billion, the average based on the pooled estimate from the pilot is
$64 billion, a difference of approximately one-third. Once the C-R distributions are
combined with the VSL distributions, not only are the mean values closer to one another, but
the distributions show considerably more overlap.
Because these distributions are the result of a Monte Carlo simulation combining the
non-normal distributions for reductions in mortality with a normal distribution for VSL, the
resulting distributions will also be non-normal, but the shape depends on the skewness of
each of the input distribution of mortality reductions. For example, the ratio of the 95th to
75th percentile of mortality reductions for Expert B is 2.9, while the same ratio for the value
of mortality reductions is 3.4, indicating the value distribution is more skewed than the
reductions distribution. In general, combining normal or left-skewed distributions in a
mulitplicative fashion will result in left-skewed distributions with greater skewness than the
input distributions. So even for the normally distributed estimates based on Pope et al.
(2002), the value distribution is somewhat skewed, because it is the result of multiplying two
normally distributed random variables.
B-32
-------
Note: Distributions labeled Expert A - Expert E are based on individual expert
responses. The distribution labeled Combined Experts is based on Monte Carlo
sampling from the individual expert distributions assuming equal weights. The
distribution labeled Pope et al. (2002) Statistical Error is based on the mean and
standard error of the C-R function from the study.
10,000
20,000 30,000 40,000 50,000
Annual Reductions in Incidence of Premature Mortality in 2015
60,000
70,000
Expert A Expert B
• Expert C
• Expert D Expert E
•Combined Experts
Pope et al (2002) Statistical Error
Figure B-5. Cumulative Distribution Functions for Annual Reductions in Premature Mortality in 2015 Associated with the
Clean Air Interstate Rule
B-33
-------
$400,000
$350,000
$300,000
$250,000
0 $200,000
to
c
o
$150,000
$100,000
$50,000
$0
Mote: Mortality distributions labeled E
expert responses. The distribution \s
Carlo sampling from the individual ex
The mortality distribution labeled Po|
mean and standard error of the C-R
Dased on a normally distributed VSL
between $1 and $10 million. The VS
ncome growth out to 2015 using an
xpert A - Expert E are based on individual
beled Combined Experts is based on Monte •
pert distributions assuming equal weights.
e et al. (2002) statistical error is based on the
unction from the study. Mortality valuation is
with a mean of $5.5 million and a 95% Cl
L distribution has then been adjsuted for
adjustment factor of 1 .15.
•
—
$76,000
•
*
<
j • $64,000
$40,000 ~
* $16.000
^^
>
•
$130,000
-r
^ $93,000
« $64,000
^
Expert A
Expert B
Expert C
Expert D
Expert E Combined Experts Pope et al (2002)
Statistical Error
Figure B-6. Results of Illustrative Application of Pilot Expert Elicitation: Dollar Value of Annual Reductions in
Premature Mortality in 2015 Associated with the Clean Air Interstate Rule
B-34
-------
0.9
.Q
re
.Q
O
3
3
O
Note: Mortality distributions labeled Expert A - Expert E are based on individual
expert responses. The distribution labeled Combined Experts is based on Monte
Carlo sampling from the individual expert distributions assuming equal weights.
The mortality distribution labeled Pope et al. (2002) statistical error is based on the
mean and standard error of the C-R function from the study. Mortality valuation is
based on a normally distributed VSL with a mean of $5.5 million and a 95% Cl
between $1 and $10 million. The VSL distribution has then been adjsuted for
income growth out to 2015 using an adjustment factor of 1.15.
$0
$100,000
$200,000
$300,000
Million ($1999)
$400,000
$500,000
$600,000
- Expert A Expert B
• Expert C Expert D Expert E
•Combined Experts
Pope et al (2002) Statistical Error
Figure B-7. Cumulative Distribution Functions for Dollar Value of Annual Reductions in Premature Mortality in 2015
Associated with the Clean Air Interstate Rule
B-35
-------
The shapes of the two distributions are more similar in this case because both reflect
the same additional information in the VSL distribution. This demonstrates that, as
additional sources of uncertainty are added to the analysis, the influence of any single source
of uncertainty will fall. Because VSL is a large source of uncertainty, the influence on
overall uncertainty relative to the distribution of the mortality reduction is also large. All of
the distributions of the value of mortality reductions have a small negative tail, this time
because of propagation of the normally distributed VSL, which has a small amount of the
distribution below zero. Again, we interpret this as a statistical artifact rather than a true
probability that the value of a statistical life is negative (implying that individuals would pay
to increase the risk of death).
We used additional Monte Carlo simulations to combine the expert-based
distributions for the dollar benefits of mortality with the distributions of dollar benefits for
the remaining health and welfare endpoints to derive estimates of the overall distribution of
total dollar benefits.10 The box plots for these distributions of overall dollar benefits
associated with CAIR in 2015 are presented in Figure B-8. Because mortality accounts for
over 90 percent of the benefits, the addition of other endpoints has little impact on the overall
distributions. The overall mean annual total dollar benefits in 2015 for the distribution
incorporating the combined expert distribution for reductions in premature mortality is $74
billion, compared to $100 billion for the results derived from the Pope et al. (2002) study.
The CDFs for total dollar benefits are provided in Figure B-9. These again suggest
that using the pilot expert elicitation-based representation of uncertainty in the relationship
between PM25 and premature mortality has a large impact on the shape and range of the
distribution of total benefits. The Pope et al. (2002) derived results have an approximately
Weibull-shaped distribution with a range from 5th to 95th percentiles of $26 billion to $210
billion, or about one order of magnitude. The distribution of total dollar benefits
incorporating the combined expert distribution for reductions in premature mortality has a
much more skewed shape with an elongated positive tail above the 75th percentile with a
range from 5th to 95th percentiles of $3 billion to $240 billion, or about two orders of
magnitude, with significant probability mass at the lower end of the range.
10Note that visibility benefits are treated as fixed for this illustrative analysis. We are working on methods to
characterize the uncertainty in visibility and other nonhealth benefits.
B-36
-------
400000
300000
o>
o>
o>
200000
Note: All non-mortality distributions are based on classical statistical error
derived from the standard errors reported in epidemiology studies and
distributions of unit values based on empirical data. Visibility benefits are
included as a constant. Mortality distributions labeled Expert A - Expert E are
based on individual expert responses. The distribution labeled Combined
Experts is based on Monte Carlo sampling from the individual expert distributions
assuming equal weights. The mortality distribution labeled Pope et al. (2002)
statistical error is based on the mean and standard error of the C-R function from
the study. Dollar benefits have been adjusted upwards to account for growth in
real income out to 2015.
o
100000
$85,000
$48,000
$24,000
T
$74,000
$140,000
$74,000
$100,000
Expert A
Expert B
Expert C
Expert D
Expert E
Combined Experts
Pope et al (2002)
Statistical Error
Figure B-8. Results of Illustrative Application of Pilot Expert Elicitation: Dollar Value of Total Annual Benefits in 2015
Associated with the Clean Air Interstate Rule
B-37
-------
Note: All non-mortality distributions are based on classical statistical error
derived from the standard errors reported in epidemiology studies and
distributions of unit values based on empirical data. Visibility benefits are
included as a constant. The distribution labeled Combined Experts is based on
Monte Carlo sampling from the individual expert distributions assuming equal
weights. The mortality distribution labeled Pope et al. (2002) statistical error is
based on the mean and standard error of the C-R function from the study. Dollar
benefits have been adjusted upwards to account for growth in real income out to
2015.
$0
$100,000 $200,000 $300,000
Million ($1999)
$400,000
$500,000
-Expert A Experts
• Expert C Expert D Expert E
•Combined Experts
Pope et al (2002) Statistical Error
Figure B-9. Cumulative Distribution Functions of Dollar Value of Total Annual PM-Related Health and Visibility Benefits
in 2015 Associated with the Clean Air Interstate Rule
B-38
-------
B. 5.3 Limitations of the Application of the Pilot Elicitation Results to the CAIR Scenario
The results presented in this section should be viewed cautiously given the limited
scope of the pilot and the limitations of the elicitation design and methods used to combine
the expert judgments discussed above. Therefore, the results presented above should be
considered "illustrative" until the methods used to interpret and apply the results of the pilot
have been peer reviewed by EPA's SAB.
Specific limitations of the illustrative application include the following:
• Extrapolation of percentile responses provided by individual experts. Each expert
provided minimum and maximum values, as well as the 5th, 25th, 50th, 75th, and
95th percentiles. To generate the continuous distributions of mortality impacts, we
had to make assumptions about the continuity of the distributions between the
reported percentiles. The use of assumptions adds uncertainty to the results.
• Interpolation of C-R relationship across PM25 levels. Expert C provided a set of
conditional distributions of the C-R relationship conditioned on the baseline level
of PM25. Because he only provided functions for a limited number of baseline
levels, we had to interpolate the values between levels, introducing additional
uncertainty. In addition, Expert C provided no information on the C-R function
for baseline PM25 levels below 8 |ig/m3 or above 20 |ig/m3. We assumed no
mortality impacts for baseline levels lower than 8 and no increase in the C-R
function above 20. If the interpolation method was incorrect, it likely biased our
results downward.
• Interpretation of Expert B's results. Expert B provided a conditional distribution
for the C-R function, conditioned on an uncertain threshold. Expert B provided
additional information about the shape of the distribution for the threshold. To
develop an applied function, we assumed that the uncertain threshold could be
incorporated into the C-R function by constructing an expected value function. If
our interpretation was flawed, the specific functions introduced may lead to a
slight overestimate of mortality impacts.
• Ranges based on individual experts should be viewed with caution because they
represent only a single individual's interpretation of the state of knowledge about
PM and mortality. Results for individual experts should not be extracted and
presented without reference to the full range of results across the five experts.
• Any range of results presented based on the application of the pilot results in the
CAIR scenario should be presented along with their relative likelihood of
occurring (i.e., the percentile represented in the distribution).
B-39
-------
B.6 References
Agency for Healthcare Research and Quality (AHRQ). 2000. HCUPnet, Healthcare Cost
and Utilization Project.
Amaral, D. 1983. Estimating Uncertainty in Policy Analysis: Health Effects from inhaled
Sulfur Oxides. Ph.D. Thesis, Department of Engineering and Public Policy, Carnegie
Mellon University, Pittsburgh, PA.
Ayyub, B.M. 2002. Elicitation of Expert Opinions for Uncertainty and Risks. CRC Press,
Florida.
Clemen, R.T., and R.L. Winkler. 1999. "Combining Probability Distributions From Experts
in Risk Analysis," Risk Analysis. 19: 187-203.
Cooke, R. 1991. Experts in Uncertainty. Oxford University Press, New York.
Cropper, M.L., and AJ. Krupnick. 1990. "The Social Costs of Chronic Heart and Lung
Disease." Resources for the Future. Washington, DC. Discussion Paper QE 89-16-
REV.
Dockery, D.W., C.A. Pope, X.P. Xu, J.D. Spengler, J.H. Ware, M.E. Fay, E.G. Ferris, and
F.E. Speizer. 1993. "An Association between Air Pollution and Mortality in Six
U.S. Cities." New England Journal of Medicine 329(24): 1753-1759.
EPA-SAB-COUNCIL-ADV-04-002. March 2004. Advisory on Plans for Health Effects
Analysis in the Analytical Plan for EPA 's Second Prospective Analysis—Benefits and
Costs of the Clean Air Act, 1990-2020: Advisory by the Health Effects Subcommittee
of the Advisory Council on Clean Air Compliance Analysis.
Evans, J.S., J.D. Graham, G.M. Gray, R.L. Sielken. 1994a. "A Distributional Approach to
Characterizing Low-Dose Cancer Risk." Risk Analysis 14(l):25-34.
Evans, J.S., G.M. Gray, R.L. Sielken, Jr., A.E. Smith, C. Valdez-Flores, and J.D. Graham.
1994b. "Use of Probabilistic Expert Judgment in Uncertainty Analysis of
Carcinogenic Potency." Regulatory Toxicology and Pharmacology 20:25-36.
Hawkins, N.C., and J.S. Evans. 1989. "Subjective Estimation of Toluene Exposures: A
Calibration Study of Industrial Hygienists." Applied. Ind. Hygiene 4:61-68.
Hawkins, N.C., and J.D. Graham. 1988. "Expert Scientific Judgment and Cancer Risk
Assessment: A Pilot Study of Pharmacokinetic Data." Risk Analysis 8(4):615-625.
Hawkins, N.C. and J.D. Graham. 1990. "Expert Scientific Judgment and Cancer Risk
Assessment: A Pilot Study of Pharmacokinetic Data," Risk Analysis. 8: 615-625.
B-40
-------
Industrial Economics, Incorporated (ffic). March 31, 1994. Memorandum to Jim
DeMocker, Office of Air and Radiation, Office of Policy Analysis and Review, U.S.
Environmental Protection Agency.
Industrial Economics, Incorporated (lEc). 2004. "An Expert Judgment Study of the
Concentration-Response Relationship Between PM2.5 Exposure and Mortality,"
available at: .
Industrial Economics, Incorporated (lEc). April 2004. "Expert Judgment Assessment of the
Relationship Between PM2 5 Exposure and Mortality." Available at:
.
Krewski, D., R.T. Burnett, M.S. Goldbert, K. Hoover, J. Siemiatycki, M. Jerrett, M.
Abrahamowicz, and W.H. White. July 2000. Reanalysis of the Harvard Six Cities
Study and the American Cancer Society Study of Paniculate Air Pollution and
Mortality. Special Report to the Health Effects Institute, Cambridge MA.
Krewski, D., S.N. Ras, J.M. Zielmski, and P.K. Hopke. 1999. "Characterization of
Uncertainty and Variability in Residential Radon Cancer Risks." Ann. N.Y. Acad.
Sci. 895:245-272.
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.
Manne, A.S., and R.G. Richels. 1994. "The Costs of Stabilizing Global CO2 Emissions: A
Probabilistic Analysis Based on Expert Judgments." The Energy Journal
Mansfield, C. 2004. "Peer Review of Expert Elicitati on." . Research Triangle Park, NC: RTI International.
McCurdy, T., and H. Richmond. 1983. Description of the OAQPS Risk Program and the
ongoing Lead NAAQS Risk Assessment Project. Paper 83-74. 1 . Presented at the 76th
Annual Meeting of the Air Pollution Control Association, June 19-24, Atlanta,
Georgia. As cited in NAS.
Morgan, M.G., S.C. Morris, M. Henrion, D.A. Amaral, and W.R. Rish. 1984. "Technical
Uncertainty in Quantitative Policy Analysis - A Sulfur Air Pollution Example." Risk
Analysis 4:201-216.
Morgan, G., and M. Henrion. 1990. Uncertainty; A Guide to Dealing with Uncertainty in
Quantitative Risk and Policy Analysis. Cambridge University Press, Cambridge.
B-41
-------
Morgan, M.G., and D.W. Keith. 1995. "Subjective Judgments by Climate Experts."
Environmental Science and Technology 29:468A-476A.
National Academy of Sciences/National Research Council (NAS). 2002. Estimating the
Public Health Benefits of Proposed Air Pollution Regulations. The National
Academies Press: Washington, D.C.
Neumann, I.E., M.T. Dickie, and R.E. Unsworth. March 31, 1994. "Linkage Between
Health Effects Estimation and Morbidity Valuation in the Section 812
Analysis—Draft Valuation Document." Industrial Economics Incorporated (TEc)
Memorandum to Jim DeMocker, U.S. Environmental Protection Agency, Office of
Air and Radiation, Office of Policy Analysis and Review.
Nordhaus, W.D. 1994. "Expert Opinion on Climatic Change." American Scientist
82:45-51.
North, W., and M.W. Merkhofer. 1976. "A Methodology for Analyzing Emission Control
Strategies." Comput. Oper. Res. 3:187-207.
Pope, C.A., III, R.T. Burnett, M.J. Thun, E.E. Calle, D. Krewski, K. Ito, and G.D. Thurston.
2002. "Lung Cancer, Cardiopulmonary Mortality, and Long-term Exposure to Fine
Particulate Air Pollution." Journal of the American Medical Association
287:1132-1141.
Reilly, J., P.H. Stone, C.E. Forest, M.D. Webster, H.E. Jacoby, and R.G. Prinn. 2001.
"Climate Change. Uncertainty and Climate Change Assessments." Science
293(5529):430-433.
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. Environmental Protection Agency, Office of
Policy Analysis. EPA-230-09-86-018. Washington, DC.
Russell, M.W., D.M. Huse, S. Drowns, E.C. Hamel, and S.C. Hartz. 1998. "Direct Medical
Costs of Coronary Artery Disease in the United States." American Journal of
Cardiology 81(9): 1110-1115.
Smith, D.H., D.C. Malone, K.A. Lawson, LJ. Okamoto, C. Battista, and W.B. Saunders.
1997. "A National Estimate of the Economic Costs of Asthma." American Journal
of Respiratory and Critical Care Medicine 156(3 Pt l):787-793.
B-42
-------
Stanford, R., T. McLaughlin, and LJ. Okamoto. 1999. "The Cost of Asthma in the
Emergency Department and Hospital." American Journal of Respiratory and Critical
Care Medicine 160(1):211-215.
Tolley, G.S. et al. January 1986. Valuation of Reductions in Human Health Symptoms and
Risks. University of Chicago. Final Report for the U.S. Environmental Protection
Agency.
U.S. Environmental Protection Agency (EPA). 1999. The Benefits and Costs of the Clean
Air Act, 1990-2010. Prepared for U.S. Congress by U.S. EPA, Office of Air and
Radiation/Office of Policy Analysis and Review, Washington, DC, November; EPA
report no. EPA-410-R-99-001.
U.S. Environmental Protection Agency (EPA). September 2000. Guidelines for Preparing
Economic Analyses. EPA 240-R-00-003.
U.S. Office of Management and Budget (OMB). October 1992. "Guidelines and Discount
Rates for Benefit-Cost Analysis of Federal Programs." Circular No. A-94.
U.S. Nuclear Regulatory Commission (NRC). 1996. "Branch Technical Position on the Use
of Expert Elicitation in the High-Level Radioactive Waste Program." November,
1996.
Viscusi, V.K., and IE. Aldy. 2003. "The Value of a Statistical Life: A Critical Review of
Market Estimates Throughout the World." Journal of Risk and Uncertainty
27(l):5-76.
Viscusi, W.K., W.A. Magat, and J. Huber. 1991. "Pricing Environmental Health Risks:
Survey Assessments of Risk-Risk and Risk-Dollar Trade-Offs for Chronic
Bronchitis." Journal of Environmental Economics and Management 21:32-51.
Walker, K.D., J.S. Evans, and D. Macintosh. 2001. "Use of Expert Judgment in Exposure
Assessment. Parti. Characterization of Personal Exposure to Benzene." J Expo
Anal Environ Epidemiol 11(4):308-22.
Walker, K.D., P. Catalano, J.K. Hammitt, and J.S. Evans. 2003. "Use of Expert Judgment in
Exposure Assessment: Part 2. Calibration of Expert Judgments about Personal
Exposures to Benzene." J Expo Anal Environ Epidemiol 13(1): 1-16.
Whitfield, R.G., and T.S. Wallsten. 1989. "A Risk Assessment for Selected Lead-induced
Health Effects: An Example of a General Methodology." Risk Analysis
9(2): 197-207.
B-43
-------
Whitfield, R.G., T.S. Wallsten, R. L. Winkler, H.M. Richmond, and S.R. Hayes. 1991.
Assessing the Risks of Chronic Lung Injury Attributable to Long-Term Ozone
Exposure. Argonne National Laboratory Report ANL/EAIS-2. NTIS/DE91016814.
Argonne, IL. July.
Winkler, R.L., T.S. Wallsten, R.G. Whitfield, H.M. Richmond, S.R. Hayes, and A.S.
Rosenbaum. 1995. "An Assessment of the Risk of Chronic Lung Injury Attributable
to Long-term Ozone Exposure." Operations Research 43(1): 19-28.
Wittels, E.H., J.W. Hay, and A.M. Gotto, Jr. 1990. "Medical Costs of Coronary Artery
Disease in the United States." American Journal of Cardiology 65(7):432-440.
Wright, G., and P. Ayton (eds.). 1994. Subjective Probability. John Wiley, Chichester.
B-44
-------
APPENDIX C
SENSITIVITY ANALYSES OF THE KEY PARAMETERS IN THE BENEFITS
ANALYSIS
The primary analysis presented in Chapter 4 is based on our current interpretation of
the scientific and economic literature. That interpretation requires judgments regarding the
best available data, models, and modeling methodologies and the assumptions that are most
appropriate to adopt in the face of important uncertainties. The majority of the analytical
assumptions used to develop the primary estimates of benefits have been reviewed and
approved by EPA's SAB. Both EPA and the SAB recognize that data and modeling
limitations as well as simplifying assumptions can introduce significant uncertainty into the
benefit results and that alternative choices exist for some inputs to the analysis, such as the
mortality C-R functions.
This appendix supplements our primary estimates of benefits with a series of
sensitivity calculations that use other sources of health effect estimates and valuation data for
key benefits categories. These supplemental estimates examine sensitivity to both valuation
issues (e.g., the appropriate income elasticity) and for physical effects issues (e.g., possible
recovery from chronic illnesses). These supplemental estimates are not meant to be
comprehensive. Rather, they reflect some of the key issues identified by EPA or
commentors as likely to have a significant impact on total benefits. The individual
adjustments in the tables should not simply be added together because 1) there may be
overlap among the alternative assumptions and 2) the joint probability among certain sets of
alternative assumptions may be low.
C.I Premature Mortality—Long-Term Exposure
Reduction in the risk of premature mortality is the most important PM-related health
outcome in terms of contribution to dollar benefits in the analysis for this rule. There are at
least three important analytical assumptions that may significantly impact the estimates of
the number and valuation of avoided premature mortalities. These include selection of the
C-R function, structure of the lag between reduced exposure and reduced mortality risk, and
effect thresholds. Results of this set of sensitivity analyses are presented in Table C-l.
C-l
-------
Table C-l. Sensitivity of Benefits of Premature Mortality Reductions to Alternative
Assumptions (Relative to Primary Estimate Benefits of the Final CAIR)
Description of Sensitivity Analysis
Avoided 1I1CKlellces
2010 |
2015
Value (million 1999$)"
2010 | 2015
Alternative Concentration-Response Functions for PM-Related Premature Mortality
Pope/ACS Study (2002 )c
Lung Cancer
Cardiopulmonary
Krewski/Harvard Six-Cities Study
2,000
9,700
29,000
2,700
13,000
38,000
$11,000
$51,000
$150,000
$14,000
$67,000
$200,000
Alternative Lag Structures for PM-Related Premature Mortality
None
8-year
15-year
Alternative
Segmented
5-Year
Distributed
Exponential
Incidences all occur in the first year
Incidences all occur in the 8th year
3% discount rate
7% discount rate
Incidences all occur in the 1 5th year
3% discount rate
7% discount rate
20 percent of incidences occur in 1st year, 50
percent in years 2 to 5, and 30 percent in
years 6 to 20
3% discount rate
7% discount rate
50 percent of incidences occur in years 1
and 2 and 50 percent in years 2 to 5
3% discount rate
7% discount rate
Incidences occur at an exponentially
declining rate following year of change in
exposure
3% discount rate
7% discount rate
13,000
13,000
13,000
13,000
13,000
13,000
13,000
13,000
13,000
13,000
13,000
17,000
17,000
17,000
17,000
17,000
17,000
17,000
17,000
17,000
17,000
17,000
$74,000
$60,000
$46,000
$49,000
$29,000
$65,000
$52,000
$71,000
$66,000
$71,000
$65,000
$102,000
$83,000
$64,000
$68,000
$40,000
$90,000
$71,000
$97,000
$91,000
$75,000
$90,000
Alternative Thresholds
No Threshold (base estimate)
5 ng/m3
10 ng/m3
15 ng/m3
20 ng/m3
25 u2/m3
13,000
13,000
11,000
780
0
0
17,000
17,000
14,000
690
0
0
$67,000
$67,000
$59,000
$4,100
$0
$0
$93,000
$93,000
$78,000
$3,800
$0
$0
Incidences rounded to two significant digits.
Dollar values rounded to two significant digits.
Note that the sum of lung cancer and Cardiopulmonary deaths will not be equal to the total all-cause death estimate. Some residual
mortality is associated with long-term exposures to PM2 5 that is not captured by the Cardiopulmonary and lung cancer categories.
C-2
-------
C.1.1 Alternative C-R Functions
Following the advice of the most recent EPA SAB-HES, we used the Pope et al.
(2002) all-cause mortality model to derive our primary estimate of avoided premature
mortality (EPA-SAB-COUNCIL-ADV-04-002, 2004). While the SAB-HES "recommends
that the base case rely on the Pope et al. (2002) study and that EPA use total mortality
concentration-response functions (C-R), rather than separate cause-specific C-R functions, to
calculate total PM mortality cases," (EPA-SAB-COUNCIL-ADV-04-002, 2004, p.2) they
also suggested that "the cause-specific estimates can be used to communicate the relative
contribution of the main air pollution related causes of death." (EPA-SAB-COUNCIL-ADV-
04-002, 2004, p. 18) As such, in Table C-l we provide the estimates of cardiopulmonary and
lung cancer deaths based on the Pope et al. (2002) study.
In addition, the SAB-HES noted that the ACS cohort used in Pope et al. (2002) "has
some inherent deficiencies, in particular the imprecise exposure data, and the
nonrepresentative (albeit very large) population" (EPA-SAB-COUNCIL-ADV-04-002, 2004,
p. 18). The SAB-HES suggests that while not necessarily a better study, the ACS is a prudent
choice for the primary estimate. They go on to note that "the Harvard Six-Cities C-R
functions are valid estimates on a more representative, although geographically selected,
population, and its updated analysis has not yet been published. The Six-Cities estimates
may be used in a sensitivity analysis to demonstrate that, with different but also plausible
selection criteria for C-R functions, benefits may be considerably larger than suggested by
the ACS study" (EPA-SAB-COUNCIL-ADV-04-002, 2004, p. 18). In previous advice, the
SAB has noted that "the [Harvard Six-Cities] study had better monitoring with less
measurement error than did most other studies" (EPA-SAB-COUNCIL-ADV-99-012, 1999).
The demographics of the ACS study population (i.e., largely white and middle to upper
middle-class) may also produce a downward bias in the estimated PM mortality coefficient,
because a variety of analyses indicate that the effects of PM tend to be significantly greater
among groups of lower socioeconomic status (Krewski et al., 2000), although the cause of
this difference has not been identified. The Harvard Six-Cities study also covered a broader
age category (25 and older compared to 30 and older in the ACS study). We emphasize that,
based on our understanding of the relative merits of the two datasets, the Pope et al. (2002)
ACS model based on mean PM2 5 levels in 63 cities is the most appropriate model for
analyzing the premature mortality impacts of CAIR. Thus it is used for our primary estimate
of this important health effect.
C-3
-------
C. 1.2 Alternative Lag Structures
Over the last ten years, there has been a continuing discussion and evolving advice
regarding the timing of changes in health effects following changes in ambient air pollution.
It has been hypothesized that some reductions in premature mortality from exposure to
ambient PM25 will occur over short periods of time in individuals with compromised health
status, but other effects are likely to occur among individuals who, at baseline, have
reasonably good health that will deteriorate because of continued exposure. No animal
models have yet been developed to quantify these cumulative effects, nor are there
epidemiologic studies bearing on this question. The SAB-HES has recognized this lack of
direct evidence. However, in early advice, they also note that "although there is substantial
evidence that a portion of the mortality effect of PM is manifest within a short period of
time, i.e., less than one year, it can be argued that, if no lag assumption is made, the entire
mortality excess observed in the cohort studies will be analyzed as immediate effects, and
this will result in an overestimate of the health benefits of improved air quality. Thus some
time lag is appropriate for distributing the cumulative mortality effect of PM in the
population" (EPA-SAB-COUNCIL-ADV-00-001, 1999, p. 9). In recent advice, the SAB-
HES suggests that appropriate lag structures may be developed based on the distribution of
cause-specific deaths within the overall all-cause estimate (EPA-SAB-COUNCIL-ADV-04-
002, 2004). They suggest that diseases with longer progressions should be characterized by
longer-term lag structures, while air pollution impacts occurring in populations with existing
disease may be characterized by shorter-term lags.
A key question is the distribution of causes of death within the relatively broad
categories analyzed in the long-term cohort studies. Although it may be reasonable to
assume the cessation lag for lung cancer deaths mirrors the long latency of the disease, it is
not at all clear what the appropriate lag structure should be for cardiopulmonary deaths,
which include both respiratory and cardiovascular causes. Some respiratory diseases may
have a long period of progression, while others, such as pneumonia, have a very short
duration. In the case of cardiovascular disease, there is an important question of whether air
pollution is causing the disease, which would imply a relatively long cessation lag, or
whether air pollution is causing premature death in individuals with preexisting heart disease,
which would imply very short cessation lags. The SAB-HES provides several
recommendations for future research that could support the development of defensible lag
structures, including using disease-specific lag models and constructing a segmented lag
distribution to combine differential lags across causes of death (EPA-SAB-COUNCIL-ADV-
04-002, 2004). The SAB-HES indicated support for using "a Weibull distribution or a
C-4
-------
simpler distributional form made up of several segments to cover the response mechanisms
outlined above, given our lack of knowledge on the specific form of the distributions" (EPA-
SAB-COUNCIL-ADV-04-002, 2004, p. 24). However, they noted that "an important
question to be resolved is what the relative magnitudes of these segments should be, and how
many of the acute effects are assumed to be included in the cohort effect estimate" (EPA-
SAB-COUNCIL-ADV-04-002, 2004, p. 24-25). Since the publication of that report in
March 2004, EPA has sought additional clarification from this committee. In its followup
advice provided in December 2004, this SAB suggested that until additional research has
been completed, EPA should assume a segmented lag structure characterized by 30 percent
of mortality reductions occurring in the first year, 50 percent occurring evenly over years 2
to 5 after the reduction in PM2 5, and 20 percent occurring evenly over the years 6 to 20 after
the reduction in PM25 (EPA-COUNCIL-LTR-05-001, 2004). The distribution of deaths over
the latency period is intended to reflect the contribution of short-term exposures in the first
year, cardiopulmonary deaths in the 2- to 5-year period, and long-term lung disease and lung
cancer in the 6- to 20-year period. Furthermore, in their advisory letter, the SAB-HES
recommended that EPA include sensitivity analyses on other possible lag structures. In this
appendix, we investigate the sensitivity of premature mortality-reduction related benefits to
alternative cessation lag structures, noting that ongoing and future research may result in
changes to the lag structure used for the primary analysis.
In previous advice from the SAB-HES, they recommended an analysis of 0-, 8-, and
15-year lags, as well as variations on the proportions of mortality allocated to each segment
in the segmented lag structure (EPA-SAB-COlMCIL-ADV-00-001, 1999,
(EPA-COUNCIL-LTR-05-001, 2004). The 0-year lag is representative of EPA's assumption
in previous RIAs. The 8- and 15-year lags are based on the study periods from the Pope et
al. (1995) and Dockery et al. (1993) studies, respectively.1 However, neither the Pope et al.
nor Dockery et al. studies assumed any lag structure when estimating the relative risks from
PM exposure. In fact, the Pope et al. and Dockery et al. analyses do not supporting or refute
the existence of a lag. Therefore, any lag structure applied to the avoided incidences
estimated from either of these studies will be an assumed structure. The 8- and 15-year lags
implicitly assume that all premature mortalities occur at the end of the study periods (i.e., at
8 and 15 years).
'Although these studies were conducted for 8 and 15 years, respectively, the choice of the duration of the study
by the authors was not likely due to observations of a lag in effects but is more likely due to the expense of
conducting long-term exposure studies or the amount of satisfactory data that could be collected during this
time period.
C-5
-------
In addition to the simple 8- and 15-year lags, we have added three additional
sensitivity analyses examining the impact of assuming different allocations of mortality to
the segmented lag of the type suggested by the SAB-HES. The first sensitivity analysis
assumes that more of the mortality impact is associated with chronic lung diseases or lung
cancer and less with acute cardiopulmonary causes. This illustrative lag structure is
characterized by 20 percent of mortality reductions occurring in the first year, 50 percent
occurring evenly over years 2 to 5 after the reduction in PM2 5, and 30 percent occurring
evenly over the years 6 to 20 after the reduction in PM2 5. The second sensitivity analysis
assumes the 5-year distributed lag structure used in previous analyses, which is equivalent to
a three-segment lag structure with 50 percent in the first 2-year segment, 50 percent in the
second 3 -year segment, and 0 percent in the 6- to 20-year segment. The third sensitivity
analysis assumes a negative exponential relationship between reduction in exposure and
reduction in mortality risk. This structure is based on an analysis by Roosli et al. (2004),
which estimates the percentage of total mortality impact in each period t as
% Mortality Reduction (t) = - - - - . (C.I)
The Roosli et al. (2004) analysis derives the lag structure by calculating the rate constant
(-0.5) for the exponential lag structure that is consistent with both the relative risk from the
cohort studies and the change in mortality observed in intervention type studies (e.g., Pope et
al. [1992] and Clancy et al. [2002]). This is the only lag structure examined that is based on
empirical data on the relationship between changes in exposure and changes in mortality.
However, the analysis has not yet been peer reviewed and is thus not yet appropriate for
adoption in the primary analysis.
The estimated impacts of alternative lag structures on the monetary benefits
associated with reductions in PM-related premature mortality (estimated with the Pope et al.
ACS impact function) are presented in Table C-l. These estimates are based on the value of
statistical lives saved approach (i.e., $5.5 million per incidence) and are presented for both a
3 and 7 percent discount rate over the lag period.
C-6
-------
C.1.3 Thresholds
Although the consistent advice from EPA's SAB2 has been to model premature
mortality associated with PM exposure as a nonthreshold effect, that is, with harmful effects
to exposed populations regardless of the absolute level of ambient PM concentrations.
EPA's most recent PM25 Criteria Document concludes that "the available evidence does not
either support or refute the existence of thresholds for the effects of PM on mortality across
the range of concentrations in the studies" (U.S. EPA, 2004, p. 9-44). Some researchers have
hypothesized the presence of a threshold relationship. The nature of the hypothesized
relationship is that there might exist a PM concentration level below which further reductions
no longer yield premature mortality reduction benefits.3
We constructed a sensitivity analysis by assigning different cutpoints below which
changes in PM2 5 are assumed to have no impact on premature mortality. The sensitivity
analysis illustrates how our estimates of the number of premature mortalities in the primary
estimate might change under a range of alternative assumptions for a PM mortality threshold.
If, for example, there were no benefits of reducing PM concentrations below the PM2 5
standard of 15 i-ig/m3, our estimate of the total number of avoided PM-related premature
mortalities in 2015 from the primary analysis would be reduced by approximately 96
percent, from approximately 17,000 annually to approximately 700 annually. The recent
NRC report stated that "for pollutants such as PM10 and PM2 5, there is no evidence for any
departure of linearity in the observed range of exposure, nor any indication of a threshold"
(NRC, 2002, p. 109). At a threshold of 10 |ig, approximately the 20th percentile of observed
concentrations in the Pope et al. (2002) study, mortality impacts would be reduced by only
16 percent to approximately 14,000 annually. Another possible sensitivity analysis that we
2The advice from the 2004 SAB-HES (EPA-SAB-COUNCIL-ADV-04-002) is characterized by the following:
"For the studies of long-term exposure, the HES notes that Krewski et al. (2000) have conducted the most
careful work on this issue. They report that the associations between PM2 5 and both all-cause and
cardiopulmonary mortality were near linear within the relevant ranges, with no apparent threshold.
Graphical analyses of these studies (Dockery et al., 1993, Figure 3, and Krewski et al., 2000, page 162) also
suggest a continuum of effects down to lower levels. Therefore, it is reasonable for EPA to assume a no
threshold model down to, at least, the low end of the concentrations reported in the studies."
3The illustrative example in Appendix B presents the potential implications of assuming some probability of a
threshold on the benefits estimate.
C-7
-------
have not conducted at this time might examine the potential for a nonlinear relationship at
lower exposure levels.4
One important assumption that we adopted for the threshold sensitivity analysis is
that no adjustments are made to the shape of the C-R function above the assumed threshold.
Instead, thresholds were applied by simply assuming that any changes in ambient
concentrations below the assumed threshold have no impacts on the incidence of premature
mortality. If there were actually a threshold, then the shape of the C-R function would likely
change and there would be no health benefits to reductions in PM below the threshold.
However, as noted by the NRC, "the assumption of a zero slope over a portion of the curve
will force the slope in the remaining segment of the positively sloped concentration-response
function to be greater than was indicated in the original study" and that "the generation of the
steeper slope in the remaining portion of the concentration-response function may fully
offset the effect of assuming a threshold." The NRC suggested that the treatment of
thresholds should be evaluated in a formal uncertainty analysis.
C.1.3 Summary of Results
The results of these sensitivity analyses demonstrate that choice of effect estimate can
have a large impact on benefits, potentially doubling benefits if the effect estimate is derived
from the HEI reanalysis of the Harvard Six-Cities data (Krewski et al., 2000). Because of
discounting of delayed benefits, the lag structure may also have a large impact on monetized
benefits, reducing benefits by 30 percent if an extreme assumption that no effects occur until
after 15 years is applied. However, for most reasonable distributed lag structures,
differences in the specific shape of the lag function have relatively small impacts on overall
benefits. For example, the overall impact of moving from the previous 5-year distributed lag
to the segmented lag recommended by the SAB-HES in 2004 in the primary estimate is
relatively modest, reducing benefits by approximately 5 percent when a 3 percent discount
rate is used and 15 percent when a 7 percent discount rate is used. If no lag is assumed,
benefits are increased by around 10 percent relative to the segmented lag with a 3 percent
discount rate and 30 percent with a 7 percent discount rate. Benefits are more sensitive to
assumptions regarding the potential for a threshold. The threshold sensitivity analysis
indicates that over 84 percent of the premature mortality-related benefits are due to changes
4The pilot expert elicitation discussed in Appendix B provides some information on the impact of applying
nonlinear and threshold-based C-R functions.
C-8
-------
in PM25 concentrations occurring above 10 i-ig/m3, and around 5 percent are due to changes
above 15 i-ig/m3, the current PM25 standard.
C.2 Other Health Endpoint Sensitivity Analyses
C.2.1 Ozone-Related Mortality
To provide the reader with a fuller understanding of the health effects associated with
reductions in air pollution associated with the final CAIR, this set of sensitivity estimates
examines the potential impact of reductions in ozone on incidence of premature mortality.
Although PM is the air pollutant most clearly associated with premature mortality, recent
research suggests that short-term ozone exposure likely contributes to premature death.
Although the CAIR analysis uses cohort studies to characterize the effect of PM25 on
premature mortality, such cohort studies have for the most part not documented an
association between ozone exposure and premature mortality. Time series studies, which
look specifically at the effects of short-term exposures, have documented effects on
premature mortality from both PM2 5 and ozone, although results have been mixed. Several
recent analyses have found consistent statistical associations between short-term ozone
exposure and increased mortality (Bell et al., 2004; Fairly et al., 2003; Thurston and Ito,
2001; Toulomi et al., 1997). The most recent of these, Bell et al. (2004), used the extensive
National Morbidity, Mortality, and Air Pollution Study database to examine associations
between ozone and premature mortality in 95 U.S. urban communities. They found that on
average, short-term changes in ozone are significantly associated with premature mortality,
and that the association is robust to adjustment for particulate matter, weather, and
seasonality.
Although they do not constitute a database as extensive as that for PM, these recent
studies provide supporting evidence for including mortality in ozone health benefits
analyses. In a 2001 analysis, Thurston and Ito reviewed previously published time-series
studies examining the effect of daily ozone levels on daily mortality. Thurston and Ito
hypothesized that much of the variability in published estimates of the ozone/mortality effect
could be explained by how well each model controlled for the influence of weather, an
important confounder and that earlier studies, which used less-sophisticated approaches to
controlling for weather, consistently underpredicted the ozone/mortality effect.
Thurston and Ito (2001) also found that models incorporating a nonlinear temperature
specification appropriate for the "U-shaped" nature of the temperature/mortality relationship
(i.e., increased deaths at both very low and very high temperatures) produced ozone/
C-9
-------
mortality effect estimates that were both more strongly positive (an increase in relative risk
over the pooled estimate for all studies evaluated, from 1.036 to 1.056) and consistently
statistically significant. Further accounting for the interaction effects between temperature
and relative humidity strengthened the positive effect. Including a PM index to control for
PM/mortality effects had little effect on these results, suggesting a relationship between
ozone and mortality independent of that for PM. However, most of the studies Thurston and
Ito examined controlled only for PM10 or broader measures of particles and did not directly
control for PM25. As such, there still may be potential for confounding of PM2 5 and ozone
mortality effects, given that ozone and PM2 5 are highly correlated during summer months in
some areas.
Two recent World Health Organization reports concluded that recent epidemiological
studies have strengthened the evidence that there are short-term O3 effects on mortality and
respiratory morbidity (Anderson et al. 2004; WHO, 2003). In addition, Levy et al. (2001)
assessed the epidemiological evidence regarding the link between short-term exposures to
ozone and premature mortality. Based on four U.S. studies (Kellsall et al., 1997;
Moolgavkar et al., 1995; Ito and Thurston, 1996; Moolgavkar, 2000), they concluded that an
appropriate pooled effect estimate is a 0.5 percent increase in premature deaths per 10 |ig/m3
increase in 24-hour average ozone concentrations, with a 95 percent confidence interval
between 0.3 percent and 0.7 percent.
In its September 2001 advisory on the draft analytical blueprint for the second
Section 812 prospective analysis, the SAB HES cited the Thurston and Ito study as a
significant advance in understanding the effects of ozone on daily mortality and
recommended re-evaluation of the ozone mortality endpoint for inclusion in the next
prospective study (EPA-SAB-COUNCIL-ADV-01-004, 2001). Based on these new analyses
and recommendations, EPA sponsored three independent meta-analyses of the
ozone-mortality epidemiology literature to inform a determination on including this
important health endpoint. Publication of these meta-analyses will significantly enhance the
scientific defensibility of benefits estimates for ozone, which include the benefits of
premature mortality reductions.
We estimate ozone mortality for this sensitivity analysis with the recognition that the
exact magnitude of the effects estimate is subject to continuing uncertainty. As we have
done in the sensitivity analyses for prior RIAs, such as for the non-road diesel rule, we used
results from three U.S. studies to calculate the base-case ozone mortality estimate
(Table C-2). We selected these studies (Ito and Thurston, 1996; Moolgavkar et al., 1995;
C-10
-------
Table C-2. Sensitivity Estimates for Ozone-Related Premature Mortality
Description of Sensitivity Analysis
Mortality from short-term ozone exposure13
Ito and Thurston( 1996)
Moolgavkar et al. (1995)
Sametetal. (1997)
Pooled estimate (random effects weights)
Avoided Incidences
2010
290
150
230
190
2015
850
410
630
520
Monetized Value
(Million 1999$)
2010
$1,700
$900
$1,400
$1,100
2015
$5,200
$2,500
$3,800
$3,200
a All estimates rounded to two significant digits.
b Mortality valued using base estimate of $5.5 million per premature statistical death, adjusted for income
growth.
Samet et al., 1997) based on the logic that the demographic and environmental conditions
existing when these studies were conducted would, on average, be most similar (relative to
international studies) to the conditions prevailing when the CAIR would be implemented.
We excluded a fourth U.S. study by Kinney et al. (1995) because, as Levy et al. (2001)
noted, that study included only a linear term for temperature. Because Kinney et al. (1995)
found no significant ozone effect, including this study would lead to an underestimate of true
mortality impacts and increase the uncertainty surrounding the estimated mortality
reductions.
We then estimated the change in mortality incidence resulting from applying the
effect estimate from each study and combined the results using a random-effects weighting
procedure that accounts for both the precision of the individual effect estimates and between-
study variability (see Appendix D for more details on this method for combining results).
However, it is important to note that this procedure only captures the uncertainty in the
underlying epidemiological work and does not capture other sources of uncertainty, such as
that in estimating air pollution exposure (Levy et al., 2001).
Table C-2 shows that if it is assumed that ozone independently affects premature
mortality, then in 2010, an additional 190 premature deaths might be avoided from
reductions in ozone concentrations due to CAIR, while in 2015, an additional 520 deaths
might be avoided. This would add an additional $1.1 billion in monetized benefits in 2010,
and an additional $3.2 billion in monetized benefits in 2015.
C-ll
-------
C.2.2 A her native and Supplementary Estimates
We also examined how the value for individual endpoints or total benefits would
change if we were to make a different assumption about specific elements of the benefits
analysis. Specifically, in Table C-3, we show the impact of alternative assumptions about
other parameters, including treatment of reversals in CB, alternative impact functions for PM
hospital and emergency room admissions, valuation of residential visibility, valuation of
recreational visibility at Class I areas outside of the study regions examined in the Chestnut
and Rowe (1990a, 1990b) study, and valuation of household soiling damages.
Table C-3. Additional Parameter Sensitivity Analyses
Alternative
Calculation
1
2
3
Treatment of
reversals in
CB
Value of
visibility
changes in all
Class I areas
Household
soiling
damage
Description of Estimate
Instead of omitting cases of CB that reverse
after a period of time, they are treated as
being cases with the lowest severity rating.
The number of avoided chronic CB in 2010
increases from 6,900 to 13,000 (88%). The
increase in 2015 is from 8,700 to 16,000
(84%).
Values of visibility changes at Class I areas in
California, the Southwest, and the Southeast
are transferred to visibility changes in Class I
areas in other regions of the country.
Value of decreases in expenditures on
cleaning are estimated using values derived
from Manuel et al. (1983).
Impact on Primary Benefit
Estimate (million 2000$)
2010
+$880
+$200
+$280
2015
+$1,100
+$200
+$340
An important assumption related to chronic conditions is the possible reversal in CB
incidences (row 1 of Table C-3). Reversals are defined as those cases where an individual
reported having CB at the beginning of the study period but reported not having CB in
follow-up interviews at a later point in the study period. Because chronic diseases are long-
lasting or permanent by definition, if the disease abates in a shorter period of time it is not
chronic. However, we have not captured the benefits of reducing incidences of bronchitis
that are somewhere in between acute and chronic. Since chronic bronchitis may be assigned
a range of severities, one way to address this is to treat reversals as cases of CB that are at the
C-12
-------
lowest severity level. These reversals of CB thus are assigned the lowest value for CB in this
sensitivity analysis, rather than omitting reversals as is the case in the primary analysis.
The alternative calculation for recreational visibility (row 2 of Table C-3) is an
estimate of the full value of visibility in the entire region affected by the CAIR emission
reductions. The Chestnut and Rowe (1990a) study from which the primary valuation
estimates are derived only examined WTP for visibility changes in the southeastern portion
of the affected region. To obtain estimates of WTP for visibility changes in the northeastern
and central portion of the affected region, we have to transfer the southeastern WTP values.
This introduces additional uncertainty into the estimates. However, we have taken steps to
adjust the WTP values to account for the possibility that a visibility improvement in parks in
one region is not necessarily the same environmental quality good as the same visibility
improvement at parks in a different region. This may be due to differences in the scenic
vistas at different parks, uniqueness of the parks, or other factors, such as public familiarity
with the park resource. To take this potential difference into account, we adjusted the WTP
being transferred by the ratio of visitor days in the two regions.
The alternative calculation for household soiling (row 5 of Table C-3) is based on the
Manuel et al. (1983) study of consumer expenditures on cleaning and household
maintenance. This study has been cited as being "the only study that measures welfare
benefits in a manner consistent with economic principals" (Desvousges et al., 1998).
However, the data used to estimate household soiling damages in the Manuel et al. study are
from a 1972 consumer expenditure survey and as such may not accurately represent
consumer preferences in 2015. EPA recognizes this limitation, but believes the Manuel et al.
estimates are still useful in providing an estimate of the likely magnitude of the benefits of
reduced PM household soiling.
C.3 Income Elasticity of Willingness to Pay
As discussed in Chapter 4, our estimates of monetized benefits account for growth in
real GDP per capita by adjusting the WTP for individual endpoints based on the central
estimate of the adjustment factor for each of the categories (minor health effects, severe and
chronic health effects, premature mortality, and visibility). We examined how sensitive the
estimate of total benefits is to alternative estimates of the income elasticities. Table C-4 lists
the ranges of elasticity values used to calculate the income adjustment factors, while
Table C-5 lists the ranges of corresponding adjustment factors. The results of this sensitivity
analysis, giving the monetized benefit subtotals for the four benefit categories, are presented
in Table C-6.
C-13
-------
Table C-4. Ranges of Elasticity Values Used to Account for Projected Real Income
Growth3
Benefit Category
Minor Health Effect
Severe and Chronic Health Effects
Premature Mortality
Visibility13
Lower Sensitivity Bound
0.04
0.25
0.08
—
Upper Sensitivity Bound
0.30
0.60
1.00
—
Derivation of these ranges can be found in Kleckner and Neumann (1999) and Chestnut (1997). COI estimates are
assigned an adjustment factor of 1.0.
No range was applied for visibility because no ranges were available in the current published literature.
Table C-5. Ranges of Adjustment Factors Used to Account for Projected Real Income
Growth3
Benefit Category
Minor Health Effect
Severe and Chronic
Health Effects
Premature Mortality
Visibility13
Lower Sensitivity Bound
2010
1.010
1.061
1.019
—
2015
1.015
1.094
1.029
—
Upper Sensitivity Bound
2010
1.074
1.153
1.269
—
2015
1.114
1.241
1.437
—
a Based on elasticity values reported in Table C-4, U.S. Census population projections, and projections of real GDP per
capita.
b No range was applied for visibility because no ranges were available in the current published literature.
C-14
-------
Table C-6. Sensitivity Analysis of Alternative Income Elasticities3
Benefit Category
Minor Health Effect
Severe and Chronic Health
Effects
Premature Mortality
Visibility and Other Welfare
Effects'3
Total Benefits
Benefits in Millions of 1999$
Lower Sensitivity Bound
2010
$640
$3,900
$63,000
$1,100
$68,000
2015
$830
$5,100
$83,000
$1,800
$91,000
Upper Sensitivity Bound
2010
$670
$4,200
$78,000
$1,100
$84,000
2015
$880
$5,600
$120,000
$1,800
$120,000
a All estimates rounded to two significant digits.
b No range was applied for visibility because no ranges were available in the current published literature.
Consistent with the impact of mortality on total benefits, the adjustment factor for
mortality has the largest impact on total benefits. The value of mortality in 2015 ranges from
90 percent to 130 percent of the primary estimate based on the lower and upper sensitivity
bounds on the income adjustment factor. The effect on the value of minor and chronic health
effects is much less pronounced, ranging from 98 percent to 105 percent of the primary
estimate for minor effects and from 93 percent to 106 percent for chronic effects.
C.4 References
Anderson, H.R., R.W. Atkinson, J.L. Peacock, L. Marston, and K. Konstantinou. 2004.
Meta-analysis of Time-series Studies and Panel Studies of Particulate Matter (PM)
and Ozone (O3): Report of a WHO Task Group. Copenhagen: World Health
Organization.
Bell, M.L., A. McDermott, S.L. Zeger, J.M. Samet, and F. Dominici. 2004. "Ozone and
Short-term Mortality in 95 U.S. Urban Communities, 1987-2000." Journal of the
American Medical Association 292:23 72-23 78.
Chestnut, L.G. 1997. "Draft Memorandum: Methodology for Estimating Values for
Changes in Visibility at National Parks." April 15.
C-15
-------
Chestnut, L.G., and R.D. Rowe. 1990a. Preservation Values for Visibility Protection at the
National Parks: Draft Final Report. Prepared for Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC and
Air Quality Management Division, National Park Service, Denver, CO.
Chestnut, L.G., and R.D. Rowe. 1990b. "A New National Park Visibility Value Estimates."
In Visibility and Fine Particles, Transactions of an AWMA/EPA International
Specialty Conference, C. V. Mathai, ed. Air and Waste Management Association,
Pittsburgh.
Clancy, L., P. Goodman, H. Sinclair, and D.W. Dockery. 2002. "Effect of Air-pollution
Control on Death Rates in Dublin, Ireland: An Intervention Study." Lancet Oct
19;360(9341):1210-4.
Desvousges, W.H., F.R. Johnson, and H.S. Banzhaf. 1998. Environmental Policy Analysis
With Limited Information: Principles and Applications of the Transfer Method (New
Horizons in Environmental Economics.) Edward Elgar Pub: London.
Dockery, D.W., C.A. Pope, X.P. Xu, J.D. Spengler, J.H. Ware, M.E. Fay, E.G. Ferris, and
F.E. Speizer. 1993. "An Association between Air Pollution and Mortality in Six
U.S. Cities." New England Journal of Medicine 329(24): 1753-1759.
EPA-SAB-COUNCIL-ADV-00-001. October 1999. The Clean Air Act Amendments
(CAAA) Section 812 Prospective Study of Costs and Benefits (1999): Advisory by the
Health and Ecological Effects Subcommittee on Initial Assessments of Health and
Ecological Effects. Part 2.
EPA-SAB-COUNCIL-ADV-99-012. July 1999. The Clean Air Act Amendments (CAAA)
Section 812 Prospective Study of Costs and Benefits (1999): Advisory by the Health
and Ecological Effects Subcommittee on Initial Assessments of Health and Ecological
Effects. Parti.
EPA-S AB-COUNCIL-ADV-01 -004. September 2001. Review of the Draft Analytical Plan
for EPA 's Second Prospective Analysis—Benefits and Costs of the Clean Air Act
1990-2020: An Advisory by a Special Panel of the Advisory Council on Clean Air
Compliance Analysis.
C-16
-------
EPA-SAB-COUNCIL-ADV-04-002. March 2004. Advisory on Plans for Health Effects
Analysis in the Analytical Plan for EPA 's Second Prospective Analysis—Benefits and
Costs of the Clean Air Act, 1990-2020: Advisory by the Health Effects Subcommittee
of the Advisory Council on Clean Air Compliance Analysis.
Fairley, D. 2003. "Mortality and Air Pollution for Santa Clara County, California, 1989-
1996." In: Revised Analysis of Time Series Studies of Air Pollution and Health.
Special Report. Boston: Health Effects Institute, 97-106.
Ito, K., and G.D. Thurston. 1996. "Daily PM10/Mortality Associations: An Investigations of
At-Risk Subpopulations." Journal of Exposure Analysis and Environmental
Epidemiology 6(l):79-95.
Kellsall, 1, J.M. Samet, and S.L. Zeger. 1997. "Air Pollution and Mortality in Philadelphia,
1974-1988." American Journal of Epidemiology 146:750-762.
Kinney, P.L., K. Ito, and G.D. Thurston. 1995. "A Sensitivity Analysis of Mortality PM-10
Associations in Los Angeles." Inhalation Toxicology 7(l):59-69.
Kleckner, N., and J. Neumann. June 3, 1999. "Recommended Approach to Adjusting WTP
Estimates to Reflect Changes in Real Income." Memorandum to Jim Democker, US
EPA/OPAR.
Krewski, D., R.T. Burnett, M.S. Goldbert, K. Hoover, J. Siemiatycki, M. Jerrett, M.
Abrahamowicz, and W.H. White. July 2000. Reanalysis of the Harvard Six Cities
Study and the American Cancer Society Study of Paniculate Air Pollution and
Mortality. Special Report to the Health Effects Institute, Cambridge MA.
Levy, J.L., Carrothers, T.J., J.T. Tuomisto, J.K. Hammitt, and J.S. Evans. 2001. "Assessing
the Public Health Benefits of Reduced Ozone Concentrations." Environmental
Health Perspectives 109:1215-1226.
Manuel, E.H., R.L. Horst, K.M. Brennan, W.N. Lanen, M.C. Duff, and J.K. Tapiero. 1983.
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.
Moolgavkar, S.H., E.G. Luebeck, T.A. Hall, and E.L. Anderson. 1995. "Air Pollution and
Daily Mortality in Philadelphia." Epidemiology 6(5):476-484.
C-17
-------
Moolgavkar, S.H. 2000. "Air Pollution and Hospital Admissions for Diseases of the
Circulatory System in Three U.S. Metropolitan Areas." Journal of the Air and Waste
Management Association 50:1199-1206.
National Research Council (NRC). 2002. Estimating the Public Health Benefits of
Proposed Air Pollution Regulations. The National Academies Press: Washington,
D.C.
Pope, C.A. Ill, J. Schwartz, and M.R. Ransom. 1992. "Daily Mortality and PM10 Pollution
in Utah Valley." Arch Environ Health 47(3):211-217.
Pope, C.A., III, MJ. 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." American Journal of Respiratory Critical Care
Medicine 151:669-674.
Pope, C.A., III, R.T. Burnett, MJ. Thun, E.E. Calle, D. Krewski, K. Ito, and G.D. Thurston.
2002. "Lung Cancer, Cardiopulmonary Mortality, and Long-term Exposure to Fine
Particulate Air Pollution." Journal of the American Medical Association
287:1132-1141.
Roosli, M., N. Kunzli, and C. Braun-Fahrlander. August 1-4, 2004. "Use of Air Pollution
'Intervention-Type' Studies in Health Risk Assessment." 16th Conference of the
International Society for Environmental Epidemiology, New York.
Samet, J.M., S.L. Zeger, I.E. Kelsall, J. Xu, and L.S. Kalkstein. March 1997. Air Pollution,
Weather, and Mortality in Philadelphia 1973-1988. Cambridge, MA: Health Effects
Institute.
Thurston, G.D., and K. Ito. 2001. "Epidemiological Studies of Acute Ozone Exposures and
Mortality." JExpo Anal Environ Epidemiol. ll(4):286-94.
Toulomi, G., K. Katsouyanni, D. Zmirou, and J. Schwartz. 1997. "Short-term Effects of
Ambient Oxidant Exposure on Mortality: A Combined Analysis within the APHEA
Project." American Journal of Epidemiology 146:177-183.
U.S. Environmental Protection Agency (EPA). 2004. Air Quality Criteria for Particulate
Matter, Volume II. Office of Research and Development. EPA/600/P-99/002bF,
October 2004.
C-18
-------
World Health Organization (WHO). 2003. Health Aspects of Air Pollution with Particulate
Matter, Ozone, and Nitrogen Dioxide: Report on a WHO Working Group.
EUR/03/5042688. Bonn, Germany: World Health Organization.
C-19
-------
APPENDIX D
SENSITIVITY ANALYSES OF KEY PARAMETERS IN THE COST AND
ECONOMIC IMPACT ANALYSIS AND A LISTING OF IPM RUNS IN SUPPORT
OF CAIR
This appendix presents results of sensitivity analyses using the IPM with alternative
assumptions for the price of natural gas and the growth rate of electricity demand. In
addition, a list of the IPM runs that were used in the various analyses done in support of the
final CAIR is provided. Model output from each of the IPM runs listed in this memo is
available in the CAIR docket and also on EPA's Web site at www.epa.gov/airmarkets/
epa-ipm.
EPA uses IPM to estimate costs and, more broadly, analyze the projected impact of
air emission control policies on the electric power sector in the 48 contiguous states and the
District of Columbia. IPM is a multiregional, dynamic, deterministic linear programming
model of the U.S. electric power sector. IPM documentation is available in the CAIR docket
and also on EPA's Web site at www.epa.gov/airmarkets/epa-ipm/.
Modeling applications of IPM produce forecasts for model plants (i.e., clusters of
real-life EGUs with similar characteristics). The model plant projections can be used to
produce parsed results, which are unit-level results derived from the model plant projections.
Projections for individual plants are based on data currently available and modeling
parameters that are simplifications of the real world. It is likely that some future actions
regarding individual plants could differ from model projections of actions. However, the
aggregate impacts are expected to be appropriately characterized by the model. Where
appropriate, EPA produced parsed results from IPM runs for use in analyzing the air quality
impacts of CAIR.
D.I Effects of Change in Assumptions for Natural Gas Prices and Electricity Growth
Sensitivity analyses were performed using projections from the 2004 Annual Energy
Outlook produced by the Energy Information Administration (EIA). EPA used EIA
estimates for the difference between natural gas prices and coal prices, which we have short-
handed as "EIA natural gas prices," as well as EIA's projection of electricity growth. These
D-l
-------
particular assumptions involve considering the higher differential between minemouth coal
and wellhead natural gas prices. For the years 2010, 2015, and 2020, there was a higher
differential of $0.25 mmBtu, $0.42 mmBtu, and $0.38 mmBtu, respectively. The electricity
growth was changed to match EIA's growth of 1.8 percent a year rather than EPA's growth
of 1.6 percent.
Total annual regional costs of CAIR with EIA assumptions are in Table D-l. The
costs of CAIR with EIA assumptions for natural gas prices and electricity growth in 2010
and 2015 are only slightly different from costs of CAIR without those assumptions and can
be attributed to the building of new and cleaner coal-fired capacity that leads to lower overall
costs (see Tables D-l and D-2). As demand continues to grow, coal-fired generation
continues to increase and requires the use of additional scrubbers. Although more pollution
controls are installed using EIA assumptions, dispatch changes lead to the use of more
efficient generation. The power sector is less inclined to use gas as a compliance option in
the region because of the higher operating cost. Once the power sector passes the point
where there is no longer excess gas capacity in the marketplace (as currently exists), new
coal-fired capacity is the logical choice to meet demand. This new capacity would be built
inside and outside the CAIR region.
The annualized regional cost of CAIR, as presented in Table D-l, is EPA's best
assessment of the cost of implementing CAIR, assuming that States adopt the model cap and
trade program. These costs are generated from rigorous economic modeling of changes in
the power sector due to CAIR. This type of analysis using IPM has undergone peer review
and been upheld in Federal courts. The direct cost includes, but is not limited to, capital
investments in pollution controls, operating expenses of the pollution controls, investments
in new generating sources, and additional fuel expenditures. EPA believes that these costs
reflect, as closely as possible, the additional costs of CAIR to industry. The relatively small
cost associated with monitoring emissions for affected sources is not included in the
annualized cost, but EPA has done a separate analysis and estimated the cost to be less than
$42 million (see Section X. B. Paperwork Reduction Act). However, there may exist certain
costs that EPA has not quantified in these estimates. These costs may include costs of
transitioning to CAIR, such as the costs associated with the retirement of smaller or less
efficient electricity generating units, employment shifts as workers are retrained at the same
company or re-employed elsewhere in the economy, and certain relatively small permitting
costs associated with Title IV that new program entrants face. Although EPA has not
quantified these costs, the Agency believes that they are small compared to the quantified
costs of the program on the power sector. The annualized cost estimates are the best and
D-2
-------
Table D-l. Annual Regional Costs of CAIR with EPA and EIA Assumptions for
Natural Gas Prices and Electric Growth (Billion $1999)
Year
EPA Assumptions
EIA Assumptions
2010
2015
2020
$2.4
$3.6
$4.4
$2.6
$3.4
$4.1
Source: Integrated Planning Model run by EPA.
Table D-2. Incremental Pollution Controls under CAIR with EPA and EIA
Assumptions for Natural Gas and Electricity Growth (Incremental GWs)
EPA Assumptions
Technology
FGD
SCR
2010
37
14
2015
64
34
2020
82
33
EIA Assumptions
2010
45
18
2015
69
39
2020
92
40
Source: Integrated Planning Model run by EPA.
most accurate based upon available information. At the macroeconomic level, the indirect
costs and impacts of higher electricity prices on the entire economy are presented in
Appendix E of the Regulatory Impact Analysis.
The marginal cost to remove additional tons of SO2 and NOX increases slightly with
EIA assumptions for natural gas and electricity growth, as more controls are installed on
coal-fired units that are slightly more expensive to control. Table D-3 compares the result of
sensitivity analysis to the CAIR case with EPA assumptions.
Table D-4 shows nationwide emissions of SO2 and NOX using EIA assumptions.
Coal-fired generation under CAIR increases using EIA assumptions for natural gas prices
and electricity growth. Table D-5 shows the generation mix with EIA assumptions.
D-3
-------
Table D-3. Marginal Cost of SO2 and NOX Reductions under CAIR with EPA and EIA
Assumptions for Natural Gas Prices and Electric Growth ($/ton, in $1999)
SO2
NOx
EPA Assumptions
EIA Assumptions
EPA Assumptions
EIA Assumptions
2010
$700
$800
$1,300
$1,400
2015
$1,000
$1,200
$1,600
$1,700
2020
$1,400
$1,500
$1,600
$1,700
Source: Integrated Planning Model run by EPA.
Table D-4. Projected Nationwide Emissions of SO2 and NOX under CAIR with EPA
and EIA Assumptions for Natural Gas and Electric Growth (Million Tons)
Base Case with EPA Assumptions
CAIR with EPA Assumptions
Base Case with EIA Assumptions
CAIR with EIA Assumptions
2010
9.7
6.1
9.7
6.1
SO2
2015
8.9
4.9
8.8
5.0
2020
8.6
4.2
8.6
4.0
NOX
2010
3.6
2.4
3.7
2.4
2015
3.7
2.1
3.7
2.1
2020
3.7
2.1
3.8
2.2
Source: Integrated Planning Model run by EPA.
D-4
-------
Table D-5. Generation Mix under CAIR with EPA and EIA Assumptions for Natural
Gas and Electric Growth (Thousand GWhs)
Base
Case
CAIR
Fuel
Coal
Oil/Natural Gas
Other
Total
Coal
Oil/Natural Gas
Other
Total
EPA
2010
2,198
111
1,223
4,198
2,163
809
1,218
4,190
Assumptions
2015
2,242
1,026
1,235
4,503
2,195
1,072
1,233
4,499
2020
2,410
1,221
1,218
4,850
2,381
1,250
1,217
4,847
EIA
2010
2,243
902
1,224
4,369
2,228
916
1,223
4,367
Assumptions
2015
2,638
867
1,235
4,739
2,632
871
1,234
4,738
2020
3,048
873
1,224
5,145
3,045
874
1,221
5,141
Note: Numbers may not add due to rounding.
Source: Integrated Planning Model run by EPA.
Coal production patterns change slightly and production for all three major coal-
producing regions is higher, because coal-fired generation is a cheaper source of electricity
than natural gas in most parts of the country with the higher EIA prices, even as more
pollution controls are added to coal-fired generation and used to meet the additional
electricity demand (see Table D-6).
Electricity prices are not greatly altered with EIA assumptions for natural gas and
electricity growth (see Tables D-7 and D-8). For the CAIR region, average electricity prices
are projected to be lower than current levels (2000) using both EPA and EIA assumptions for
natural gas and electricity growth.
The IPM sensitivities listed in this appendix (Table D-9) that are not discussed in this
document were also used in support of CAIR, and relevant discussion can be found in the
appropriate section of the CAIR preamble.
D-5
-------
Table D-6. Coal Production for the Electric Power Sector under CAIR with EPA and
EIA Assumptions for Natural Gas and Electricity Growth (Million Tons)
Base
Case
CAIR
Supply
Area
Appalachia
Interior
West
National
Appalachia
Interior
West
National
2000
299
131
475
905
299
131
475
905
2003
275
135
526
936
275
135
526
936
EPA
2010
325
161
603
1,089
306
164
607
1,077
Assumptions
2015
315
162
631
1,109
310
193
579
1,082
2020
301
173
714
1,188
331
219
607
1,156
EIA
2010
328
161
626
1,115
320
174
614
1,109
Assumptions
2015
341
182
748
1,271
367
207
676
1,250
2020
340
247
840
1,428
390
260
765
1,415
Source: 2000 and 2003 data are from EIA. All projections are from the Integrated Planning Model run by EPA.
Table D-7. Retail Electricity Prices by NERC Region with the Base Case (No Further
Controls) and with CAIR Using EPA Assumptions for Natural Gas and Electricity
Growth (Mills/kWh)
Power
Region
ECAR
ERCOT
MAAC
MAIN
MAPP
NY
NE
FRCC
STV
SPP
Primary States Included
OH, MI, IN, KY, WV, PA
TX
PA, NJ, MD, DC, DE
IL, MO, WI
MN, IA, SD, ND, NE
NY
VT, NH, ME, MA, CT, RI
FL
VA, NC, SC, GA, AL, MS, TN, AR, LA
KS, OK, MO
Regionwide
2000
57.4
65.1
80.4
61.2
57.4
104.3
89.9
67.9
59.3
59.3
66.0
Base Case
2010
51.7
57.9
59.3
52.6
52.8
82.8
77.4
71.2
56.2
54.2
58.0
2015
55.2
64.4
69.4
57.8
49.3
87.9
83.9
71.3
55.1
57.0
60.8
2020
56.1
62.6
72.2
61.0
47.6
88.1
82.8
69.5
55.3
56.7
61.0
CAIR
2010
53.8
59.3
61.2
54.0
52.9
83.3
77.5
71.7
57.0
54.6
59.2
2015
58.5
64.6
71.7
60.3
49.6
88.8
84.7
72.3
56.2
57.5
62.4
2020
58.0
63.3
72.8
62.0
48.0
88.4
83.0
70.5
56.6
57.0
62.1
Percent Change
2010
4.0%
2.5%
3.2%
2.6%
0.2%
0.5%
0.1%
0.8%
1.5%
0.7%
2.0%
2015
5.9%
0.2%
3.4%
4.3%
0.7%
1.0%
1.0%
1.3%
2.1%
0.9%
2.7%
2020
3.4%
1.2%
0.8%
1.7%
0.8%
0.3%
0.2%
1.5%
2.3%
0.6%
1.8%
Source: Retail Electricity Price Model run by EPA. 2000 prices are fromEIA's AEO 2003.
D-6
-------
Table D-8. Retail Electricity Prices by NERC Region with the Base Case (No Further
Controls) and with CAIR Using EIA Assumptions for Natural Gas and Electricity
Growth (Mills/kWh)
Power
Region
ECAR
ERCOT
MAAC
MAIN
MAPP
NY
NE
FRCC
STV
SPP
Primary States Included
OH, MI, IN, KY, WV, PA
TX
PA, NJ, MD, DC, DE
IL, MO, WI
MN, IA, SD, ND, NE
NY
VT, NH, ME, MA, CT, RI
FL
VA, NC, SC, GA, AL, MS, TN, AR, LA
KS, OK, MO
Regionwide
2000
57.4
65.1
80.4
61.2
57.4
104.3
89.9
67.9
59.3
59.3
66.0
Base Case
2010
53.5
63.3
63.1
54.9
52.9
89.0
85.1
72.5
57.1
56.2
60.5
2015
59.8
66.0
74.7
63.8
49.6
91.3
85.5
74.6
57.1
59.5
64.0
2020
57.1
64.4
72.8
62.4
48.1
87.8
81.2
73.7
57.1
57.9
62.5
CAIR
2010
55.3
63.6
64.0
55.9
53.1
89.1
84.7
73.3
57.8
56.7
61.3
2015
61.5
66.6
75.4
65.2
49.9
91.9
85.9
75.3
58.3
59.7
65.1
2020
58.8
65.0
73.7
63.3
48.6
88.8
81.8
74.3
58.6
58.1
63.6
Percent Change
2010
3.5%
0.5%
1.4%
1.7%
0.4%
0.1%
-0.4%
1.1%
1.1%
0.9%
1.4%
2015
3.0%
0.8%
1.0%
2.2%
0.6%
0.6%
0.5%
0.9%
2.0%
0.3%
1.6%
2020
3.0%
0.9%
1.3%
1.5%
0.9%
1.1%
0.9%
0.9%
2.6%
0.3%
1.8%
Source: Retail Electricity Price Model run by EPA. 2000 prices are fromEIA's AEO 2003.
D-7
-------
Table D-9. Listing of Runs from the Integrated Planning Model Used in Analyses Done
in Support of the CAIR Final Rule Analyses
Run Name
Base Case 2004
CAIR 2004_Analysis
CAIR 2004_Final
CAIR 2004_Final_DE and NJ
BART 2004_Nationwide
CAIR + BART 2004
CAIR 2004_CSP
Base Case 2004_EIA
CAIR 2004_EIA
CAIR 2004_EIA_One Phase
CAIR 2004_EIA_SCR Costs
CAIR 2004_SCR Bypass_NOx SIP Call
Run Description
Base case model ran, which includes the national Title IV
SO2 cap-and-trade program; NOX SIP Call regional ozone
season cap-and-trade program; and state-specific programs
in Connecticut, Illinois, Maine, Massachusetts, Minnesota,
Missouri, New Hampshire, New York, North Carolina,
Oregon, Texas, and Wisconsin. This ran represents
conditions without the proposed CAIR.
CAIR control strategy used for much of the analytical work
for the final CAIR (includes AR/DE/NJ for annual controls
and no ozone season cap and is the IPM ran used for air
quality modeling)
Final CAIR policy (includes annual and ozone season caps
for the States who contribute to PM2.5 and/or ozone
nonattainment)
CAIR Final policy with DE and NJ included in the annual
program (based off CAIR 2004_Final)
Nationwide BART control strategy
CAIR Analysis control strategy, with BART requirements in
non-CAIR states
CAIR Final policy with annual NOX compliance supplement
pool
Base Case run with EIA assumptions for the difference
between natural gas prices and coal prices, as well as EIA's
projection of electricity growth
CAIR Analysis ran with EIA assumptions for the difference
between natural gas prices and coal prices, as well as EIA's
projection of electricity growth
CAIR Analysis ran with EIA assumptions for the difference
between natural gas prices and coal prices, EIA's projection
of electricity growth, and Phase II caps pushed to 2010
CAIR Analysis ran with EIA assumptions for the difference
between natural gas prices and coal prices, EIA's projection
of electricity growth, and SCR capital costs and fixed O&M
costs scaled up 30 percent
CAIR Analysis ran with additional SCR bypass cost for
existing NOX SIP Call units only
(Continued)
D-8
-------
Table D-9. Listing of Runs from the Integrated Planning Model Used in Analyses Done
in Support of the CAIR Final Rule Analyses (continued)
Run Name
CAIR 2004_No NOX
CAIR 2004_No SO2
CAIR 2004_No SO2_Summer NO^l
CAIR 2004_No SO2_ Summer NOL2
BART 2004_No NOX
BART 2004_No SO2
CAIR 2004_No Retirement Ratios
Run Description
CAIR Analysis SO2 policy, with base case NOX
CAIR Analysis NOX policy, with base case SO2
CAIR NOX control during ozone season only in all CAIR
states, with base case SO2
CAIR NOX control during ozone season in 8-hour ozone
states, with base case SO2
Nationwide BART SO2 limits, with base case NOX
Nationwide BART NOX limits, with base case SO2
CAIR Analysis run but with an SO2 cap rather than use of
Title IV allowance retirement ratios for use in CAIR
Parsed Files
EPA base case parsed for year 20 10
EPA base case parsed for year 2015
EPA base case parsed for year 2020
EPA CAIR parsed for year 20 10
EPA CAIR parsed for year 2015
EPA CAIR parsed for year 2020
EPA BART parsed for year 2015
EPA CAIR + BART parsed for year 2015
D-9
-------
APPENDIX E
CAIR INDUSTRY-SECTOR IMPACTS
EPA estimates the direct costs of implementing CAIR at $3.6 billion in 2015 in the
CAIR region. Given the impact of this rule on electricity generators, we believe it is
important to gauge the extent to which the rule might affect other industry sectors. To do so,
we conducted a limited analysis of the economy-wide effects of implementing CAIR.
We were particularly interested in learning how anticipated changes in electricity
prices might affect industry sectors that are large electricity users. The models we employed
indicated those impacts would be small, even without incorporating the beneficial economic
effects of CAIR-related air quality improvements such as improved worker health and
productivity. Rather, our analyses continue to show that the value of even the limited subset
of CAIR benefits we were able to quantify substantially outweigh implementation costs. The
degree to which projected benefits exceed projected costs would be even greater if we were
able to include a number of other beneficial effects, such as a reduction in acid rain damage
and lowering of nitrogen deposition.
By focusing only on cost-side spillover effects on the economy, the industry-sector
impacts projected by our macroeconomic models are likely overstated, primarily because the
positive market impacts of the CAIR on labor availability and productivity are excluded. In
this regard, an independent panel of experts has encouraged EPA to work toward
incorporating both beneficial and costly effects when modeling the economy-wide
consequences of regulation. EPA is actively working to develop this capability.
Although neither model has yet been configured to include the indirect economic
benefits of air quality improvements, EPA employed two distinct computable general
equilibrium models to gauge the potential magnitude of the economy-wide effects of CAIR
implementation costs. The first model, known as Intertemporal General Equilibrium Model
(IGEM), has a long track record and was used by the Agency for the first of the two Clean
Air Act Section 812 studies. The other model, called EMPAX-CGE, is currently in peer
review and has the advantage of disaggregating the U.S. into multiple regions. As with all
models, these tools have their respective strengths and weaknesses, and differences in data
and choice of functional form imply that the models are likely to show slightly different
E-l
-------
results. Despite the differences between the models, the results of the respective analyses
show similarly small impacts of CAIR on energy intensive industries. For example,
production changes for the chemical manufacturing industry are estimated at -0.01 percent to
-0.04 percent in 2010. Furthermore, neither model was configured to capture the beneficial
economic consequences of the increased labor availability and productivity expected to result
from CAIR-related air quality improvements. If these labor productivity improvements were
included, the small production output decreases projected by both models might be partially
or entirely offset. EPA continues to investigate the feasibility of incorporating labor
productivity gains and other beneficial effects of air quality improvements in computable
general equilibrium models. The individual analyses of CAIR by the two general
equilibrium models follow.
E.I IGEM Economy-wide Analysis of CAIR: Analysis of Electric-Sector Impacts
CAIR is designed to improve air quality in nonattainment areas by achieving
reductions in SO2 and/or NOX emissions from the electric power industry in 29 eastern states
and the District of Columbia.
EPA has modeled the macroeconomic impacts of the electric-sector changes expected
from this rule. This analysis used the IGEM developed by Dale Jorgenson, Peter Wilcoxen,
and Mun Sing Ho and maintained by Dale Jorgenson Associates. This appendix discusses
the economy-wide model and the approach used to analyze the rule, along with the results of
this modeling.
IGEM is a dynamic computable general equilibrium model of the U.S. economy. The
model is a stylized representation of the entire economy, in which supply, demand, prices,
and quantities for goods and services reach equilibria for each year of the simulation's time
horizon. IGEM represents the U.S. economy as 35 distinct sectors, roughly corresponding to
the two-digit levels of the North American Industry Classification System (NAICS). This
level of disaggregation allows for insight into the economy-wide effects of policies that
directly affect only a limited number of sectors.
The model has been used in peer-reviewed academic studies and in government,
private-sector, and nonprofit policy analyses, including the first study of the Clean Air Act
under Section 812. In the 812 process, the model was subjected to peer review by the EPA's
SAB.
E-2
-------
E.1.1 Modeling Approach
EPA believes that the best tool currently available to analyze rules and policies
affecting the electric power sector is IPM developed by ICF Consulting. (For information on
this model, see www.epa.gov/airmarkets/epa-ipm.) IPM is a highly disaggregated model of
the electric utility industry that provides a much more detailed view of changes in the sector
than would be possible in the framework of an economy-wide model such as IGEM.
However, IPM currently has no representation of other sectors and cannot analyze the
economy-wide impacts of the rule.
EPA has chosen to use the results of IPM regarding electric sector costs and fuel
quantity changes and introduce them as inputs into IGEM. Introducing these changes as
inputs causes IGEM to find new equilibria, as the relative prices of factor inputs adjust and
economic agents change their demand for goods and services. A change of this nature in an
economy-wide model shows that, compared to a reference case (base case), some sectors
experience increased demand for their outputs, while other sectors face reduced demand.
The model accounts for the overall changes in economic activity, reporting the respective
impacts on the output of each of the 35 individual producing industries, as well as on
consumer prices, labor supply and demand, and GDP. IGEM, therefore, complements the
sector-specific IPM analysis by showing how the changes in the electric sector affect the
other sectors of the economy in some detail.
Please note that this analysis only accounts for CAIR impacts to the electric power
sector itself as modeled by IPM. It does not account for other economic effects, including
the substantial economic benefits of reduced emissions of atmospheric pollutants. The
annual benefits to which EPA can assign a dollar valuation are estimated to be $73.3 billion
in 2010 and $60.4 billion (3 percent and 7 percent discount rates, respectively) in 2010 and
$101 billion or $86.3 billion (3 percent and 7 percent discount rates, respectively) in 2015.
There are additional benefits, including environmental benefits and some health
improvements for which EPA could not estimate dollar values. See Chapter 4 for more
information on benefits analyses. The dollar figures cited above and throughout this
appendix are in 1999 dollars.
E.I.2 Modeling Methodology
Economic models have different embedded assumptions and model structures; thus, it
is always challenging to link two models together. Because IGEM is an aggregate,
econometrically derived model of the entire U.S. economy and IPM is a detailed, technology
E-3
-------
rich model of a single sector, these challenges are certainly present. EPA developed the
methodology of linking the IPM outputs as IGEM inputs by working closely with Dale
Jorgenson Associates.
The approach used in the present analysis consists of two steps. The first is to use
IPM to calculate the total incremental resource costs of C AIR to the electric sector compared
to a reference case. These costs include changes in capital costs, operating and maintenance
(O&M) costs, and fuel costs associated with electricity generation. The incremental costs
then are added to the resource cost of inputs for the IGEM electricity sector. The simulation
shows that these additional costs result in changes in productivity and, hence, lower supply
of and demand for electricity in the rest of the economy. For businesses, governments, and
households, electricity becomes relatively more expensive, and these "consumers" adjust
their purchasing behaviors, substituting away from electricity.
The second step is to account for the changes IPM projects for the coal sector. More
than 90 percent of coal produced in the United States is consumed in the electric sector
(DOE, 2004). EPA considers the IPM projections of coal production for the power sector to
be sufficiently accurate to incorporate them into the economy-wide representation of IGEM.
These impacts are introduced into IGEM by adjusting the productivity of the coal sector to
account for the IPM projection of the quantity change of coal consumed. The specific
percentage change to the IGEM coal-sector productivity is calculated as the percentage
reduction in coal consumption projected by IPM, multiplied by the fraction of total coal
output consumed by the electric sector. This adjustment changes the price of coal, and all
sectors then adjust, demanding relatively less coal.
Because of the differences in underlying data, model structures, and estimated
parameters between IGEM and IPM, it is not possible to match both the price and quantity
changes arising in each methodology. EPA and Dale Jorgenson Associates believe that more
meaningful estimates for the economy arise by explicitly matching the coal quantity impacts
projected by IPM rather than IPM's projected price impacts. Furthermore, although the
majority of coal produced in the United States is consumed in the electric sector, this is not
the case for natural gas. Therefore, IPM projections of gas usage for the electric sector are
not necessarily indicative of the economy-wide demand for this fuel, and EPA has chosen
not to model the impacts of CAIR on natural gas using the same approach as that used for
coal. Rather, IGEM is allowed to solve for all changes in natural gas markets as the
economy responds to the rule's impacts on electricity and coal markets.
E-4
-------
E.I.3 Projected Impacts on Specific Industries
The productivity changes incurred by the electric sector and coal sector contribute to
changes in output for virtually all sectors. This happens because, as noted above, relative
price levels for electricity and for energy sources which may substitute for electricity change
under CAIR. In this simulation, consumers demand for energy intensive goods tends, in
general, to fall slightly, while demand for other goods tends to rise.
For the vast majority of the 35 IGEM sectors, the impact is less than 0.05 percent
(five one hundredths of 1 percent) of forecasted domestic output measured in sales and may
be positive, representing increased production, or negative, representing loss of output.
For some sectors, however, the results are more pronounced. IGEM shows increases
in natural gas output, because natural gas is substituted for coal and electricity; this increase
is 0.16 percent in 2010 and 0.20 percent in 2015. Other sectors benefit as well, as consumers
substitute away from energy consumption to other products, particularly agriculture and
food.
Sectors that face relatively larger output changes also include the electric and coal
sectors. The loss in the coal sector is very similar to that predicted by IPM, at 1.27 percent
in 2010 and 1.63 percent in 2015. The electric sector itself faces a loss of output of 1.01
percent in 2010 and 1.27 percent in 2015. Other energy-intensive industries also lose
output, ranging from 0.01 percent for the chemical industry to 0.10 percent for petroleum
refining. Figure E-l details the 2015 results. (The complete results for 2010 and 2015 for
each of the 35 sectors represented in IGEM are included later in this appendix in
Section E.I.9.)
E. 1.4 Projected Impacts on Consumer Prices
Indices of consumer prices are well-accepted measures of price levels within an
economy. The indices increase when consumers face higher prices for the goods they
purchase. These indices are calculated as weighted-average prices for particular "baskets" of
goods commonly purchased by households. Within IGEM, "the basket" is based on
household purchases of the 35 commodities available in its various markets for goods and
services. The electric sector impacts of the CAIR have effects on IGEM's aggregate
consumer price level. There is a direct effect as electricity becomes relatively more
expensive and an indirect effect as industries that consume electricity face a higher price for
E-5
-------
Agriculture, forestry, fisheries
Metal mining
Coal mining
Crude oil and gas extraction
Non-metallic mineral mining
C onstruction
Food and kindred products
Tobacco manufactures
Textile mill products
Apparel and other textile products
Lumber and wood products
Furniture and fixtures
Paper and allied products
Printing and publishing
Chemicals and allied products
Petroleum refining
Rubber and plastic products
Leather and leather products
Stone, clay and glass products
Primary metals
Fabricated metal products
N on-electrical machinery
Electrical machinery
Motor vehicles
Other transportation equipment
Instruments
Miscellaneous manufacturing
Transportation and warehousing
C ommunications
Electric utilities (services)
Gas utilities (services)
Wholesale and retail trade
Finance, insurance and real estate
Personal and business services
Government enterprises
-5.0% -4.0% -3.0% -2.0% -1.0% 0.0%
Percentage Change from Reference Case
1.0%
Source: Intertemporal General Equilibrium Model
Figure E-l. Impact on Domestic Output, by Sector, 2015
E-6
-------
a factor input, and then pass on that price increase by charging higher prices for goods and
services. Figure E-2 shows the percentage change in consumer prices (approximately 0.032
percent in 2010 and 0.036 percent 2015) over the reference case.
Percent Change From Reference Case
0.10% -,
008%
On^%
0.04% -
On? %.
Onn% -i
n n? %.
n HJ%
n n^%
n 08%.
-n 10% -
• • •
2010
2015
Source: Intertemporal General Equilibrium Model
Figure E-2. Change in Consumer Prices Compared to Reference Case
E. 1.5 Projected Impacts on Labor Markets
IGEM projects changes in labor utilization under the CAIR scenario (see Figure E-3).
Note that this is not an explicit statement about employment or jobs, as the IGEM model
does not calculate estimates of employment or job creation. Rather, the model accounts for
labor supply and demand in terms of "hours." An increase in "labor input" associated with
the CAIR implies an increase in the number of quality-adjusted hours supplied by
households and demanded by employers. In the IGEM framework, as energy prices rise,
firms substitute other inputs, which include capital, materials, and labor. For details on the
representation of labor within IGEM, and information on interpreting the results of the
model, please see Section E. 1.8. The change in labor input is less than 0.01 percent in 2010
and 2015.
E-7
-------
0>
ISl
«
u
0>
u
=
0>
£
u
«
s
o
£
u
W)
=
«
f
U
•^
u
u
u
PH
0.05%
On A o/
On i *>A
On 9 *>A
On 1 o/
On 1 o/
On 9 *>A
On i *>A
Ond«/«
0.05%
• I
2010 2015
Source: Intertemporal General Equilibrium Model
Figure E-3. Change in Labor Input Compared to Reference Case
E.I. 6 Projected Impacts on GDP
In this analysis, the costs incurred by the electric sector contribute to a change in
GDP. EPA believes that these GDP results are incomplete, as they do not represent an
economy-wide modeling of the substantial economic benefits expected to accrue due to
decreased health impacts and other effects of reduced emissions. Because these results are
incomplete, the impacts on GDP should not be construed as the costs of the rule. The results
here show only the impact of resource cost changes to the electricity and coal sectors. These
changes cause the other IGEM sectors and households to adjust to accommodate higher
energy prices, and the economy incurs net losses. Almost all general equilibrium models can
be expected to show that a change in resource costs to one sector tends to be transmitted to
other sectors. IGEM is no exception. As shown in Figure E-4, IGEM projects the overall
cost as a result of the direct effects on coal and electricity to be approximately 0.03 percent
of GDP (three one hundredths of one percent) in 2010 and 2015.
E-8
-------
0.10% n
0)
-------
$16,000 -
va
ON
ON
ON
e
o
•= $15,000 -
^
u
•a
o
L.
0.
u
i
%
&e
2
O
$13,C
IGEM Reference Case
CAIR Policy Case
$16,153 - Reference.*1 $16,147 - CAIR
$13,893 - Reference $13,889 - CAIR
2010
2015
Source: IntertemporalGeneral Equilibrium Model
Figure E-5. U.S. Gross Domestic Product (GDP): Reference Case vs. CAIR
E. 1.7 /GEM Mo
-------
consumers' relative demands for goods and services, and trends in productivity
improvement.
IGEM is characterized by "perfect foresight," in which economic agents act
rationally with complete knowledge of the future to maximize the long-term net benefits
arising from their decisions. It is also important to note that IGEM represents capital as
"perfectly mobile," so that infrastructure is easily converted to other uses as dictated by
demand considerations within the context of overall availability (supply).
E. 1.8 Labor Representation in IGEM
IGEM is a "full employment" model, in which all inputs are fully used, including the
labor market. This assumption follows from IGEM's focus on market equilibria (supply and
demand balances) in all markets, including labor, and from the fact that labor is represented
by quality-adjusted hours and not by persons. Labor supply is endogenously determined in
the model, so that the amount of labor supply and demand and the real (i.e., inflation-
adjusted) wage rate adjust in response to policy changes. Labor use, the labor variable
reported by IGEM, is a proxy measurement for "employment," but there are distinct
differences. In the full employment framework, households decide the quantity of labor
services to provide. In IGEM, workers have the option to enjoy leisure or provide labor and
do so based on the prevailing real wage rate. An increase in the labor utilization rate implies
that individuals are choosing to provide more labor to the market in response to firms'
demand for labor as a substitute for electricity and coal.
E.1.9 Sector Detail
Thirty-five sectors comprise the U.S. economy and are represented in IGEM. The list
of these sectors, along with the response of the sectors to the costs of C AIR, are presented in
Table E-l. The table shows the percentage change in output from each sector (the change is
in "constant - dollar sales" as a measure of output).
E.2 EMPAX-CGE Regional Macroeconomic Analysis of CAIR
CAIR reduces emissions of SO2 and NOX from electricity generation to improve air
quality.1 To complement the analysis of CAIR effects on electricity generation conducted by
the EPA using the IPM,2 the macroeconomic implications of this rule have been estimated
'See for details.
2See for complete IPM documentation.
E-ll
-------
Table E-l. Change in Sector Output in Response to CAIR
Sector
Agriculture, forestry, fisheries
Metal mining
Coal mining
Crude oil and gas extraction
Non-metallic mineral mining
Construction
Food and kindred products
Tobacco manufactures
Textile mill products
Apparel and other textile products
Lumber and wood products
Furniture and fixtures
Paper and allied products
Printing and publishing
Chemicals and allied products
Petroleum refining
Rubber and plastic products
Leather and leather products
Stone, clay and glass products
Primary metals
Fabricated metal products
Non-electrical machinery
Electrical machinery
Motor vehicles
Other transportation equipment
Instruments
Miscellaneous manufacturing
Transportation and warehousing
Communications
Electric utilities (services)
Gas utilities (services)
Wholesale and retail trade
Finance, insurance and real estate
Personal and business services
Government enterprises
Percent Change from Reference
2010
0.07%
-0.04%
-1.27%
-0.04%
-0.03%
-0.01%
0.11%
0.08%
-0.01%
-0.05%
0.00%
-0.02%
0.01%
0.01%
-0.01%
-0.07%
-0.02%
-0.05%
0.00%
-0.07%
-0.01%
-0.02%
-0.02%
-0.03%
-0.01%
-0.03%
-0.03%
0.02%
0.02%
-1.01%
0.16%
0.01%
0.03%
0.02%
0.00%
2015
0.10%
-0.06%
-1.63%
-0.06%
-0.04%
-0.01%
0.16%
0.12%
0.01%
-0.06%
0.01%
-0.03%
0.01%
0.02%
-0.01%
-0.10%
-0.02%
-0.06%
0.01%
-0.09%
-0.01%
-0.02%
-0.02%
-0.05%
-0.02%
-0.02%
-0.04%
0.03%
0.02%
-1.27%
0.20%
0.01%
0.03%
0.02%
-0.01%
Source: Intertemporal General Equilbrium Model
E-12
-------
using EPA's EMPAX-CGE model. EMPAX-CGE is a macroeconomic simulation model
developed by RTI International (RTI) for EPA's Office of Air Quality Planning and
Standards (OAQPS). The focus of this analysis of CAIR is examining the sectoral and
regional distribution of economic effects across the U.S. economy. This appendix section
discusses the EMPAX model, the approach used to incorporate electricity-sector results from
IPM, and the results of the macroeconomic analysis. Please note that this analysis focuses on
electricity-sector impacts of CAIR as estimated by IPM. It does not account for other
economic and noneconomic effects, especially the substantial economic and health benefits
associated with reduced emissions.
E. 2.1 Background and Summary of EMPAX-CGE Model3
EMPAX was first developed in 2000 to support the economic analysis of EPA's
maximum achievable control technology (MACT) rules controlling emissions from three
categories of combustion sources (reciprocating internal combustion engines, boilers, and
turbines). The initial framework consisted of a national multimarket partial-equilibrium
model with linkages between manufacturing industries and the energy sector. Effects of
combustion rules on these industries were estimated through their influence on energy prices
and output. Modified versions of EMPAX were subsequently used to analyze economic
impacts of strategies for improving air quality in the Southern Appalachian mountain region.
Recent work on EMPAX has extended its scope to cover all aspects of the U.S.
economy at a regional level in either static or dynamic modes (the dynamic version of
EMPAX is used in this analysis). Although major regulations directly affect a large number
of industries, substantial indirect impacts can also result from changes in production, input
use, income, and household consumption patterns. Consequently, EMPAX now includes
economic linkages among all industrial and energy sectors as well as households that supply
factors of production and purchase goods (i.e., a computable general equilibrium, or CGE,
framework). This gives the version of EMPAX called EMPAX-CGE the ability to trace
economic impacts as they are transmitted throughout the economy and allows it to provide
critical insights to policy makers evaluating the magnitude and distribution of costs
associated with environmental policies. The EMPAX-CGE model was used to investigate
macroeconomic impacts of the Interstate Air Quality Rule, predecessor to CAIR.4
3See Section E.2.8 for additional details on the EMPAX-CGE model.
4See .
E-13
-------
The dynamic version of EMPAX-CGE employed in this analysis of CAIR is an
intertemporally optimizing model. Agents have perfect foresight and maximize utility across
all time periods subject to budget constraints, while firms maximize profits subject to
technology constraints. Nested constant elasticity of substitution (CES) functions are used to
portray substitution possibilities available to producers and consumers. Along with the
underlying data, the nesting structures and associated substitution elasticities define current
production technologies and possible alternatives. Most industries have constant returns to
scale with the exception of fossil-fuel and agriculture industries that have decreasing returns
to scale due to use of factors in fixed supply (land and inputs of primary fuels, respectively).
The current tax structure of the United States is included in EMPAX-CGE to account for
"tax interaction" effects, where interactions between tax distortions and environmental
policies can affect macroeconomic costs of policies.
The economic data in this CGE model come from state-level information provided by
the Minnesota IMPLAN Group, and the energy data come from the DOE's Energy
Information Agency (EIA). In the dynamic version of EMPAX-CGE, these data are used to
define five regions within the United States, each containing 17 industries and four types of
households classified by income.5 The five regions have been selected to preserve important
regional differences in electricity generation technologies, and 17 industries are included that
cover five important types of energy (coal, crude oil, electricity from fossil and nonfossil
generation, natural gas, and refined petroleum), the energy-intensive industries most likely to
be affected by environmental policies, and the remaining sectors of the economy.
Four sources of economic growth are included: technological change from
improvements in energy efficiency, growth in the available labor supply from population
growth and changes in labor productivity, increases in stocks of natural resources, and capital
accumulation. Changes in energy use per unit of output are modeled through exogenous
autonomous energy efficiency improvements (AEEI). The baseline solution in
EMPAX-CGE matches, as closely as possible, EIA forecasts for energy production by fuel
type, energy prices, fuel consumption by industry, industrial output, and regional economic
growth through 2025.6
5 Static versions of EMPAX-CGE have more industries and households because they do not have to solve for
multiple time periods simultaneously and, consequently, have few computational constraints on the number
of industries and households.
6EIA forecasts from the Annual Energy Outlook 2003 (AEO) are used in this analysis.
E-14
-------
Distortions associated with the existing tax structure in the United States have been
included in EMPAX-CGE. A wide range of theoretical and empirical literature has
examined "tax interactions" and found that they can substantially alter costs of
environmental (and other) policies. The IMPLAN economic database used by EMPAX-CGE
includes information on some types of taxes, which have been combined with other sources
to cover important distortions from capital and income taxes.
E. 2.2 Modeling Approach for Electricity Policies
EMPAX-CGE can be used to analyze a wide array of policy issues and is capable of
estimating how a change in a single part (or multiple parts) of the economy will influence
producers and consumers across the United States. However, although CGE models have
been used extensively to analyze climate policies that limit carbon emissions from electricity
production,7 some other types of utility emissions policies are more difficult to consider.
Unlike carbon dioxide, emissions of pollutants such as SO2, NOX, and mercury are not
necessarily proportional to fuel use.
These types of emissions can be lowered by a variety of methods: fuel switching
from high- to low-sulfur coal, moving from coal- to gas-fired generation, and/or installing
retrofit equipment designed to reduce emissions. However, the boiler-specific nature of
these decisions, and their costs and effects, cannot be adequately captured by the more
general structure of a CGE model. In addition, because of the ways that retrofits and
construction of new generating units affect electricity prices and fuel use, a detailed
characterization of electricity markets is preferable when estimating implications of policies
like CAIR. For these reasons, we developed an interface that allows a linkage between
EMPAX-CGE and the IPM model.
IPM is a comprehensive model of electricity generation and transmission in the
United States. The model contains data on all generating units available to dispatch
electricity to the national grid, their existing equipment configurations and fuel consumption,
transmission constraints, and generating costs. It includes characteristics of new units and
retrofits that can be built and/or installed. IPM is capable of estimating how electric utilities
will respond to policies by determining the least-cost methods of generating sufficient
electricity to meet demands, while meeting emissions reduction (and other) objectives.
7See, for example, the analyses of energy/climate using CGE models organized by the Stanford University
Energy Modeling Forum (http://www.stanford.edu/group/EMF/home/index.htm).
E-15
-------
However, IPM does not fully consider how changes in the electricity sector, or
electricity prices, will affect the rest of the U.S. economy. Combining the strengths of IPM
(disaggregated unit-level analyses of electricity policies) with the strengths of CGE models
(macroeconomic effects of environmental policies) allows investigation of economy-wide
implications of policies that would normally be hard to estimate consistently and effectively.
For electricity-generation regulations like CAIR that require a very disaggregated level of
analysis, IPM can determine for EMPAX-CGE a number of electricity market outcomes
needed to evaluate macroeconomic implications of policies. The linkage with IPM then
allows EMPAX-CGE to take these findings and use them in "counterfactual" policy
evaluations.
Among the many results provided by IPM, several can potentially have significant
implications for the rest of the economy including changes in electricity prices, fuel
consumption by utilities, fuel prices, and changes in electricity production expenditures.
EMPAX-CGE is capable of simultaneously incorporating some or all of these IPM findings,
depending on the desired type and degree of linkage between the two models. At the
regional level, EMPAX-CGE can match changes estimated by IPM for the following
variables:
• electricity prices (percentage change in retail prices)
• coal and gas consumption for electricity (percentage changes in Btus)
• coal and gas prices (percentage changes in prices)
• coal and gas expenditures ($ changes—Btus of energy input times $/MMBtu)
• capital costs ($ changes)
• fixed operating costs ($ changes)
• variable operating costs ($ changes)
For EMPAX-CGE to effectively incorporate IPM data on changes in costs, they have
to be expressed in terms of the productive inputs used in CGE models (i.e., capital, labor, and
material inputs produced by other industries). Rather than assume these costs represent a
proportional scaling up of all inputs to the electricity industry in EMPAX-CGE, we use
Nestor and Pasurka (1995) data on purchases made by industries for environmental
protection reasons to allocate these additional expenditures across inputs within
E-16
-------
EMPAX-CGE. Once these expenditures are specified, the incremental costs from IPM can
be used to adjust the production technologies and input purchases by electricity generation in
the CGE model.
E. 2.3 Modeling Methodology for CAIR
The macroeconomic impacts of CAIR, as simulated by CGE models, will be a
function of the methodology used to link IPM to EMPAX-CGE and the economic
interactions accounted for by the CGE model. Initial effects will revolve around how using
additional resources in electricity generation draws some capital, labor, and materials from
other sectors of the economy. This, in turn, may affect prices in markets supplying these
inputs to utilities. Similarly, changes in coal and gas use in electricity and associated impacts
on their prices will have implications for the rest of the economy (although any spillovers in
coal markets will have limited effects outside of electricity because most is used for
generation). In addition, any electricity price increases associated with these initial effects
will encourage improvements in energy efficiency, switching to alternate forms of energy
(increases in natural gas prices may mitigate this effect), and lower consumption of
electricity in general (these demand decreases would lead to lower production levels with
associated benefits for the environment). The magnitude of these adjustments will be a
function of the structure of the CGE model and the elasticities in it that control the ease of
these substitutions.
The macroeconomic effects beyond energy production and consumption decisions
also depend on the theoretical structure of the CGE model used in the analysis. Similar to
the perfect-foresight nature of IPM, CGE models like EMPAX-CGE assume that firms and
consumers will observe and anticipate policies to be enacted in the future. This causes them
to adjust their behavior and investment decisions in all time periods in the model (including
the starting year of the model). As a result, anticipation of changes in production and
consumption costs in the future will cause shifts in behavior in all model years as people
prepare ahead of policy enactment. The aggregate implications of these changes will also be
influenced by income effects (how people will alter consumption levels in response to having
more or less money) and substitution effects (how people will alter their patterns of
consumption purchases in response to changes in relative prices of goods). For example, an
anticipated decrease in labor productivity in the future (leading to lower wages) may cause
an increase in work effort today, while labor is more productive. Alternatively, it may lead
E-17
-------
to a decrease in work effort today because, in part, labor today is used to generate capital
goods for tomorrow, which will be used to augment less-productive labor in the future.
The methodology used to link IPM and EMPAX-CGE for this CAIR analysis focuses
on CAIR resource costs and implications for coal use by utilities.8 IPM estimates of
additional resources used by electric utilities (the capital, fixed, and variable costs) are used
to adjust generation technologies in EMPAX-CGE. The same procedure is used to account
for incremental increases in natural gas purchases by utilities (although all these costs are
applied exclusively to natural gas inputs to electricity generation in EMPAX-CGE, rather
than being apportioned using the Nestor and Pasurka data). Given that approximately 90
percent of all coal is consumed in generation, we use IPM estimates of changes in coal use
directly within EMPAX-CGE, expressed as percentage changes in Btus, rather than adopting
a less direct linkage. However, for both coal and natural gas, EMPAX-CGE is allowed to
estimate the impacts on commodity prices faced by the rest of the economy.
E. 2.4 Projected Impacts on Specific Industries
Impacts of CAIR on electricity-generation costs and their subsequent effects on
electricity prices, along with coal market changes, will affect output and prices of all
industries in EMPAX-CGE. These effects may increase or decrease output and/or revenue,
depending on their implications for production costs and technologies, and shifts in
household demands. However, as shown in Figure E-6, estimates for output changes from
CAIR (outside of electricity and coal) are generally around 0.05 percent and may be positive
or negative.
Some industries are affected more than others and others may increase production.
Natural gas output increases as electric utilities switch out of coal generation and as higher
electricity prices cause other businesses and consumers to move to alternate energy sources.
Energy-intensive sectors of the economy are relatively more affected than other firms
because they rely more heavily on electricity and other fuels. However, the largest of these
declines in output (aluminum) is approximately two-tenths of 1 percent.
8See Section E.2.15 for EMPAX-CGE results using other linkages to IPM.
E-18
-------
-5.0% -4.0% -3.0% -2.0% -1.0% 0.0%
Percent Change from Reference
Coal
Crude Oil
Electricity
Natural Gas
Petroleum
Agriculture
EIS: Food and Kindred
HS: Paper and Allied
eS: Chemicals
HS: Glass
HS: Cement
EIS: Iron and Steel
EIS: Aluminum
Other Manufacturing
Services
Transportation
1.0%
Figure E-6. CAIR Impacts on U.S. Domestic Output, 2015
Source: EMPAX-CGE.
Regional effects tend to show variation that does not appear at the national level.
Figure E-7 shows these regional results for energy markets and highlights the aggregate U.S.
results with a solid bar. The largest differences are in electricity generation because the
western part of the United States is relatively unaffected by the policy, other than through
spillover effects reflected in both IPM and EMPAX-CGE. Increases in natural gas output
and declines in coal production are also distributed unevenly across the country.
E-19
-------
l.U/0 -
o
u
c
2 0.0% -
!a
* -1.0% -
g
t -2.0% -
M
e
cS
g -3.0% -
e
8 4.0% -
0-
-5.0% -
in:
u
1
• u
u
ri
u
KJ
o
J3
tl
z
o
GO
§
-d
2
cS
1
^
Coal
C3
O
^3
g
o
GO
1
-d
2
cS
1
^
Crude Oil
C3
O
,c
1
c
V
^M CO J3^M GO _^^M GO
|||P|||||P|||||P
-< h-7-
-------
u.iuyo -
n n^o/^
n nno/
-U.lD%
n i no/*
n i ^o/
n ^Cfl/r.
n o^o/
•i
ii ,
BHI
1
i
1
IP L
U II
1 IIJ
U
U
U|
- I 1 •
IB ~
T~
cS "3 » --2 S £3 cS
^) ^T ^ ra ^^ 0}
•S GO "2 S ^ 'S &
Agriculture Ehe
M
-H -(— > t/3
3 M fl
? £ 3
0 13 PL,
S
rgy-Inte
inufactu
^
u
nsh
rin]
GO
ID
/e
j
« ^ « g tS 2
c3 ^ P •--" OJ) i-
_§ o S ^ >
•S GO 13 PH ^
I 2
Other
Manufacturing
Northeast
South
cc
S| Midwest
a
J3
ice
3
GO
ID
« ^ -s g ts g
c3 ^ P •--" OJ) t— '
_§ o S ^ >
•S GO 13 PH ^
1 2
Transportation
Figure E-8. CAIR Impacts on Regional Industrial Output, 2015
Source: EMPAX-CGE.
experience an advantage over similar firms in other regions that face proportionately larger
price increases. Although the West sees the greatest improvement in comparative advantage,
output also rises in other regions (see Figure E-9).
E. 2.5 Projected Impacts on Consumer Prices
Changes in consumer price levels are used to measure price effects of policies and
any resulting implications for average purchase prices paid by households. EMPAX-CGE
calculates an overall price level across the "basket" of goods and services bought by
consumers. For a policy like CAIR, consumer price levels will be affected directly by
changes in electricity prices faced by households and indirectly by changes in goods prices
that have been produced using electricity. Figure E-10 shows percentage changes in average
consumer prices on average ranging from 0.02 to 0.04 percent.
E-21
-------
l.uyo "
g 0.8% -
e
2 0.6% -
=2
£ 0.4% -
S 0.2% -
* 0.0% -
60
| -0.2% -
u -0.4% -
a
8 -0.6% -
•—
£ -0.8% -
-1.0% -
F
||
™ m - - » »» o B
ail ™ o l i • ~
1
u
n
D
• "-1
III
111
1
LJ
19
I u
III
0
1
i
•
III
D
1
1 Hi;
•U
1
™ II Ei
-4— » (~i •*— ' ^ i_> C/l -4— » r* ~ C/l
2 5 " fl'S^M'S M fl"S&<
gg |^|P 2 g |^|p
^M-^S^ ^M-^S^
1 ^ 1 ^
Food and Paper and
Kindred Allied
Northeast
£
"3
&
Midwest
M
a
&
S
^
1
GO
Chemicals
Northeast
^
"3
5
Midwest
f/3
g
3
S
^
u
D
Glass
Northeast
"3
£'£
"2 p
3 "S
3|
D
Cement
O
!
Midwest
M
a
'3
S
to
1
GO
p
Iron and Steel
Northeast
^
"3
&
Midwest
M
a
'a
s
to
1
GO
Aluminum
Figure E-9. CAIR Impacts on Regional Energy-Intensive Output, 2015
E. 2.6 Projected Impacts on Labor Markets
CGE models like EMPAX-CGE typically consider how policies may influence labor
markets through how they alter the number of productivity-adjusted hours of labor supplied
by households (this is not the same as estimating jobs or employment). Empirical estimates
of lab or-supply elasticities are used by EMPAX-CGE to simulate how demands by firms and
supply decisions by households are made, along with resulting implications for real wages.
EMPAX-CGE, similar to IGEM, is a full-employment model in which households choose
between labor and leisure time, based on both income and substitution effects.
E-22
-------
Percent Change from Reference
0.10% -,
A AQO/
A AGO/
A ATO/
A A/;0/
U.UO/0
A A«0/
U.UJ /O
A A/10/
A AOO/
U.UJ /o
A AOO/
n mo/
U.Ul/0
_^
^^^^
^-^^
2010
2015
Figure E-10. Change in Consumer Prices Compared to Reference Case
Source: EMPAX-CGE.
Figure E-l 1 gives EMPAX-CGE's projected impacts of CAIR on labor markets. The
results indicate that people are choosing to work slightly fewer hours in response to small
declines in real wage rates, rather than work more hours to offset additional costs of
purchasing goods. All of these effects are extremely small, however, on the order of two
one-thousandths of 1 percent.
E.2.7 Projected Impacts on GDP
The combination of all economic interactions as described earlier will be reflected in
the changes in GDP estimated by a CGE model. Given that this cost-based approach to
analyzing CAIR does not reflect its benefits to the environment, public health, and labor
productivity, CGE models (including EMPAX-CGE) will tend to estimate declines in total
E-23
-------
Percent Change from Reference
u.uiuyo -
A AAOO/
U.UUo /o
A AA/:O/
A A A/10/
U.UU4/0
A AAOO/
A AAAO/ _.
A AAOO/
-U.OOz/o
A AA/IO/
A AA/;0/
-U.UUO /O
A AAOO/
-0.010% -
•,
'^^^^_^
2010
2015
Figure E-ll. Change in Labor Inputs Compared to Reference Case
Source: EMPAX-CGE
production in the United States, as shown in Figure E-12. Because these results are
incomplete and do not reflect potential benefits of CAIR, the impacts on GDP should not be
construed as the costs of the rule. EMPAX-CGE projects decreases in GDP of between 0.03
percent and 0.05 percent (five one-hundredths of 1 percent).
Overall, it should be noted that the estimated implications of CAIR for U.S. GDP are
extremely small relative to the total size of the economy. Figure E-13 illustrates GDP in the
model baseline and CAIR policy cases. As shown, the GDP impact is negligible and, in fact,
it is not possible to adjust the scale of the graph to the point where the two lines do not
overlap. Even these small costs could be reversed if the CGE analyses were extended to
include benefits associated with CAIR such as improvements in labor productivity from
environmental improvements.
E-24
-------
0.10% n
A AGO/
iV A A/:O/
ox u.uo/o
OS
•S A A/10/
QJ U.U4/0
OX
XT A mo/
,g U.UZ/o
B
CJ A AAO/ _
•_ U.UU/o 1
PH
^"^ n mo/
Q
O A A/10/
-U.U4/0
fe/i n n./:o/
~ -U.UO/o
OS
U A AGO/
-U.Uo/o
-0.10% -
B
— ^
2010
2015
Figure E-12. Change in GDP Compared to Reference Case
Source: EMPAX-CGE
National GDP effects like those in Figure E-13 may tend to obscure variation at a
regional or local level. Several potential sources of divergences in regional impacts exist:
• differences in IPM regional results based on regional mixes of generation
technologies (coal, gas, oil, and nonfossil use), which may be averaged out at a
national level;
• differences in regional production and consumption patterns for electricity and
nonelectricity energy goods;
• differences in industrial composition of regional economies;
• differences in household consumption patterns; and
• differences in regional growth forecasts.
E-25
-------
$15,000 -,
$14,500
$14,000
B
o
CQ
a
•o
o
O
$12,500
-------
0.10%
0.05%
0.00% ir
60
e
8
-0.15%
-0.20%
-0.25%
2010
2015
•Northeast
•South A Midwest
•Plains
West D US
Figure E-14. Change in Regional GDP Compared to Reference Case
Source: EMPAX-CGE.
Tables E-2 and E-3 show EMPAX-CGE estimates of changes in revenue and output
quantities for 2010 and 2015.
E.2.8 EMPAX-CGE Model Description: General Model Structure
This section provides additional details on the EMPAX-CGE model structure, data
sources, and assumptions. The version of EMPAX-CGE used in this CAIR analysis is a
dynamic, intertemporally optimizing model that solves in 5-year intervals from 2005 to
2050. It uses the classical Arrow-Debreu general equilibrium framework wherein
households maximize utility subject to budget constraints, and firms maximize profits
subject to technology constraints. The model structure, in which agents are assumed to have
perfect foresight and maximize utility across all time periods, allows agents to modify
behavior in anticipation of future policy changes, unlike dynamic recursive models that
assume agents do not react until a policy has been implemented.
E-27
-------
Table E-2. U.S. Domestic Output Changes
Model Run Industry
Percentage Change in Revenue (%) Coal
Grade oil
Electricity
Natural gas
Petroleum
Agriculture
Energy-intensive manufacturing
Other manufacturing
Services
Transportation
Percentage Change in Quantity (%) Coal
Grade oil
Electricity
Natural gas
Petroleum
Agriculture
Energy-intensive manufacturing
Other manufacturing
Services
Transportation
2010
-2.73%
-0.04%
0.60%
0.38%
-0.03%
-0.04%
-0.04%
-0.03%
-0.04%
-0.04%
-0.81%
-0.02%
-1.33%
0.08%
-0.02%
-0.03%
-0.04%
0.00%
-0.01%
-0.01%
2015
-3.13%
-0.08%
0.93%
0.16%
-0.06%
-0.07%
-0.07%
-0.07%
-0.06%
-0.09%
-1.17%
-0.04%
-2.01%
0.05%
-0.04%
-0.05%
-0.07%
-0.04%
-0.02%
-0.04%
Source: EMPAX-CGE.
E-28
-------
Table E-3. U.S. Domestic Energy-Intensive Sector Output Changes
Model Run
Percentage Change in Revenue (%)
Percentage Change in Quantity (%)
Industry
Food and Kindred
Paper and Allied
Chemicals
Glass
Cement
Iron and Steel
Aluminum
Food and Kindred
Paper and Allied
Chemicals
Glass
Cement
Iron and Steel
Aluminum
2010
-0.04%
-0.03%
-0.04%
-0.04%
-0.01%
-0.02%
0.02%
-0.03%
-0.03%
-0.04%
-0.05%
-0.02%
-0.02%
-0.12%
2015
-0.06%
-0.07%
-0.08%
-0.09%
-0.02%
-0.06%
-0.02%
-0.04%
-0.07%
-0.08%
-0.11%
-0.09%
-0.08%
-0.20%
Source: EMPAX-CGE.
Nested CES functions are used to portray substitution possibilities available to
producers and consumers. Figure E-15 illustrates this general framework and gives a broad
characterization of the model.10 Along with the underlying data, these nesting structures and
associated substitution elasticities determine the effects that will be estimated for policies.
These nesting structures and elasticities used in EMPAX-CGE are generally based on the
Emissions Prediction and Policy Analysis (EPPA) Model developed at the Massachusetts
Institute of Technology (Babiker et al., 2001). Although the two models are quite different
(EPPA is a recursive-dynamic, international model focused on national-level climate-change
policies), both are intended to simulate how agents will respond to environmental policies.
'"Although it is not illustrated in Figure E-15, some differences across industries exist in their handling of
energy inputs. In addition, the agriculture and fossil-fuel sectors in EMPAX-CGE contain equations that
account for the presence of fixed inputs to production (land and fossil-fuel resources, respectively).
E-29
-------
Utility
Consumption Leisure
Foreign
Local
Output
Regional
Output
Household utility is a CES function
of consumption and leisure.
Consumption is a Cobb-Douglas
composite of the 16 types of goods.
Each consumption good is a CES
composite of foreign and
domestically produced goods.
Domestic goods are a CES composite
of locally produced goods and goods
from other regions.
Most producer goods use fixed pro-
portions of intermediate inputs and a
capital-labor-energy (KLE) composite.
KLE
Intermediates
//1V
Energy
Value Added
Energy
(5 Types) Capitel
Labor
Intermediate materials inputs are the 11 types of non-
energy goods, in fixed proportion for each industry.
The KLE composite is a CES function
of energy and value-added (KL).
Energy is a CES composite of5 types of fuel. The
structure of this function varies across industries.
Value added is a Cobb-Douglas
composite of capital and labor.
Figure E-15. General Production and Consumption Nesting Structure in EMPAX-
CGE
Given this basic similarity, EMPAX-CGE has adopted a comparable structure.
EMPAX-CGE is programmed in the GAMS11 language (Generalized Algebraic Modeling
System) and solved as a mixed complementarity problem (MCP)12 using MPSGE software
(Mathematical Programming Subsystem for General Equilibrium).13 The PATH solver from
GAMS is used to solve the MCP equations generated by MPSGE.
"See Brooke et al. (1998) for a description of GAMS (http://www.gams.com/).
12Solving EMPAX-CGE as a MCP problem implies that complementary slackness is a feature of the equilibrium
solution. In other words, any firm in operation will earn zero economic profits and any unprofitable firms
will cease operations. Similarly, for any commodity with a positive price, supply will equal demand, or
conversely any good in excess supply will have a zero price.
13See Rutherford (1999) for MPSGE documentation (http://debreu.colorado.edu).
E-30
-------
E.2.9 Data Sources
The economic data come from state-level information provided by the Minnesota
IMPLAN Group14 and energy data come from EIA.15 Although IMPLAN data contain
information on the value of energy production and consumption in dollars, these data are
replaced with EIA data for several reasons. First, the policies being investigated typically
focus on energy markets, making it essential to include the best possible characterization of
these markets in the model. Although the IMPLAN data are developed from a variety of
government data sources at the U.S. Bureau of Economic Analysis and U.S. Bureau of Labor
Statistics, these data do not always agree with energy information collected by EIA directly
from manufacturers and electric utilities. Second, it is necessary to have physical quantities
for energy consumption in the model to portray effects of environmental policies. EIA
reports physical quantities, while IMPLAN does not. Finally, although the IMPLAN data
reflect the year 2000, the initial baseline year for the model is 2005. Thus, AEO energy
production and consumption, output, and economic growth forecasts for 2005 are used to
adjust the year 2000 IMPLAN data.
EMPAX-CGE combines these economic and energy data to create a balanced social
accounting matrix (SAM) that provides a baseline characterization of the economy. The
SAM contains data on the value of output in each sector, payments for factors of production
and intermediate inputs by each sector, household income and consumption, government
purchases, investment, and trade flows. A balanced SAM for the year 2005 consistent with
the desired sectoral and regional aggregation is produced using procedures developed by
Babiker and Rutherford (1997) and described in Rutherford and Paltsev (2000). The
methodology relies on standard optimization techniques to maintain the calculated energy
statistics while minimizing the changes needed in the economic data to create a new
balanced SAM that matches AEO forecasts for the baseline model year of 2005.
These data are used to define 10 regions within the United States, each containing 40
industries. Regions have been selected to capture important differences across the country in
electricity-generation technologies, while industry aggregations are controlled by available
energy consumption data. Prior to solving EMPAX-CGE, these regions and industries are
aggregated up to the categories to be included in the analysis.
14See for a description of the Minnesota IMPLAN Group and its data.
15These EIA sources include AEO 2003, the Manufacturing Energy Consumption Survey, State Energy Data
Report, State Energy Price and Expenditure Report, and various annual industry profiles.
E-31
-------
Table E-4 presents the industry categories included in EMPAX-CGE for the CAIR
analysis. Their focus is on maintaining as much detail in the energy-intensive sectors16 as is
allowed by available energy consumption data and computational limits of dynamic CGE
models. In addition, the electricity industry is separated into fossil-fuel generation and
nonfossil generation, which is necessary because many electricity policies like CAIR affect
only fossil-fired electricity.
Table E-4. EMPAX-CGE Industries
EMPAX Industry NAICS Classifications
Coal 2121
Crude Oil" 211111
Electricity (fossil and nonfossil) 2211
Natural Gas 211112, 2212, 4862
Petroleum Refining 324
Agriculture 11
Energy-Intensive Sector: Food 311
Energy-Intensive Sector: Paper and Allied 322
Energy-Intensive Sector: Chemicals 325
Energy-Intensive Sector: Glass 3272
Energy-Intensive Sector: Cement 3273
Energy-Intensive Sector: Iron and Steel 3311
Energy-Intensive Sector: Aluminum 3313
Other Manufacturing 312-316, 321, 323, 326-327, 331-339
Services All Others
Transportationb 481-488
a Although NAICS 211111 covers crude oil and gas extraction, the gas component of this sector is moved to
the Natural Gas industry.
b Transportation does not include NAICS 4862 (natural gas distribution), which is part of the natural gas
industry.
16Energy-intensive sectors industry categories are based onEIA definitions of energy-intensive manufacturers in
the Assumptions for the Annual Energy Outlook 2003.
E-32
-------
Figure E-16 shows the five regions used in this analysis, which have been defined
based on the expected regional distribution of policy impacts, availability of economic and
energy data, and computational limits on model size. These regions have been constructed
from the underlying 10-region database designed to follow, as closely as possible, the
electricity market regions defined by the North American Electric Reliability Council
(NERC).17
""West" also includes
Alaska and Hawaii.
Figure E-16. Regions Defined in EMPAX-CGE for the CAIR Analysis
"Economic data and information on nonelectricity energy markets are generally available only at the state level,
which necessitates an approximation of the NERC regions that follows state boundaries. For the CAIR
analysis, these approximations include Northeast = NPCC + MAAC, Southeast = SERC + FERC, Midwest
= ECAR + MAIN, Plains = MAPP + SPP + ERCOT, and West = WSCC. See for
further discussion of these regions.
E-33
-------
E.2.10 Production Functions
All productive markets are assumed to be perfectly competitive and have production
technologies that exhibit constant returns to scale, except for the agriculture and natural
resource extracting sectors, which have decreasing returns to scale because they use factors
in fixed supply (land and fossil fuels, respectively). The electricity industry is separated into
two distinct sectors: fossil-fuel generation and nonfossil generation. This allows tracking of
variables such as heat rates for fossil-fired utilities (Btus of energy input per kilowatt hour of
electricity output).
All markets must clear (i.e., supply must equal demand in every sector) in every
period, and the income of each agent in the model must equal their factor endowments plus
any net transfers. Along with the underlying data, the nesting structures shown in
Figure E-15 and associated substitution elasticities define current production technologies
and possible alternatives.
E.2.11 Utility Functions
Each region in the dynamic version of EMPAX-CGE contains four representative
households, classified by income, that maximize intertemporal utility over all time periods in
the model subject to budget constraints, where the income groups are
• $0 to $14,999,
• $15,000 to $29,999,
• $30,000 to $49,999, and
• $50,000 and above.
These representative households are endowed with factors of production including
labor, capital, natural resources, and land inputs to agricultural production. Factor prices are
equal to the marginal revenue received by firms from employing an additional unit of labor
or capital. The value of factors owned by each representative household depends on factor
use implied by production within each region. Income from sales of these productive factors
is allocated to purchases of consumption goods to maximize welfare.
Within each time period, intratemporal utility received by a household is formed from
consumption of goods and leisure. All consumption goods are combined using a
Cobb-Douglas structure to form an aggregate consumption good. This composite good is
then combined with leisure time to produce household utility. The elasticity of substitution
E-34
-------
between consumption goods and leisure depends on empirical estimates of lab or-supply
elasticities and indicates how willing households are to trade off leisure time for
consumption. Over time, households consider the discounted present value of utility
received from all periods' consumption of goods and leisure.
Following standard conventions of CGE models, factors of production are assumed to
be intersectorally mobile within regions, but migration of productive factors is not allowed
across regions. This assumption is necessary to calculate welfare changes for the
representative household located in each region in EMPAX-CGE. EMPAX-CGE also
assumes that ownership of natural resources and capital embodied in nonfossil electricity
generation is spread across the United States through capital markets.
E.2.12 Trade
In EMPAX-CGE, all goods and services are assumed to be composite, differentiated
" Armington" goods made up of locally manufactured commodities and imported goods.
Output of local industries is initially separated into output destined for local consumption by
producers or households and output destined for export. This local output is then combined
with goods from other regions in the United States using Armington-trade elasticities that
indicate agents make relatively little distinction between output from firms located within
their region and output from firms in other regions within the United States. Finally, the
domestic composite goods are aggregated with imports from foreign sources using lower
trade elasticities to capture the fact that foreign imports are more differentiated from
domestic output than are imports from other regional suppliers in the United States.
E.2.13 Tax Rates and Distortions
Taxes and associated distortions in economic behavior have been included in
EMPAX-CGE because theoretical and empirical literature found that taxes can substantially
alter estimated policy costs. The IMPLAN economic database used by EMPAX-CGE
includes information on taxes such as indirect business taxes (all sales and excise taxes) and
Social Security taxes. However, IMPLAN reports factor payments for labor and capital at
their gross-of-tax values, which necessitates use of additional data sources to determine
personal income and capital tax rates. Information from the TAXSEVI model at the National
Bureau of Economic Research (Feenberg and Courts, 1993), along with user-cost-of-capital
calculations from Fullerton and Rogers (1993), are used to establish tax rates.
Along with these rates, distortions associated with taxes are a function of labor
supply decisions of households. As with other CGE models focused on interactions between
E-35
-------
tax and environmental policies (e.g., Bovenberg and Goulder [1996]; Goulder and Williams
[2003]), an important feature of EMPAX-CGE is its inclusion of a labor-leisure
choice—how people decide between working and leisure time. Labor supply elasticities
related to this choice determine, to a large extent, how distortionary taxes are in a CGE
model. Elasticities based on the relevant literature have been included in EMPAX-CGE (i.e.,
0.4 for the compensated labor supply elasticity and 0.15 for the uncompensated labor supply
elasticity). These elasticity values give an overall marginal excess burden associated with
the existing tax structure of approximately 0.3.
E.2.14 Intertemporal Dynamics and Economic Growth
There are four sources of economic growth in EMPAX-CGE: technological change
from improvements in energy efficiency, growth in the available labor supply (from both
population growth and changes in labor productivity), increases in stocks of natural
resources, and capital accumulation. Energy consumption per unit of output tends to decline
over time because of improvements in production technologies and energy conservation.
These changes in energy use per unit of output are modeled as AEEIs, which are used to
replicate energy consumption forecasts by industry and fuel from EIA.18 The AEEI values
provide the means for matching expected trends in energy consumption that have been taken
from the AEO forecasts. They alter the amount of energy needed to produce a given quantity
of output by incorporating improvements in energy efficiency and conservation. Labor force
and regional economic growth, electricity generation, changes in available natural resources,
and resource prices are also based on the AEO forecasts.
Savings provide the basis for capital formation and are motivated through people's
expectations about future needs for capital. Savings and investment decisions made by
households determine aggregate capital stocks in EMPAX-CGE. The IMPLAN dataset
provides details on the types of goods and services used to produce the investment goods
underlying each region's capital stocks. Adjustment dynamics associated with formation of
capital are controlled by using quadratic adjustment costs experienced when installing new
capital, which imply that real costs are experienced to build and install new capital
equipment.
18See Babiker et al. (2001) for a discussion of how this methodology was used in the EPPA model (EPPA
assumes that AEEI parameters are the same across all industries in a country, while AEEI values in
EMPAX-CGE are industry specific).
E-36
-------
Prior to investigating policy scenarios, it is necessary to establish a baseline path for
the economy that incorporates economic growth and technology changes that are expected to
occur in the absence of the policy actions. Beginning from the initial balanced SAM dataset,
a "steady-state" growth path is first specified for the economy to ensure that the model
remains in equilibrium in future years.19 Once the model is able to replicate a steady-state
growth path, the assumption of a constant growth rate is replaced by actual forecasts from
AEO. After incorporating these forecasts, EMPAX-CGE is solved to generate a baseline
consistent with them through 2025. Once this baseline is established, it is possible to run
"counterfactual" policy experiments.
E.2.15 Alternative IPM-to-EMPAXLinkages
As discussed in Section E.2.2, EMPAX-CGE is capable of incorporating a variety of
results from IPM, depending on the desired type of linkage between the two models. This
section presents macroeconomic impacts of CAIR, as shown by changes in GDP, from two
alternative methods of using IPM results. These findings are contrasted to the "Central
Case" (i.e., the results presented above) to demonstrate how the methodology used to link the
two models can influence results. One alternative, referred to as the "IPM Price & Fuel
Case," places a higher degree of reliance on IPM results than the "Central Case." In the
other alternative, referred to as the "Unconstrained Case," EMPAX-CGE is allowed to
determine more market outcomes than in the "Central Case." These scenarios provide a
range of results that illustrate the macroeconomic implications of different methods for
linking macroeconomic models with the IPM results.
Specifically, the three alternative approaches are as follows:
• "Central Case"—This case from the main analysis incorporates IPM estimates of
resource costs (capital costs, and fixed and variable operating costs), along with
percentage changes in coal use (expressed in Btus), in EMPAX-CGE. The IPM
resource costs are used to adjust electricity generation costs within EMPAX-CGE
by requiring additional purchases of capital, labor, and material inputs. Natural
gas expenditures are also adjusted based on IPM findings by requiring additional
purchases of gas by utilities.
19A steady-state growth path requires all variables in the model to grow at a constant rate over time, including
labor, output, inputs to production, and consumption. If the model has been properly specified, the
steady-state replication check will show that the economy remains in equilibrium in each year along this
path.
E-37
-------
• "Unconstrained Case"—This case incorporates IPM estimates of both resource
costs and fuel expenditures for coal and gas into EMPAX-CGE. Unlike the
"Central Case," this case allows changes in the quantity of coal use in the
electricity sector to be estimated by EMPAX-CGE, once declines in the dollar
value of coal purchases from the IPM model have been incorporated into the
model. These declines in coal purchases are integrated using the same
methodology applied to the capital, labor, and material inputs needed to generate
electricity—in this case, by requiring fewer purchases of coal to produce
electricity.
• "IPMPrice & Fuel Case"—This case places the most reliance on IPM findings.
Rather than allowing EMPAX-CGE to determine electricity price outcomes based
on IPM resource costs, it replicates the IPM price results in EMPAX-CGE and
concentrates on examining their implications for the rest of the economy.20
Similarly, this case uses IPM data on changes in coal and gas use (in physical
units) by electricity sector instead of allowing the CGE model to make these
decisions. Market prices for coal and gas are still determined by EMPAX-CGE.
This case takes into consideration the fact that, although most resource costs of
electricity policies are borne by coal-fired generation, electricity prices are
typically determined by the marginal unit in operation. Because of this, there
may not be a direct correlation between policy costs and implications for
electricity prices, although the economy outside of the electricity industry will
respond to both electricity prices and any effects from drawing additional
resources into electricity production.
Figure E-17 illustrates the implications of these alternative linkages between IPM and
EMPAX-CGE for estimates of CAIR GDP effects. The "Central Case" from the main
analysis and "Unconstrained Case" follow similar paths; however, the "Unconstrained Case"
is uniformly less expensive. It provides more degrees of freedom to adjust coal consumption
by utilities in response to demand changes estimated by EMPAX-CGE and also has added
flexibility to shift among production inputs. This results in GDP impacts to around 20
percent lower than in the "Central Case."
20IPM's estimated wholesale prices are converted into changes in retail prices for EMPAX-CGE using EIA
estimates of transmission and distribution costs of approximately $27 per megawatt hour.
E-38
-------
O
B
0.00%
-0.02%
-0.04%
-0.06%
-0.08%
-0.10%
2010
2015
-EMPAX-Central Case
•EMPAX - Unconstrained
•EMPAX - IPM Price & Fuel
Figure E-17. GDP Impacts of Alternative Linkages (in %)
The "IPM Price & Fuel Case" shows changes in GDP that follow the same pattern as
the "Central Case," but are generally lower. The methodologies of these two cases are
substantially different: many of the effects in the "IPM Price & Fuel Case" are driven by
IPM's estimated changes in electricity prices (and to a lesser degree by impacts on gas prices
from increased demand by generators), while in the "Central Case" electricity prices
predicted by EMPAX-CGE are controlled by how many additional resource costs are
entering electricity production. IPM results show moderate increases in electricity prices in
2010 and 2015, leading to smaller GDP effects in the "IPM Price & Fuel Case" than in the
"Central Case."
E-39
-------
E.3 References
Babiker, M.H., and T.F. Rutherford. 1997. "Input-Output and General Equilibrium
Estimates of Embodied CO2: A Data Set and Static Framework for Assessment."
University of Colorado at Boulder, Working Paper 97-2. Available at
http ://debreu. Colorado, edu/papers/gtaptext.html.
Babiker, M.H., J.M. Reilly, M. Mayer, R.S. Eckaus, IS. Wing, and R.C. Hyman. 2001.
"The MIT Emissions Prediction and CO2 Policy Analysis (EPPA) Model: Revisions,
Sensitivities, and Comparisons of Results." MIT Joint Program on the Science and
Policy of Global Change, Report No. 71. Available at
http ://web .mit. edu/globalchange/www/eppa.html.
Bovenberg, L.A., and L.H. Goulder. 1996. "Optimal Environmental Taxation in the
Presence of Other Taxes: General-Equilibrium Analysis." American Economic
Review 86(4):985-1000.
Brooke, A., D. Kendrick, A. Meeraus, and R. Raman. 1998. GAMS: A User's Guide.
GAMS Development Corporation. Available at http://www.gams.com.
Feenberg, D., and E. Courts. 1993. "An Introduction to the TAXSIM Model." Journal of
Policy Analysis and Management 12(1): 189-194. Available at
http://www.nber.org/~taxsim/.
Fullerton, D., and D. Rogers. 1993. "Who Bears the Lifetime Tax Burden?" Washington,
DC: The Brookings Institute. Available at http://bookstore.brookings.edu/
book_details.asp?product%5Fid=10403.
Goulder, L.H., and R.C. Williams. 2003. "The Substantial Bias from Ignoring General
Equilibrium Effects in Estimating Excess Burden, and a Practical Solution." Journal
of Political Economy 111:898-927.
Minnesota EVIPLAN Group. 2003. State-Level Data for 2000. Available from
http://www.implan.com/index.html.
Nestor, D.V., and C.A. Pasurka. 1995. The U.S. Environmental Protection Industry: A
Proposed Framework for Assessment. U.S. Environmental Protection Agency, Office
of Policy, Planning, and Evaluation. EPA 230-R-95-001. Available at
http://yosemite.epa.gov/ee/epa/eermfile.nsf/llf680ff78df42f585256b45007e6235/41
b8b642ab9371df852564500004b543/$FILE/EE-0217A-l.pdf.
E-40
-------
Rutherford, T.F. 1999. "Applied General Equilibrium Modeling with MPSGE as a GAMS
Subsystem: An Overview of the Modeling Framework and Syntax." Computational
Economics 14(l):l-46. Also available at
http://www.gams.com/solvers/mpsge/syntax.htm.
Rutherford, T.F., and S.V. Paltsev. 2000. "GTAP-Energy in GAMS: The Dataset and Static
Model." University of Colorado at Boulder, Working Paper 00-2. Available at
http ://debreu. Colorado, edu/papers/gtaptext.html.
U.S. Department of Energy, Energy Information Administration. State Energy Data Report.
Washington DC. Available at http://www.eia.doe.gov/emeu/states/
_use_multi state, html.
U.S. Department of Energy, Energy Information Administration. State Energy Price and
Expenditure Report. Washington DC. Available at
http://www.eia.doe.gov/emeu/states/price_multi state.html.
U.S. Department of Energy, Energy Information Administration. 2001. Manufacturing
Energy Consumption Survey 1998. Washington DC. Available at
http://www.eia.doe.gov/emeu/mecs/.
U.S. Department of Energy, Energy Information Administration. January 2003. Annual
Energy Outlook 2003. DOE/EIA-0383(2003). Washington DC. Available at
http://www.eia.doe.gov/oiaf/archive/aeo03/pdf/03 83(2003).pdf
U.S. Department of Energy, Energy Information Administration. January 2004. Annual
Energy Outlook 2004. DOE/EIA-0383(2003). Washington, DC. Available at
http://www.EOA.DOE.GOV/OIAF/AOE/PDF/0383 (2001).
E-41
-------
APPENDIX F
ADDITIONAL TECHNICAL INFORMATION SUPPORTING
THE BENEFITS ANALYSIS
This appendix provides additional technical details about several important elements
of the benefits analysis, including the spatial interpolation method and health effect pooling
methods. Additional details on benefits methods can be found in the BenMAP User's
Manual, available in the docket and online at http://www.epa.gov/ttn/ecas/benmodels.html.
F.I Voronoi Neighbor Averaging
In calculating the base year concentrations of PM species and ozone at model grid
cells prior to scaling with model outputs, we used a spatial interpolation method known as
Voronoi Neighbor Averaging (VNA).
The first step in VNA is to identify the set of neighboring monitors for each of the
grid cells in the continental United States. The figure below presents nine grid cells and
seven monitors, with the focus on identifying the set of neighboring monitors for grid cell E.
# = Center Grid-Cell "E"
*
= Air Pollution Monitor
F-l
-------
In particular, BenMAP identifies the nearest monitors, or "neighbors," by drawing a
polygon, or Voronoi cell, around the center of each grid cell. The polygons have the special
property that the boundaries are the same distance from the two closest points.
# = Center Grid-Cell "E"
*
- Air Pollution Monitor
We then chose those monitors that share a boundary with the center of grid cell E. These are
the nearest neighbors, and we used these monitors to estimate the air pollution level for this
grid cell.
To estimate the air pollution level in each grid cell, BenMAP calculates the air
pollution metrics for each of the neighboring monitors and then calculates an
inverse-distance weighted average of the metrics. The further the monitor is from the grid
cell center, the smaller the weight.
The weight for the monitor 20 kilometers from the center of grid cell E is calculated as
follows:
J_
on
= 0.27 .
20 16 14.
F-2
-------
15 \x
10 miles *-
15 miles
*
\ 20 miles
\
# = Center Grid-Cell "E"
*
= Air Pollution Monitor
The weights for the other monitors would be calculated in a similar fashion.
F.2 The Random/Fixed Effect Pooling Procedure
Often more than one study has estimated a C-R function for a given pollutant-health
endpoint combination. Each study provides an estimate of the pollutant coefficient, P, in the
C-R function, along with a measure of the uncertainty of the estimate. Because uncertainty
decreases as sample size increases, combining data sets is expected to yield more reliable
estimates of p and therefore more reliable estimates of the incidence change predicted using
P. Combining data from several comparable studies to analyze them together is often
referred to as meta-analysis.
For a number of reasons, including data confidentiality, it is often impractical or
impossible to combine the original data sets. Combining the results of studies to produce
better estimates of p provides a second-best but still valuable way to synthesize information
(DerSimonian and Laird, 1986). This is referred to as pooling. Pooling PS requires that all
of the studies contributing estimates of P use the same functional form for the C-R function.
That is, the PS must be measuring the same thing.
It is also possible to pool the study-specific estimates of incidence change derived
from the C-R functions, instead of pooling the underlying PS themselves. For a variety of
reasons, this is often possible when it is not feasible to pool the underlying PS. For example,
F-3
-------
if one study is log-linear and another is linear, we could not pool the PS because they are not
different estimates of a coefficient in the same C-R function but are instead estimates of
coefficients in different C-R functions. We can, however, calculate the incidence change
predicted by each C-R function (for a given change in pollutant concentration and, for the
log-linear function, a given baseline incidence rate) and pool these incidence changes.
BenMAP allows the pooling of incidence changes predicted by several studies for the same
pollutant-health endpoint group combination. It also allows the pooling of the corresponding
study-specific estimates of monetary benefits.
As with estimates based on only a single study, BenMAP allows you to characterize
the uncertainty surrounding pooled estimates of incidence change and/or monetary benefit.
To do this, BenMAP pools the study-specific distributions of incidence changes (or monetary
benefit) to derive a pooled distribution. This pooled distribution incorporates information
from all the studies used in the pooling procedure.
F.2.1 Weights Used for Pooling
The relative contribution of any one study in the pooling process depends on the
weight assigned to that study. A key component of the pooling process, then, is the
determination of the weight given to each study. Various methods can be used to assign
weights to studies. Below we discuss the possible weighting schemes that are available in
BenMAP.
Subjective (User-specified) Weights
BenMAP allows the user the option of specifying the weights to be used. Suppose,
for example, the user wants to simply average all study-specific results. He would then
assign a weight of 1/N to each of the N study-specific distributions that are to be pooled.
Note that subjective weights are limited to two decimal places and are normalized if they do
not sum to one.
Automatically Generated Weights
A simple average has the advantage of simplicity but the disadvantage of not taking
into account the uncertainty of each of the estimates. Estimates with great uncertainty
surrounding them are given the same weight as estimates with very little uncertainty. A
common method for weighting estimates involves using their variances. Variance takes into
account both the consistency of data and the sample size used to obtain the estimate, two key
F-4
-------
factors that influence the reliability of results. BenMAP has two methods of automatically
generating pooling weights using the variances of the input distributions—fixed effects
pooling and random/fixed effects pooling.
The discussion of these two weighting schemes is first presented in terms of pooling
the pollutant coefficients (the PS), because that most closely matches the discussion of the
method for pooling study results as it was originally presented by DerSimonian and Laird
(1986). We then give an overview of the analogous weighting process used within BenMAP
to generate weights for incidence changes rather than PS.
F.3 Fixed Effects Weights
The fixed effects model assumes that there is a single true C-R relationship and
therefore a single true value for the parameter p that applies everywhere. Differences among
PS reported by different studies are therefore simply the result of sampling error. That is,
each reported P is an estimate of the same under lying parameter. The certainty of an
estimate is reflected in its variance (the larger the variance, the less certain the estimate).
Fixed effects pooling therefore weights each estimate under consideration in proportion to
the inverse of its variance.
Suppose there are n studies, with the ith study providing an estimate p; with variance
v;(I=l, ..., n). Let
denote the sum of the inverse variances. Then the weight, w;, given to the ith estimate, p;, is
1/v,
w, = —-
S
This means that estimates with small variances (i.e., estimates with relatively little
uncertainty surrounding them) receive large weights and those with large variances receive
small weights.
F-5
-------
The estimate produced by pooling based on a fixed effects model, then, is just a
weighted average of the estimates from the studies being considered, with the weights as
defined above. That is,
P/e = £ *, * P, •
The variance associated with this pooled estimate is the inverse of the sum of the inverse
variances:
Table F-l shows the relevant calculations for this pooling for three sample studies.
Table F-l. Example of Fixed Effects Model Calculations
Study
1
2
3
Sum
Pi
0.75
1.25
1.00
V,
0.1225
0.0025
0.0100
1/V,
8.16
400
100
£ = 508.16
w,
0.016
0.787
0.197
£ = 1.000
w,*P,
0.012
0.984
0.197
£=1.193
The sum of weighted contributions in the last column is the pooled estimate of P
based on the fixed effects model. This estimate (1.193) is considerably closer to the estimate
from study 2 (1.25) than is the estimate (1.0) that simply averages the study estimates. This
reflects the fact that the estimate from study 2 has a much smaller variance than the estimates
from the other two studies and is therefore more heavily weighted in the pooling.
The variance of the pooled estimate, vfe, is the inverse of the sum of the variances, or
0.00197. (The sums of the p; and v; are not shown, because they are of no importance. The
sum of the l/v; is S, used to calculate the weights. The sum of the weights, w;, i=l, ..., n, is
1.0, as expected).
F-6
-------
F.4 Random/Fixed Effects Weights
An alternative to the fixed effects model is the random effects model, which allows
the possibility that the estimates p; from the different studies may in fact be estimates of
different parameters, rather than just different estimates of a single underlying parameter. In
studies of the effects of PM10 on mortality, for example, if the composition of PM10 varies
among study locations the underlying relationship between mortality and PM10 may be
different from one study location to another. For example, fine particles make up a greater
fraction of PM10 in Philadelphia than in El Paso. If fine particles are disproportionately
responsible for mortality relative to coarse particles, then one would expect the true value of
P in Philadelphia to be greater than the true value of P in El Paso. This would violate the
assumption of the fixed effects model.
The following procedure can test whether it is appropriate to base the pooling on the
random effects model (vs. the fixed effects model). A test statistic, Qw, the weighted sum of
squared differences of the separate study estimates from the pooled estimate based on the
fixed effects model, is calculated as
Q = £ 1
*~-W 2-ii
vi
Under the null hypothesis that there is a single underlying parameter, p, of which all the p;s
are estimates, Qw has a chi-squared distribution with n-1 degrees of freedom. (Recall that n
is the number of studies in the meta-analysis.) If Qw is greater than the critical value
corresponding to the desired confidence level, the null hypothesis is rejected. That is, in this
case the evidence does not support the fixed effects model, and the random effects model is
assumed, allowing the possibility that each study is estimating a different p. (BenMAP uses
a 5 percent one-tailed test.)
The weights used in a pooling based on the random effects model must take into
account not only the within-study variances (used in a meta-analysis based on the fixed
effects model) but the between-study variances as well. These weights are calculated as
follows:
Using Qw, the between-study variance, r|2, is
F-7
-------
Qw ~ (n~
;1/V,2
£
1/v
It can be shown that the denominator is always positive. Therefore, if the numerator
is negative (i.e., if Qw < n-1), then r\2 is a negative number, and it is not possible to calculate
a random effects estimate. In this case, however, the small value of Qw would presumably
have led to accepting the null hypothesis described above, and the meta-analysis would be
based on the fixed effects model. The remaining discussion therefore assumes that r2 is
positive.
Given a value for r\2, the random effects estimate is calculated in almost the same
way as the fixed effects estimate. However, the weights now incorporate both the within-
study variance (v;) and the between-study variance (r|2). Whereas the weights implied by the
fixed effects model used only v;, the within-study variance, the weights implied by the
random effects model use v; +r|2.
Let v;* = v; +r\2. Then
o* _ 1
S - E — '
and
w, =
The estimate produced by pooling based on the random effects model, then, is just a
weighted average of the estimates from the studies being considered, with the weights as
defined above. That is,
F-8
-------
rand
w* * p. .
The variance associated with this random effects pooled estimate is, as it was for the
fixed effects pooled estimate, the inverse of the sum of the inverse variances:
rand
1
\lv*
The weighting scheme used in a pooling based on the random effects model is
basically the same as that used if a fixed effects model is assumed, but the variances used in
the calculations are different. This is because a fixed effects model assumes that the
variability among the estimates from different studies is due only to sampling error (i.e., each
study is thought of as representing just another sample from the same underlying
population), while the random effects model assumes that there is not only sampling error
associated with each study, but that there is also between-study variability—each study is
estimating a different underlying p. Therefore, the sum of the within-study variance and the
between-study variance yields an overall variance estimate.
F.5 Fixed Effects and Random/Fixed Effects Weighting to Pool Incidence Change
Distributions and Dollar Benefit Distributions
Weights can be derived for pooling incidence changes predicted by different studies,
using either the fixed effects or the fixed/random effects model, in a way that is analogous to
the derivation of weights for pooling the PS in the C-R functions. As described above,
BenMAP generates a Latin hypercube representation of the distribution of incidence change
corresponding to each C-R function selected. The means of those study-specific Latin
hypercube distributions of incidence change are used in exactly the same way as the reported
PS are used to calculate fixed effects and random effects weights described above. The
variances of incidence change are used in the same way as the variances of the PS. The
formulas above for calculating fixed effects weights, for testing the fixed effects hypothesis,
and for calculating random effects weights can all be used by substituting the mean incidence
F-9
-------
change for the ith C-R function for p; and the variance of incidence change for the ith C-R
function for V;.1
Similarly, weights can be derived for dollar benefit distributions. As described
above, BenMAP generates a Latin hypercube representation of the distribution of dollar
benefits. The means of those Latin hypercube distributions are used in exactly the same way
as the reported PS are used to calculate the fixed effects and random effects weights
described above. The variances of dollar benefits are used in the same way as the variances
of the PS. The formulas above for calculating fixed effects weights, for testing the fixed
effects hypothesis, and for calculating random effects weights can all be used by substituting
the mean dollar benefit change for the ith valuation for p; and the variance of dollar benefits
for the ith valuation for v;.
BenMAP always derives fixed effects and random/fixed effects weights using
nationally aggregated results, and uses those weights for pooling at each grid cell (or county,
etc., if the user chooses to aggregate results prior to pooling). This is done because BenMAP
does not include any regionally based uncertainty—that is, all uncertainty is at the national
level in BenMAP, and all regional differences (e.g., population) are treated as certain.
F.6 Reference
DerSimonian, R. and N. Laird, "Meta-analysis in Clinical Trials," Controlled Clinical Trials
7:3 (1986), 177-188.
1 There may be a problem with transferring the fixed effects hypothesis test to "incidence change space." The
test statistic to test the fixed effects model is a chi-squared random variable. In the original paper on this
pooling method, DerSimonian and Laird (1986) were discussing the pooling of estimates of parameters,
which are generally normally distributed. The incidence changes predicted from a C-R function will not be
normally distributed if the C-R function is not a linear function of the pollutant coefficient, which, in most
cases it is not. (Most C-R functions are log-linear.) In that case, the test statistic may not be chi-square
distributed. However, most log-linear C-R functions are nearly linear because their coefficients are very
small. In that case the test statistic is likely to be nearly chi-square distributed.
F-10
-------
APPENDIX G
HEALTH-BASED COST-EFFECTIVENESS OF REDUCTIONS IN AMBIENT PM2 5
ASSOCIATED WITH CAIR
G.I Summary
Health-based cost-effectiveness analysis (CEA) and cost-utility analysis (CUA) have
been used to analyze numerous health interventions but have not been widely adopted as
tools to analyze environmental policies. The Office of Management and Budget (OMB)
recently issued Circular A-4 guidance on regulatory analyses, requiring federal agencies to
"prepare a CEA for all major rulemakings for which the primary benefits are improved
public health and safety to the extent that a valid effectiveness measure can be developed to
represent expected health and safety outcomes." Environmental quality improvements may
have multiple health and ecological benefits, making application of CEA more difficult and
less straightforward. For CAIR CEA may provide a useful framework for evaluation: non-
health benefits are substantial, but the majority of quantified benefits come from health
effects. Therefore, EPA is including in the CAIR RIA a preliminary and experimental
application of one type of CEA—a modified quality-adjusted life-years (QALYs) approach.
In this instance, the direct usefulness of cost-effectiveness analysis is mitigated by the lack of
rule alternatives to compare relative effectiveness, but one can still make some comparisons
to other benchmarks bearing in mind methodological differences.
QALYs were developed to evaluate the effectiveness of individual medical
treatments, and EPA is still evaluating the appropriate methods for CEA for environmental
regulations. Agency concerns with the standard QALY methodology include the treatment
of people with fewer years to live (the elderly); fairness to people with preexisting conditions
that may lead to reduced life expectancy and reduced quality of life; and how the analysis
should best account for nonhealth benefits, such as improved visibility.
The Institute of Medicine (a member institution of the National Academies of
Science) has established the Committee to Evaluate Measures of Health Benefits for
Environmental, Health, and Safety Regulation to assess the scientific validity, ethical
implications, and practical utility of a wide range of effectiveness measures used or proposed
in CEA. This committee is expected to produce a report by the end of 2005. In the interim,
G-l
-------
however, agencies are expected to provide CEAs for rules covered by Circular A-4
requirements.
The methodology presented in this appendix is not intended to stand as precedent
either for future air pollution regulations or for other EPA regulations where it may be
inappropriate. It is intended solely to demonstrate one particular approach to estimating the
cost-effectiveness of reductions in ambient PM25 in achieving improvements in public
health. Reductions in ambient PM2 5 likely will have other health and environmental benefits
that will not be reflected in this CEA. Other EPA regulations affecting other aspects of
environmental quality and public health may require additional data and models that may
preclude the development of similar health-based CEAs. A number of additional
methodological issues must be considered when conducting CEAs for environmental
policies, including treatment of nonhealth effects, aggregation of acute and long-term health
impacts, and aggregation of life extensions and quality-of-life improvements in different
populations. The appropriateness of health-based CEA should be evaluated on a case-by-
case basis subject to the availability of appropriate data and models, among other factors.
CAIR is expected to result in substantial reductions in potential population exposure
to ambient concentrations of PM. The benefit-cost analysis presented in Chapter 4 shows
that CAIR achieves substantial health benefits whose monetized value far exceeds costs (net
benefits are over $100 billion in 2015). Despite the risk of oversimplifying benefits,
cautiously-interpreted cost-effectiveness calculations may provide further evidence of
whether the costs associated with CAIR are a reasonable health investment for the nation.
This analysis provides estimates of commonly used health-based effectiveness
measures, including lives saved, life years saved (from reductions in mortality risk), and
QALYs saved (from reductions in morbidity risk) associated with the reduction of ambient
PM2 5 due to CAIR. In addition, a new aggregate effectiveness metric, Morbidity Inclusive
Life Years (MILY) is introduced to address some of the concerns about aggregation of life
extension and quality-of-life impacts.1 It represents the sum of life years gained due to
reductions in premature mortality and the QALY gained due to reductions in chronic
morbidity. This measure may be preferred to existing QALY aggregation approaches
because it does not devalue life extensions in individuals with preexisting illnesses that
reduce quality of life. However, the MILY measure is still based on life years and thus still
'This metric is also referred to as a "fair QALY" in Hubbell (2004b). The metric is also being considered by the
Institute of Medicine Committee to Evaluate Measures of Health Benefits for Environmental, Health, and
Safety Regulation under the term "fair QALY."
G-2
-------
inherently gives more weight to interventions that reduce mortality and morbidity impacts
for younger populations with higher remaining life expectancy. This analysis focuses on life
extensions and improvements in quality of life through reductions in two diseases with
chronic impacts: chronic bronchitis (CB) and nonfatal acute myocardial infarctions. Monte
Carlo simulations are used to propagate uncertainty in several analytical parameters and
characterize the distribution of estimated impacts.
Presented in three different metrics, the analysis suggests the following:
• In 2010 CAIR will result in:
- 13,000 (95% CI: 5,400 - 20,000) premature deaths avoided, or
- 140,000 (95% CI: 104,000 - 184,000) life years gained (discounted at 3
percent), or
- 190,000 (95% CI: 140,000 - 250,000) MILYs gained (discounted at 3
percent).
• In 2015, CAIR will result in:
- 17,000 (95% CI: 7,200 - 26,000) premature deaths avoided, or
- 190,000 (95% CI: 140,000 - 240,000) life years gained (discounted at 3
percent), or
- 250,000 (95% CI: 180,000 - 330,000) MIL Ys gained (discounted at 3
percent).
• Using a 7 percent discount rate, mean discounted life years gained are 110,000 in
2010 and 140,000 in 2015; mean MILYs gained are 140,000 in 2010 and 180,000
in 2015. (The estimates of premature deaths avoided are not affected by the
discount rate.)
• The associated reductions in CB and nonfatal acute myocardial infarctions will
reduce medical costs by approximately $2.1 billion in 2010 and $2.7 billion in
2015 based on a 3 percent discount rate, or $1.8 billion in 2010 and $2.3 billion in
2015 based on a 7 percent discount rate.
• Other health and nonhealth benefits are valued at $1.9 billion in 2010 and $2.8
billion in 2015.
Direct private compliance costs of the CAIR rule are $2.4 billion in 2010 and $3.6
billion in 2015. Therefore, the net costs (costs minus avoided cost of illness minus other
benefits) are negative, indicating that CAIR results in cost savings. As such, traditional cost-
G-3
-------
effectiveness ratios are not informative. However, it is possible to calculate the maximum
costs for the rule that would still result in cost-effective improvements in public health
compared with standard benchmarks of $50,000 and $100,000 per QALY:
1. Taking into account avoided medical costs and other benefits, annual costs of
CAIR would need to exceed $14 billion (95% CI: $9.8 billion - $18 billion)
in 2010 and $18 billion (95% CI: $13 billion - $23 billion) in 2015 to have a
cost per MILY that exceeds a benchmark of $50,000, based on a 3 percent
discount rate.
2. Annual costs of CAIR would need to exceed $23 billion (95% CI: $17 billion
- $31 billion) in 2010 and $30 billion (95% CI: $22 billion - $40 billion) in
2015 to have a cost per MILY that exceeds a benchmark of $100,000, based
on a 3 percent discount rate.
3. Using a 7 percent discount rate, annual costs of CAIR would need to exceed
$11 billion in 2010 and $14 billion in 2015 to have a cost per MILY that
exceeds a benchmark of $50,000, and would need to exceed $18 billion in
2010 and $23 billion in 2015 to have a cost per MILY that exceeds a
benchmark of $100,000.
Given costs of $2.4 billion and $3.6 billion in 2010 and 2015, respectively, CAIR is
clearly a very cost-effective way to achieve improvements in public health.
G.2 Introduction
Analyses of environmental regulations have typically used benefit-cost analysis to
characterize impacts on social welfare. Benefit-cost analyses allow for aggregation of the
benefits of reducing mortality risks with other monetized benefits of reducing air pollution,
including acute and chronic morbidity, and nonhealth benefits such as improved visibility.
One of the great advantages of the benefit-cost paradigm is that a wide range of quantifiable
benefits can be compared to costs to evaluate the economic efficiency of particular actions.
However, alternative paradigms such as CEA and CUA analyses may also provide useful
insights. CEA involves estimation of the costs per unit of benefit (e.g., lives or life years
saved). CUA is a special type of CEA using preference-based measures of effectiveness,
such as QALYs.
CEA and CUA are most useful for comparing programs that have similar goals, for
example, alternative medical interventions or treatments that can save a life or cure a disease.
G-4
-------
They are less readily applicable to programs with multiple categories of benefits, such as
those reducing ambient air pollution, because the cost-effectiveness calculation is based on
the quantity of a single benefit category. In other words, we cannot readily convert
improvements in nonhealth benefits such as visibility to a health metric such as life years
saved. For these reasons, environmental economists prefer to present results in terms of
monetary benefits and net benefits.
However, QALY-based CUA has been widely adopted within the health economics
literature (Neumann, 2003; Gold et al., 1996) and in the analysis of public health
interventions (US FDA, 2004). QALY-based analyses have not been as accepted in the
environmental economics literature because of concerns about the theoretical consistency of
QALYs with individual preferences (Hammitt, 2002), treatment of nonhuman health
benefits, and a number of other factors (Freeman, Hammitt, and De Civita, 2002). For
environmental regulations, benefit-cost analysis has been the preferred method of choosing
among regulatory alternatives in terms of economic efficiency. Recently several academic
analyses have proposed the use of life years-based benefit-cost or CEAs of air pollution
regulations (Cohen, Hammitt, and Levy, 2003; Coyle et al., 2003; Rabl, 2003; Carrothers,
Evans, and Graham, 2002). In addition, the World Health Organization has adopted the use
of disability-adjusted life years, a variant on QALYs, to assess the global burden of disease
due to different causes, including environmental pollution (Murray et al., 2002; de Hollander
etal., 1999).
Recently, the U.S. OMB (Circular A-4, 2003) issued new guidance requiring federal
agencies to provide both CEA and benefit-cost analyses for major regulations. The OMB
Circular A-4 directs agencies to "prepare a CEA for all major rulemakings for which the
primary benefits are improved public health and safety to the extent that a valid effectiveness
measure can be developed to represent expected health and safety outcomes." We are
including a CEA for CAIR to illustrate one potential approach for conducting a CEA. This is
an experimental application, and EPA is still evaluating the appropriate methods for CEA for
environmental regulations with multiple outcomes. The Institute of Medicine (a member
institution of the National Academies of Science) has empaneled the Committee to Evaluate
Measures of Health Benefits for Environmental, Health, and Safety Regulation to assess the
scientific validity, ethical implications, and practical utility of a wide range of effectiveness
measures used or proposed in CEA. This committee is expected to produce a report by the
end of 2005. However, in the interim, are required to provide CEAs to comply with Circular
A-4. As such, EPA has begun developing methodologies for estimating the cost-
effectiveness of air pollution regulations. The methodology presented in this appendix is not
G-5
-------
intended to stand as precedent either for future air pollution regulations or for other EPA
regulations governing water, solid waste, or other regulatory objectives. It is intended solely
to demonstrate one particular approach to estimating the effectiveness of reductions in
ambient PM2 5 in achieving improvements in public health. This analysis focuses on
effectiveness measured by improvements in life expectancy and reductions in the incidence
of two diseases with chronic impacts on quality of life: CB and nonfatal acute myocardial
infarctions. Other EPA regulations affecting other aspects of environmental quality and
public health may require additional data and models that may preclude the development of
similar QALY-based analyses. The appropriateness of QALY-based CEA should be
evaluated on a case-by-case basis subject to the availability of appropriate data and models.
Preparation of a CEA requires identification of an appropriate measure of rule
effectiveness. Given the significant impact of reductions in ambient PM25 on reductions in
the risk of mortality, lives saved is an important measure of effectiveness. However, one of
the ongoing controversies in health impact assessment regards whether reductions in
mortality risk should be reported and valued in terms of statistical lives saved or in terms of
statistical life years saved. Life years saved measures differentiate among premature
mortalities based on the remaining life expectancy of affected individuals. In general, under
the life years approach, older individuals will gain fewer life years than younger individuals
for the same reduction in mortality risk during a given time period, making interventions that
benefit older individuals seem less beneficial relative to similar interventions benefitting
younger individuals. A further complication in the debate is whether to apply quality
adjustments to life years lost. Under this approach, individuals with preexisting health
conditions would have fewer QALYs lost relative to healthy individuals for the same loss in
life expectancy, making interventions that primarily benefit individuals with poor health
seem less beneficial to similar interventions affecting primarily healthy individuals.
In addition to substantial mortality risk reduction benefits, CAIR will also result in
significant reductions in chronic and acute morbidity. Several approaches have been
developed to incorporate both morbidity and mortality into a single effectiveness metric.
The most common of these is the QALY approach, which expresses all morbidity and
mortality impacts in terms of quality of life multiplied by the duration of time with that
quality of life. The QALY approach has some appealing characteristics. For example, it can
account for morbidity effects as well as losses in life expectancy without requiring the
assignment of dollar values to calculate total benefits. By doing so it provides an alternative
framework to benefit-cost analysis for aggregating quantitative measures of health impacts.
G-6
-------
While used extensively in the economic evaluation of medical interventions (Gold et
al., 1996), QALYs have not been widely used in evaluating environmental health regulations.
A number of specific issues arise with the use of QALYs in evaluating environmental
programs that affect a broad and heterogeneous population and that provide both health and
nonhealth benefits. The U.S. Public Health Service report on cost-effectiveness in health and
medicine notes the following:
For decisions that involve greater diversity in interventions and the people to whom
they apply, cost-effectiveness ratios continue to provide essential information, but
that information must, to a greater degree, be evaluated in light of circumstances and
values that cannot be included in the analysis. Individuals in the population will
differ widely in their health and disability before the intervention, or in age, wealth,
or other characteristics, raising questions about how society values gains for the more
and less health, for young and old, for rich and poor, and so on. The assumption that
all QALYs are of equal value is less likely to be reasonable in this context. (Gold et
al., 1996, p. 11)
Use of QALYs as a measure of effectiveness for environmental regulations is still
developing, and while this analysis provides one framework for using QALYs to evaluate
environmental regulations, there are clearly many issues, both scientific and ethical, that
need to be addressed with additional research. The Institute of Medicine panel evaluating
QALYs and other effectiveness measures will develop criteria for choosing among the
measures that potentially are useful for regulatory impact analysis and will make
recommendations regarding measures appropriate for assessing the health benefits of
regulatory interventions and propose criteria for identifying regulations for which CEA is
appropriate and informative.
This appendix presents cost-effectiveness methodologies for evaluating programs
such as CAIR that are intended to reduce ambient PM2 5 starting from the standard QALY
literature and seeking a parallel structure to benefit-cost analysis in the use of air quality and
health inputs (see Hubbell [2004a] for a discussion of some of the issues that arise in
comparing QALY and benefit-cost frameworks in analyzing air pollution impacts). For the
purposes of this analysis, we calculate effectiveness using several different metrics, including
lives prolonged, life years gained, and modified QALYs. For the life years and QALY-type
approaches, we use life table methods to calculate the change in life expectancy expected to
result from changes in mortality risk from PM. We use existing estimates of preferences for
different health states to obtain QALY weights for morbidity endpoints associated with air
G-7
-------
pollution. In general, consistent with the Gold et al. (1996) recommendations, we use
weights obtained from a societal perspective when available. We explore several different
sources for these weights to characterize some of the potential uncertainty in the QALY
estimates. We follow many of the principles of the reference case analysis as defined in
Gold et al. (1996), although in some cases we depart from the reference case approach when
data limitations require us to do so (primarily in the selection of quality-of-life weights for
morbidity endpoints). We also depart from the reference case in the method of combining
life expectancy and quality-of-life gains.
Results in most tables are presented only at a discount rate of 3 percent, rather than at
both 3 percent and 7 percent as recommended in EPA and OMB guidance. This is strictly
for ease of presentation in EPA's first demonstration of this approach to cost-effectiveness.
Aggregate results at 7 percent are presented in the summary, and the impact of using a 7
percent discount rate instead of 3 percent rate is summarized in a sensitivity analysis.
Monte Carlo simulation methods are used to propagate uncertainty in several of the
model parameters throughout the analysis. We characterize overall uncertainty in the results
with 95 percent confidence intervals based on the Monte Carlo simulations. In addition, we
examine the impacts of changing key parameters, such as the discount rate, on the
effectiveness measures and the cost-effectiveness metrics.
The remainder of this appendix provides an overview of the key issues involved in
life year- and QALY-based approaches for evaluating the health impacts of air pollution
regulations, provides detailed discussions of the steps required for each type of effectiveness
calculation, and presents the CEA for CAIR. Section G.3 introduces the various
effectiveness measures and discusses some of the assumptions required for each.
Section G.4 details the methodology used to calculate changes in life years and quality
adjustments for mortality and morbidity endpoints. Section G.5 provides the results for
CAIR and discusses their implications for cost-effectiveness of CAIR.
G.3 Effectiveness Measures
Three major classes of benefits are associated with reductions in air pollution:
mortality, morbidity, and nonhealth (welfare). For the purposes of benefit-cost analysis,
EPA has presented mortality-related benefits using estimates of avoided premature
mortalities, representing the cumulative result of reducing the risk of premature mortality
from long-term exposure to PM25 for a large portion of the U.S. population. Morbidity
benefits have been characterized by numbers of new incidences avoided for chronic diseases
G-8
-------
such as CB, avoided admissions for hospitalizations associated with acute and chronic
conditions, and avoided days with symptoms for minor illnesses. Nonhealth benefits are
characterized by the monetary value of reducing the impact (e.g., the dollar value of
improvements in visibility at national parks).
For the purposes of CEA, we focus the effectiveness measure on the quantifiable
health impacts of the reduction in PM25. Treatment of nonhealth benefits is important and is
discussed in some detail later in this section. If the main impact of interest is reductions in
mortality risk from air pollution, the effectiveness measures are relatively straightforward to
develop. Mortality impacts can be characterized similar to the benefits analysis, by counting
the number of premature mortalities avoided, or can be characterized in terms of increases in
life expectancy or life years.2 Estimates of premature mortality have the benefit of being
relatively simple to calculate, are consistent with the benefit-cost analysis, and do not impose
additional assumptions on the degree of life shortening. However, some have argued that
counts of premature mortalities avoided are problematic because a gain in life of only a few
months would be considered equivalent to a gain of a many life years, and the true
effectiveness of an intervention is the gain in life expectancy or life years (Rabl, 2003; Miller
and Hurley, 2003).
Calculations of changes in life years and life expectancy can be accomplished using
standard life table methods (Miller and Hurley, 2003). However, the calculations require
assumptions about the baseline mortality risks for each age cohort affected by air pollution.
A general assumption may be that air pollution mortality risks affect the general mortality
risk of the population in a proportional manner. However, some concerns have been raised
that air pollution affects mainly those individuals with preexisting cardiovascular and
respiratory disease, who may have reduced life expectancy relative to the general population.
This issue is explored in more detail below.
Air pollution is also associated with a number of significant chronic and acute
morbidity endpoints. Failure to consider these morbidity effects may understate the cost-
2Life expectancy is an ex ante concept, indicating the impact on an entire population's expectation of the
number of life years they have remaining, before knowing which individuals will be affected. Life
expectancy thus incorporates both the probability of an effect and the impact of the effect if realized. Life
years is an ex post concept, indicating the impact on individuals who actually die from exposure to air
pollution. Changes in population life expectancy will always be substantially smaller than changes in life
years per premature mortality avoided, although the total life years gained in the population will be the
same. This is because life expectancy gains average expected life years gained over the entire population,
while life years gained measures life years gained only for those experiencing the life extension.
G-9
-------
effectiveness of air pollution regulations or give too little weight to reductions in particular
pollutants that have large morbidity impacts but no effect on life expectancy. The QALY
approach explicitly incorporates morbidity impacts into measures of life years gained and is
often used in health economics to assess the cost-effectiveness of medical spending programs
(Gold et al., 1996). Using a QALY rating system, health quality ranges from 0 to 1, where 1
may represent full health, 0 death, and some number in between (e.g., 0.8) an impaired
condition. QALYs thus measure morbidity as a reduction in quality of life over a period of
life. QALYs assume that duration and quality of life are equivalent, so that 1 year spent in
perfect health is equivalent to 2 years spent with quality of life half that of perfect health.
QALYs can be used to evaluate environmental rules under certain circumstances, although
some very strong assumptions (detailed below) are associated with QALYs. The U.S. Public
Health Service Panel on Cost Effectiveness in Health and Medicine recommended using
QALYs when evaluating medical and public health programs that primarily reduce both
mortality and morbidity (Gold et al., 1996). Although there are significant nonhealth
benefits associated with air pollution regulations, over 90 percent of quantifiable monetized
benefits are health-related, as is the case with CAIR. Thus, it can be argued that QALYs are
more applicable for these types of regulations than for other environmental policies.
However, the value of nonhealth benefits should not be ignored. As discussed below, we
have chosen to subtract the value of nonhealth benefits from the costs in the numerator of the
cost-effectiveness ratio.
In the following sections, we lay out a phased approach to describing effectiveness.
We begin by discussing how the life-extending benefits of air pollution reductions are
calculated, and then we incorporate morbidity effects using the QALY approach. We also
introduce an alternative aggregated health metric, Morbidity Inclusive Life Years (MILY) to
address some of the ethical concerns about aggregating life extension impacts in populations
with preexisting disabling conditions.
The use of QALYs is predicated on the assumptions embedded in the QALY
analytical framework. As noted in the QALY literature, QALYs are consistent with the
utility theory that underlies most of economics only if one imposes several restrictive
assumptions, including independence between longevity and quality of life in the utility
function, risk neutrality with respect to years of life (which implies that the utility function is
linear), and constant proportionality in trade-offs between quality and quantity of life
(Pliskin, Shepard, and Weinstein, 1980; Bleichrodt, Wakker, and Johannesson, 1996). To
the extent that these assumptions do not represent actual preferences, the QALY approach
will not provide results that are consistent with a benefit-cost analysis based on the Kaldor-
G-10
-------
Hicks criterion.3 Even if the assumptions are reasonably consistent with reality, because
QALYs represent an average valuation of health states rather than the sum of societal WTP,
there are no guarantees that the option with the highest QALY per dollar of cost will satisfy
the Kaldor-Hicks criterion (i.e., generate a potential Pareto improvement [Garber and Phelps,
1997]).
Benefit-cost analysis based on WTP is not without potentially troubling underlying
structures as well, incorporating ability to pay (and thus the potential for equity concerns)
and the notion of consumer sovereignty (which emphasizes wealth effects). Table G-l
compares the two approaches across a number of parameters. For the most part, WTP allows
parameters to be determined empirically, while the QALY approach imposes some
conditions a priori.
Table G-l. Comparison of QALY and WTP Approaches
Parameter
QALY
WTP
Risk aversion
Relation of duration and quality
Proportionality of duration/
quality trade-off
Treatment of time/age in utility
function
Preferences
Source of preference data
Treatment of income and prices
Risk neutral
Independent
Constant
Utility linear in time
Community/Individual
Stated
Not explicitly considered
Empirically determined
Empirically determined
Variable
Empirically determined
Individual
Revealed and stated
Constrains choices
3The Kaldor-Hicks efficiency criterion requires that the "winners" in a particular case be potentially able to
compensate the "losers" such that total societal welfare improves. In this case, it is sufficient that total
benefits exceed total costs of the regulation. This is also known as a potential Pareto improvement, because
gains could be allocated such that at least one person in society would be better off while no one would be
worse off.
G-ll
-------
G.4 Changes in Premature Death, Life Years, and Quality of Life
To generate health outcomes, we used the same framework as for the benefit-cost
analysis described in Chapter 4. For convenience, we summarize the basic methodologies
here. For more details, see Chapter 4 and the BenMAP user's manual
(http://www.epa.gov/ttn/ecas/benmodels.html).
BenMAP uses health impact functions to generate changes in the incidence of health
effects. Health impact functions are derived from the epidemiology literature. A standard
health impact function has four components: an effect estimate from a particular
epidemiological study, a baseline incidence rate for the health effect (obtained from either
the epidemiology study or a source of public health statistics like CDC), the affected
population, and the estimated change in the relevant PM summary measure.
A typical health impact function might look like this:
where y0 is the baseline incidence, equal to the baseline incidence rate times the potentially
affected population; p is the effect estimate; and Ax is the estimated change in PM2 5. There
are other functional forms, but the basic elements remain the same.
G. 4. 1 Calculating Reductions in Premature Deaths
As in several recent air pollution health impact assessments (e.g., Kunzli et al., 2000;
EPA, 2004), we focus on the prospective cohort long-term exposure studies in deriving the
health impact function for the estimate of premature mortality. Cohort analyses are better
able to capture the full public health impact of exposure to air pollution over time (Kunzli et
al., 2001; NRC, 2002). We selected an effect estimate from the extended analysis of the
ACS cohort (Pope et al., 2002). This latest re-analysis of the ACS cohort data provides
additional refinements to the analysis of PM-related mortality by (a) extending the follow-up
period for the ACS study subjects to 16 years, which triples the size of the mortality data set;
(b) substantially increasing exposure data, including consideration for cohort exposure to
PM25 following implementation of PM25 standard in 1999; (c) controlling for a variety of
personal risk factors including occupational exposure and diet; and (d) using advanced
statistical methods to evaluate specific issues that can adversely affect risk estimates,
including the possibility of spatial autocorrelation of survival times in communities located
near each other. The effect estimate from Pope et al. (2002) quantifies the relationship
between annual mean PM2 5 levels and all-cause mortality in adults 30 and older. We
G-12
-------
selected the effect estimate estimated using the measure of PM representing average
exposure over the follow-up period, calculated as the average of 1979-1984 and 1999-2000
PM2 5 levels. The effect estimate from this study is 0.0058, which is equivalent to a relative
risk of 1.06 for a 10 |ig change in PM25. Although there are other cohort-based studies of the
relationship between PM25 and mortality, none provide the same level of population and
geographic coverage as the ACS study.
Age, cause, and county-specific mortality rates were obtained from CDC for the years
1996 through 1998. CDC maintains an online data repository of health statistics, CDC
Wonder, accessible at http://wonder.cdc.gov/. The mortality rates provided are derived from
U.S. death records and U.S. Census Bureau postcensal population estimates. Mortality rates
were averaged across 3 years (1996 through 1998) to provide more stable estimates. When
estimating rates for age groups that differed from the CDC Wonder groupings, we assumed
that rates were uniform across all ages in the reported age group. For example, to estimate
mortality rates for individuals ages 30 and up, we scaled the 25- to 34-year old death count
and population by one-half and then generated a population-weighted mortality rate using
data for the older age groups.
The reductions in incidence of premature mortality within each age group associated
with the CAIR reductions in PM2 5 in 2010 and 2015 are summarized in Table G-2.
G.4.2 Calculating Changes in Life Years from Direct Reductions in PM2 ^-Related
Mortality Risk
To calculate changes in life years associated with a given change in air pollution, we
used a life table approach coupled with age-specific estimates of reductions in premature
mortality. We began with the complete unabridged life table for the United States in 2000,
obtained from CDC (CDC, 2002). For each 1-year age interval (e.g., zero to one, one to two)
the life table provides estimates of the baseline probability of dying during the interval,
person years lived in the interval, and remaining life expectancy. From this unabridged life
table, we constructed an abridged life table to match the age intervals for which we have
predictions of changes in incidence of premature mortality. We used the abridgement
method described in CDC (2002). Table G-3 presents the abridged life table for 10-year age
intervals for adults over 30 (to match the Pope et al. [2002] study population). Note that the
abridgement actually includes one 5-year interval, covering adults 30 to 34, with the
remaining age intervals covering 10 years each. This is to provide conformity with the age
intervals available for mortality rates.
G-13
-------
Table G-2. Estimated Reduction in Incidence of All-cause Premature Mortality
Associated with the Clean Air Interstate Rule
Age Interval
30-34
35-44
45-54
55-64
65-74
75-84
85+
Total
Reduction in All-Cause Premature Mortality
(95% CI)
2010
96
(33 - 160)
370
(120-610)
870
(300- 1,400)
1,700
(580-2,800)
2,400
(830-4,000)
3,400
(1,100-5,600)
3,800
(1,300-6,400)
12,700
(4,300-21,000)
2015
120
(42 - 200)
420
(140-700)
1,000
(340- 1,700)
2,300
(790-3,800)
3,600
(1,200-6,000)
4,100
(1,400-6,800)
5,100
(1,700-8,400)
17,000
(5,700-28,000)
From the abridged life table (Table G-3), we obtained the remaining life expectancy
for each age cohort, conditional on surviving to that age. This is then the number of life
years lost for an individual in the general population dying during that age interval. This
information can then be combined with the estimated number of premature deaths in each
age interval calculated with BenMAP (see previous subsection). Total life years gained will
then be the sum of life years gained in each age interval:
Total Life Years = LEt x Mt
where LE; is the remaining life expectancy for age interval /', M; is the change in incidence of
mortality in age interval /', and N is the number of age intervals.
G-14
-------
Table G-3. Abridged Life Table for the Total Population, United States, 2000
Age Interval
Start Age End Age
30 35
35 45
45 55
55 65
65 75
75 85
85 95
95 100
100+
Probability of
Dying
Between Ages
x to x+1
Qx
0.00577
0.01979
0.04303
0.09858
0.21779
0.45584
0.79256
0.75441
1.00000
Number
Surviving to
Agex
Ix
97,696
97,132
95,210
91,113
82,131
64,244
34,959
7,252
1,781
Number
Dying
Between Ages
x to x+1
dx
564
1,922
4,097
8,982
17,887
29,285
27,707
5,471
1,781
Person Years
Lived
Between Ages
x to x+1
Lx
487,130
962,882
934,026
872,003
740,927
505,278
196,269
20,388
4,636
Total Number
of Person
Years Lived
Above Age x
Tx
4,723,539
4,236,409
3,273,527
2,339,501
1,467,498
726,571
221,293
25,024
4,636
Expectation
of Life at
Agex
ex
48.3
43.6
34.4
25.7
17.9
11.3
6.3
3.5
2.6
For the purposes of determining cost-effectiveness, it is also necessary to consider the
time-dependent nature of the gains in life years. Standard economic theory suggests that
benefits occurring in future years should be discounted relative to benefits occurring in the
present. OMB and EPA guidance suggest discount rates of three and seven percent. As
noted earlier, we present gains in future life years discounted at 3 percent. Results based on
7 percent are included in the summary and the overall impact of a 7 percent rate is
summarized in Table G-16. Selection of a 3 percent discount rate is also consistent with
recommendations from the U.S. Public Health Service Panel on Cost Effectiveness in Health
and Medicine (Gold et al., 1996).
Discounted total life years gained is calculated as follows:
tLE
Discounted LY = e~rtdt,
Jo
where r is the discount rate, equal to 0.03 in this case, t indicates time, and LE is the life
expectancy at the time when the premature death would have occurred. Life years are further
discounted to account for the lag between the reduction in ambient PM2 5 and the reduction in
mortality risk. We use the same 20-year segmented lag structure that is used in the benefit-
cost analysis (see Chapter 4).
G-15
-------
The most complete estimate of the impacts of PM25 on life years is calculated using
the Pope et al. (2002) C-R function relating all-cause mortality in adults 30 and over with
ambient PM2 5 concentrations averaged over the periods 1979-1983 and 1999-2000. Use of
all-cause mortality is appropriate if there are no differences in the life expectancy of
individuals dying from air pollution-related causes and those dying from other causes. The
argument that long-term exposure to PM2 5 may affect mainly individuals with serious
preexisting illnesses is not supported by current empirical studies. For example, the Krewski
et al. (2000) ACS reanalysis suggests that the mortality risk is no greater for those with
preexisting illness at time of enrollment in the study. Life expectancy for the general
population in fact includes individuals with serious chronic illness. Mortality rates for the
general population then reflect prevalence of chronic disease, and as populations age the
prevalence of chronic disease increases.
The only reason one might use a lower life expectancy is if the population at risk
from air pollution was limited solely to those with preexisting disease. Also, note that the
OMB Circular A-4 notes that "if QALYs are used to evaluate a lifesaving rule aimed at a
population that happens to experience a high rate of disability (i.e., where the rule is not
designed to affect the disability), the number of life years saved should not necessarily be
diminished simply because the rule saves lives of people with life-shortening disabilities.
Both analytic simplicity and fairness suggest that the estimate number of life years saved for
the disabled population should be based on average life expectancy information for the
relevant age cohorts." As such, use of a general population life expectancy is preferred over
disability-specific life expectancies. Our primary life years calculations are thus consistent
with the concept of not penalizing individuals with disabling chronic health conditions by
assessing them reduced benefits of mortality risk reductions.
For this analysis, direct impacts on life expectancy are measured only through the
estimated change in mortality risk based on the Pope et al. (2002) C-R function. The SAB-
FIES has advised against including additional gains in life expectancy due to reductions in
incidence of chronic disease or nonfatal heart attacks (EPA-SAB-COUNCIL-ADV-04-002).
Although reductions in these endpoints are likely to result in increased life expectancy, the
FIES has suggested that the cohort design and relatively long follow-up period in the Pope et
al. study should capture any life-prolonging impacts associated with those endpoints.
Impacts of CB and nonfatal heart attacks on quality of life will be captured separately in the
QALY calculation as years lived with improved quality of life. The methods for calculating
this benefit are discussed below.
G-16
-------
G.4.2.1 Should Life Years Gained Be Adjusted for Initial Health Status ?
The methods outlined above provide estimates of the total number of life years
gained in a population, regardless of the quality of those life years, or equivalently, assuming
that all life years gained are in perfect health. In some CEAs (Cohen, Hammitt, and Levy,
2003; Coyle et al., 2003), analysts have adjusted the number of life years gained to reflect the
fact that 1) the general public is not in perfect health and thus "healthy" life years are less
than total life years gained and 2) those affected by air pollution may be in a worse health
state than the general population and therefore will not gain as many "healthy" life years
adjusted for quality, from an air pollution reduction. This adjustment, which converts life
years gained into QALYs, raises a number of serious ethical issues. Proponents of QALYs
have promoted the nondiscriminatory nature of QALYs in evaluating improvements in
quality of life (e.g., an improvement from a score of 0.2 to 0.4 is equivalent to an
improvement from 0.8 to 1.0), so the starting health status does not affect the evaluation of
interventions that improve quality of life. However, for life-extending interventions, the
gains in QALY will be directly proportional to the baseline health state (e.g., an individual
with a 30-year life expectancy and a starting health status of 0.5 will gain exactly half the
QALYs of an individual with the same life expectancy and a starting health status of 1.0 for a
similar life-extending intervention). This is troubling because it imposes an additional
penalty for those already suffering from disabling conditions. Brock (2002) notes that "the
problem of disability discrimination represents a deep and unresolved problem for resource
prioritization."
OMB (2003) has recognized this issue in their Circular A-4 guidance, which includes
the following statement:
When CEA is performed in specific rulemaking contexts, you should be prepared to
make appropriate adjustments to ensure fair treatment of all segments of the
population. Fairness is important in the choice and execution of effectiveness
measures. For example, if QALYs are used to evaluate a lifesaving rule aimed at a
population that happens to experience a high rate of disability (i.e., where the rule is
not designed to affect the disability), the number of life years saved should not
necessarily be diminished simply because the rule saves the lives of people with
life-shortening disabilities. Both analytic simplicity and fairness suggest that the
estimated number of life years saved for the disabled population should be based on
average life expectancy information for the relevant age cohorts. More generally,
when numeric adjustments are made for life expectancy or quality of life, analysts
G-17
-------
should prefer use of population averages rather than information derived from
subgroups dominated by a particular demographic or income group, (p. 13)
This suggests two adjustments to the standard QALY methodology: one adjusting the
relevant life expectancy of the affected population, and the other affecting the baseline
quality of life for the affected population.
In addition to the issue of fairness, potential measurement issues are specific to the air
pollution context that might argue for caution in applying quality-of-life adjustments to life
years gained due to air pollution reductions. A number of epidemiological and toxicological
studies link exposure to air pollution with chronic diseases, such as CB and atherosclerosis
(Abbey et al., 1995; Schwartz, 1993; Suwa et al., 2002). If these same individuals with
chronic disease caused by exposure to air pollution are then at increased risk of premature
death from air pollution, there is an important dimension of "double jeopardy" involved in
determining the correct baseline for assessing QALYs lost to air pollution (see Singer et al.
[1995] for a broader discussion of the double-jeopardy argument).
Analyses estimating mortality from acute exposures that ignore the effects of long-
term exposure on morbidity may understate the health impacts of reducing air pollution.
Individuals exposed to chronically elevated levels of air pollution may realize an increased
risk of death and chronic disease throughout life. If at some age they contract heart (or some
other chronic) disease as a result of the exposure to air pollution, they will from that point
forward have both reduced life expectancy and reduced quality of life. The benefit to that
individual from reducing lifetime exposure to air pollution would be the increase in life
expectancy plus the increase in quality of life over the full period of increased life
expectancy. If the QALY loss is determined based on the underlying chronic condition and
life expectancy without regard to the fact that the person would never have been in that state
without long-term exposure to elevated air pollution, then the person is placed in double
jeopardy. In other words, air pollution has placed more people in the susceptible pool, but
then we penalize those people in evaluating policies by treating their subsequent deaths as
less valuable, adding insult to injury, and potentially downplaying the importance of life
expectancy losses due to air pollution. If the risk of chronic disease and risk of death are
considered together, then there is no conceptual problem with measuring QALYs, but this
has not been the case in recent applications of QALYs to air pollution (Carrothers, Evans,
and Graham, 2002; Coyle et al., 2003). The use of QALYs thus highlights the need for a
better understanding of the relationship between chronic disease and long-term exposure and
G-18
-------
suggests that analyses need to consider morbidity and mortality jointly, rather than treating
each as a separate endpoint (this is an issue for current benefit-cost approaches as well).
Because of the fairness and measurement concerns discussed above, for the purposes
of this analysis, we do not reduce the number of life years gained to reflect any differences in
underlying health status that might reduce quality of life in remaining years. Thus, we
maintain the assumption that all direct gains in life years resulting from mortality risk
reductions will be assigned a weight of 1.0. The U.S. Public Health Service Panel on Cost
Effectiveness in Health and Medicine recommends that "since lives saved or extended by an
intervention will not be in perfect health, a saved life year will count as less than 1 full
QALY" (Gold et al., 1996). However, for the purposes of this analysis, we propose an
alternative to the traditional aggregate QALY metric that keeps separate quality adjustments
to life expectancy and gains in life expectancy. As such, we do not make any adjustments to
life years gained to reflect the less than perfect health of the general population. Gains in
quality of life will be addressed as they accrue because of reductions in the incidence of
chronic diseases. This is an explicit equity choice in the treatment of issues associated with
quality-of-life adjustments for increases in life expectancy that still capitalizes on the ability
of QALYs to capture both morbidity and mortality impacts in a single effectiveness measure.
G.5 Calculating Changes in the Quality of Life Years (Morbidity)
In addition to directly measuring the quantity of life gained, measured by life years, it
may also be informative to measure gains in the quality of life. Reducing air pollution also
leads to reductions in serious illnesses that affect quality of life. These include CB and
cardiovascular disease, for which we are able to quantify changes in the incidence of nonfatal
heart attacks. To capture these important benefits in the measure of effectiveness, they must
first be converted into a life-year equivalent so that they can be combined with the direct
gains in life expectancy.
For this analysis, we developed estimates of the QALYs gained from reductions in
the incidence of CB and nonfatal heart attacks associated with reductions in ambient PM25.
In general, QALY calculations require four elements:
1. the estimated change in incidence of the health condition,
2. the duration of the health condition,
3. the quality-of-life weight with the health condition, and
G-19
-------
4. the quality-of-life weight without the health condition (i.e., the baseline health
state).
The first element is derived using the health impact function approach. The second element
is based on the medical literature for each health condition. The third and fourth elements
are derived from the medical cost-effectiveness and cost-utility literature. In the following
two subsections, we discuss the choices of elements for CB and nonfatal heart attacks.
The preferred source of quality-of-life weights are those based on community
preferences, rather than patient or clinician ratings (Gold et al., 1996). Several methods are
used to estimate quality-of-life weights. These include rating scale, standard gamble, time
trade-off, and person trade-off approaches (Gold, Stevenson, and Fryback, 2002). Only the
standard gamble approach is completely consistent with utility theory. However, the time
trade-off method has also been widely applied in eliciting community preferences (Gold,
Stevenson, and Fryback, 2002).
Quality-of-life weights can be directly elicited for individual specific health states or
for a more general set of activity restrictions and health states that can then be used to
construct QALY weights for specific conditions (Horsman et al., 2003; Kind, 1996). For this
analysis, we used weights based on community-based preferences, using time trade-off or
standard gamble when available. In some cases, we used patient or clinician ratings when no
community preference-based weights were available. Sources for weights are discussed in
more detail below. Table G-4 summarizes the key inputs for calculating QALYs associated
with chronic health endpoints.
G. 5.1 Calculating QAL Ys Associated with Reductions in the Incidence of Chronic
Bronchitis
CB is characterized by mucus in the lungs and a persistent wet cough for at least 3
months a year for several years in a row. CB affects an estimated 5 percent of the U.S.
population (American Lung Association, 1999). For gains in quality of life resulting from
reduced incidences of PM-induced CB, discounted QALYs are calculated as
DISCOUNTED QALY GAINED = £ ACS,, x D* x (w;. - wf
G-20
-------
Table G-4. Summary of Key Parameters Used in QALY Calculations for Chronic
Disease Endpoints
Parameter
Discount rate
Quality of life
preference score for
chronic bronchitis
Duration of acute phase
of acute myocardial
infarction (AMI)
Probability of CHF post
AMI
Probability of angina
post AMI
Quality-of-life
preference score for
post- AMI with CHF (no
angina)
Quality-of-life
preference score for
post-AMI with CHF and
angina
Quality-of-life
preference score for
post-AMI with angina
(no CHF)
Quality-of-life
preference score for
post-AMI (no angina, no
CHF)
Value(s)
0.03 (0.07
sensitivity
analysis)
0.5-0.7
5.5 days -22
days
0.2
0.51
0.80-0.89
0.76-0.85
0.7-0.89
0.93
Source(s)
Gold et al. (1996), U.S. EPA (2000), U.S. OMB (2003)
Triangular distribution centered at 0.7 with upper bound at 0.9
(Vos, 1999a) (slightly better than a mild/moderate case) and a
lower bound at 0.5 (average weight for a severe case based on
Vos [1999a] and Smith and Peske [1994])
Uniform distribution with lower bound based on average
length of stay for an AMI (AHRQ, 2000) and upper bound
based on Vos (1999b).
Vos, 1999a (WHO Burden of Disease Study, based on Cowie
etal., 1997)
American Heart Association, 2003
(Calculated as the population with angina divided by the total
population with heart disease)
Uniform distribution with lower bound at 0.80 (Stinnett et al.,
1996) and upper bound at 0.89 (Kuntz et al., 1996). Both
studies used the time trade-off elicitation method.
Uniform distribution with lower bound at 0.76 (Stinnett et al.,
1996, adjusted for severity) and upper bound at 0.85 (Kuntz et
al., 1996). Both studies used the time trade-off elicitation
method.
Uniform distribution with lower bound at 0.7, based on the
standard gamble elicitation method (Pliskin, Stason, and
Weinstein, 1981) and upper bound at 0.89, based on the time
trade-off method (Kuntz et al., 1996).
Only one value available from the literature. Thus, no
distribution is specified. Source of value is Kuntz et al.
(1996).
G-21
-------
where ACB; is the number of incidences of CB avoided in age interval i, w; is the average
QALY weight for age interval i, w^B is the QALY weight associated with CB, Dt is the
discounted duration of life with CB for individuals with onset of disease in age interval i,
f A
equal to e~n' dt, where D;is the duration of life with CB for individuals with onset of
J/=i
disease in age interval i.
A limited number of studies have estimated the impact of air pollution on new
incidences of CB. Schwartz (1993) and Abbey et al. (1995) provide evidence that long-term
PM exposure gives rise to the development of CB in the United States. Because this analysis
focuses on the impacts of reducing ambient PM2 5, only the Abbey et al. (1995) study is used,
because it is the only study focusing on the relationship between PM2 5 and new incidences of
CB. The number of cases of CB in each age interval is derived from applying the impact
function from Abbey et al. (1995), to the population in each age interval with the appropriate
baseline incidence rate.4 The effect estimate from the Abbey et al. (1995) study is 0.0137,
which, based on the logistic specification of the model, is equivalent to a relative risk of 1.15
for a 10 |ig change in PM25. Table G-5 presents the estimated reduction in new incidences of
CB associated with CAIR in 2010 and 2015.
CB is assumed to persist for the remainder of an affected individual's lifespan.
Duration of CB will thus equal life expectancy conditioned on having CB. CDC has
estimated that COPD (of which CB is one element) results in an average loss of life years
equal to 4.26 per COPD death, relative to a reference life expectancy of 75 years (CDC,
2003). Thus, we subtract 4.26 from the remaining life expectancy for each age group, up to
age 75. For age groups over 75, we apply the ratio of 4.26 to the life expectancy for the 65 to
74 year group (0.237) to the life expectancy for the 75 to 84 and 85 and up age groups to
estimate potential life years lost and then subtract that value from the base life expectancy.
Quality of life with chronic lung diseases has been examined in several studies. In an
analysis of the impacts of environmental exposures to contaminants, de Hollander et al.
(1999) assigned a weight of 0.69 to years lived with CB. This weight was based on
"Prevalence rates for CB were obtained from the 1999 National Health Interview Survey (American Lung
Association, 2002). Prevalence rates were available for three age groups: 18-44, 45-64, and 65 and older.
Prevalence rates per person for these groups were 0.0367 for 18-44, 0.0505 for 45-64, and 0.0587 for 65
and older. The incidence rate for new cases of CB (0.00378 per person) was taken directly from Abbey et
al. (1995).
G-22
-------
Table G-5. Estimated Reduction in Incidence of Chronic Bronchitis Associated with
the Clean Air Interstate Rule
Reduction in Incidence (95% Confidence Interval)
Age Interval
25-34
35-44
45-54
55-64
65-74
75-84
85+
Total
2010
1,100(31
1,400(41
1,600 (46
1,300(36
- 2,200)
- 2,800)
- 3,200)
- 2,500)
770(21- 1,500)
470(13
210(6
6,900 (190
-910)
-400)
- 14,000)
2015
1,400 (40 -
1,700 (47 -
1,900 (52 -
1,800 (49 -
1,100(32-
570(16-
270 (8 -
8,700 (240 -
2,800)
3,300)
3,700)
3,400)
2,200)
1,100)
530)
17,000)
physicians' evaluations of health states similar to CB. Salomon and Murray (2003)
estimated a pooled weight of 0.77 based on visual analogue scale, time trade-off, standard
gamble, and person trade-off techniques applied to a convenience sample of health
professionals. The Harvard Center for Risk Analysis catalog of preference scores reports a
weight of 0.40 for severe COPD, with a range from 0.2 to 0.8, based on the judgments of the
study's authors (Bell et al., 2001). The Victoria Burden of Disease (BoD) study used a
weight of 0.47 for severe COPD and 0.83 for mild to moderate COPD, based on an analysis
by Stouthard et al. (1997) of chronic diseases in Dutch populations (Vos, 1999a). Based on
the recommendations of Gold et al. (1996), quality-of-life weights based on community
preferences are preferred for CEA of interventions affecting broad populations. Use of
weights based on health professionals is not recommended. It is not clear from the Victoria
BoD study whether the weights used for COPD are based on community preferences or
judgments of health professionals. The Harvard catalog score is clearly identified as based
on author judgment. Given the lack of a clear preferred weight, we select a triangular
distribution centered at 0.7 with an upper bound at 0.9 (slightly better than a mild/moderate
case defined by the Victoria BoD study) and a lower bound at 0.5 based on the Victoria BoD
study. We will need additional empirical data on quality of life with chronic respiratory
diseases based on community preferences to improve our estimates.
G-23
-------
Selection of a reference weight for the general population without CB is somewhat
uncertain. It is clear that the general population is not in perfect health; however, there is
some uncertainty as to whether individuals' ratings of health states are in reference to a
perfect health state or to a generally achievable "normal" health state given age and general
health status. The U.S. Public Health Service Panel on Cost Effectiveness in Health and
Medicine recommends that "since lives saved or extended by an intervention will not be in
perfect health, a saved life year will count as less than 1 full QALY" (Gold et al., 1996).
Following Carrothers, Evans, and Graham (2002), we assumed that the reference weight for
the general population without CB is 0.95. To allow for uncertainty in this parameter, we
assigned a triangular distribution around this weight, bounded by 0.9 and 1.0. Note that the
reference weight for the general population is used solely to determine the incremental
quality-of-life improvement applied to the duration of life that would have been lived with
the chronic disease. For example, if CB has a quality-of-life weight of 0.7 relative to a
reference quality-of-life weight of 0.9, then the incremental quality-of-life improvement in
0.2. If the reference quality-of-life weight is 0.95, then the incremental quality-of-life
improvement is 0.25. As noted above, the population is assumed to have a reference weight
of 1.0 for all life years gained due to mortality risk reductions.
We present discounted QALYs over the duration of the lifespan with CB using a 3
percent discount rate. Based on the assumptions defined above, we used Monte Carlo
simulation methods as implemented in the Crystal Ball™ software program to develop the
distribution of QALYs gained per incidence of CB for each age interval.5 Based on the
assumptions defined above, the mean 3 percent discounted QALY gained per incidence of
CB for each age interval along with the 95 percent confidence interval resulting from the
Monte Carlo simulation is presented in Table G-6. Table G-6 presents both the undiscounted
and discounted QALYs gained per incidence.
G. 5.2 Calculating QAL Ys Associated with Reductions in the Incidence ofNonfatal
Myocardial Infarctions
Nonfatal heart attacks, or acute myocardial infarctions, require more complicated
calculations to derive estimates of QALY impacts. The actual heart attack, which results
when an area of the heart muscle dies or is permanently damaged because of oxygen
deprivation, and subsequent emergency care are of relatively short duration. Many heart
5Monte Carlo simulation uses random sampling from distributions of parameters to characterize the effects of
uncertainty on output variables. For more details, see Gentile (1998).
G-24
-------
Table G-6. QALYs Gained per Avoided Incidence of CB
Age Interval
Start Age
25
35
45
55
65
75
85+
End Age
34
44
54
64
74
84
QALYs Gained per Incidence
Undiscounted
12.15
(4.40-19.95)
9.91
(3.54-16.10)
7.49
(2.71-12.34)
5.36
(1.95-8.80)
3.40
(1.22-5.64)
2.15
(0.77-3.49)
0.79
(0.27-1.29)
Discounted (3%)
6.52
(2.36-10.71)
5.94
(2.12-9.66)
5.03
(1.82-8.29)
4.03
(1.47-6.61)
2.84
(1.02-4.71)
1.92
(0.69-3.13)
0.77
(0.26-1.25)
attacks result in sudden death. However, for survivors, the long-term impacts of advanced
CHD are potentially of long duration and can result in significant losses in quality of life and
life expectancy.
In this phase of the analysis, we did not independently estimate the gains in life
expectancy associated with reductions in nonfatal heart attacks. Based on recommendations
from the SAB-HES, we assumed that all gains in life expectancy are captured in the
estimates of reduced mortality risk provided by the Pope et al. (2002) analysis. We only
estimate the change in quality of life over the period of life affected by the occurrence of a
heart attack. This may understate the QALY impacts of nonfatal heart attacks but ensures
that the overall QALY impact estimates across endpoints do not double-count potential life-
year gains.
Our approach adapts a CHD model developed for the Victoria Burden of Disease
study (Vos, 1999b). This model accounts for the lost quality of life during the heart attack
and the possible health states following the heart attack. Figure G-l shows the heart attack
QALY model in diagrammatic form. The total gain in QALYs is calculated as:
G-25
-------
Acute Treatment Stage
Chronic Post-AMI Follow up Stage
Post AMI QALY with Angina and CHF
Nonfatal AMI
Post AMI QALY with CHF without Angina
Post AMI QALY with Angina without CHF
Post AMI QALY without Angina or CHF
Figure G-l. Decision Tree Used in Modeling Gains in QALYs from Reduced Incidence
of Nonfatal Acute Myocardial Infarctions
DISCOUNTED AMI QAL Y GAINED =
4
V V
AAMI x
x W -
postAMI
where AAMI; is the number of nonfatal acute myocardial infarctions avoided in age interval
/', wf^1 is the QALY weight associated with the acute phase of the AMI, PJ is the probability
of being in they'th post-AMI status, wjost is the QALY weight associated with post-AMI
G-26
-------
health status/, w; is the average QALY weight for age interval i, D = e~n dt, the
J/=i
discounted value of Z)/4^7, the duration of the acute phase of the AMI, and
_ r\poslAMI
Y)_p°s - I e~rtdt, is the discounted value of Df"Am^ the duration of post-AMI
health status/
Nonfatal heart attacks have been linked with short-term exposures to PM2 5 in the
United States (Peters et al., 2001) and other countries (Poloniecki et al., 1997). We used a
recent study by Peters et al. (2001) as the basis for the impact function estimating the
relationship between PM25 and nonfatal heart attacks. Peters et al. is the only available U.S.
study to provide a specific estimate for heart attacks. Other studies, such as Samet et al.
(2000) and Moolgavkar (2000), show a consistent relationship between all cardiovascular
hospital admissions, including for nonfatal heart attacks, and PM. Given the lasting impact
of a heart attack on longer-term health costs and earnings, we chose to provide a separate
estimate for nonfatal heart attacks based on the single available U.S. effect estimate. The
finding of a specific impact on heart attacks is consistent with hospital admission and other
studies showing relationships between fine particles and cardiovascular effects both within
and outside the United States. These studies provide a weight of evidence for this type of
effect. Several epidemiologic studies (Liao et al., 1999; Gold et al., 2000; Magari et al.,
2001) have shown that heart rate variability (an indicator of how much the heart is able to
speed up or slow down in response to momentary stresses) is negatively related to PM levels.
Heart rate variability is a risk factor for heart attacks and other CHDs (Carthenon et al.,
2002; Dekker et al., 2000; Liao et al., 1997, Tsuji et al., 1996). As such, significant impacts
of PM on heart rate variability are consistent with an increased risk of heart attacks.
The number of avoided nonfatal AMI in each age interval is derived from applying
the impact function from Peters et al. (2001) to the population in each age interval with the
appropriate baseline incidence rate.6 The effect estimate from the Peters et al. (2001) study
is 0.0241, which, based on the logistic specification of the model, is equivalent to a relative
6Daily nonfatal myocardial infarction incidence rates per person were obtained from the 1999 National Hospital
Discharge Survey (assuming all diagnosed nonfatal AMI visit the hospital). Age-specific rates for four
regions are used in the analysis. Regional averages for populations 18 and older are 0.0000159 for the
Northeast, 0.0000135 for the Midwest, 0.0000111 for the South, and 0.0000100 for the West.
G-27
-------
risk of 1.27 for a 10 |ig change in PM25. Table G-7 presents the estimated reduction in
nonfatal AMI associated with CAIR in 2010 and 2015.
Table G-7. Estimated Reduction in Nonfatal Acute Myocardial Infarctions Associated
with the Clean Air Interstate Rule
Age Interval
18-24
25-34
35-44
45-54
55-64
65-74
75-84
85+
Total
Reduction in Incidence*(95% Confidence Interval)
2010
9(2- 16)
92 (23 - 160)
630(160-1,100)
2,400 (600 - 4,200)
4,000 (990 - 6,900)
4,000 (990 - 6,900)
3,800 (940 - 6,500)
2,300 (570 - 4,000)
17,000(4,300-30,000)
2015
11(3- 18)
120 (29 - 200)
700(180- 1,200)
2,800 (690 - 4,800)
5,300(1,300-9,200)
5,800(1,400- 10,000)
4,600(1,100-7,900)
3,000 (750 - 5,200)
22,000 (5,600 - 39,000)
Acute myocardial infarction results in significant loss of quality of life for a relatively
short duration. The WHO Global Burden of Disease study, as reported in Vos (1999b),
assumes that the acute phase of an acute myocardial infarction lasts for 0.06 years, or around
22 days. An alternative assumption is the acute phase is characterized by the average length
of hospital stay for an AMI in the United States, which is 5.5 days, based on data from the
Agency for Healthcare Research and Quality's Healthcare Cost and Utilization Project
(HCUP).7 We assumed a distribution of acute phase duration characterized by a uniform
distribution between 5.5 and 22 days, noting that due to earlier discharges and in-home
therapy available in the United States, duration of reduced quality of life may continue after
discharge from the hospital. In the period during and directly following an AMI (the acute
phase), we assigned a quality of life weight equal to 0.605, consistent with the weight for the
period in treatment during and immediately after an attack (Vos, 1999b).
7Average length of stay estimated from the HCUP data includes all discharges, including those due to death. As
such, the 5.5-day average length of stay is likely an underestimate of the average length of stay for AMI
admissions where the patient is discharged alive.
G-28
-------
During the post-AMI period, a number of different health states can determine the
loss in quality of life. We chose to classify post-AMI health status into four states defined by
the presence or absence of angina and congestive heart failure (CHF). This makes a very
explicit assumption that without the occurrence of an AMI, individuals would not experience
either angina or CHF. If in fact individuals already have CHF or angina, then the quality of
life gained will be overstated. We do not have information about the percentage of the
population have been diagnosed with angina or CHF with no occurrence of an AMI. Nor do
we have information on what proportion of the heart attacks occurring due to PM exposure
are first heart attacks versus repeat attacks. Probabilities for the four post-AMI health states
sum to one.
Given the occurrence of a nonfatal AMI, the probability of congestive heart failure is
set at 0.2, following the heart disease model developed by Vos (1999b). The probability is
based on a study by Cowie et al. (1997), which estimated that 20 percent of those surviving
AMI develop heart failure, based on an analysis of the results of the Framingham Heart
Study.
The probability of angina is based on the prevalence rate of angina in the U.S.
population. Using data from the American Heart Association, we calculated the prevalence
rate for angina by dividing the estimated number of people with angina (6.6 million) by the
estimated number of people with CHD of all types (12.9 million). We then assumed that the
prevalence of angina in the population surviving an AMI is similar to the prevalence of
angina in the total population with CHD. The estimated prevalence rate is 51 percent, so the
probability of angina is 0.51.
Combining these factors leads to the probabilities for each of the four health states as
follows:
I. Post AMI with CHF and angina = 0.102
II. Post AMI with CHF without angina = 0.098
III. Post AMI with angina without CHF = 0.408
IV. Post AMI without angina or CHF = 0.392
Duration of post-AMI health states varies, based in part on assumptions regarding life
expectancy with post-AMI complicating health conditions. Based on the model used for
established market economies (EME) in the WHO Global Burden of Disease study, as
reported in Vos (1999b), we assumed that individuals with CHF have a relatively short
G-29
-------
remaining life expectancy and thus a relatively short period with reduced quality of life
(recall that gains in life expectancy are assumed to be captured by the cohort estimates of
reduced mortality risk). Table G-8 provides the duration (both discounted and undiscounted)
of CHF assumed for post-AMI cases by age interval.
Table G-8. Assumed Duration of Congestive Heart Failure
Age Interval
Start Age End Age
18 24
25 34
35 44
45 54
55 64
65 74
75 84
85+
Duration
Undiscounted
7.11
6.98
6.49
5.31
1.96
1.71
1.52
1.52
of Heart Failure
Discounted (3%)
6.51
6.40
6.00
4.99
1.93
1.69
1.50
1.50
Duration of health states without CHF is assumed to be equal to the life expectancy
of individuals conditional on surviving an AMI. Ganz et al. (2000) note that "Because
patients with a history of myocardial infarction have a higher chance of dying of CHD that is
unrelated to recurrent myocardial infarction (for example, arrhythmia), this cohort has a
higher risk for death from causes other than myocardial infarction or stroke than does an
unselected population." They go on to specify a mortality risk ratio of 1.52 for mortality
from other causes for the cohort of individuals with a previous (nonfatal) AMI. The risk
ratio is relative to all-cause mortality for an age-matched unselected population (i.e., general
population). We adopted the same ratios and applied them to each age-specific all-cause
mortality rate to derive life expectancies (both discounted and undiscounted) for each age
group after an AMI, presented in Table G-9. These life expectancies are then used to
represent the duration of non-CHF post-AMI health states (III and IV).
For the four post-AMI health states, we used QALY weights based on preferences for
the combined conditions characterizing each health state. A number of estimates of QALY
weights are available for post-AMI health conditions.
G-30
-------
Table G-9. Assumed Duration of Non-CHF Post-AMI Health States
Age Interval
Start Age End Age
18 24
25 34
35 44
45 54
55 64
65 74
75 84
85+
Post-AMI Life
Undiscounted
55.5
46.1
36.8
27.9
19.8
12.8
7.4
3.6
Expectancy (non-CHF)
Discounted (3%)
27.68
25.54
22.76
19.28
15.21
10.82
6.75
3.47
The first two health states are characterized by the presence of CHF, with or without
angina. The Harvard Center for Risk Analysis catalog of preference scores provides several
specific weights for CHF with and without mild or severe angina and one set specific to
post-AMI CHF. Following the Victoria Burden of Disease model, we assumed that most
cases of angina will be treated and thus kept at a mild to moderate state. We thus focused
our selection on QALY weights for mild to moderate angina. The Harvard database includes
two sets of community preference-based scores for CHF (Stinnett et al., 1996; Kuntz et al.,
1996). The scores for CHF with angina range from 0.736 to 0.85. The lower of the two
scores is based on angina in general with no delineation by severity. Based on the range of
the scores for mild to severe cases of angina in the second study, one can infer that an
average case of angina has a score around 0.96 of the score for a mild case. Applying this
adjustment raises the lower end of the range of preference scores for a mild case of angina to
0.76. We selected a uniform distribution over the range 0.76 to 0.85 for CHF with mild
angina, with a midpoint of 0.81. The same two studies in the Harvard catalog also provide
weights for CHF without angina. These scores range from 0.801 to 0.89. We selected a
uniform distribution over this range, with a midpoint of 0.85.
The third health state is characterized by angina, without the presence of CHF. The
Harvard catalog includes five sets of community preference-based scores for angina, one that
specifies scores for both mild and severe angina (Kuntz et al., 1996), one that specifies mild
angina only (Pliskin, Stason, and Weinstein, 1981), one that specifies severe angina only
(Cohen, Breall, and Ho, 1994), and two that specify angina with no severity classification
G-31
-------
(Salkeld, Phongsavan, and Oldenburg, 1997; Stinnett et al., 1996). With the exception of the
Pliskin, Stason, and Weinstein score, all of the angina scores are based on the time trade-off
method of elicitation. The Pliskin, Stason, and Weinstein score is based on the standard
gamble elicitation method. The scores for the nonspecific severity angina fall within the
range of the two scores for mild angina specifically. Thus, we used the range of mild angina
scores as the endpoints of a uniform distribution. The range of mild angina scores is from
0.7 to 0.89, with a midpoint of 0.80.
For the fourth health state, characterized by the absence of CHF and/or angina, there
is only one relevant community preference score available from the Harvard catalog. This
score is 0.93, derived from a time trade-off elicitation (Kuntz et al., 1996). Insufficient
information is available to provide a distribution for this weight; therefore, it is treated as a
fixed value.
Similar to CB, we assumed that the reference weight for the general population
without AMI is 0.95. To allow for uncertainty in this parameter, we assigned a triangular
distribution around this weight, bounded by 0.9 and 1.0.
Based on the assumptions defined above, we used Monte Carlo simulation methods
as implemented in the Crystal Ball™ software program to develop the distribution of
QALYs gained per incidence of nonfatal AMI for each age interval. For the Monte Carlo
simulation, all distributions were assumed to be independent. The mean QALYs gained per
incidence of nonfatal AMI for each age interval is presented in Table G-10, along with the 95
percent confidence interval resulting from the Monte Carlo simulation. Table G-10 presents
both the undiscounted and discounted QALYs gained per incidence.
G.6 Cost-Effectiveness Analysis
Given the estimates of changes in life expectancy and quality of life, the next step is
to aggregate life expectancy and quality-of-life gains to form an effectiveness measure that
can be compared to costs to develop cost-effectiveness ratios. This section discusses the
proper characterization of the combined effectiveness measure and the appropriate
calculation of the numerator of the cost-effectiveness ratio.
G-32
-------
Table G-10. QALYs Gained per Avoided Nonfatal Myocardial Infarction
Age Interval
Start Age
18
25
35
45
55
65
75
85+
End Age
24
34
44
54
64
74
84
QALYs Gained
Undiscounted
4.18
(1.24-7.09)
3.48
(1.09-5.87)
2.81
(0.88-4.74)
2.14
(0.67-3.61)
1.49
(0.42-2.52)
0.97
(0.30-1.64)
0.59
(0.20-0.97)
0.32
(0.13-0.50)
per Incidence"
Discounted (3%)
2.17
(0.70-3.62)
2.00
(0.68-3.33)
1.79
(0.60-2.99)
1.52
(0.51-2.53)
1.16
(0.34-1.95)
0.83
(0.26-1.39)
0.54
(0.19-0.89)
0.31
(0.13-0.49)
Mean of Monte Carlo generated distribution; 95% confidence interval presented in parentheses.
G. 6.1 Aggregating Life Expectancy and Quality-of-Life Gains
To develop an integrated measure of changes in health, we simply sum together the
gains in life years from reduced mortality risk in each age interval with the gains in QALYs
from reductions in incidence of CB and acute myocardial infarctions. The resulting measure
of effectiveness then forms the denominator in the cost-effectiveness ratio. What is this
combined measure of effectiveness? It is not a QALY measure in a strict sense, because we
have not adjusted life-expectancy gains for preexisting health status (quality of life). It is
however, an effectiveness measure that adds to the standard life years calculation a scaled
morbidity equivalent. Thus, we term the aggregate measure morbidity inclusive life years, or
MILYs. Alternatively, the combined measure could be considered as QALYs with an
assumption that the community preference weight for all life-expectancy gains is 1.0. If one
considers that this weight might be considered to be a "fair" treatment of those with
G-33
-------
preexisting disabilities, the effectiveness measure might be termed "fair QALY" gained.
However, this implies that all aspects of fairness have been addressed, and there are clearly
other issues with the fairness of QALYs (or other effectiveness measures) that are not
addressed in this simple adjustment. The MILY measure violates some of the properties
used in deriving QALY weights, such as linear substitution between quality of life and
quantity of life. However, in aggregating life expectancy and quality-of-life gains, it merely
represents an alternative social weighting that is consistent with the spirit of the recent OMB
guidance on CEA. The guidance notes that "fairness is important in the choice and execution
of effectiveness measures" (OMB, 2003). The resulting aggregate measure of effectiveness
will not be consistent with a strict utility interpretation of QALYs; however, it may still be a
useful index of effectiveness.
Applying the life expectancies and distributions of QALYs per incidence for CB and
AMI to estimated distributions of incidences yields distributions of life expectancy and
QALYs gained due to CAIR. These distributions reflect both the quantified uncertainty in
incidence estimates and the quantified uncertainty in QALYs gained per incidence.
For the CAIR 2010 analysis year, Table G-l 1 presents the mean 3 percent discounted
MILYs gained for each age interval, broken out by life expectancy and quality-of-life
categories. Note that quality-of-life gains occur from age 18 and up, while life expectancy
gains accrue only after age 29. This is based on the ages of the study populations in the
underlying epidemiological studies. It is unlikely that such discontinuities exist in reality,
but to avoid overstating effectiveness, we chose to limit the life-expectancy gains to those
occurring in the population 30 and over and the morbidity gains to the specific adult
populations examined in the studies. Table G-l2 provides the same information for the 2015
analysis year.
It is worth noting that around a third of mortality-related benefits are due to
reductions in premature deaths among those 75 and older, while only 7 percent of morbidity
benefits occur in this age group. This is due to two factors: (1) the relatively low baseline
mortality rates in populations under 75, and (2) the relatively constant baseline rates of
chronic disease coupled with the relatively long period of life that is lived with increased
quality of life without CB and advanced heart disease.
G-34
-------
Table G-ll. Estimated Gains in 3 Percent Discounted MILYs Associated with the
Clean Air Interstate Rule in 2010a
Age
18-24
25-34
35^4
45-54
55-64
65-74
75-84
85+
Total
Life Years Gained
from Mortality Risk
Reductions
(95% CI)
—
2,500
(880-4,100)
9,100
(3,100-15,000)
19,000
(6,600-31,000)
31,000
(11,000-51,000)
34,000
(12,000 - 57,000)
33,000
(11,000-54,000)
15,000
(5,200 - 25,000)
140,000
(100,000 - 180,000)
QALY Gained from
Reductions in
Chronic Bronchitis
(95% CI)
—
7,300
(430 - 18,000)
8,500
(330-21,000)
8,200
(410-20,000)
5,200
(310-13,000)
2,200
(110-5,100)
900
(43-2,100)
160
(7 - 370)
33,000
(1,700-78,000)
QALY Gained from
Reductions in Acute
Myocardial Infarctions
(95% CI)
20
(4-42)
180
(34-410)
1,100
(200 - 2,500)
3,600
(640-8,100)
4,500
(760 - 10,000)
3,200
(590 - 7,300)
2,000
(380 - 4,400)
680
(130-1,500)
15,000
(2,900 - 34,000)
Total Gain in
MILYs
(95% CI)
20
(4-42)
10,000
(3,000-20,000)
19,000
(8,200 - 32,000)
31,000
(15,000-48,000)
41,000
(19,000-63,000)
40,000
(17,000-63,000)
36,000
(14,000 - 57,000)
16,000
(6,000 - 26,000)
190,000
(140,000-250,000)
Note that all estimates have been rounded to two significant digits.
The relationship between age and the distribution of MILYs gained from mortality
and morbidity is shown in Figure G-2. Because the baseline mortality rate is increasing in
age at a much faster rate than the prevalence rate for CB, the share of MILYs gained
accounted for by mortality is proportional to age. At the oldest age interval, avoiding
incidences of CB leads to only a few MILYs gained, due to the lower number of years lived
with CB. MILYs gained from avoided premature mortality is low in the youngest age
intervals because of the low overall mortality rates in these intervals, although the number of
MILYs per incidence is high. In later years, even though the MILYs gained per incidence
avoided is low, the number of cases is very high due to higher baseline mortality rates.
G-35
-------
Table G-12. Estimated Gains in 3 Percent Discounted MILYs Associated with the
Clean Air Interstate Rule in 2015a
Age
18-24
25-34
35^4
45-54
55-64
65-74
75-84
85+
Total
Life Years Gained
from Mortality Risk
Reductions
(95% CI)
—
3,200
(1,100-5,300)
11,000
(3,800 - 17,000)
22,000
(8,200 - 36,000)
42,000
(15,000-69,000)
51,000
(18,000 - 84,000)
40,000
(15,000-65,000)
20,000
(7,400 - 33,000)
190,000
(140,000-240,000)
QALY Gained from
Reductions in
Chronic Bronchitis
(95% CI)
—
9,400
(510-22,000)
9,800
(670 - 24,000)
9,400
(430 - 23,000)
7,100
(440 - 17,000)
3,200
(230 - 7,800)
1,100
(68 - 2,700)
210
(14-500)
40,000
(2,400 - 96,000)
QALY Gained from
Reductions in Acute
Myocardial Infarctions
(95% CI)
22 (4 - 52)
230
(42 - 520)
1,250
(230 - 2,800)
4,100
(780 - 9,300)
5,900
(1,000 - 14,000)
4,700
(870-11,000)
2,400
(430 -5,400)
890
(180 - 1,900)
20,000
(3,600 - 44,000)
Total Gain in
MILYs
(95% CI)
22 (4 - 52)
13,000
(3,800-26,000)
22,000
(9,600 - 37,000)
36,000
(18,000 - 55,000)
55,000
(27,000 - 84,000)
59,000
(26,000 - 92,000)
44,000
(18,000-69,000)
21,000
(8,400 - 34,000)
250,000
(180,000 - 330,000)
a Note that all estimates have been rounded to two significant digits.
Summing over the age intervals provides estimates of total MILYs gained for CAIR
in 2010 and 2015. The total number of discounted (3 percent) MILYs gained in 2010 is
190,000 (95% CI: 140,000 - 250,000) and in 2015 is 250,000 (95% CI: 180,000 - 330,000).
G. 6.2 Dealing with Acute Health Effects andNonhealth Effects
Health effects from exposure to particulate air pollution encompass a wide array of
chronic and acute conditions in addition to premature mortality (EPA, 1996). Although
chronic conditions and premature mortality generally account for the majority of monetized
benefits, acute symptoms can affect a broad population or sensitive populations (e.g., asthma
exacerbations in asthmatic children. In addition, reductions in air pollution may result in a
G-36
-------
35-44
45-54
55-64 65-74
Age Group
75-84
Figure G-2. Distribution of Mortality and Morbidity Related MILY Across Age
Groups for the CAIR in 2015 (3 percent Discount Rate)
broad set of nonhealth environmental benefits, including improved visibility in national
parks, increased agricultural and forestry yields, reduced acid damage to buildings, and a
host of other impacts. QALYs address only health impacts, and the OMB guidance notes
that "where regulation may yield several different beneficial outcomes, a cost-effectiveness
comparison becomes more difficult to interpret because there is more than one measure of
effectiveness to incorporate in the analysis."
With regard to acute health impacts, Bala and Zarkin (2000) suggest that QALYs are
not appropriate for valuing acute symptoms, because of problems with both measuring utility
for acute health states and applying QALYs in a linear fashion to very short duration health
states. Johnson and Lievense (2000) suggest using conjoint analysis to get healthy-utility
time equivalences that can be compared across acute effects, but it is not clear how these can
G-37
-------
be combined with QALYs for chronic effects and loss of life expectancy. There is also a
class of effects that EPA has traditionally treated as acute, such as hospital admissions,
which may also result in a loss of quality of life for a period of time following the effect. For
example, life after asthma hospitalization has been estimated with a utility weight of 0.93
(Bell et al., 2001; Kerridge, Glasziou, and Hillman, 1995).
How should these effects be combined with QALYs for chronic and mortality
effects? One method would be to convert the acute effects to QALYs; however, as noted
above, there are problems with the linearity assumption (i.e., if a year with asthma symptoms
is equivalent to 0.7 year without asthma symptoms, then 1 day without asthma symptoms is
equivalent to 0.0019 QALY gained). This is troubling from both a conceptual basis and a
presentation basis. An alternative approach is simply to treat acute health effects like
nonhealth benefits and subtract the dollar value (based on WTP or COI) from compliance
costs in the CEA.
To address the issues of incorporating acute morbidity and nonhealth benefits, OMB
suggests that agencies "subtract the monetary estimate of the ancillary benefits from the
gross cost estimate to yield an estimated net cost." As with benefit-cost analysis, any
unqualified benefits and/or costs should be noted and an indication of how they might affect
the cost-effectiveness ratio should be described. We will follow this recommended "net
cost" approach in the illustrative exercise, specifically in netting out the benefits of health
improvements other than reduced mortality and chronic morbidity, and the benefits of
improvements in visibility at national parks (see Chapter 4 for more details on these benefit
categories).
G.6.3 Cost-Effectiveness Ratios
Construction of cost-effectiveness ratios requires estimates of effectiveness (in this
case measured by lives saved, life years gained, or MILYs gained) in the denominator and
estimates of costs in the numerator. The estimate of costs in the numerator should include
both the direct costs of the controls necessary to achieve the reduction in ambient PM25 and
the avoided costs (cost savings) associated with the reductions in morbidity (Gold et al.,
1996). In general, because reductions in air pollution do not require direct actions by the
affected populations, there are no specific costs to affected individuals (aside from the
overall increases in prices that might be expected to occur as control costs are passed on by
affected industries). Likewise, because individuals do not engage in any specific actions to
realize the health benefit of the pollution reduction, there are no decreases in utility (as might
occur from a medical intervention) that need to be adjusted for in the denominator. Thus, the
G-38
-------
elements of the numerator are direct costs of controls minus the avoided COI associated with
CB and nonfatal AMI. In addition, to account for the value of reductions in acute health
impacts and nonhealth benefits, we net out the monetized value of these benefits from the
numerator to yield a "net cost" estimate. For the MILY aggregate effectiveness measure, the
denominator is simply the sum of life years gained from increased life expectancy and the
sum of QALYs gained from the reductions in CB and nonfatal AMI.
Avoided costs for CB and nonfatal AMI are based on estimates of lost earnings and
medical costs.8 Using age-specific annual lost earnings and medical costs estimated by
Cropper and Krupnick (1990) and a 3 percent discount rate, we estimated a lifetime present
discounted value (in 2000$) due to CB of $150,542 for someone between the ages of 27 and
44; $97,610 for someone between the ages of 45 and 64; and $11,088 for someone over 65.
The corresponding age-specific estimates of lifetime present discounted value (in 2000$)
using a 7 percent discount rate are $86,026, $72,261, and $9,030, respectively. These
estimates assumed that 1) lost earnings continue only until age 65, 2) medical expenditures
are incurred until death, and 3) life expectancy is unchanged by CB.
Because the costs associated with a myocardial infarction extend beyond the initial
event itself, we consider costs incurred over several years. Using age-specific annual lost
earnings estimated by Cropper and Krupnick (1990) and a 3 percent discount rate, we
estimated a present discounted value in lost earnings (in 2000$) over 5 years due to a
myocardial infarction of $8,774 for someone between the ages of 25 and 44, $12,932 for
someone between the ages of 45 and 54, and $74,746 for someone between the ages of 55
and 65. The corresponding age-specific estimates of lost earnings (in 2000$) using a 7
percent discount rate are $7,855, $11,578, and $66,920, respectively. Cropper and Krupnick
(1990) do not provide lost earnings estimates for populations under 25 or over 65. Thus, we
do not include lost earnings in the cost estimates for these age groups.
Two estimates of the direct medical costs of myocardial infarction are used. The first
estimate is from Wittels, Hay, and Gotto (1990), which estimated expected total medical
costs of MI over 5 years to be $51,211 (in 1986$) for people who were admitted to the
8Gold et al. (1996) recommend not including lost earnings in the cost-of-illness estimates, suggesting that in
some cases, they may be already be counted in the effectiveness measures. However, this requires that
individuals fully incorporate the value of lost earnings and reduced labor force participation opportunities
into their responses to time-tradeoff or standard-gamble questions. For the purposes of this analysis and for
consistency with the way costs-of-illness are calculated for the benefit-cost analysis, we have assumed that
individuals do not incorporate lost earnings in responses to these questions. This assumption can be relaxed
in future analyses with improved understanding of how lost earnings are treated in preference elicitations.
G-39
-------
hospital and survived hospitalization (there does not appear to be any discounting used).
Using the CPI-U for medical care, the Wittels estimate is $109,474 in year 2000$. This
estimated cost is based on a medical cost model, which incorporated therapeutic options,
projected outcomes, and prices (using "knowledgeable cardiologists" as consultants). The
model used medical data and medical decision algorithms to estimate the probabilities of
certain events and/or medical procedures being used. The second estimate is from Russell et
al. (1998), which estimated first-year direct medical costs of treating nonfatal myocardial
infarction of $15,540 (in 1995$), and $1,051 annually thereafter. Converting to year 2000$,
that would be $23,353 for a 5-year period (without discounting).
The two estimates from these studies are substantially different, and we have not
adequately resolved the sources of differences in the estimates. Because the wage-related
opportunity cost estimates from Cropper and Krupnick (1990) cover a 5-year period, we used
estimates for medical costs that similarly cover a 5-year period. We used a simple average of
the two 5-year estimates, or $65,902, and add it to the 5-year opportunity cost estimate. The
resulting estimates are given in Table G-13.
Table G-13. Estimated Costs Over a 5-Year Period (in 2000$) of a Nonfatal Myocardial
Infarction
Age Group
0-24
25-44
45-54
55-65
>65
Opportunity Cost
$0
$8,774b
$12,253b
$70,619b
$0
Medical Cost3
$65,902
$65,902
$65,902
$65,902
$65,902
Total Cost
$65,902
$74,676
$78,834
$140,649
$65,902
a An average of the 5-year costs estimated by Wittels, Hay, and Gotto (1990) and Russell et al. (1998).
b From Cropper and Krupnick (1990), using a 3 percent discount rate.
The total avoided COI by age group associated with the reductions in CB and
nonfatal acute myocardial infarctions is provided in Table G-14. Note that the total avoided
COI associated with CAIR is $2.1 billion in 2010 and $2.7 billion in 2015. Note that this
does not include any direct avoided medical costs associated with premature mortality. Nor
does it include any medical costs that occur more than 5 years from the onset of a nonfatal
AMI. Therefore, this is likely an underestimate of the true avoided COI associated with
CAIR.
G-40
-------
Table G-14. Avoided Costs of Illness Associated with Reductions in Chronic Bronchitis
and Nonfatal Acute Myocardial Infarctions Associated with CAIR in 2010 and 2015
Age Range
18-24
25-34
35-44
45-54
55-64
65-74
75-84
85+
Total
Avoided
Cost of Illness (in millions of 2000$) (95% confidence interval)3
Chronic Bronchitis
2010
—
$166
($5 - $325)
$215
($6 -$421)
$156
($4 - $305)
$125
($3 - $244)
$8
($0-$16)
$5
($0-$10)
$2
($0 - $4)
$677
($360 -$991)
2015
—
$212
($6 -$41 5)
$247
($7 - $484)
$179
($5 - $350)
$168
($5 - $328)
$12
($0 - $24)
$6
($0-$12)
$3
($0 - $6)
$827
($447 -$1,213)
Nonfatal Acute
2010
$1
($0 - $2)
$7
($1-$17)
$46
($7 -$118)
$184
($29 - $467)
$547
($133 -$1,171)
$251
($30 - $684)
$238
($28 - $649)
$145
($17 -$396)
$1,418
($680 -$2,3 10)
Myocardial Infarction
2015
$1
($0 - $2)
$8
($1 - $22)
$52
($8 -$134)
$211
($33 -$535)
$728
($177 -$1,558)
$367
($44 - $999)
$289
($34 - $788)
$190
($23 -$516)
$1,846
($870 - $3,007)
a Note that the confidence intervals for avoided COI include both the uncertainty in the unit values for each
health effect and the uncertainty in the estimated change in incidence for each health effect. Uncertainties
are combined using Monte Carlo simulation methods.
In a traditional cost-effectiveness analysis, net costs of the intervention would be
divided by the effectiveness measure to calculate a cost per life year or cost per QALY.
However, for both of the years of analysis, net costs of CAIR are negative, implying that
CAIR is a cost-saving rule. However, it is possible to calculate the costs that would be
necessary for the cost-effectiveness of CAIR to exceed various thresholds. Cost-
effectiveness ratios are usually interpreted in a relative sense, because there is no universally
agreed on cost-effectiveness cutoff for environmental health interventions. Although the
U.S. Public Health Service Panel on Cost Effectiveness in Health and Medicine did not
recommend a cost-effectiveness threshold for generalized use, it may be useful to identify
cost thresholds that would make CAIR cost-ineffective relative to other life-saving or quality
G-41
-------
of life-improving interventions. The Harvard Cost Utility Analysis database suggests a
median cost-utility ratio of $31,000 per QALY (2002$) for respiratory and cardiovascular
interventions, while Teng et al. (1995) report a median cost per life-year saved for live-
saving interventions of $48,000 (1993$). The health economics literature often uses either
$50,000 or $100,000 per QALY as a de facto cut point with ratios less than these values
considered cost-effective. For the purposes of this analysis, we computed the costs necessary
to exceed the $50,000 and $100,000 cost-effectiveness thresholds, without endorsing
$50,000 or $100,000 as an absolute threshold beyond which interventions should not be
implemented. Decisions as to whether a specific control strategy is justified should be based
on a complete comparison of benefits and costs.
Table G-15 summarizes the effectiveness measures and avoided costs associated with
CAIR in 2010 and 2015 and presents the implicit costs of C AIR that would be necessary for
the cost-effectiveness ratio to exceed the $50,000 and $100,000 threshold.
G.7 Discount Rate Sensitivity Analysis
A large number of parameters and assumptions are necessary in conducting a CEA.
Where appropriate and supported by data, we have included distributions of parameter values
that were used in generating the reported confidence intervals. For the assumed discount
rate, we felt it more appropriate to examine the impact of the assumption using a sensitivity
analysis rather than through the integrated probabilistic uncertainty analysis.
The choice of a discount rate, and its associated conceptual basis, is a topic of
ongoing discussion within the academic community. OMB and EPA guidance require using
both a 7 percent rate and a 3 percent rate. In the most recent benefit-cost analyses of air
pollution regulations, a 3 and 7 percent discount rate have been adopted in the primary
analysis. A 3 percent discount rate reflects a "social rate of time preference" discounting
concept. A 3 percent discount rate is also consistent with the recommendations of the NAS
panel on CEA (Gold et al., 1996), which suggests that "a real annual (riskless) rate of 3
percent should be used in the Reference Case analysis." We have also calculated MILYs and
the implicit cost thresholds using a 7 percent rate consistent with an "opportunity cost of
capital" concept to reflect the time value of resources directed to meet regulatory
requirements. Further discussion of this topic appears in Chapter 7 of Gold et al. (1996), in
Chapter 6 of the EPA Guidelines for Economic Analysis, and in OMB Circular A-4.
G-42
-------
Table G-15. Summary of Results for the Clean Air Interstate Rule"
Result Using 3% Discount Rate (95% Confidence Interval)
2010
2015
Life years gained from mortality
risk reductions
QALY gained from reductions in
chronic bronchitis
QALY gained from reductions in
acute myocardial infarctions
Total gain in MILYs
Avoided cost of illness
Chronic bronchitis
Nonfatal AMI
CAIR rule costsb
Net cost per MILYs
Implied annual cost necessary to
exceed $50,000/QALY threshold
Implied annual cost necessary to
exceed $100,000/QALY threshold
140,000
(100,000 - 180,000)
33,000
(1,700 - 78,000)
15,000
(2,900 - 34,000)
190,000
(140,000-250,000)
$680 million
($360 million- $990 million)
$1,400 million
($680 million- $2,300 million)
$2.4 billion
Cost saving
$13 billion
($10 billion-$18 billion)
$23 billion
($17 billion-$31 billion)
190,000
(140,000-240,000)
40,000
(2,400 - 96,000)
20,000
(3,600 - 44,000)
250,000
(180,000 - 330,000)
$830 million
($450 million- $1,200 million)
$1,800 million
($870 million- $3,000 million)
$3.6 billion
Cost saving
$18 billion
($13 billion-$23 billion)
$30 billion
($22 billion-$40 billion)
a Consistent with recommendations of Gold et al. (1996), all summary results are reported at a precision level
of two significant digits to reflect limits in the precision of the underlying elements.
b Costs are the private firm costs of compliance derived from the Integrated Planning Model, as discussed in
Chapter 2, and reflect discounting using firm specific costs of capital.
Table G-16 presents a summary of results using the 7 percent discount rate and the
percentage difference between the 7 percent results and the base case 3 percent results.
Adoption of a 7 percent discount rate decreases the estimated life years and QALYs gained
from implementing the CAIR. Adopting a discount rate of 7 percent results in a 26 percent
reduction in the estimated total MILYs gained in each year, while the implicit cost necessary
to exceed the $50,000 cost-effectiveness threshold is reduced by 20 percent in each year.
G-43
-------
Table G-16. Impacts of Using a 7 Percent Discount Rate on Cost Effectiveness Analysis
for the Clean Air Interstate Rule in 2015
Result Using 7 Percent
Discount Rate
Percentage Change
Relative to Result Using 3
Percent Discount Rate
Life years gained from mortality
risk reductions
QALY gained from reductions in
chronic bronchitis
QALY gained from reductions in
acute myocardial infarctions
Total gain in MILYs
Avoided cost of illness
Chronic bronchitis
Nonfatal AMI
Implied cost necessary to exceed
$50,000/QALY threshold
Implied cost necessary to exceed
$100,000/QALY threshold
140,000
26,000
15,000
180,000
$540 million
$1,800 million
$14 billion
$24 billion
-25
-35
-22
-26
-35
-3a
-20
-23
There is a 3 percent difference in estimated avoided costs of nonfatal AMI; however, because of rounding,
the reported cost for both 3 and 7 percent discount rates is $1,800 million.
G.8 Conclusions
We calculated the effectiveness of CAIR based on reductions in premature deaths and
incidence of chronic disease. We measured effectiveness using several different metrics,
including lives saved, life years saved, and QALYs (for improvements in quality of life due
to reductions in incidence of chronic disease). We suggested a new metric for aggregating
life years saved and improvements in quality of life, morbidity inclusive life years (MTLY)
which assumes that society assigns a weight of one to years of life extended regardless of
preexisting disabilities or chronic health conditions.
Using the MILYs metric, we estimated that CAIR could cost up to $14 billion
annually in 2010 and up to $18 billion annually in 2015 and would still likely be cost-
G-44
-------
effective relative to other health interventions for cardiovascular and respiratory disease.
Given costs of $2.4 billion and $3.6 billion in 2010 and 2015, respectively, CAIR is clearly a
very cost-effective way to achieve improvements in public health.
CEA of environmental regulations that have substantial public health impacts may be
informative in identifying programs that have achieved cost-effective reductions in health
impacts and can suggest areas where additional controls may be justified. However, the
overall efficiency of a regulatory action can only be judged through a complete benefit-cost
analysis that takes into account all benefits and costs, including both health and nonhealth
effects. The benefit-cost analysis for CAIR, provided in Chapter 4, shows that CAIR has
large net benefits, indicating that CAIR will likely result in improvements in overall public
welfare and will provide health benefits in a highly cost-effective manner.
G.9 References
Abbey, D.E., B.L. Hwang, RJ. Burchette, T. Vancuren, and P.K. Mills. 1995. "Estimated
Long-Term Ambient Concentrations of PM(10) and Development of Respiratory
Symptoms in a Nonsmoking Population." Archives of Environmental Health 50(2):
139-152.
Agency for Healthcare Research and Quality. 2000. HCUPnet, Healthcare Cost and
Utilization Project.
American Heart Association. 2003. Heart Disease and Stroke Statistics—2003 Update.
Dallas, Texas: American Heart Association.
American Lung Association. 2002. Trends in Chronic Bronchitis and Emphysema:
Morbidity and Mortality. American Lung Association, Best Practices and Program
Services, Epidemiology and Statistics Unit.
Bala, M.V., and G.A. Zarkin. 2000. "Are QALYs an Appropriate Measure for Valuing
Morbidity in Acute Diseases?" Health Economics 9:177-180.
Bell, C.M., R.H. Chapman, P.W. Stone, E.A. Sandberg, and PJ. Neumann. 2001. "An Off-
the-Shelf Help List: A Comprehensive Catalog of Preference Scores from Published
Cost-Utility Analyses." Medical Decision Making 21:288-94.
Bleichrodt, H., P.P. Wakker, and M. Johannesson. 1996. "Characterizing QALYs by Risk
Neutrality." Journal of Risk and Uncertainty 15:107-114.
G-45
-------
Brock, D.W. 2002. "Chapter 14.3: Fairness and Health." In Summary Measures of
Population Health: Concepts, Ethics, Measurement and Applications., C.J.L. Murray,
J. A. Salomon, C.D. Mathers, and A.D. Lopez (eds.). Geneva: World Health
Organization.
Carrothers, T. J., J.S. Evans, and J.D. Graham. 2002. "The Lifesaving Benefits of Enhanced
Air Quality." Harvard Center for Risk Analysis Working Paper.
Carnethon, M.R., D. Liao, G.W. Evans, W.E. Cascio, L.E. Chambless, W.D. Rosamond, and
G. Heiss. 2002. "Does the Cardiac Autonomic Response to Postural Change Predict
Incident Coronary Heart Disease and Mortality? The Atherosclerosis Risk in
Communities Study." American Journal of Epidemiology 155(l):48-56.
Centers for Disease Control and Prevention (CDC) Health, United States. 2003. Table 30.
Years of Potential Life Lost before Age 75 for Selected Causes of Death, According
to Sex, Race, and Hispanic Origin: United States, Selected Years 1980-2000.
Centers for Disease Control and Prevention (CDC). 2002. National Vital Statistics Reports
51(3).
Cohen, D.J., J.A. Breall, and K.K. Ho. 1994. "Evaluating the Potential Cost-Effectiveness
of Stenting as a Treatment for Symptomatic Single-Vessel Coronary Disease: Use of
a Decision-Analytic Model." Circulation 89(4): 1859-1874.
Cohen, J.T., J.K. Hammitt, and J.I. Levy. 2003. "Fuels for Urban Transit Buses: A Cost-
effectiveness Analysis." Environmental Science and Technology 37:1477-1484.
Cowie, M.R., A. Mosterd, D.A. Wood, et al. 1997. "The Epidemiology of Heart Failure."
European Heart Journal 18:208-25.
Coyle, D., D. Stieb, R.T. Burnett, P. DeCivita, D. Krewski, Y. Chen, and M.J. Thun. 2003.
"Impact of Particulate Air Pollution on Quality Adjusted Life Expectancy in
Canada." Journal of Toxicology and Environmental Health, Part A 66:1847-1863.
Cropper, M.L., and AJ. Krupnick. 1990. "The Social Costs of Chronic Heart and Lung
Disease." Discussion Paper QE 89-16-REV. Washington, DC: Resources for the
Future.
de Hollander, A.E.M., J.M. Melse, E. Lebret, and P.G.N. Kramers. 1999. "An Aggregate
Public Health Indicator to Represent the Impact of Multiple Environmental
Exposures." Epidemiology 10:606-617.
G-46
-------
Dekker, J.M., R.S. Crow, A.R. Folsom, PJ. Hannan, D. Liao, C.A. Swenne, and E.G.
Schouten. 2000. "Low Heart Rate Variability in a 2-Minute Rhythm Strip Predicts
Risk of Coronary Heart Disease and Mortality From Several Causes: The ARIC
Study." Circulation 2000 102:1239-1244.
EPA-SAB-COUNCIL-ADV-04-002. March 2004. Advisory on Plans for Health Effects
Analysis in the Analytical Plan for EPA 's Second Prospective Analysis—Benefits and
Costs of the Clean Air Act, 1990-2020: Advisory by the Health Effects Subcommittee
of the Advisory Council on Clean Air Compliance Analysis.
Freeman, A.M., J.K. Hammitt, and P. De Civita. 2002. "On Quality Adjusted Life Years
(QALYs) and Environmental/Consumer Safety Valuation." AERE Newsletter
22(1):7-12.
Ganz, D.A., K.M. Kuntz, G.A. Jacobson, and J. Avorn. 2000. "Cost-Effectiveness of
3-Hydroxy-3-Methylglutaryl Coenzyme A Reductase Inhibitor Therapy in Older
Patients with Myocardial Infarction." Annals of Internal Medicine 132:780-787.
Garber, A.M., and C.E. Phelps. 1997. "Economic Foundations of Cost-Effectiveness
Analysis." Journal of Health Economics 16:1-31.
Gentile, I.E. 1998. Random Number Generation andMonte Carlo Methods. New York,
NY: Springer Verlag.
Gold, M.R., I.E. Siegel, L.B. Russell, and M.C. Weinstein. 1996. Cost-effectiveness in
Health and Medicine. New York: Oxford University Press.
Gold, D.R., A. Litonjua, J. Schwartz, E. Lovett, A. Larson, B. Nearing, G. Allen, M. Verrier,
R. Cherry, and R. Verrier. 2000. "Ambient Pollution and Heart Rate Variability."
Circulation 101(11): 1267-1273.
Gold, M.R., D. Stevenson, and D.G. Fryback. 2002. "HALYS and QALYS and DALYS,
Oh My: Similarities and Differences in Summary Measures of Population Health."
Annual Review of Public Health 23:115-34.
Hammitt, J.K. 2002. "QALYs versus WTP" Risk Analysis 22(5):985-1001.
Harvard Center for Risk Analysis, Catalog of Preference Scores. Available at:
http://www.hcra.harvard.edu/pdf/preferencescores.pdf.
G-47
-------
Horsman, J., W. Furlong, D. Feeny, and G. Torrance. 2003. "The Health Utilities Index
(FUJI): Concepts, Measurement Properties, and Applications." Health and Quality of
Life Outcomes 1:54.
Hubbell, B. 2004a. "Implementing QALYs in the Analysis of Air Pollution Regulations."
Environmental and Resource Economics (forthcoming).
Hubbell, B. 2004b. "Health Based Cost Effectiveness of Ambient PM2.5 Reductions"
Unpublished working paper.
Johnson, F.R., and K. Lievense. 2000. Stated-Preference Indirect Utility and Quality-
Adjusted Life Years. Report prepared for Health Canada, July.
Kerridge, R.K., P.P. Glasziou, and K.M. Hillman. 1995. "The Use of "Quality-Adjusted
Life Years" (QALYs) to Evaluate Treatment in Intensive Care." Anaesthia and
Intensive Care 23:322-31.
Kind, P. 1996. The EuroQoL Instrument: An Index of Health-Related Quality of Life.
Quality of Life and Pharmacoeconomics in Clinical Trials, Second Edition, B.
Spilker (ed.), pp. 191-201. Philadelphia, PA: Lippincott-Raven Publishers.
Krewski, D., R.T. Burnett, M.S. Goldbert, K. Hoover, J. Siemiatycki, M. Jerrett, M.
Abrahamowicz, and W.H. White. 2000. Reanalysis of the Harvard Six Cities Study
and the American Cancer Society Study of Paniculate Air Pollution and Mortality.
Special Report to the Health Effects Institute, Cambridge, MA.
Kuntz, K.M, J. Tsevant, L. Goldman, and M.C. Weinstein. 1996. "Cost-Effectiveness of
Routine Coronary Angiography after Acute Myocardial Infarction." Circulation
94(5):957-65.
Kunzli, N., R. Kaiser, S. Medina, M. Studnicka, O. Chanel, P. Filliger, M. Kerry, F. Horak
Jr., V. Puybonnieux-Texier, P. Quenel, J. Schneider, R. Seethaler, J-C Vergnaud, and
H. Sommer. 2000. "Public-Health Impact of Outdoor and Traffic-Related Air
Pollution: A European Assessment." The Lancet 356:795-801.
Kunzli, N., S. Medina, R. Kaiser, P. Quenel, F. Horak Jr, and M. Studnicka. 2001.
"Assessment of Deaths Attributable to Air Pollution: Should We Use Risk Estimates
Based on Time Series or on Cohort Studies?" American Journal of Epidemiology
153(ll):1050-55.
G-48
-------
Liao, D., J. Cai, W.D. Rosamond, R.W. Barnes, R.G. Hutchinson, E.A. Whitsel, P.
Rautaharju, and G. Heiss. 1997. "Cardiac Autonomic Function and Incident
Coronary Heart Disease: A Population-Based Case-Cohort Study. The ARIC Study.
Atherosclerosis Risk in Communities Study." American Journal of Epidemiology
145(8):696-706.
Liao, D., J. Creason, C. Shy, R. Williams, R. Watts, and R. Zweidinger. 1999. "Daily
Variation of Particulate Air Pollution and Poor Cardiac Autonomic Control in the
Elderly." Environmental Health Perspectives 107:521-5.
Magari, S.R., R. Hauser, J. Schwartz, P.L. Williams, T.J. Smith, and D.C. Christiani. 2001.
"Association of Heart Rate Variability with Occupational and Environmental
Exposure to Particulate Air Pollution." Circulation 104(9):986-91.
Miller, E.G., and J.F. Hurley. 2003. "Life Table Methods for Quantitative Impact
Assessments in Chronic Mortality." Journal of Epidemiology and Community Health
57:200-206.
Moolgavkar, S.H. 2000. "Air Pollution and Hospital Admissions for Diseases of the
Circulatory System in Three U.S. Metropolitan Areas." Journal of the Air Waste
Management Association 50:1199-206.
Murray, C.J.L., J.A. Salomon, C.D. Mathers, and A.D. Lopez. 2002. "Summary Measures
of Population Health: Concepts, Ethics, Measurement and Applications." Geneva:
World Health Organization.
National Research Council (NRC). 2002. Estimating the Public Health Benefits of
Proposed Air Pollution Regulations. Washington, DC: The National Academies
Press.
Neumann, P. 2003. "A Web-based Registry of Cost-utility Analyses." Risk in Perspective
11(3).
Peters, A., D.W. Dockery, I.E. Muller, and M.A. Mittleman. 2001. "Increased Particulate
Air Pollution and the Triggering of Myocardial Infarction." Circulation
103:2810-2815.
Pliskin, J.S., D.S. Shepard, and M.C. Weinstein. 1980. "Utility Functions for Life Years and
Health Status." Operations Research 28:206-224.
G-49
-------
Pliskin, J.S., W.B. Stason, and M.C. Weinstein. 1981. "Coronary Artery Bypass Graft
Surgery: Clinical Decision Making and Cost-Effective Analysis." Medical Decision
Making 1(1): 10-28.
Poloniecki, J.D., R.W. Atkinson, A.P. de Leon, and H.R. Anderson. 1997. "Daily Time
Series for Cardiovascular Hospital Admissions and Previous Day's Air Pollution in
London, UK." Occupational and Environmental Medicine 54(8):535-40.
Pope, D.A., R.T. Burnett, MJ. Thun, E.E. Cale, D. Krewski, K. Ito, and G.D. Thurston.
2002. "Lung Cancer, Cardiopulmonary Mortality, and Long-Term Exposure to Fine
Particulate Air Pollution." Journal of the American Medical Association
287:1132-1141.
Rabl, A. 2003. "Interpretation of Air Pollution Mortality: Number of Deaths or Years of
Life Lost?" Journal of the Air and Waste Management Association 53:41-50.
Russell, M.W., D.M. Huse, S. Drowns, E.C. Hamel, and S.C. Hartz. 1998. "Direct Medical
Costs of Coronary Artery Disease in the United States." American Journal of
Cardiology %\(9}:ll\Q-5.
Salkeld, G., P. Phongsavan, and B. Oldenburg. 1997. "The Cost-Effectiveness of a
Cardiovascular Risk Reduction Program in General Practice." Health Policy
41:105-19.
Salomon, J.A., and C.J.L. Murray. 2003. "A Multi-Method Approach to Measuring
Health-State Valuations." Health Economics, online.
Samet, J.M., S.L. Zeger, F. Dominici, F. Curriero, I. Coursac, D.W. Dockery, J. Schwartz,
and A. Zanobetti. June 2000. The National Morbidity, Mortality and Air Pollution
Study: Part II: Morbidity, Mortality and Air Pollution in the United States.
Research Report No. 94, Part II. Cambridge, MA: Health Effects Institute.
Schwartz, J. 1993. "Particulate Air Pollution and Chronic Respiratory Disease."
Environmental Research 62:7-13.
Singer, P., J. McKie, H. Kuhse, J. Richardson, and J. Harris. 1995. "Double Jeopardy and
the Use of QALYs in Health Care Allocation." Journal of Medical Ethics
21:144-157.
G-50
-------
Smith, K.J., and R.R. Pesce. 1994. "Pulmonary Artery Catheterization in Exacerbations of
COPD Requiring Mechanical Ventilation: A Cost-Effectiveness Analysis."
Respiratory Care 39:961-7.
Stinnett, A.A., M.A. Mittleman, M.C. Weinstein, K.M. Kuntz, DJ. Cohen, L.W. Williams,
P.A. Goldman, D.O. Staiger, M.G.M. Hunink, J. Tsevat, A.N.A. Tosteson, and L.
Goldman. 1996. "The Cost-effectiveness of Dietary and Pharmacologic Therapy for
Cholesterol Reduction in Adults." In Cost-effectiveness in Health and Medicine,
Gold, M.E., J. E. Siegel, L.B. Russell, and M.C. Weinstein (eds.). New York:
Oxford University Press.
Stouthard, M.E.A., J.L. Essink-Bot, GJ. Bonsel, JJ. Barendregt, P.G.N. Kramers, H.P.A.
Van de Water, and LJ. Gunning-Schepers. 1997. "Disability Weights for Diseases
in the Netherlands." Rotterdam: Department of Public Health, Erasmus University
Rotterdam.
Suwa, T. J.C. Hogg, K.B. Quinlan, A. Ohgami, R. Vincent, and S.F. van Eeden. 2002.
"Parti culate Air Pollution Induces Progression of Atherosclerosis." Journal of the
American College of Cardiology 39:935-942.
Teng, T.O., M.E. Adams, J.S. Pliskin, D.G. Safiran, I.E. Siegel, M.C. Weinstein, and J.D.
Graham. 1995. "Five Hundred Life-Saving Interventions and Their
Cost-Effectiveness." Risk Analysis 15:369-390.
Tsuji, H., M.G. Larson, FJ. Venditti, Jr., E.S. Manders, J.C. Evans, C.L. Feldman, and D.
Levy. 1996. "Impact of Reduced Heart Rate Variability on Risk for Cardiac Events.
The Framingham Heart Study." Circulation 94(11):2850-2855.
U.S. Environmental Protection Agency (EPA). 1996. Review of the National Ambient Air
Quality Standards for P articulate Matter: Assessment of Scientific and Technical
Information. Report no. EPA/4521R-96-013. Research Triangle Park, NC: Office of
Air Quality Planning and Standards.
U.S. Environmental Protection Agency (EPA). 2000. Guidelines for Preparing Economic
Analyses. EPA240-R-00-003.
U.S. Environmental Protection Agency (EPA). 2004. Final Regulatory Analysis: Control of
Emissions from Nonroad Diesel Engines.
G-51
-------
U.S. Food and Drug Administration (FDA). 2004. Bar Code Label Requirements for Human
Drug Products and Biological Products; Final Rule. Federal Register: February 26,
2004 Volume 69 (38): 9119-9171.
U.S. Office of Management and Budget (OMB). 2003. Circular A-4: Regulatory Analysis,
September 17, 2003. Available at: http://www.whitehouse.gov/omb/
circulars/a004/a-4.pdf.
Vos, T. 1999a. Final worksheet VM1 COPD.xls, prepared for Victoria Burden of Disease
Study. Available at: http://www.dhs.vic.gov.au/phd/bod/daly.htm.
Vos, T. 1999b. Final worksheet VL2 IFID.xls, prepared for Victoria Burden of Disease
Study. Available at: http://www.dhs.vic.gov.au/phd/bod/daly.htm.
Wittels, E.H., J.W. Hay, and A.M. Gotto, Jr. 1990. "Medical Costs of Coronary Artery
Disease in the United States." American Journal of Cardiology 65(7):432-40.
G-52
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-452/R-05-002
Environmental Protection Air Quality Strategies and Standards Division March 2005
Agency Research Triangle Park, NC
------- |