The Benefits and Cleats
of the Clean Air
197O to 199O
Prepared for
US Congress
by
US Environmental Protection Agency
October 1996-DRAFT
-------
-------
Abstract
In Section 812 of the Clean Air Act Amendments of 1990, Congress required EPA to conduct periodic „
assessments of the benefits and costs of the Act. The first was to be a retrospective study, followed by a series of
prospective assessments. This study represents the retrospective study. The study was designed to answer the
following specific question: "How do the overall health, welfare, ecological, and economic benefits of Clean Air Act
programs compare to the costs of these programs?" To address the question, EPA developed and contrasted two
scenarios: a "control scenario" which reflected actual historical implementation of the 1970 and 1977 Clean Air
Acts, and a "no-control scenario" which reflected the assumption that air pollution control programs were frozen at
their 1970 levels of scope and stringency. The differences between these two scenarios in terms of economic and
environmental outcomes were taken to represent the incremental costs and benefits of the Clean Air Act from 1970
to 1990. "" .
A complex sequence of analytical steps was required to construct and evaluate the control and no-control
scenarios and the differences in their respective outcomes. These steps included: (1) direct cost estimation, (2)
macroeconomic modeling, (3) emissions modeling, (4) air quality modeling, (5) physical effects estimation, (6)
economic valuation, and (7) results aggregation and uncertainty modeling. Each of the first six steps required the
adoption of important assumptions and contributed to overall uncertainty in the bottom line estimate of net benefits
, of the Clean Air Act. The effects of key uncertainties are described, and additional analyses were performed to test
the sensitivity of the net benefit results to changes in important assumptions.
Direct expenditures made to comply with 1970 to 1990 Clean Air Act requirements rose steadily in inflation-
adjusted terms throughout the study period. Annualizing this stream of expenditures to correct for accrual of costs
and benefits beyond 1990 and adjusting the individual year costs for inflation to reflect their equivalent value in
1990 yielded a direct compliance cost estimate for the 1970 to 1990 period equal to 523 billion dollars. This
estimate of direct compliance cost is presented as a single point estimate because potentially important uncertainties
in the cost estimates could not be quantified. Potential additional costs not reflected in this point estimate include (a)
indirect economic effects, such as adverse effects on capital formation, and (b) potential adverse effects on
production-related technological innovation, net of potentially beneficial pollution control-related technological
innovation.
Direct benefits of the 1970 to 1990 Clean Air Act include reductions in airborne emissions of a variety of pollutants
and associated improvements in air quality, human health, human welfare, and ecosystem functioning. For the year 1990,
sulfur dioxide (SOj) emissions were 40 percent lower, primarily due to utilities installing scrubbers and/or switching to .
lower sulfur fuels. Nitrogen oxides (NOJ emissions were 30 percent lower, mostly because of the installation of catalytic
converters on highway vehicles. Volatile organic compound (VOC) emissions were also lower primarily due to motor
vehicle controls, with VOC levels 45 percent lower under the control scenario than under the no-control scenario. Motor
vehicle controls were also primarily responsible for a 50 percent reduction in carbon monoxide (CO) emissions under the
control scenario. Control scenario lead emissions in 1990 were projected to be about 3,000 tons, reflecting a 99 percent
reduction from the projected no-control level of 237,000 tons. These emissions reductions were projected to lead to
significant improvements in air quality, including lower rates of acid deposition, lower ambient concentrations of a broad
range of pollutants, and improved visibility. Using concentration-response functions for a subset of the potential benefits of
these improvements in air quality yielded ranges of quantitative estimates of reduced adverse effects on human health and
welfare. Ecosystem changes and a large number of additional human health and welfare could not be quantified in this
analysis.
~ Applying economic valuation functions from the economics literature to a number of these adverse effect changes
provided measures of the economic value of air quality-related improvements. Because uncertainties in the physical effects
estimation and economic valuation steps were modeled quantitatively, the total direct benefit estimate is presented as a
range with associated probabilities. The central case modeling results indicate a range for overall monetizable benefits of
; 10.5 to 40.6 trillion dollars for the 1970 to 1990 period, a range which is estimated to have a 90 percent probability of
including the "true" estimate of total direct monetized benefits. The mean of the distribution of the benefits estimate is 23
trillion dollars (expressed in terms of 1990 value, using a five percent discount rate). This mean estimate of direct monetized
benefits is approximately 45 times the estimate of direct costs for the 1970 to 1990 period. -
There are numerous important uncertainties associated with these results, particularly on the benefits side. Additional
research in critical areas of uncertainty, such as particulate matter-related mortality incidence and economic valuation,
might lead to more refined estimates. In addition, further research in the areas of ecosystem effects and effects of air
toxics, among others, would allow more comprehensive treatment of benefit categories in future assessments.
-------
-------
Acknowledgments , .-. ix
Executive Summary ,~ ,....» xi
Purpose of the Study <...-. , xi
Study Design .,'. ... ,^. _, . 1 xi
Summary of Results ... „ sv. T xii
Direct Costs . .^... ^ xii
Indirect Economic Effects ~.. ..7.Y..-- . xiii
Emissions xiv
Air Quality _.,.-.... /.---. xv
Physical Effects „.. .„ „-. .,,.1. xvi
Economic Valuation ." xx
Monetized Benefits and Costs „ » xxi
Conclusions and Future Directions ... .r. , » xxiii
Tables ....................i^....* xxv
Figures xxxi
/ , '
\ , " * .
Acronyms ana Abbreviations xxxiii
Introduction i 1
, Background and Purpose 1
QeSalt Air Act Requirements, 1970 to 1990 ; 1
Section 812 of the Clean Air Act Amendments of 1990 2
Analytical Design and Review .-.' 2
Target Variable 2
Key Assumptions .....' 3
Analytical Sequence 4
Review Process ... 7
Report Organization i 8
Cost and Macroeconomic Effects 9
Direct Compliance Costs 10
Indirect Effects of the CAA ... . , 11
Sectoral Impacts ....;... 12
Aggregate Effects ., .-...' ; 12
Uncertainties and Sensitivities in the Cost and Macroeconomic Analysis ..... . 13
Productivity and Technical Change ;. ..-1 13
-------
Discount Rates . ... 14
Exclusion of Health Benefits from the Macroeconomic Model .... ..... 15
Emissions ...» 17
Sector-Specific Approach 19
Summary of Results 20
Uncertainty in the Emissions Estimates 23
AirQnalHy 25
General Methodology 27
Sample Results ,. ~- • •"^ 28
Carbon Monoxide i - 28
Sulfur Dioxide ".'....: 29
Nitrogen Dioxide _. 30
Paniculate Matter "..... •. .30
Ozone .-v. . «..:-»....,>,.,. • 30
Urban Ozone T 30
Rural Ozone 31
Acid Deposition ;..;..... ....: 31
Visibility .32
Physical Effects :;> ;> 37
Human Health and Welfare Effects Modeling Approach .. 37
Air Quality ., i'...... ./f. 37
Population •,*£. 37
Health and Welfare Effects ............ ;^.,, i 38
Key Analytical Assumptions ..:..„. .'.... 40
Mapping Populations to Monitors 41
Choice of Study 42
Variance Within Studies • ., .42
Health Effects Modeling Results 43
Mortality 43
Short-Term Exposure Studies 43
Long-Term Exposure Studies 44
Avoided Excess Premature Mortality Estimates 44
Non-Fatal Health Impacts :. - 45
Other Physical Effects 46
Ecological Effects 46
Quantified Agricultural Effects 46
Effects of Air Toxics , 46
Economic Valuation 51
Willingness-to-Pay Estimates '..'. 51
Mortality ..,,... • 52
Survey-Based Values • 53
Chronic Bronchitis 54
Respiratory-Related Ailments 54
Minor Restricted Activity Days 55
11
-------
Visibility 55
Avoided Cost Estimates 55
Hypertension and Hospital Admissions 55
Household Soiling , 56
Other Valuation Estimates , 56
Changes in Children's IQ .- , .A-........ 56
Work Loss Days and Worker Productivity , 57
Valuation Uncertainties ;- *' 57
Mortality Risk Benefits Transfer -. 57
Aggregate Results and Uncertainty »»... ,^...... .61
Aggregate Monetary Benefits - ..">._..-„„-.. 62
Benefits and Costs .... ".."..; -.-.-. 64
Quantified Uncertainty and Sensitivities ,..._., vvv .• 65
Quantified Uncertainty in the Benefits Analysis „:-'. ^.). .„ 65
Sources of Quantified Uncertainty 66
Potential Sensitivities in the Cost-Benefit Analysis ; 67
Discount Rates ;. -—..- 67
PM-Related Mortality Valuation *.,.!-. -.„.•• ;.- 67
Second-Order Macroeconomic Impacts .. 68
Appendix A: Cost and Macroeconomic Modeling 71
Macroeconomic Modeling ... ., 71
Structure of the Jorgenson-WilCdxen Model 74
The Business Sector 75
The Household Sector 75
The Government Sector ....... 75
The Rest-of-the-Wbrld Sector .76
Environmental Regulation, Investment, and Capital Formation 76
The General Equilibrium -v.. 77
Capital Costs - Stationary Sources...... 79
Operating and Maintenance Costs - Stationary Sources 79
Capital Costs-Mobile Sources 80
.Operating and Maintenance - Mobile Sources . 80
Direct Compliance Expenditures Data . ,.. 80
Cost of Clean Data 80
EPAData ...... 80
Commerce Data 81
•'''-' Capital Expenditures Data 82
Operation and Maintenance Expenditures Data 82
-RecoveredCosts . , '.- 83
. Capital Expenditures Data f 85
Operation and Maintenance Expenditures Data i 85
Assessment Results -.. 89
GNP and Personal Consumption .. 91
Prices — . ....... 92
Sectoral Effects: Changes in Prices and Output by Industry 93
Changes in Employment Across Industries ...; 94
Potential Sources of Error in the Cost Data 97
m
-------
Mobile Source Costs ., ;• 99
Endogenous Productivity Growth in the Macro Model.. 100
Stationary Source Cost Estimate Revisions ..- ,. 100
Amortization Period for Stationary Source Plant and Equipment 101
Cost and Macroeconomic Modeling References ..... • • 102
Appendix B: Emissions Modeling 105
Industrial Boilers and Processes 107
Overview of Approach ..~ .". 107
Industrial Boilers : ..:'. /*>.,: jf>7
Industrial Processes and In-Process Fuel Combustion :, — ".„ €08
~- , .•". : r. sf. .r. ;,.>. ?..: 108
Establishment of Control Scenario Emissions • • •"-.-^ • l^8
Development of Economic Driver Data for the Control Scenario - Industrial Boilers and Processes
;. .W.v., 112
No-control Scenario Emissions 114
Industrial Boiler Emissions of SOj, NO,, and TSP 114
Industrial Boiler Emissions of CO and VOC 114
Industrial Process Emissions >. ;..- 115
Lead Emissions .'.... % .\. 115
Off-Highway Vehicles y- 117
Overview of Approach >. 117
Development of Control Scenario 117
No-control Scenario Emissions Estimates 117
National and State-Level Off-Highway Emission Estimates 118
On-Highway ...,'. 120
Overview of Approach ....;.. 120
Personal Travel .-. 120
Goods Movement 123
Oflier Transportation Activities •. 123
Lead Emissions '•,••' 123
Estimation of No-control Scenario Emissions 124
Development of Emission Factors ..;... 124
Allocation of Highway Activity to States 125
Development of Highway Pollutant Estimates 125
Utilities .....-.' - 134
Overview of Approach , 134
Establishment of Control Scenario Emissions ." 134
No-control Scenario Emissions J... 139
ARGUS No-control Scenario 141
Estimation of Lead Emissions from Utilities .— 141
CEUM Sensitivity Case 142
Commercial/Residential '. 142
Control Scenario Emissions 143
Emissions Data • • • 145
Energy Data 147
Economic/Demographic Data 148
No-control Scenario Emissions , 148
Emissions Data 148
IV
-------
Energy Data . 149
Economic/Demographic Data ........ . ... s 150
Emissions Modeling References .... 154
Appendix C: Air Quality Modeling 159
Introduction :.-.... .,tf% 159
Carbon Monoxide 160
Control scenario carbon monoxide profiles ". .^.... .„" 160
No-control scenario carbon monoxide profiles , „. .". 161
Summary differences in carbon monoxide air quality „,._._."- 164
Key caveats and uncertainties for carbon monoxide ".^ r-.-i-v.v "164
Sulfur Dioxide V;.v"X.,..,... 165
Controlscenario sulfur dioxide profiles „.. .^ v.lI-.?V:... 165
No-control scenario sulfur dioxide profiles i--. i~. : * 166
Summary differences in sulfur dioxide air quality „. 166
Key caveats and uncertainties for sulfur dioxide 167
Nitrogen Oxides , 167
Control scenario nitrogen oxides profiles * 168
No-control scenario nitrogen oxides profiles 169
Summary differences in nitrogen oxides air quality 169
Key caveats and uncertainties for nitrogen oxides ». 170
Acid Deposition .,. ,v. „ 170
Control scenario acid deposition profiles ...-..' 171
No-control scenario acid deposition profiles ... £f 173
Summary differences in acid deposition .; 174
Key caveats and uncertainties for acid deposition 175
Particulate Matter . ,>..... ,,,... .v,^. 176
Control scenario particulate matter profiles .. 178
No-control scenario particulate matter profiles 181
Summary differences in particulate matter air quality 183
Key caveats and uncertainties for particulate matter ......; 183
Ozone i .. 184
Control scenario ozone profiles 187
No-control scenario ozone profiles 187
Summary differences in ozone air quality 190
Key caveats and uncertainties for ozone ...'..' 191
Visibility ,,,'.* 192
Control scenario visibility ; 193
No-control scenario visibility '. 194
Summary differences in visibility ......... 197
Deciview Haze Index 197
Modeling Results 198
Key caveats and uncertainties forvisibility ...:.. f 200
f Air Quality Modeling References 200
Appendix D: Human Health and Visibility Effects of Criteria Pollutants 203
Introduction and Overview 4 203
Principles for the §812 Benefits 203
Health Effects Studies .. -.; 204
-------
Epidemiological Studies .. 204
Human Clinical Studies 205
Issues In Selecting Studies To Estimate Health Effects 206
Types of Studies Considered , 207
Confounding Factors < - 207
Thresholds . 207
Exposure Modeling 207
Uncertainty .'.". 208
Links to Air Quality Modeling '.,'. , 209
Target Population , 209
Exposure Ranges :210
Peer Reviewed Research .:.......... 210
Application Of Health Science Research And Air Quality Results To EstimateTIealth Effects
'...: h::...... 211
Concentration-Response Functions .~..'-... - 212
Issues Common to Several Pollutants .- 212
Pollutants Using Dose-response Information from Epidemiological Studies 213
Ozone : 213
Particulate Matter ". 225
Nitrogen Oxides . .'. 236
Pollutants Using Dose-response Information from Laboratory Studies 238
Carbon Monoxide 238
Sulfur Dioxide 241
Visibility: Selection and Use of Concentration-Response Functions 243
Health Effects Model , 245
Air Quality ....... v ^ i.".... 245
Data Coverage ;:. ^i.. 246
Population Distribution .. .-,*..-....-.. 248
Census Data '. 249
Gridding U.S. Population 249
Allocating Exposure Estimates to the Population 249
Health And Welfare Effects 251
Health Effects Model Results I 253
Health Modeling References 257
Appendix E: Ecological Effects of Criteria Pollutants . 263
Introduction 263
Benefits From Avoidance of Damages to Aquatic Ecosystems 263
Acid Deposition _. ; 264
Background 264
Current Impacts of Acid Deposition 267
Benefits From Acid Deposition Avoidance Under the CAA 269
Eutrophication 271,
Atmospheric Deposition and Eutrophication 271
Valuing Potential Benefits from Eutrophication Avoidance Under the CAA ,... 272
Mercury 273
Benefits from Avoided Damages to Wetland Ecosystems 274
Introduction 274
Effects of Acidification 275
VI
-------
\
Effects of Nutrient Loading : . 276
Summary of Wetland Ecosystem Effects -... 277
Benefits from Avoided Damages to Forests .., -i........ 277
Introduction . . 277
Current Air Pollutant Effects on Forests 278
Acid Deposition Impacts T. 278
Ozone Impacts . 279
Valuation of Benefits From CAA-Avoided Damages to Forests 280
Background ^. k 280
Commercial Timber Harvesting -.". .-. 281
Non-marketed Forest Services -. I* _. -282
Ecosystem Effects References , ; , ...,.;. 283
Appendix F: Effects of Criteria Pollutants on Agriculture 287
Introduction ... ^.~,- 287
- Ozone Concentration Data _ 288
Control and No-control Scenario Ozone Concentration Data 288
Calculation of the W126 Statistic 289
Aggregating Ozone Data to the County Level ; 290
Yield Change Estimates .^ ....;;..; 291
Exposure-Response Functions .v. 291
Minimum/Maximum Exposure-Response Functions 291
Calculation of Ozone Indices ... v..., .,-.' 292
Calculations of County Weights .>.. .. ...» 293
Calculation of Percent Changeiin Yield ,:.... ;i. 293
Economic Impact Estimates >,.,... ... 295
Agricultural Simulation Model (AGSIM) ,.. «. 296
Conclusions : 298
Agricultural Effects References : 299
f
Appendix G: Lead Benefits Analysis 301
Methods Used to Measure and Value Health Effects 301
Health Benefits to Children .. 3p2
Changes in IQ 302
Children with IQs Less Than 70 '. 308
Changes in Neonatal Mortality 309
Hypertension .' 311
Changes In Coronary Heart Disease 312
Changes in Initial Cerebrovascular Accidents and Initial Atherothrombotic Brain Infarctions
.... v- • • • • • •••••• •••••• 316
CHangesin Premature Mortality : 317
Health Benefits to Women 318
Changes in Coronary Heart Disease 319
Changes in Atherothrombotic Brain Infarctions and Initial Cerebrovascular Accidents ...... 320
Changes in Premature Mortality ..— 320
Industrial Processes'and Boilers and Electric Utilities : 321
TRIData 321
Derivation of Industrial Process Emissions Differentials 1970-1990 322
Matching TRI Data to Industrial Process Emissions Differentials 324
Vll
-------
TRIData ...325
Derivation of Industrial Combustion Emissions 1970-1990 325
Matching TRI Data to Industrial Combustion Emissions Data 327
Coal-Use Data ; 328
The EPA Interim Emissions Inventory .... 328
Matching the Coal-Use Data to the Interim Emissions Inventory 328
Emissions Factors and Control Efficiencies 328
Relationship Between Air Lead Concentrations and Blood Lead Levels 332
Estimates of Initial Blood Lead Concentrations .../,.. 334
Reduction in Health Effects Attributable to Gasoline Lead Reductions ; 340
1970-Forward and 1990-Backward Approaches 341
Lead Benefits Analysis References .... 346
- yt •;
AppendixH: AirToxics ~. ~;>. 353
Introduction »... .r. 353
Limited Scope of this Assessment ~.,. 353
Historyof Air Toxics Standards under the Clean Air Act of 1970 355
Quantifiable Stationary Source Air Toxics Benefits 355
EPA Analyses of Cancer Risks from Selected Air-Tpxic Pollutants 356
Cancer Risk Estimates from NESHAP Risk Assessments 357
Non-utility Stationary Source Cancer Incidence Reductions 358
PES Study 359
Methodology '. ..;...... 359
Findings .'...• ;.> 360
ICFRe-analysis ...•' .*. 361
Methodology ; ,,.;.T. 361
Findings .. .->>•, 362
Mobile Source HAP Exposure Reductions 364
Methodology 365
Results ... 366
Non-Cancer Health Effects 366
Ecological Effects , .'. 367
Conclusions — Research Needs 368
Health Effects ... 368
Exposure Assessment ,,. 369
Ecosystem Effects . 369
Economic Valuation 370
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants 371
Methods Used to Value Health Effects 371
Results of Valuation of Health and Welfare Effects .. 381
Sensitivity Analyses 385
Economic Valuation References 393
AppendixJ: Future Directions „ 395
Research Implications 395
Future Section 812 Analyses 398
vm
-------
This project is managed under the direction of Robert D. Brenner, Directotof the UjiFEPA Office of
Air and Radiation/Office of Policy Analysis and Review and Richard D. Morgenstern, Associate Assistant
Administrator for Policy Planning and Evaluation, U.S. EPA (currently on leave as Visiting Scholar,
Resources for the Future). The principal project managers are JinrDeMocker, EPA/OAR/OPAR; Al
McGartiand, Director, EPA/OPPE/OEE; and Tom Gillis, EPA/OPPE/OEE. ~"
Many EPA staff contributed or reviewed portions of this draft document, including Joel Schwartz,
Michael Shapiro, Peter Preuss, Tracey Woodruff, Diane DeWitC^^rAxekad, Joel Scheraga, Anne
Grambsch, Jenny Weinberger, Allyson Siwik, Richard Scheffe,vVasu~Kilaru, Amy Vasu, Kathy
Kaufmann, Mary Ann Stewart, Eric Smith, Dennis J. Kotchmar, Warren Freas, Tom Braverman, Bruce
Polkowsky, David Mobley, Sharon Nizich, David Meisenheimer, FredDimmick, Harvey Richmond, John
Haines, John Bachmann, Ron Evans, Tom McMullen, Dan Mussatti, Bill Vatavuk, Larry Sorrels, Dave
McKee, Susan Stone, Melissa McCullough, Rosalina Rodriguez, Vickie Boothe, Tom Walton, Michele
McKeever, Vicki Atwell, Kelly Rimer, Bob Fegle£,~Aparna Koppikar, Les Grant, Judy Graham, Robin
Dennis, Dennis Leaf, Ann Watkins, Penny Carey, Joe Somers, Pam Brodowicz, Byron Bunger, Allen
Basala, David Lee, Bill O'Neill, Susan Herrqd, and Susan Stendebach. Allyson Siwik of
EPA/OAR/OAQPS and Bob Fegley of EPA/ORD/OSPRE played particularly important roles in
coordinating substantive and review contributions from their respective offices.
A number of contractors developed key elements of the analysis and supporting documents. These
contractors include Bob Unswjrth and Jim Neumann of Industrial Economics, Incorporated (ffic); Leland
Deck, Lisa|^|OTn, Brad ;Fjf|||~Susan Keane,TCathleen Cunningham, and John Voyzey of Abt Associates;
Bruce Bfffiil^pkia Kirnl^&deep^Kohli, Anne Button, Barry Galef, Cynde Sears, and Tony Bansal of
ICF Resol^^M^ingstaft Mch^ Woolfolk, Shelly Eberly, Chris Emery, Till Stoekenius, and
Andy Gray ofl^^|ms Appficli^ffifernational (ICF/SAT); Dale Jorgenson, Peter Wilcoxen, and
Richard Goettlej^^fenAssocffii^llm Lockhart of the Environmental Law Institute (ELI); Beverly
Goodrich, Reh^^^^ft^%oberts, and Lucille Bender of Computer Sciences Corporation; Margaret
Sexsmith of ^nalyticai'Sg|nces,^ Incorporated; Ken Meardon of Pacific Environmental Services (PES);
David South, Gale Boyd, Melalnie Tomkins, and K. Guziel of Argonne National Laboratory (ANL); Don
Gamer; Rex Brown and Jacob Ulvila of Decision Science Consortium; and Jim Wilson and Dianne P.
Crocker of Pechan Associates. John Pitcher and H. Glenn Court of STRA managed the technical
tSoduction of the document. The SARMAP AQM runs were provided by Carol Bohnenkamp of EPA
Region 9 and Saffet Tanrikulu of the California Air Resources Board.
Science Advisory Board review of this report is supervised by Donald G- Barnes, Director of the SAB
Staff. SABrstaff coordinating the reviews have included Jack Kopyoomjian, Sam Rondberg, Fred Talcott,
and Randall Bond. Diana Pozun provided administrative support.
- i •.''",., - •
The SAB ACCACA is chaired by Richard Schmalensee of MTT. Members include Morton Lippmann
of New York University Medical Center, William Nordhaus of Yale University, Paul Portney of
Resources for the Future, Kip Viscusi of Duke University, A, Myrick Freeman of Bowdoin College,
Maureen Cropper of the World Bank, Ronald Cummings of Georgia State University, Daniel Dudek of me
Environmental Defense Fund, Robert Mendelsohn of Yale University, Wayne Kachel of MELE
Associates, William Cooper of Michigan State University, Thomas Tietenberg of Colby College, Paul lioy
t
ix
-------
Acknowledgments
of the Robert Wood Johnson School of Medicine, Roger McClellan of the Chemical Industry Institute of
Toxicology, Richard Conway of Union Carbide Corporation, and Wallace Gates of the University of
Maryland.
The SAB ACCACA Physical Effects Subcommittee is chaired by Morton Lippmann. Members
included David V. Bates of the University of British Columbia, A|Myrick Freeman of Bowdoin.College,
Gardner Brown, Jr. of the University of Washington, Timothy Larson of the University of Washington,
Lester Lave of Carnegie Mellon University, Joseph Meyer of the University of Wyoming, Robert Rowe of
Hagler Bailly, Incorporated, George Taylor of the University of Nevada, Bernard Weiss of the University
of Rochester Medical Center, and George Wolff of the General Motors Research Laboratory.
The SAB ACCACA Air Quality Subcommittee is chaired by George Wolff. Members Have included
Benjamin Liu of the University of Minnesota, Peter Mueller of the Electric Power Research Institute,
Warren White of Washington University, Joe Mauderly of the Lovelace Biomedical & Environmental
Research Institute, Philip Hopke of Clarkson University, Paulette Middleton of Science Policy Associates,
James H. Price, Jr. of the Texas Natural Resource Conservation Commission, and Harvey Jeffries of the
University of North Carolina, Chapel Hill.
A number of interagency review, meetings have been held during the course of the development of this
analysis. Agencies and Departments which Save been-represerited at*these meetings include the Council
oa Environmental Quality, the Council of Economic Advisors, the Department of Energy, the National
Acid Precipitation Assessment Program, me Department of Commerce, the Department of Labor, and the
Office of Management and Budget ,;i .
This report could not have been produced without the support of key administrative support staff. The
project managers are grateful to l^oria.Smoke, Carolyn Hicks, Eunice Javis, Gloria Booker, Thelma Butler,
Wanda Farrar, Ladonya Langston, Michelle Olawuyi, and Eileen Pritchard for their timely and tireless
support on this project.
-------
Executive Summary
Purpose of the Study
Throughout the history of the federal Clean Air Act, questions have been raised as to whether the
health and environmental benefits of air pollution control justify the costs incurred'by industry, taxpayers,
and consumers. For the most part, questions about the costs and benefits of individual regulatory standards
continue to be addressed during the regulatory development process through Regulatory Impact Analyses
(RIAs) and other analyses which evaluate regulatory costs, benefits, and design variables such as scope,
stringency, and timing. There have never been, however, any comprehensive, long-term; scientifically ,
valid and reliable studies which answered the broader question:
"How do the overall health,-welfare, ecological, and economic benefits of Clean Air Act
programs compare to the costs of these programs?"
M
To address this void, Congress added to thjft990 Clean Air Act Amendments a requirement under
section 812 that EPA conduct periodic, stierilftcally reviewed studies to assess the benefits and the costs
of the Clean Air Act. Congress further required EPAJto conductthe assessments to reflect central
tendency, or "best estimate," assumptions rather than'the conservative assumptions sometimes deemed
appropriate for setting protective standards.
This report is the first in this ongoing series of Reportsto Congress. By examining the benefits and
costs of the 1970 and 1977 Amendments, this report addresses the question of the overall value of
America's historical investment in cleaner air. TWfirst Prospective Study, now in progress, will evaluate
the benefits and coste of the 1990 Amendments.
Study Design
-Designing a study to evaluate effectively the benefits and costs of the entire Clean Air Act over a 20-
yealPperiod throughout thirU.S. was a formidable challenge, particularly given limitations in historical
data, scientific tools, and available resources. As a result of these constraints, the levels of geographical,
industry-specific and pollutant-specific detail which could be incorporated in the assessment were limited.
For example, the eniissions data and models available for this study supported development of emissions
prqfec^^^^^Site, rather than county, level. While relying on state-level data allows for development
of reasOT|Wlslational-level estimates of the benefits and costs of the Clean Air Act, it is not possible use
the results of this study to identify costs and benefits of the Clean Air Act for particular locations. The
need to rely on national aggregate compliance expenditure data similarly precludes estimation of costs and
benefits by individual company or industry. Finally, atmospheric transport and transformation of
pollutants from one species to another (e.g., transformation of gaseous sulfur dioxide to participate :
sulfates) make it difficult to estimate benefits and costs by individual pollutant.
XI
-------
Executive Summary
Another important feature of the current study is that it assesses the benefits and costs of all air
pollution control measures implemented from 1970 to 1990 regardless of the motivation for their adoption.
This approach makes it unnecessary to speculate about which actions were taken by states arid local
governments in response to Clean Air Act mandates and which actions were taken for other reasons.
By allowing for acceptable limitations on the disaggregation of results by government sector, industrial
sector, pollutant, and location, the study could be designed and data and models could be selected which
allowed EPA to measure the total benefits and costs of historical air pollution controls. In fact, even with
these limitations, the present study represents the most extensive evaluation ever completed of the Clean
AirAct . , ,
f
Nevertheless, significant uncertainties usually pervade assessments of environmental program costs
and benefits, and the present study is no exception. While there is uncertainty associated with each step of
the analysis, resources permitted quantitative assessment of only the physical effects estimation and
economic valuation steps. Potential biases and errors in other steps of the analysis are evaluated
qualitatively, however, and the Project Team believes that the physical effects and valuation steps are the
ones which contribute most to aggregate uncertainty in the estimate of net benefits.
The study derived the benefit and cost estimates by examining the differences in economic, human
health, and environmental outcomes under two alternative scenarios: a "control scenario" and a "no-control
scenario." The control scenario reflects actual historical implementation of clean air programs and is based
largely on historical data. The no-control scenario is a hypothetical scenario which reflects the assumption
that no air pollution controls were established beyond those in place prior to enactment of the 1970
Amendments. Each of the two scenarios were men evaluated by a sequence of economic, emissions, air
quality, physical effect, economic valuation, and uncertainty models to yield the differences between the
scenarios in economic, human health, and environmental outcomes. Details of this analytical sequence are
presented in Chapter 1 and are summarized in Figure 4 of that chapter.
Summary of Results
Direct Costs
To comply with the Clean Air Act,
businesses, consumers, and government
entities all incurred higher costs for many
goods and services. The costs of providing
goods and services to the economy were
higher primarily due to requirements to
install, operate, and maintain pollution
abatement equipment. In addition, costs
were incurred to design and implement
regulations, monitor and report regulatory
compliance, and invest in research and
development Ultimately, these higher costs
Figure 1. 1990 Control and No-control Scenario Emissions
(in millions of short tons).
Xll
-------
Executive Summary
of production were borne by stockholders, business owners, consumers, and taxpayers.
To tally total compliance expenditures made pursuant to the Clean Air Act over the period from 1970
to 1990, actual dollars spent in a given year were adjusted to reflect their value in 1990. The purpose of
this adjustment was to correct for the effects of inflation so that expenditure levels in different years could
be compared. In addition, an adjustment was made to the raw expenditure data to take account of the fact
- ' " ' j^*?'"*1!
that some expenditures in any given year were for control equipment which provided benefits for more *
than one year. This "annualization" adjustment, which is similar to calculating the value of a year's worth
of mortgage payments for a house, allowed for a comparison of costs in a,gjven year with the value of the
emissions reductions achieved in that same year. Figure 2 summarizes the historical data on Clean Air Act
compliance costs, adjusted both for inflation and for the value of long-term investments in equipment.
These direct cost data show that, even on an inflation-adjusted basis, direct costs to meet requirements
of the Clean Air Act increased steadily over the period of the analysis at a rate of increase equal to a little
under four percent per year. This compares to a growth rate in Real Gross Domestic Product (GDP) over
the same period of about 334 percent. The implication of mis comparison with GDP is that air pollution
control expenditures during the study period increased at a slightly faster rate than total output by the U.S.
economy.
Indirect Economic Effects.
As part of the process of developing input data to model emissions under the control and no-control
scenarios, a macroeconomic model was run to estimate changes in key economic variables such as total
electricity sales; In addition to providing economic indicators needed for the emissions modeling, -the
macroecoSomic model provided estimates of other; indirect economic effects of the Clean Air Act. For
example, thPmacroeconomic model runs for the control and no-control scenario provided estimates of
overall econofflc^pr^duction, prices'for goods and services from various sectors, levels and patterns of
investment, and patterns of employment
Whilejthese indirect effect estimates are reported in Chapter 2 and Appendix A, EPA concluded that
these projections do not provide a complete and satisfactory indication of the overall economic effects of
the Clean Air Act and are not presented as primary results of this analysis. The principal limitation of the
macroeconomic results isjjpat, while they do reflect some of the indirect economic costs of air pollution
control expenditures, they do not reflect the indirect economic benefits of the pollution control achieved.
For example, the macroeconomic model results suggest there was a decrease in total U.S. production due
k> displacement of capital investment caused by pollution control expenditures; however, the
macro^conomc modeling does not capture the increase in total U.S. production achieved by reductions in
ak pollution-related worker illness and absenteeism. The macroeconomic modeling also omits the
economic expenditure reductions achieved by the Clean Air Act, such as reductions in medical expenses
for air pollution-related illness and disease. These savings would offset, at least in part, the capital
displacement effect of compliance expenditures. This imbalanced treatment by the macroeconomic model
of the effects of the Clean Air Act led EPA to conclude that the most appropriate comparisons in the
context of this study are between the direct costs and the direct benefits of the Clean Air Act.
Xlll
-------
Executive Summary
Figure 2. Total Direct Costs of the CAA (in billions of
inflation-adjusted 1990 dollars.) .
1975
Emissions
Emissions were substantially lower by 1990 under the control scenario than under the no-control
scenario, as shown in Figure 2. Sulfur dioxide (SO^ emissions were 40 percent lower, primarily due to
utilities installing scrubbers and/or switching to lower sulfur fuels. Nitrogen oxides (NOJ emissions were
30 percent lower by 1990, mostly because of
the installation of catalytic converters on
highway vehicles. Volatile organic compound
(VOC) emissions were also lower primarily
due to motor vehicle controls, with 1990 VOC
levels 45 percent lower under the control
scenario than under the no-control scenario.
Motor vehicle controls were also primarily
responsible for a 50 percent reduction in 1990
carbon monoxide (CO) emissions under the
control scenario.
For particulate matter, it is important to
recognize the distinction between reductions
in directly emitted particulate matter arid
reductions in ambient concentrations of
particulate matter in the atmosphere; As
discussed further in the next section; changes •
in particulate matter air quality depend both
on changes in emissions of total suspended particulates (TSP) and on changes in emissions of gaseous
pollutants, such as sulfur dioxide and nitrogen oxides, which can be converted to particulate matter
through chemical transformation in the atmosphere. Emissions of primary, or directly emitted, total
suspended particulates were 75 percent lower under the control scenario by 1990 than under the no-control
scenario. This substantial difference is primarily due to vigorous efforts in the 1970s to reduce visible
emissions from utility and industrial smokestacks.
Control scenario lead emissions for 1990 were projected to be about 3,000 tons, reflecting a 99 percent
reduction from the projected no-control level of 237,000 tons. The vast majority of the difference in lead
emissions under the two scenarios is attributable to reductions in the use of leaded gasoline. By 1990,
reductions in highway vehicle emissions accounted for 221,000 of the total 234,000 ton reduction in total
airborne lead emissions under the control scenario. (Results for lead are not shown in Figure 2 because the
absolute levels of lead emissions are measured in thousands, not millions, of tons, and will not be
discernible on a graph of this scale.)
As discussed in Chapter 3, there are important uncertainties in the emissions estimates. These include
(a) potential errors in the macroeconomic model results used to configure the emissions models and (b)
potential overestimation of no-control scenario emissions resulting from the use of 1970 emission factors
throughout the 1970 to 1990 period.
xiv
-------
Executive Summary
Air Quality
The substantial reductions in air pollutant emissions achieved by the Clean Air Act translate into
significantly improved air quality throughout the U.S. For sulfur dioxide, nitrogen oxides, and carbon
monoxide, the improvements in air quality under the control scenario were treated as proportional to the
estimated reduction in emissions. This is because, for these polTiitalts, changes in ambient concentrations
in a particular area are strongly related to changes in emissions in that area.. While the differences in
control and no-control scenario air quality for each of these pollutants do vary from place to place because
of local variability in emissions reductions, the overall national average improvements in air quality for
these pollutants achieved by 1990 were: 40 percent improvement for sulfur dioxide, 30 percent
improvement for nitrogen oxides, and 50 percent improvement for carbon monoxide.
—' ? V-
The differences in ambient concentrations of ground-level ozone and participate matter under the two
scenarios, however^ are not so straightforward. I/jng-rangeJransport through the atmosphere, non-linear
formation mechanisms, and other complexities meant that air quality models had to be used to translate
changes in emissions of precursors of these pollutants to "changes in ambient concentrations of ozone or
participate matter. Estimating differences in acid deposition and visibility in the Eastern U.S. also required
the use of air quality models since they are not necessarily directly proportional to changes in emissions of
pollutants which cause these effects.
Reductions in ground-level ozone were achieved through reductions in its precursor pollutants,
particularly volatile organic compounds and nitrogen oxides/ The differences in ambient ozone
concentrations estimated under the control scenario varied significantly from one location to another,
primarily because of local differences in the relative proportion of VOC and NO*, in meteorology and in
precursor emissions reductions. On a national average basis, however, ozone concentrations in 1990 under
the control scenario were about 15 percent lower than projected under the no-control scenario. For several
reasons, this overall reduction in ozone is significantly less than the estimated 30 percent reduction in
precursor nitrogen oxides and the estimated 45 percent reduction in precursor volatile organic compounds.
First, significant natural (i.e., biogenic) sources of VOGs limit the level of ozone reduction achieved.by
reductions in man-made (i.e., anthropogenic) VOCs. Second, current knowledge of atmospheric
photochemistry suggests that ozone* reductions will tend to be proportionally smaller than reductions in
precursor emissions. Finally, the plume model system used to estimate changes in urban ozone for this
study is incapable of handling long-range transport of ozone from upwind areas, complex flow phenomena,
and multi-day pollution^events in a realistic manner. While regional scale ozone models address these
phenomena more effectively, resource limitations precluded using these more sophisticated but costly
models Jor this study.
* ~ •* ,
~ jThere are many pollutants which contribute to ambient concentrations of particulate matter. The
relative contributions of these individual pollutant species to ambient particulate matter concentrations vary
from one region of the country to the next, and from urban areas to rural areas. The most important
particle species, from a human health standpoint, are the fine particles which can be respired deep into the
lungs. While some fine particles are directly emitted by sources, the most important fine particle species
. 'Ambient NO2 concentrations are driven by anthropogenic emissions whereas ambient VOCs result from both anthropogenic and biogenic
sources (e.g., terpenes emitted by trees).
XV
-------
Executive Summary
are actually formed in the atmosphere through chemical conversion of gaseous air pollutants. These
species are referred to as secondary particles. The three most important secondary particles are (1) sulfates,
which derive primarily from sulfur dioxide emissions; (2) nitrates, which derive primarily from nitrogen
oxides emissions; and (3) organic aerosols, which can be directly emitted as well as a result of volatile
organic compound emissions. This highlights an important and unique feature of particulate matter as an
ambient pollutant: more than any other pollutant, reductions in particulate matter are actually achieved
through reductions in a wide variety of air pollutants. In other words, controlling particulate matter means
controlling "air pollution" in a very broad sense.
Because of its comprehensive coverage of pollutants contributing to ambient particulate matter, the
present study is uniquely configured to illustrate the broader range of benefits achieved by historical
reductions in "air pollution." The results of this analysis indicate that the reductions in sulfur dioxide,
nitrogen oxides, volatile organic compounds, and directly-emitted pardcles achieved by the Clean Air Act
resulted in an overall, national average reduction in total ambient particulate matter of about 45 percent by
1990.
Reductions in sulfur dioxide and nitrogen oxides also translate into reductions in formation, transport,
and deposition of secondarily formed acidic compounds such as sulfate and nitric acid. These are the
principal pollutants responsible for acid precipitation, or "acid rain." Regional acid deposition modeling of
the control and no-control scenarios indicatesthat, by 1990, sulfur and nitrogen deposition were
significantly higher under the no-control scenario throughout the 31 Eastern States covered by the model.,
Percentage increases in sulfur deposition under the no-control scenario ranged up to more than 40 percent
in the upper Great Lakes and Florida-Southeast Atlantic Coast areas and were due, primarily, to significant
projected increases in the use of high-sulfur fuels by utilities in the upper Great Lakes and Gulf Coast
states. Nitrogen deposition aiso;was significantly higher under the no-control scenario, with percentage
increases reaching levels of 25 percent or higher^albtig the Eastern Seaboard. This higher level of nitrogen
deposition can be attributed primarily to higher projected emissions of nitrogen oxides from motor
vehicles. ,; ,
• »••/ • •'
Finally, decreases in ambient concentrations of light-scattering pollutants, such as sulfates and nitrates,
under the control scenario.wereestimated to lead to perceptible improvements in visibility throughout the
Eastern states and Southwestern urban areas modeled for this study.
The assumptions required to estimate hypothetical no-control scenario air quality conditions are
particularly significant sources of uncertainty in the estimates of air quality change, especially for those
pollutants which are not linearly related to changes in associated emissions such as ozone and particulate
matter. As discussed in Chapter 4, limitations on geographic coverage and lack of access to state-of-the-art
models, among other factors, led to significant uncertainties in the estimates of air quality change between
the scenarios. Most of these uncertainties are expected to lead to biases which are either unknown in
direction or which tend toward underestimation of air quality improvements due to the Clean Air Act.
Physical Effects
The lower ambient concentrations of sulfur dioxide, nitrogen oxides, particulate matter, carbon
monoxide, ozone and lead projected under the control scenario yielded a substantial variety of human
health, welfare and ecological benefits. For a number of these benefit categories, quantitative functions
-xvi
-------
Executive Summary
were available from the scientific
literature which allowed estimation of
the reduction in incidence of adverse
effects.' Examples of these categories
include the human mortality and
morbidity effects of a number of
pollutants, changes in
neurobehavioral effects among
children caused by exposure to lead
and changes in visibility impairment.
\
A number of benefit categories,
however, could not be quantified
and/or monetized for a variety of
reasons. In some cases, strong
scientific evidence of an effect
existed, but data were still too limited
to support quantitative estimates of
incidence reduction (e.g., changes in
lung function associated with long-
term exposure to ozone). In other
cases, substantial scientific'
uncertainties prevailed regarding the
existence and magnitude of adverse
effect (e.g., contribution of ozone to
air pollution^related mortality).
Finally, ™
there was
estimate incidenl
Table 1,/Selected Health Benefits of the CAA, 1970-1990 (in
thousands of cases reduced per year, except as noted).
H«
Effect
Mortali\ > 'lug
(PM10,Oj,§p2,Pb) mid
(thousands^ low
Heart Attacks
(thousands)
Strokes
'(P«>)
(thousands)
X
high
mid
lowy
• jr
Respiratory symptoms
,(SO^ ,
(thousands)
/
Respiratory Ulnefe
(Ncy "X
(millions) /
high
mid
1975 198ft
1985
38
20
11
140
79
45
19 < 24*
'14' V18
; 10 \13
5
4
"3
10
,8
" 6
43
10
7
66
165' „ 146
15
6
5'
4
12
10
>8
fi''*- A
effects for which
information to
ridiiction. but for
. SKMSSfy _ '
which there \Ygig3i§lpailable economic value measures; thus reductions in adverse effects could not be
expressed mlnonetarytenns. Examples of this latter category include pulmonary function decrements
caused by acute exposures to ozone and reduced time to onset of angina pain caused by carbon monoxide
exposure.
"Table 1 provides a summary of some of the key differences in estimated human health outcomes under
the control and no-control scenarios. Results are presented as thousands of cases avoided in 1990 due to
control of the pollutants listed in the table and reflect reductions estimated for the entire U.S. population
living in the 48 continental states.
, -F • ' -
Numerous scientific studies were used as sources of the concentration-response functions applied to
estimate differences in premature mortality, morbidity, and other outcomes under the two scenarios. These
different studies often implied different ranges of estimates for each of these outcomes, contributing to
uncertainty in the quantitative estimates of incidence. These uncertainties were captured quantitatively
through the use of Monte Carlo computer modeling, which is briefly described later in this Executive
Summary and described in more detail in Chapter 7. Table 1 Summarizes the ranges of estimates generated
by the Monte Carlo modeling for each of the monetizable benefits of the CAA.
XV11
-------
Executive Summary
In Table 1, the 5th and 95* percentile estimates for each effect reflect the ends of a range which, based
on the available studies, is 90 percent likely to include the true estimated incidence for that effect. The
mean estimate for each effect is the average of all of the Monte Carlo-generated estimates for that effect.
Perhaps the most important uncertainty associated with the results presented in Table 1 is in the
estimation of particulate matter-related mortality. For this particular endpoint, considerable scientific
uncertainty remains concerning the influence of elevated particulate concentrations on prematurity of
death. Table 1 presents results based on both long-term and sjiort-term studies. The short-term studies*;?v
provide a basis for estimating incidences of premature mortality associated with acute episoHes of exposure
to elevated particulate matter, but the magnitude of life-shortening and the effect of long-term exposures to
elevated particulate matter are not discernible from these studlesrTiielbng-tenn studies a3dress these two
specific deficiencies in the short-term studies but may be confounl&dby the effects of pollutant exposures
occurring prior to the period of the long-term study. The estimates presented in Table 1 are therefore not
additive but should be viewed as an expression of the plausible, overall range of avoided premature
mortality associated with particulate matter exposure. The estimates of premature mortalities avoided due
to reduced lead exposure ere additive with the PM-related short-term and long-term exposure estimates.
Adverse human health effects of the Clean Air Act "criteria pollutants" sulfur dioxide, nitrogen oxides,
ozone, particulate matter, carbon monoxide, and lead dominate the quantitative estimates presented in this
report in part because knowledge of physical relationships is greatest for these pollutants. The Clean Air
Act yielded other benefits, however, which are important even though they may have been uncertain and/or
difficult to quantify. The other principal benefit categories which, for a variety of reasons, could not be
satisfactorily quantified include all benefits accruing from reductions in hazardous air pollutants (also
referred to as air toxics); reductions in damage to cultural resources, buildings, and other materials;
reductions in adverse effects on wetland, forest, arid aquatic ecosystems; and a variety of additional human
health and welfare effects of criteria pollutants. A complete list of the major benefits which could not be
monetizedI is provided in Table 2.
In addition to controiling the six criteria pollutants, the 1970 and 1977 Clean Air Act Amendments led
to reductions in emissions of a small number of hazardous air pollutants. Control of these pollutants
resulted both from regulatory standards set specifically to control hazardous air pollutants and from
incidental reductions achieved through programs aimed at controlling criteria pollutants. Due to resource
and data constraints, however* reliable quantitative estimates of reductions in adverse effects of hazardous
air pollutants could not be developed for this report.
XVlll
-------
Executive Summary
< fable 2. Major,NoAoroaefeed Benefits of the dean An; Act.
PoUutant
Nomnonetized Benefits
Partkulate
Matter
Changes in Pulmonary Function
Other Chronic Respiratory Diseases
' Inflammation of the Lung
Chronic Asthma and Bronchitis
Ozone
Changes in PuInKinaiy Function „', -
Increased Airway Responsiveness to Stimuli ^ •
CentroacinarFibrpsis _ t' •*
Inflammation of the Lung ' ,
Immunblpgical Changes
Chronic Respiratory Diseases > •
Extrapulmonary Effects (i.e., other organ systems)
Forest and other Ecological Effects
Materials Damage
<.
Carbon •
Monoxide
Decreased Time to Onset of Angina
Behavioral Effects
Other Cardiovascular Effects
Developmental Effects -
Sulfur
Dioxide
Reductions in both hazardous air
pollutants and criteria pollutants likely
yielded widespread improvements in
the functioning and quality of aquatic
and terrestrial ecosystems. In addition
to any intrinsic value to be attributed
to these ecological systems, human
welfare is enhanced through
improvements in a variety of
ecological services. For example,
protection of freshwater ecosystems
achieved through reductions in
deposition of acidic air pollutants may
have improved commercial and
recreational fishing. Other potential
ecological benefits of reduced acid
deposition include improved wildlife
viewing, maintenance of biodiversity,
and nutrient cycling. Increased
growth and productivity of U.S.
forests may have resulted from
reduced emissions of ozone-forming
precursors, especially volatile organic
compounds and nitrogen oxides.
More vigorous forest ecosystems in
turn yieldliiraiiety of benefits,
including increased timber
production; improved forest aesthetics
for people enjoying outdoor activities
such as hunting, fishing, and
camping; and improvements in
ecological services such as nutrient
cycling and temporary sequestration
of global warming gases. Again, due
to resource and data limitations these
improvements in ecological
conditions have hot been quantified in
this assessment.
In considering the quantitative
results of this analysis, it is important •••••^^^^••^^^^••i TI " '"' '—*—i
to remember the potential significance
of the benefits which could not be
quantified in physical terms. This caveat is even more significant for consideration of the monetary benefit
results presented below. The range of benefits included in the economic assessment is further reduced
Nitrogen
Oxides
Lead
Air Toxics
Respiratory" Symptoms m Non-Asthmatics
Hospital Admissions
Agricultural Effects
Materials Damage
EcologcalJBffects - ' ~"
Increased Airway Responsiveness to Stimuli"
Decreased Pulmonary Function
Inflammation of the Lung
Immunological Changes
Eye Irritation
Materials Damage
Eutrophication (e.g., Chesapeake Bay)
Acid Deposition * ;
Cancer \
Cardiovascular Diseases .
Reproductive Effects in Women
Other Neurological, Metabolic Effects in Children
Fetal Effects from Maternal Exposure, me IQ Loss
Ecological Effects
All Human Health Effects
Ecological Effects
XIX
-------
Executive Summary
because some effects, such as carbon monoxide-related reduction in time to onset of angina attack, could
be quantified in physical terms but not valued in economic terms.
Economic Valuation
Estimated reductions in the incidence of adverse effects represent an important, and to some people a -
sufficient, basis for characterizing the benefits of air pollution control. However, to aggregate the full
range of benefits and compare them to costs, a common unit of measure must be applied. The metric most
commonly used for cost-benefit analysis is dollars. Assigning monetary values to all effects which are
amenable to economic valuation permits a summation of monetized direct benefits and subsequent
comparison to the direct compliance costs of the Clean Air Act.
Before proceeding through this step, it is important to recognize the substantial controversies and
uncertainties which pervade attempts to characterize adverse human health and ecological effects of
pollution in dollar terms. To many, dollar-based estimates of the value of avoiding outcomes such as loss
of human life, pain and suffering, or ecological degradation do not capture the full and true value to society
as a whole of avoiding or reducing these effects. Adherents to this view tend^o favor assessment
procedures which (a) adopt the most technically defensible dollar-based valuation estimates for analytical
purposes but (b) leave the moral dimensions of policy evaluation to those who must decide whether, and
how, to use cost-benefit results in making public policy decisions. This is the paradigm adopted in the
present study. Given the Congressional mandate to perform a cost-benefit study of the Clean Air Act, the
Project Team has endeavored to apply widely-recognized, customary techniques of Applied Economics to
perform this cost-benefit analysis. However, EPA believes there are social and personal values furthered
by the Clean Air Act which have not been effectively captured by the dollar-based measures used in this
study. Therefore, EPA strongly encourages readers to look beyond the dollar-based comparison of costs
and benefits of the Clean Air Act and consider the broad social value of the reductions in adverse health
and environmental effects which have been achieved as well as any additional adverse social consequences
of regulation which may not be reflected in the cost estimates reported herein.
For this report, unit valuation estimates were derived from the economic literature and reported in
dollars per case reduced for health effects, and dollars per unit of avoided damage for welfare effects.
Similar to estimates of physical effects provided by health studies, the monetary values of benefits for this
study are reported both in terms of mean value and as a range of estimates. These value ranges, and
approaches used for each of the effects monetized in this study, are described in Chapter 6.
As with other components of the assessment, there may be significant uncertainties concerning
estimates of economic values. Potential errors in the estimates arise for a variety, of reasons. One
important example is the uncertainty caused by application of monetary values derived from one
circumstance (e.g., wage compensation demanded for accepting riskier jobs) to a different circumstance
(e.g., the value of reductions in risk from exposure to air pollution). Substantial uncertainties also apply in
cases where survey-based approaches, such as "contingent valuation," are used to estimate the monetary
value of reductions in risk of an adverse effect
These uncertainties are discussed in detail in Chapter 6 and in Appendix I. As for physical effect
estimates, the ranges of unit values were utilized in the Monte Carlo modeling to evaluate the effect of
valuation uncertainty on the overall monetary benefit estimates.
XX
-------
Executive Summary
Monetized
and Costs
benefits
The total monetized economic
benefit attributable to the Clean Air Act
was derived by applying the unit values
(or ranges of values) to the stream of
monetizable physical effects estimated
for the 1970 to 1990 period. In
developing these estimates, steps were
taken to avoid double-counting of
benefits. In particular, it was assumed
that studies providing different measures
of a given effect were providing
alternative estimates of the same effect.
For example, the Project Team had
access to several studies estimating air
pollution-related "restricted activity
days" of various degrees of severity, as
well as studies which estimated "lost
work days." To avoid double-counting,"
the various "restricted activity^ health
effects were^considered to be alternative
measures for a single, combined
endpoint.
To evaluate the total, aggregate effect
of uncertainties in the analysis, a Monte
Carlo computer model was used to
calculate the estimates of total monetary
benefit.
jTabie 3.'iC&irrai Estimates of Economic Value perTJrdt of 7
Avowed Effect(in I999;a6uars> -' « ,,\ '••"' "„
' Endpoint
Mortality - ( ""'
Heart Attacks'' _ ~
Strokes f *" "' v ,,
Hospital Admissions
Respiratory
Valuation (mid-estimate)
-„ '$4,^0,000 per case" '\
$587,000 per«ase ""
tfpper Respiratory niness *
Lower Respiratory Dlness
'Acute Bronchitis
Acute Respiratory ' , ;,
Symptoms'
Work Loss Days '
Restricted Activity Days
Asthma Attacks
[Q Changes , -"*
Lost IQ Points
Incidence of IQ < 70
Hypertension
Decreased Wbrker'Prbdnctiyity
Visibility - - v ,
Household Soiling
Agriculture (Net Surplus) "
'$73,00 per casef,"
.- $10,000 per case^
$8,000 per case " '^
s
^ 5. s t. (^ ft •*, t
" $18percase "">
™ \$10'per case
s '$45 per case
• -$17-per case
•$83-perday' ' "
$38perday
, ' ,$32pe'rcase ,j
"'^52,700 per case, < ,
, $680 per year per case
< *• rl j
' ' ' " '^ •, -' " *
* Direct Economic Valuation
"•."<•<• "-
"Direct Economic Valuation
. Direct Economic Valuation
' Est Change in Econ. Surplus
First, the model selected an estimate of the incidence of a given physical effect from the range of
possible incidence estiffiates calculated during the physical effects step. Second, the model selected an
estimate of the economic value of reducing that effect from the range of values derived from the economics
literature. TJurd,"me model multiplied the reduction in incidence by the unit value to derive one estimate
of themonetary benefit of reductions in that effect. The model then repeated this procedure over and over
again sampling new values from the range of estimated incidence and the range of economic value. After
running through this procedure thousands of times, a range of possible outcomes was compiled, with each
outcome carrying with it a specific probability of occurrence.
The idea behind this approach is to allow presentation of the bottom-line results in a way which
captures the combined effect of uncertainties in key variables, such as the concentration-response functions
and the unit economic values. It must be emphasized that the results presented herein reflect only the
uncertainties in those two particular variables because physical effects estimation and economic valuation
xxi
-------
Executive Summary
were the only steps for which uncertainty could be captured quantitatively. Quantifying the uncertainties
associated with earlier steps in the analytical sequence would have required multiple runs of the
macroeconomic, emissions, and/or air quality models. This was far beyond the scope of the.resources
available for this project However, prior experience in regulatory analysis has demonstrated repeatedly
that the physical effects estimation and economic valuation steps contribute much more to overall
uncertainty than these earlier steps.
It is also important to emphasize that the central case results presented below reflect the specific ~
analytical assumptions and methodologies adopted by the Project Team. The potential effects of adopting
alternative assumptions for two important variables, the discount rate and the valuation method for
particulate-related mortality, were assessed separately by performing additional sets of Monte Carlo model
runs. Varying one of these factors at a time, the sensitivity ofthe net benefit estimates to changes in each
of these factors was evaluated quantitatively. These sensitivity analyses, summarized in Chapter 7,
indicate that (1) adopting discount rates of three or seven percent in lieu ofthe central case five percent'
makes little difference in the mean or range of the net benefit estimates and (2) adjusting the value of
particulate-related mortality downward to reflect the relatively advanced age and/or poor health status of
many victims may have a potentially significant influence on the benefit estimates but would not be
expected to change the basic finding of substantial positive net benefits.
By adding up the ranges of values for each effect obtained by the procedure described above, a range
of possible economic values for the estimated differences in physical outcomes under the control and no-
control scenarios could be derived. Again, each ofthe possibly outcomes in this range carries with it a
certain probability of occurring. Based on this probability information, and adding up the results obtained
for the entire 20 year period of the study, the following statement can be made regarding the central case
results: '. .. :,f-/ "• ; v^--'*'••''-"'/""" '
The Project Team found that 90 percent ofthe credible estimates of the total monetized
benefits of the Clean Air Act realized during the period from 1970 to 1990 were within the
range of 10.5 to 40.6 trillion dollars, with a central estimate of 23.0 trillion dollars (in 1990-
value dollars, discounted at 5 percent). By comparison, annualized direct costs of
compliance equal approximately 523 billion dollars over the same period. Subtracting total
costs from total benefits, net direct benefits of 1970 to 1990 controls attributed in this
analysis to the Clean Air Act were within a 90 percent probability internval of 10.0 to 40.1
trillion dollars, with a central estimate of 22.5 trillion dollars. However, even this broad
range of estimated net benefits would be expected to expand further if analytical
uncertainties associated with compliance cots, macroeconomic effects, emissions projections,
and air quality modeling could be quantified and incorporated in the uncertainty analysis.
Finally,rthe central estimate of 23.0 trillion dollars hi benefits may be significantly
underestimated due to the exclusion of large numbers of benefits from the monetized benefit
estimate (e^*., all air toxics effects, ecosystem effects, numerous human health effects).
Figure 3 provides a graphical representation of the estimated range of total direct monetarized benefits
and compares this range to estimated direct compliance costs. Clearly, even the lower bound estimate of
monetized benefits substantially exceeds the costs of the historical Clean Air Act. As shown in the more
detailed, year-to-year results presented in Chapter 7, monetized benefits consistently and substantially
xxu
-------
Executive Summary
exceeded costs throughout the 1970 to 1990
period.
Future Directions
Figures. Total Direct Costs and Monetized Direct
Benefits of the Clean Air Act, 1970 to 1990 (in trillions
of 1990 dollars). -
, 50
,40-
'30-
I
020
H10-
O-1
95th percentife
Mean
5th perccntUe
Costs
Benefits
First and foremost, the conclusion to be
drawn from these results is that the benefits of
the Clean Air Act and associated control
programs substantially exceeded costs. Even
considering the large number of important
uncertainties permeating each step of the
analysis, it is extremely unlikely mat the
converse could be true.
A second important implication of this
study is that a large proportion of the
monetized benefits of the historical Clean Air
Act derive from reducing two pollutants: lead
and particulate matter2 (see Table 1). Some
readers may argue that, while programs to
control these two pollutants may have been worthwhile, many other historical Clean Air Act programs
would not pass a benefit-cost test when consideredln isolation. While this may or may not be true, this
analysisprbvides no evidenceTto support or reject such conjectures. On the cost side, the historical
expenditure datanised in this analysis are not structured in ways which allow attribution of control costs to
specific programs or standards. On the benefit side, most control programs yielded a variety of benefits,
many of which included reductions in other pollutants such as ambient particulate matter. For example,
new source performance standards for sulfur dioxide emissions from coal-fired utility plants yielded
benefits beyond those associated with reducing exposures to gaseous sulfur dioxide. The reductions in
sulfur dioxide emissions also led to reductions in ambient fine particle sulfates, yielding human health,
ecological, and visibility benefits. Even so, it is likely that some specific historical programs or standards
may not have yielded monetized benefits in excess of costs.
^Although this study indicates that the 1970 to 1990 cumulative direct benefits pf the Clean Air Act
-clearly exceeded the cumulative direct cots of those controls, these results do not inform the question of
whether additional future controls would be cost-beneficial. Examining this issue is the central purpose of
tfie first Section 812 Prospective Study now in progress.3 As the first in an ongoing series of benefit-cost
2Ambient particulate matter results from emissions of a wide array of precursor pollutants, including sulfur dioxide, nitrogen oxides, and
organic compounds. ,
'Nevertheless, a recent peer-reviewed study examining sulfate-related health benefits alone of the 1990 Amendments pertaining to acid rain
control found benefits .well in excess of the total estimated costs of the 1900 Amendments. See hagler Bailly Consulting, Human Health Benefits
from Sulfate ReductionsUnder Title IVofthe 1990 Clean Air ActAmendments, submitted to U.S. Environment Protection Agency/Office of Air
and Radiation/Office of Atmospheric Programs/Acid Rain Division, November 1995.
XX111
-------
Executive Summary
studies of future Clean Air Act programs, this study will evaluate the specific, incremental costs and
benefits of new control programs initiated subsequent to the 1990 Amendments.
f
As for the Retrospective Study, the first Prospective Study is designed to complement, not substitute
for, the regulatory analyses developed for individual programs or standards. Nevertheless, the first
Prospective Study will differ from the Retrospective Study in thatit will be structured to allow
comparisons of the costs and benefits of major program areas, rather than just"the totality of the 1990
Amendments. This will be feasible largely because, unlike the Retrospective Study, substantial high-
quality, program-specific benefit and cost data have already been developed which are consistent wittfthe
specific control and baseline scenarios to be analyzed.
Another important element of the first Prospective Studyis that the Project Team will seek, resources
permitting, to expand the range and depth of the benefits analysis. In-particular, efforts will be made to
develop valid and reliable estimates of the benefits of future reductions in hazardous air pollutants, one of
the key program areas addressed more fully by the 1990 Amendments than by the 1970 and 1977
Amendments. In addition, efforts will be made to provide amore thorough and informative assessment of
the ecological effects of reductions in both criteria pollutants and hazardous air pollutants. EPA's view is
that ecological benefits of air pollution controls were substantial, even if they could not be quantified for
this study. In addition, ecological effects have social and political significance far in excess of any
apparent dollar-based, economic measures of value. As such, the Project Team places a high priority on
improving and expanding assessment of ecological effects of air pollution controls.
This retrospective study highughts important areas of uncertainty associated with many of the
monetized benefits included in the quantitative analysis (see Appendix J). The sensitivity analyses
presented in Chapter 7 demonstrate that additional research in the areas of participate matter-related
mortality incidence, valuation of premature mortality, and valuation of particulate-related chronic -
bronchitis, among others, might fedu<:e critical uncertainties in future assessments.
Finally, the results of this retrospective study provide useful lessons with respect to the value and the
limitations of cost-benefit analysis as a ipbl for evaluating environmental programs. Cost-benefit analysis
can provide a valuable framework for organizing and evaluating information on the effects of
environmental programs. When used properly, cost-benefit analysis can help illuminate important effects
of changes in policy and can help set priorities for closing information gaps and reducing uncertainty. .
Such proper use, however, requires that sufficient levels of time and resources be provided to permit
careful, thorough, and technically and scientifically sound data-gathering and analysis. When cost-benefit
analyses are presented without effective characterization of the uncertainties associated with the results,
Cost-benefit studies can be used in highly misleading and damaging ways. Given the substantial
uncertainties which permeate cost-benefit assessment of environmental programs, as demonstrated by the
broad range of estimated benefits presented in this study, cost-benefit analysis is best used to inform, but
not dictate, decisions related to environmental protection policies, programs, and research.
" xxiv
-------
Tables
Table 1. Selected Health Benefits of the CAA, 1970-1990 (in thousands of causes reduceffier year, except
as noted).
.... • • • xvii
Table 2. Major Nonmonetized Benefits of the Clean Air Act «^ xix
Table 3. Central Estimates of Economic Value per Unit of Avoided Effect (in 1990 dollars) xxi
Table 4. Estimated Annual CAA Compliance Costs (Sbillions) .\. „. /.' 10
Table 5. Compliance Cost, GNP, and Consumption Impacts Discounted to 1990 ($1990 billions) 15
Tabled. Summary of Sector-Specific Emission Modeling Approaches 20
Table 7. Uncertainties Associated with Emissions Modeling.. .~.~.V. 24
Table 8. Key Uncertainties Associated with Air Quality Modeling. 34
Table 8 (con't). Key Uncertainties Associated with Air Quality Modeling. 35
Table 10. Human Health Effects of Criteria Pollutants 39
Table 11. Selected Welfare Effects of Criteria Pollutants. ^ 40
Table 12. Percent of Population (of the Continental US) within 50km of a monitor (or in a County with
PMmonitors), 1970-1990 .. 42
Table 13. Criteria Pollutants Health Benefits -.Distributions of 1990 Avoided Premature Mortalities
(thousands of cases reduced) for 48 State Population .. 44
Table 14. Criteria Pollutants Health Benefits - Distributions of 1990 Non-Fatal Avoided Incidence
(thousands of cases reduced) for4;fpState Population 45
Table 15. Health and Welfare Effects of Hazardous Air Pollutants. 48
Table 16. Uncertainties AsscjcSttedfyHth Physic^tEffect&^xJeling............ 49
Table 17.Hlsi$handWelfa*e!Efl^tsUnitValu^6ir|i?9^0dollars) ..... 52
Table ISfSuiamaiy of MolMii^ Valuation Estimates (mffi . 53
Table 19. ^^a^Mortal%^^3Sased on Wage-Risk Studies: Potential Sources and Likely Direction
• ' of Bias. '''feivvkv..*,., .,..:..,,'.•,. 59
Table 20. Presenf^lue oil970 to 1990 Monetized Benefits by Endpoint Category for 48 State
Population and total Monetized Benefits (billions of $1990, discounted to 1990 at 5 percent)
'.. .*&.........:.^.-.>i;v>.'......' '. .... ........ •...- 63
Table 21. Monetized Annual Benefits and Costs, 1970-1990 (in billions of 1990-value dollars).
- - .yi • i> • •.....,..'...'.., 64
Table 22. Quantified Uncertainty Ranges for Monetized Annual Benefits and Benefit/Cost Ratios, 1970-
1990 (in billions of 19l>0-value dollars) .. 66
Table 23. Effect of Alternative Discount Rates on Present Value of Total Monetized Benefits/Costs for
1970 to 1990 (itftnllions of 1990 dollars). 67
Table 24. Results of Mortality Benefits Sensitivity Analyses, Monetized Benefits for 1970 to 1990 (in
trillions of 1990 dollars, discounted at 5 percent) 68
Table 25. Compliance Costs, Some Avoided Costs, and Productivity Improvements, 1970-1990 (in billions
of $1990). 69
Table 26. Key Distinguishing Characteristics of the Jorgenson-Wilcoxen Model.
' » • • • .......... * 73
Table 27. Definitions of Industries Within the J/W Model.
'•' • 74
Table 28. Estimated Capital and O&M Expenditures for Stationary Source Air Pollution Control (millions
of current dollars). ... g2
xxv
-------
: Tables
Table 29. Estimated Recovered Costs for Stationary Source Air Pollution Control (millions of current
dollars) .'.... • • • • • •....84
Table 30. Estimated Capital and Operation and Maintenance Expenditures for Mobile Source Air
Pollution Control (millions of current dollars) 85
Table 31. O&M Costs and Credits (millions of current dollars) .. 86
Table 32. Other Air Pollution Control Expenditures (millions of current dollars) 88
Table 33. Estimated Annual CAA Compliance Expenditures (Smillions) , 89
Table 34. Compliance Expenditures and Annualized Costs, 1973 to "1990 ($1990 millions) 90
Table 35. Costs Discounted to 1990 ($1990 millions) V.; 90
Table 36. Differences in Gross National Product Between the Control and No-control Scenarios
: 91
Table 37. GNP and Consumption Impacts Discounted to 1990 ($1990 billions) 92
Table 38. Difference in Personal Consumption Between the Control -and J«fo-Control Scenarios.
93
Table 39. Percentage Difference in Energy Prices Between the Control and No-control Scenarios.
'. 94
Table 40. Potential Sources of Error and Their Effect oh Total Costs of Compliance. 98
Table41. BEA Estimates of Mobile Source Costs •.-. "99
Table 42. Comparison of EPA and BEA Stationary Source Expenditure Estimates (millions of current
dollars) 100
Table 43. Annualized Costs Assuming 40-Year Stationary Source Capital Amortization Period, 1973 to
1990 ($1990millions) • 101
Table 44. Effect of Amortization Periods on AMualized Costs Discounted to 1990 (billions of $1990)
, ......;...... 101
Table 45. Correspondence Between Process Emissions Categories Used by MSCET, Trends, and J/W
Industrial Sectors and Identifier Codes. , 109
Table 45 (continued). Correspondence Between Process Emissions Categories Used by MSCET, Trends,
and J/W Industrial Sectors and Identifier Codes. 110
Table 46. Fuel Use Changes Between Control and No-control Scenarios 115
Table 47. Difference m Control and Ncncdntrol Scenario Off-Highway Mobile Source Emissions.
. 119
Table 48. Sources of Data for Transportation Sector Control Scenario Activity Projection.
, 127
Table 49. Distribution of Households by Demographic Attributes for Control Scenario.
128
Table 50. Economic and Vehicle Usage Data for Vehicle Ownership Projection -
Control Scenario. • 129
Table 51. Control Scenario Personal Characteristics.* , 130
Table 52. Distribution of Households by Income Class for No-control Scenario.
131
Table 53. Economic and Vehicle Usage Data for Vehicle Ownership Projection -
No-control Scenario • 132
Table 54. Percent Changes in Key Vehicle Characteristics Between the Control and No-control Scenarios.
133
Table 55. J/W Estimates of Percentage Increases in National Electricity Generation Under No-control
Scenario • • • 141
xxvi
-------
•' " ; • • Tables
Table 56. Trends Source Categories and (1975 to 1985) Scaling Factors for TSP and CO.
••••''• 147
Table57. Percentage Change in Real Energy Demand by Households from Control to No-control
Scenario ... 149
Table 58. Percentage Change in Commercial Energy Demand from Control to No-controTScenario.
.......'...• -....- ... ;... .r: iso
Table 59. JW Percent Differential in Economic Variables Used in CRESS 150
Table 60. TSP Emissions Under the Control and No-control ScSB^aribs by Target Year (in thousands of
short tons). . .- 151
Table 61. SOX Emissions Under the Control and No-control Scenarios by Target Year (in thousands of
short tons) ; .- .^.,'... 151
Table 62. NOX Emissions Under the Control and No-control Scenarios by Target Year (in thousands of
short tons) i -._„.. >.... 152
Table 63. VOC Emissions Under the Control and No-control Scenarios by.Target Year (in thousands of
short tons). :...... 152
Table 64. CO Emissions Under the Control and No-control Scenarios by TargefYear (in thousands of
short tons). . ., ... 153
Table 65. Lead (Pb) Emissions Under the Control and No-control ScenariosTjy Target Year (in thousands
of short tons) ...',". 153
Table 66. Summary of CO Monitoring Data 160
Table67. Format of Air Quality Profile Databases .k 163
Table 68. Summary of SO2 Monitoring Data. .,. 166
Table 69. Summary of NO2 Monitoring Data. ..•;,:; ,.....' 1... 168
Table 70. Summary of NO Monitoring Data , 168
Table 71. Summary of TSP Monitoring Data ;.....; .v. .. 178
Table 72. SJpumaryof PM^Ii|^toring Data. ,^ .. > .. 178
Table 73^|FinefirticIe (PM^E^^ 180
Table 74. Goarse^article (PM^toppgliChemical Composition by U.S. Region. 181
Table 75. PM Control Scenario Air ^aBi^Profile Filenames 182
Table 76. PM NoHCoritrol^Scenario Air Quality Profile Filenames. 182
Table 77. Urban Areas Modeled with OZIPM4. ,.' 185
Table TSilSummary of Ozone Monitoring Data 187
Table 791 Apportionment of Emissions Inventories for SAQM Runs, 188
Table 80. 1990 Control Scenario Visibility Conditions for 30 Southwestern U.S. Cities 195
Table 81. 1990 No-control Scenario Visibility Conditions for 30 Southwestern U.S. Cities 196
Table 82. Summary of Relative Change in Visual Range and DeciView Between 1990 Control and No-
control Scenario Visibility Conditions for 30 Southwestern U.S. Cities 199
Table 83. Summary of Dose Response Functions for Ozone 216
Table84. Summary of Dose Response Functions for Particulate Matter .. 227
Table 85. Summary of Dose Response* Functions for NO2 ^ 237
Table 86. Summary of Dose Response Functions for Carbon Monoxide 240
Table 87. Summary of Dose Response Functions for Sulfur Dioxide ...'.' 242
Table 88. Summary of Concentration-Response Functions for Visibility* 244
Table 89. Criteria Air Pollutant Monitors in the U.S., 1970 - 1990. 248
Table 90. Population Coverage in the "Within 50 km" Model Runs (percent of continental U.S.
population). 249
xxvn
-------
Tables
Table 91. Population Coverage for "Extrapolated to All U.S." Model Runs (percent of continental U.S.
population) • • 250
Table92. Human Health Effects of Criteria Pollutants 252
Table 93. Quantified Benefits Which Could Not Be Monetized - Extrapolated to the Entire 48 State
Population - -• • 254
Table 94. Criteria Pollutants Health Effects - Extrapolated to 48 State U.S. Population (Cases per year)
..255
Table 95. Summary of Biological Changes with Surface Water Acidification........ ;• 266
Table 96. Comparison of Population of Acidic National Surface Water Survey (NSWS) by Chemical
Category1 , , :l.. r-...." .. 268
Table 97. Results from Benefits Assessments of Aquatic Ecosystem Use Values from Acid Deposition
Avoidance • 271
Table 98. Agriculture Exposure-Response Functions 292
Table 99. Relative No-control to Control Percent Yield Change (harvested acres) for the Minimum
Scenario '• .. 294
Table 100. Relative No-control to Control Percent Yield Change (harvestedlacres) for the Maximum
Scenario ' 295
Table 101. Change in Farm Program Payments, Net Crop Income, Consumer Surplus, and Net Surplus
Due to the CAA (millions 1990 $) : 298
Table 102. Quantified and Unquantified Health Effects of Lead 301
Table 103. Elements of Piecewise Linear Function for Estimating Probability of IQ < 70 as a Function of
Blood Lead (PbB) Range -; 308
Table 104. Air Modeling Parameters. 331
Table 105. Estimated Indirect Intake Slopes: Increment of Blood Lead Concentration (in ^g/dL) per Unit
of Air Lead Concentration (wg/m3) 334
Table 106. Estimated Lead Emissions from Electric Utilities, Industrial Processes, and Industrial
Combustion (in Tons). .••••' 337
Table 107. Yearly Differences in Number of Health Effects Between the Controlled and Uncontrolled
Scenarios: Industrial Processes, Boilers, and Electric Utilities (Holding Other Lead Sources at
Constant 1970 Levels) . 338
Table 108. Yearly Differences in Number of Health Effects Between the Controlled and Uncontrolled
Scenarios: Industrial Processes, Boilers, and Electric Utilities (Holding Other Lead Sources at
Constant 1990 Levels) , 339
Table 109. Lead Burned in Gasoline (in tons) 343
Table 110. Yearly Differences in Number of Health Effects Between the Controlled and Uncontrolled
Scenarios: Lead in Gasoline only (Holding Other Lead Sources at Constant 1970 Levels).
344
Table 111. Yearly Differences in Number of Health Effects Between the Controlled and Uncontrolled
Scenarios: Lead in Gasoline only (Holding Other Lead Sources at Constant 1990 Levels).
: .- 345
Table 112. Health and Welfare Effects of Hazardous Air Pollutants 354
Table 113. Cancer Incidence Reductions and Monetized Benefits for NESHAPs 358
Table 114. Unit Values Used for Economically Valuing Endpoints 374
Table 115. Criteria Pollutants Health and Welfare Benefits - Extrapolated to Entire 48 State Population
Present Value (in 1990 using 5% discount rate) of Benefits from 1970 -1990 (in billions of
1990 dollars) 382
XXVlll
-------
Tables
Table 116. Present Value of 1970 to 1990 Monetized Benefits by Endpoint Category for 48 State
Population and Total Monetized Benefits (discount rate of 5 percent) 383
Table 117. Monte Carlo Simulation Model Results for Target Years, Plus Present Value in 1990 Terms of
Total Monetized Benefits for Entire 1970 to 1990 Period (in billions of 1990-value dollars).
• • •'!..•' ' . . • ' .'.«:„ • • 383
Table 118. Comparison of 1990 (Single Year) Monetized Benefits by Endpoint for 48 State Population
and Monitored Areas (in millions of 1990 dollars) .*. ; .• 385
Table 119. Effect of Alternative Discount Rates on Present Value of Total Monetized Benefits for 1970 :
to 1990 (in trillions of 1990 dollars). .- 386
Table 120. Alternative Estimates of the Present Value of Mortality Associated With" PM
(based on Pope etal., 1996, in trillions of 1990 dollars) .1.. 391
xxix
-------
XXX
-------
Figures
Figure i. 1990 Control and No-control Scenario Emissions (in millions of short tons). xii
Figure 2. Total Direct Costs of the CAA (in billions of inflation-adjusted 1990 dollars.) xiv
Figure 3. Total Direct Costs and Monetized Direct Benefits of the Clean Air Ac^l970 to 1990 (in
trillions of 1990 dollars). „ xxiii
Figure 4. Summary of Analytical Sequence and Modeled versus Historical Data Basis.
' * ~ ff
Figure 5. Control and No-control Scenario Total SOX Emission Estimates , 21
Figure 6. Control and No-control Scenario Total NOX Emission Estimates 21
Figure 7. Control and No-control Scenario Total VOC Emission Estimates 21
Figure 8. Control and No-control Scenario Total CO Emission Estimates. 22
Figure 9. Control and No-control Scenario Total TSP EmisskHriEstimates 22
Figure 10. Control and No-control Scenario Total Pb Emission Estimates 22
Figure 11. Frequency Distribution of Estimated Ratios for 1990 Control to No-control Scenario 95th
Percentile l*Hour Average CO Concentrations, by Monitor. 28
Figure 12. Frequency Distribution of Estimated Ratios for 1990 Control to No-control Scenario 95th
Percentile 1-Hour Average SO2 CMnicehtrations, by Monitor. 29
Figure 13. Frequency Distribution of Estimated Ratios for 1990 Control to No-control Scenario 95th
Percentile 1-Hour Average NOj Concentrations, by Monitor. 30
Figure 14. Distribution of Estimated Ratios for 1990 Control to No-Control Annual Mean TSP
Concentrations, by Monitored County . 30
Figure 15. Distribution of Estimated Ratios for 1990 Control to No-control OZEPM4 Simulated 1-Hour
Peak Ozone Concentrations, by Urban Area. •.....- , 31
Figure 16. Distribution of Estimated Ratios for 1990 Control to No-control RADM Simulated Daytime
Average Ozone Concentrations, by RADM Grid CelL .... 31
Figure 17. Distribution of Estimated Ratios for 1990 Control to No-control SAQM Simulated Daytime
Average Ozone Concentrations, by SAQM Monitor.. .. , ......... 31
Figure 18. RADM-Predicted Percent Increase in Total Sulfur Deposition (Wet + Dry) Under the No-
control Scenario.
;.................. 32
Figure 19. RADM-Predicted Percent Increase in Total Nitrogen Deposition (Wet + Dry) Under the No-
control Scenario. ,
_ .. .g,^^;x&.' • " ., '.,..'. ........ '32
Figure ^^'Rl^DM-Predicted Increase in Visibility Degradation, Expressed in DeciViews, for Poor
Visibility Conditions (90th Percentile) Under the No-control Scenario. ... 33
Figure 21. Monte Carlo Simulation Model Results for Target Years (in billions of 1990 dollars) . 64
Figure 22. Distribution of 1990 Monetized Benefits of CAA (in billions of 1990 dollars) 65
Figure 23. Uncertainty Ranges Deriving From Individual Uncertainty Factors .. . 66
Figure 24. Percent Difference in Real Investment Between Control and No-control Scenarios 95
Figure 25. Percent Difference in Price of Output by Sector Between Control and No-control Scenario for
1990 ......r. 96
Figure 26. Percent Difference in Quantity of Output by Sector Between Control and No-control Scenario
for 1990 , 96
XXXI
-------
Figure 27. Percent Difference in Employment by Sector Between Control and No-control Scenario for
1990 ,....97
Figure 28. Comparison of Control, No-control, and Trends SOX Emission Estimates 105
Figure 29. Comparison of Control, No-control, and Trends NOX Emission Estimates 105
Figure 30. Comparison of Control, No-control, and Trends VOC Emission Estimates. 106
Figure 31. Comparison of Control, No-control, and Trends CO Emission Estimates. ..-,.r 106
Figure 32. Frequency Distribution of Estimated Ratios for 1990 Control to No-control Scenario 95th
Percentile 1-Hour Average CO Concentrations, by Monitor. 164
Figure 33. Frequency Distribution of Estimated Ratios for 1990 Control to No-control Scenario 95th
Percentile 1-Hour Average SO2 Concentrations, by Monitor. 167
Figure 34. Frequency Distribution of Estimated Ratios for 1990 Control to No-cohtroT Scenario 95th
Percentile 1-Hour Average NO2 Concentrations, by Monitor. ., .. 169
Figure 35. Location of the High Resolution RADM 20-km Grid Nested Inside the 80-km RADM
Domain. • 171
Figure 36. RADM-Predicted 1990 Total Sulfur Deposition (Wet + Dry; in kg/ha) Under the Control
Scenario. * , 172
Figure 37. RADM-Predicted 1990 Total Nitrogen Deposition (Wet + Dry; in kg/ha) Under the Control
Scenario ' • • 173
Figure 38. RADM-Predicted 1990 Total Sulfur Deposition (Wet + Dry; in kg/ha) Under the No-control
Scenario '... •••'• .'.' 173
Figure 39. RADM-Predicted 1990 Total Nitrogen Deposition (Wet + Dry; in kg/ha) Under the No-control
Scenario 174
Figure 40. RADM-Predicted Percent Increase in Total Sulfur Deposition (Wet + Dry; in kg/ha) Under the
No-control Scenario i .-. - 174
Figure 41. RADM-Predicted Percent Increase in Total Nitrogen Deposition (Wet + Dry; in kg/ha) Under
the No-control Scenario. i.. • 176
Figure 42. Distribution of Estimated Ratios for 1990 Control to No-Control Annual Mean TSP
Concentrations, by Monitored County. ......; 183
Figure 43. RADM and SAQM Modeling Domains, with Rural Ozone Monitor Locations. ..... 186
Figure 44. Distribution of Estimated Ratios for 1990 Control to No-control RADM-Simulated Daytime
Average Rural Ozone Concentrations, by RADM Grid Cell 189
Figure 45. Distribution of Estimated Ratios for 1990 Control to No-control SAQM-Simulated Daytime
Average Ozone Concentrations, by SAQM Monitor. 190
Figure 46. Distribution of Estimated Ratios for 1990 Control to No-control OZIPM4 Simulated 1-Hour
Peak Ozone Concentrations, by Urban Area. • • 190
Figure 47. RADM-Predicted Visibility Degradation, Expressed in Annual Average DeciView, for Poor
Visibility Conditions (90th Percentile) Under the Control Scenario 193
Figure 48. RADM-Predicted Visibility Degradation, Expressed in Annual Average DeciView, for Poor
Visibility Conditions (90th Percentile) Under the No-control Scenario 197
Figure 49. RADM-Predicted Increase in Visibility Degradation, Expressed in Annual Average DeciView,
for Poor Visibility Conditions (90th Percentile) Under the No-control Scenario 198
Figure 50. PES Estimated Reductions in HAP-Related Cancer Cases. "... 360
Figure 51. ICF Estimated Reductions in Total HAP-Related Cancer Cases Using Upper Bound Asbestos
Incidence and Lower Bound Non-Asbestos HAP Incidence. .. 363
Figure 52. ICF Estimated Reduction in Total HAP-Related Cancer Cases Using Upper Bound Incidence
for AUHAPs 363
Figure 53. National Annual Average Motor Vehicle HAP Exposures («g/m3) 366
Figure 54. Monte Carlo Simulation Model Results for Target Years (in Millions of 1990 dollars) 384
Figure 55. Uncertainty Ranges Deriving From Individual Uncertainty Factors 387
xxxu
-------
Acronyms and Abbreviations
fieq/L
ACCACA
ACCACAPERS
AGSIM
AIRS
Al3*
ANC
ANL
APPI
AQCR
ARGUS
ASI
ATERIS
ATLAS
AUSM
BEA »-.'
b~~ "*"-•>,,
ext v . '
EG/ED
BI 1
BID
BP
BTlf
GAA
CAAA90
'CAPMS -
CAR!
CAS AC
CDC
CERL
CEUM
CHD
CIPP
microequivalents per liter
micrograms per cubic meter
, micrograms
micrometers, also referred to as microns
SAB Advisory Council on Clean Air Compliance Analysis
SAB ACCACA Physical Effects Subcommittee T ;
AGricultural Simulation Model _ 5
EPA Aerometric Information Retrieval System
aluminum
acid neutralizing capacity
Argonne National Laboratories ,
Argonne Power Plant Inventory"
Air Quality Control Region
Argonne Utility Simulation Model
Acid Stress Index
Air Toxic Exposure and Risklnformation System
Aggregate Timberland Assessment System
Advanced Utility Simulation Model
Bureau~of Economic Analysis"
total light extinction '
^Block Group"/Enumeration District
' / _
atherothrombotic brain infarction
Background Mormftion Document
blood pressure '
British Thermal Unit
confidence interval
cerebrovascular accident
Clean Air Act
Qean Air Act Amendments of 1990
EPA's Criteria Air Pollutant Modeling System
California Air Resources Board
SAB Clean Air Scientific Advisory Committee
Centers for Disease Control (now CDCP, Centers for Disease Control and
Prevention)
EPA/ORD Corvallis Environmental Research Laboratory (old name; see NERL)
ICF Coal and Electric Utility Model
coronary heart disease r
changes hi production processes
xxxin
-------
Acronyms and Abbreviations
CO
CO2
COH
COHb
COPD
CPUE
CR
CRESS
CSTM
CTG
CV
CVM
D.C.
DBF
DDE
DDT '
DFEVj
dL
DOC
DOE
DOI
DRI
dV
DVSAM
EC
EDB
EDC
EFI
El
EIA
EKMA
ELI
EOL
EPA
EPRI
ESEERCO
ESP
FERC
FGD
carbon monoxide
carbon dioxide
coefficient of haze
blood level of carboxyhemoglobin
chronic obstractive pulmonary disease
catch per unit effort
concentration-response
Commercial and Residential Simulation System model
Coal Supply and Transportation Model
Control Techniques Guidelines
contingent valuation
contingent valuation method
District of Columbia
diastolic blood pressure
dichlorodiphenyldichloroethylene
dichlorodiphenyltrichloroethane
decrement of forced expiratory volume (in one second)
deciliter
Department of Commerce
Department of Energy
Department of Interior : ^;
Data Resoutoes Incorporated
DeciViewlffaze Index
Ibisaggregate Vehicle jStpck Allocation Model
extinction coefficient
ethylene dibromide
ethylene dichloride
Electronic Fuel Injection
Electronic Ignition
Energy Information Administration
Empirical Kinetics Modeling Approach
Environmental Law Institute
end-of-line
Environmental Protection Agency
Electric Power Research Institute
Empire State Electric Energy Research Corporation
electrostatic precipitator ,
Federal Energy Regulatory Commission
forced expiratory volume (in one second)
flue gas desulfurization
xxxrv
-------
Acronyms and Abbreviations
FHWA Federal Highway Administration • ,
FIFRA Federal Insecticide, Fungicide, and Rodenticide Act
FIP -- Federal Information Processing System
FR Federal Register
FRP ' Forest Response Program •
GDP , gross domestic product t
GEMS Graphical Exposure Modeling System
GM geometric mean
GNP Gross National Product
GSD geometric standard deviation _ _; " r
H2SO4 sulfuricacid
ha . hectares
HAP Hazardous Air Pollutant
HAPEM-MS Hazardous Air Pollutant Exposure Model - Mobile Source
HNO3 nitric acid ,
hp horsepower
HTCM Hedonic Travel-Cost Model
ICARUS Investigation of Costs and ReljaBility in Utility Systems
ICD-9 International Classpcation of Diseases, Ninth Version (1975 Revision)
ICE Industrial Comtiustion Emissions model
lEc Industrial Economics, Incorporated
lEUBK^^, "> EPA's Integrated Exposure Uptake Biokinetic model
IMS ~" ~". Integrated Model Set
IFF "~"~ iterative proportional fitting
IQ intelligence quotient
ISCLT ' Industrial Source Complex Long Term air quality model
J/W "' Jorg(BnsonY\Vilcoxen
kg kilograms
km kilometers
Jbs- pounds '
XRT J°wer respiratory illness
m/s „ - " meters per second
m " meters
m3 cubic meters
Mm megameters
MMBTU millionBTU
MOBILESa EPA's mobile source emission factor model
mpg miles per gallon
Mpls Minneapolis
MRAD minor restricted activity day
XXXV
-------
Acronyms and Abbreviations
MSCET Month and State Current Emission Trends
MTD metric tons per day
MVATS EPA's Motor Vehicle-Related Air Toxics Study
MVMA Motor Vehicle Manufacturers Association , -
Mwe megawatt equivalent
N nitrogen
NA not available
NAAQS National Ambient Air Quality Standard^
NAFAP National Acid Precipitation Assessment Program
NARSTO North American Research Strategy for Tropospheric Ozone
NATTCH National Air Toxics Information aearirigSouse
NCLAN National Crop Loss Assessment Network
NEA National Energy Accounts
NBRA National Economic Research Associates
NERC North American Electric Reliability Council
NERL EPA/ORD National Exposure Research Laboratory (new name for CERL)
NESHAP National Emission Standard for Hazardous Air Pollutants
NHANES First National Health and Nutrition Examination Survey
NHANESII Second National Health and Nutrition Examination Survey
NIPA National Income and Product Accounts
NMOCs nonmethane organic compounds
NO nitric oxide }
NO2 nitrogen dioxide
NO3* nitrate ion .
NOX nitrogen oxides
NFTS Nationwide Personal Transportation Survey'
NSPS New Source Performance Standards .
NSWS • National Surface Water Survey
O&M operating and maintenance
O3 ozone
OAQPS EPA/OAR Office of Air Quality Planning and Standards
OAR EPA Office of Air and Radiation
OMS EPA/OAR Office of Mobile Sources
OPAR EPA/OAR Office of Policy Analysis and Review
OPPE EPA Office of Policy Planning and Evaluation _
ORD EPA Office of Research and Development
OZEPM4 Ozone Isopleth Plotting with Optional Mechanism-IV
PACE Pollution Abatement Costs and Expenditures survey
PAN peroxyacetyl nitrate ,
PAPE Pollution Abatement Plant and Equipment survey
xxxvi
-------
Acronyms and Abbreviations
Pb
PbB
PCB
PES
pH .
PIC
PM10
POP
ppb
PPH
pphm
ppm
PPRG
PRYL
PURHAPS
PVC
r2
RAD
RADM ~
RADM/iKJ
RfD
RIA
ROM
RRAD
RUM
s.e»
SAI "_^
SAQI2
SARA
SARMAP
SCC
SEDS
SIC
SIP
SJVAQS
lead
blood lead level
polychlorinated biphenyl
Pacific Environmental Services •">„••"'-,
the logarithm of the reciprocal of hydrogen ion concentration, a measure of acidity
product of incomplete combustion. _^
particulates less than or equal to 10 microns in aerometric diameter
particulates less than or equal to 2.5 microns in aerometric diameter ' ff
population >|i
exposed population of exercising mild asthmatics
exposed population of exercising moderate asthmatics
parts per billion
people per household
parts per hundred million
parts per million :
Pooling Project Research Group
percentage relative yield loss
PURchasedHeatAniPower
polyvihyl chloride
statistical correlation coefficient, squared
restrictedjactivity day
Regional Acid Deposition Model
_ RADM Engineermg Model
TResource Allocation and Mine Costing model
reference dose
Regulatory Impact Analysis
Regional Oxidant Model
respiratory restricted activity day
Randlin Utility Model
• . .Si' . • - • " .
standard error
-fir . • •. .
.-Science Advisory Board
Systems Apjplications International
SARMAP Air Quality Model
Superfund Amendment Reauthorizatidn Act
SJVAQS/AUSPEX Regional Modeling Adaptation Project
Source Classification Code
State Energy Data System
Standard Industrial Classification
State Implementation Plan
San Joaquin Valley Air Quality Study
XXXVll
-------
Acronyms and Abbreviations
SMS A Standard Metropolitan Statistical Area
SO2 sulfur dioxide
SO42" sulfateion
SOS/T State of Science and Technology (refers to a series of NAPAP reports)
SRaw Specific Airway Resistance
STAR Stability Array weather database
TAMM90 Timber Assessment Market Model (revised version)
TEEMS Transportation Energy and Emissions Modeling System
TIUS Truck Inventory and Use Surveys "
TRI Toxic Release Inventory
TSP total suspended particulate ",.-,,
U.S. , United States
UAM Urban Airshed Model
URI upper respiratory illness
USDA United States Department of Agriculture " .
USEPA United States Environmental Protection Agency
VC vinyl chloride ; j $
VMT vehicle miles traveled
VOC volatile organic compounds
VOP Vehicle Ownership Projection
VR visual range
W126 index of peak weighted average of cumulative ozone concentrations
WLD Work Loss Day
WTP willingness to pay
xxxvm
-------
1
Introduction
As part of the Clean Air Act Amendments of 1990, Congress established a requirement under Section
812 that EPA develop periodic Reports to Congress estimating the benefits and costs of the Clean Air Act
itself. The first such report was to be a retrospective analysis, with a series of prospective analyses to
follow every two years thereafter. This report represents theiretrospective study, covering the period
beginning with passage of the Clean Air ActlAmendments of 1970, until 1990 when Congress enacted the
most recent comprehensive amendments to the Act.
Since the legislative history associated with section 812 is sparse, there is considerable uncertainty
regarding Congressional intent behind the requirement for periodic cost-benefit evaluations of the Clean
Air Act (CAA). However, EPA believes the principal goal of these amendments was that EPA should
develop, and periodically exercise, the ability to-providei Congress and the public with,up-to-date,
comprehensive information about the economic costs, economic benefits, and health, welfare, and
ecological effects of CAA programs. The results of such analyses might then provide useful information
for refinement of CAA programs during future reauthorizations of the Act.
The retrospective analysis presented in this Report to Congress has been designed to provide an
unprecedented examination "of the overall costs and benefits of the historical Clean Air Act. Many other
analyses have attempted to identifjrthe isolated effects of individual standards or programs, but no analysis
with the present degree of validity, breadth and integration has ever been successfully developed. Despite
data limitations, considerable scientific uncertainties, and severe resource constraints; the EPA Project
Team was able to developfia broad assessment of the costs and benefits associated with the major CAA
programs of the 1970 tcj§990 period. Beyond the statutory goals of Section 812, EPA intends to use the
results of this stady/lplielp support decisions on future investments in air pollution research. Finally, many
of the methodologies and modeling systems developed for the Retrospective Study may be applied in the
future to the ongoing series of section 812 prospective studies.
Clean Air Act Requirements, 197Oto 19BO
The Clean Air Act establishes a framework for the attainment and maintenance of clean and healthful
air quality levels. The Clean Air Act was enacted in 1970 and amended twice - in 1977 and most recently
in 1990. ,
-------
Chapter 1: Introduction
The 1970 Clean Air Act contained a number of key provisions. First, EPA was directed to establish
national ambient air quality standards for the major criteria air pollutants. The states were required to
develop implementation plans describing how they would control emission limits from individual sources
to meet and maintain the national standards. Second, the 1970 CAA contained deadlines and strengthened
enforcement of emission limitations and state plans with measures involving both the states and the federal
government. Third, the 1970 Act forced new sources to. meet standards based on the best available
technology. Finally, the Clean Air Act of 1970 addressed hazardous pollutants and automobile exhausts.
- t -~;~, >,
The 1977 Clean Air Act Amendments also set new requirements on clean areas already in attainmei
with the national ambient air quality standards. In addition, 1^1977 Amendments set out provisions^
help areas that failed to comply with deadlines for achievement of the national ambient air quaHtJp
standards. For example, permits for new major sources and modifications were required.
The 1990 Clean Air Act Amendments considerably strengthened the earlier versions of the Act. With
respect to nonattainment, the Act set forth a detailed and graduated program, Teflecting the fact that
problems in some areas are more difficult and complex than others. The 1990 Act also established a list of
189 regulated hazardous air pollutants and a multi-step program for controlling emissions of these toxic air
pollutants. Significant control programs were also established for emissions of acid rain precursors and
stratospheric ozone-depleting chemicals. The biggest regulatory procedural change in the Act is the new
permit program where all major sources are now required to obtain an operating permit. Finally, the
amendments considerably expanded the enforcement provisions of the Clean Air Act, adding
administrative penalties and increasing potential civil penalties.
Section Q12 of the Glean Air Act Amendments
of '
Section 812 of the Clean Air Act /^ndments of 1990 requires the EPA to perform a "retrospective"
analysis which assesses the costs and fenefite to the public health, economy and the environment of clean
air legislation enacted prior to me 1990'amendments. Section 812 directs that EPA shall measure the
effects on "employment, productivity, cost of living, economic growth, and the overall economy of the
United States" of the Clean Air .Act Section 812 also requires that EPA consider all of the economic,
public health, and environmental benefits of efforts to comply with air pollution standards. Finally, section
812 requires EPA to evaluate the prospective costs and benefits of the Clean Air Act every two years.
Analytical Design and Review
' • i ;.,•"-•• . • , ' ' , '
Target Variable
The retrospective analysis was designed to answer the following question:
"How do the overall health, welfare, ecological, and economic benefits of Clean Air Act
programs compare to the costs of these programs?"
By examining the overall effects of the Clean Air Act, this analysis complements the Regulatory
Impact Analyses (RIAs) developed by EPA over the years to evaluate individual regulations. Resources
-------
Chapter 1: Introduction
were used more efficiently by recognizing that these RlAs, and other EPA analyses, provide complete
information about the costs and benefits of specific rules.
Focusing on the broader target variables of "overall costs" and "overall benefits" of ffirciean Air Act,
the EPA Project Team adopted an approach based on construction and comparison of two distinct
scenarios: a "no-control scenario" and a "control scenario." The no-control scenario essentially freezes
federal, state, and local air pollution controls at the levels of strinjpfey and effectiveness which prevailed,
in 1970. The control scenario assumes that all federal, state, affSfocal rules promulgated pursuant to, or in
support of, the CAAduring 1970 to 1990 were enacted. Thiffhalysis thenestimates the differences
between the economic and environmental outcomes associated with these two scenarios.
Key Assumptions
Two key assumptions were made during the scenario design process to avoid miring the analytical
process in endless speculation. First, the "no-control" scenario was defined to reflect the assumption that
no additional air pollution controls were imposed by any level of government or voluntarily initiated by
private entities after 1970. Second, it is assumed (hat the geographic distribution of population and
economic activity remains the same betweeirthe two scenarios.
The first assumption is an obvious simplification. In the absence of the CAA, one would expect to see
some air pollution abatement activity, either voluntary or due to state or local regulation. It is conceivable
that state and local regulation would have required air pollution abatement equal to -or even greater than-
. that required by the CAA; particularly since some states, most notably California, have done so. If one
were to assume that state and local regulations would have been equivalent to CAA standards, then a cost-
benefit analysis of the CAArwould be a meaningless exercise since both costs and benefits would equal
zero. Any attempt to predict how states' and localities' regulations would have differed from the CAA
would be too speculative to support the credibility of the ensuing analysis. Instead, the no-control scenario
has been structured to reflect the assumption that states and localities would not have invested further in air
pollution control programs after 1970 in the absence of the federal CAA. That is, this analysis accounts
for all costs and benefits of air pollution control from 1970 to 1990. Speculation about the fraction of costs
and benefits attributable exclusively to the federal CAA is left to others.
,. The second assumption concerns changing demographic patterns in response to air pollution. In the
hypothetical no-control world, air quality is worse than that in the historical "control" world particularly
in urban industrial areas. It is possible that in the no-control case more people, relative to the control
case, "would move away from the most heavily polluted areas. Rather than speculate on the scale of
population S^ment, the analysis assumes no differences in demographic patterns between the two
scenarios. Similarly, the analysis assumes no changes in the spatial pattern of economic activity. For
example: if, in the no-control case, an industry is expected to produce greater output than it did in the
control case, that increased output is produced by actual historical plants, avoiding the need to speculate
about the location or other characteristics of new plants providing additional productive capacity.
Analytical Sequence
-------
Chapter 1: Introduction
The analysis was designed and implemented in a sequential manner following seven basic steps which
are summarized below and described in detail later in this report. The seven major steps were:
direct cost estimation
macroeconomic modeling :
emissions modeling ,
air quality modeling ; ^ x
health and environmental effects estimation '" ~\ ~ ' ^f"
economic valuation :."
results aggregation and uncertainty characterization
? ^
By necessity, each of these components had to be completed in a sequential manner. The emissions
modeling effort had to be completed entirely before the air quality models could be configured and run; the
air quality modeling results had to be completed before the health and environmental consequences of air
quality changes could be derived; and so on. The analytical sequence, and the modeled versus actual data
basis for each analytical component, are summarized in Figure 4 and described in the remainder of this
section.
The first step of the analysis was to estimate the total direct costs incurred by public and private
entities to comply with post-1970 CAA requirements. These data were obtained directly from Census
Bureau and Bureau of Economic Analysis'(BEA) data on compliance expenditures reported by sources,
and from EPA analyses. These direct cost data were then adopted as inputs to the macroeconomic model
used to project economic conditions-such as production levels, prices, employment patterns, and other
economic indicators-under the two scenarios. To ensure a consistent basis for scenario comparison, the
analysis applied the same macroeconomic modeling system to estimate control and no-control scenario
economic conditions.4 First, a control scenario was constructed by running the macroeconomic model
using actual historical data for input factors such as economic growth rates during the 1970 to 1990 period.
The model was ffien re-run for the no-control scenario by, in essence, returning all post-1970 CAA
compliance expenditures to the economy. With these additional resources available for capital formation,
personal consumption, and other purposes, overall economic conditions under the no-control scenario
differed from those of the control scenario. In addition to providing estimates of the difference in overall
economic growth and other outcomes under the two scenarios, these first two analytical steps were used to
define specific economic conditions used as inputs to the emissions modeling effort, the first step in the
estimation of CAA benefits*5
4 Using modeled economic conditions for both scenarios has both advantages and disadvantages. The principal disadvantage is that historical
economic conditions "predicted" by a macroeconomic model will not precisely duplicate actual historical events and conditions. However, this
disadvantage is outweighed by the avoidance of distortions and biases which would result from comparing a modeled no-control scenario with
actual historical conditions. By using the same macroeconomic model for both scenarios, model errors and biases essentially cancel out, yielding
more robust estimates of scenario differences, which are what this analysis seeks to evaluate.
* For example, the macroeconomic model projected different electricity sales levels under the two scenarios, and these sales levels were used
is key input assumptions by the utility sector emissions model.
-------
Chapter 1: Introduction
Figure 4. Summary of Analytical Sequence and Modeled versus Historical Data Basis.
Control Scenario No-control Scenario
Compile historical compliance expenditure
data
Develop modeled macroeconomic scenario
based on actual historical economic data
Develop modeled macroeconomic scenario
by rerunning control scenario with •
compliance expenditures added back to the
economy
Project emissions by year, pollutant, and
sector using control scenario
macroeconomic projection as input to
sector-specific emissions models
Re-run sector-specific emissions models
using no-control scenario macroeconomic
projection
Develop statistical profiles of historical air
quality for each pollutant based on
historical monitoring data (plus
extrapolations to cover unmonitored areas)
Derive no-control air quality profiles by
adjusting control scenario profiles based on
differences in air quality modeling of
control scenario and no-control scenario
emissions inventories
Estimate physical effects based on
application of concentration-response
functions to historical air quality profiles
Y
Estimate physical effects based on
application of concentration-response
functions to no-control scenario air quality
profiles
Calculate differences in physical outcomes
between control and no-control scenario
j
Estimate economic value of differences in
physical outcomes between the two
scenarios*
Compare historical, direct compliance costs
with estimated economic value of
monetized benefits, considering additional
benefits which could not be quantified
and/or monetized
Cliccia
i viciu
-------
Chapter 1: Introduction
Using appropriate economic indicators from the macroeconomic model results as inputs, a variety of
emissions models were run to estimate emissions levels under the two scenarios. These emissions models
provided estimates of emissions of six major pollutants6 from each of six key emitting sectors: utilities,
industrial processes, industrial combustion, on-highway vehicles, off-highway vehicles, and ,
commercial/residential sources. The resulting emissions profiles reflect state-wide total emissions from
each pollutant-sector combination for the years 1975,1980,1985, and 1990.7
* ~ -^
The next step toward estimation of benefits involved translating these emissions inventories into ~~
estimates of air quality conditions under each scenario. Given the complexity, data requirements, and
operating costs of state-of-the-art air quality models-and the afore-mentioned resource constraints-the
EPA Project Team adopted simplified, linear scaling approacfies fora number of pollutants. However, for
ozone and other pollutants or air quality conditions which mvolvejsubstantial non-linear formation effects
and/or long-range atmospheric transport and transformation, the EPA Project Team invested the time and
resources needed to use more sophisticated modeling systems. For example, urban area-specific ozone
modeling was conducted for 147 urban areas throughout the 48 contiguous states.
Up to this point of the analysis, both the control and nb-confrol scenarid were based on modeled
conditions and outcomes. However, at the airquality modeling step, the analysis returns to a foundation
based on actual historical conditions and data. Specifically, actual historical air quality monitoring data
from 1970 to 1990 are used to define the bbntrol scenario. Air quality conditions under the no-control
scenario are then derived by scaling the historical data adopted for the control scenario by the ratio of the
modeled control and no-control scenario air quality. This approach takes advantage of the richness of the
historical data on air quality, provides a realistic grounding for the benefit measures, and yet retains the
analytical consistency conferred by using the same modeling approach for both scenarios. The outputs of
this step of the analysis are statistical profiles for each pollutant characterizing air quality conditions at
each monitoring sitein the lower 48! states.8
The control and no-control scenario airquality profiles were then used as inputs to a modeling system
which translates afr qualitytop^sicai outcomes -such as mortality, emergency room visits, or crop yield
losses- through the use of concentration-response functions. These concentration-response functions are
in turn derived from studies'fpund in the scientific literature on the health and ecological effects of air
pollutants. At this point, estimates were derived of the differences between the two scenarios in terms of
incidence rates for a broad range of human health and other effects of air pollution by year, by pollutant,
and by monitor.9 .
* These six pollutants are total suspended particulates (TSP), sulfur dioxide (SOj), nitrogen oxides (NOJ, caiixm monoxide (CO), volatile
oiganic compounds (VOCs), and lead (Pb). The other CAA criteria pollutant, ozone (O3), is formed in the atmosphere through the interaction of
sunlight and ozone precursor pollutants such as NO, and VOCs.
7 By definition, 1970 emissions under the two scenarios are identical.
1 The one exception is particulate matter (PM). For PM, air quality profiles for both Total Suspended Particulates (TSP) and particulates less
than or equal to 10 microns in diameter (PM:0) were constructed at the county level rather than the individual monitor level.
* Or, for PM, by county.
-------
Chapter 1: Introduction
In the next step, economic valuation models or coefficients were used to estimate the economic value
of the reduction in incidence of those adverse effects which were amenable to such monetization. For
example, a distribution of unit values derived from the economic literature was used to estimate the value
of reductions in mortality risk associated with exposure to participate matter. In addition, benefits which
could not be expressed in economic terms were compiled and are presented herein. In some cases,
quantitative estimates of scenario differences in the incidence of a nonmonetized effect were calculated.10
In many other cases, available data and techniques were insufficient to support anything more than a
qualitative characterization of the change in effects.
Finally, the costs and monetized benefits were combined to provide a range of estimates for the partial,
net economic benefit of the CAA with the range reflecting quantified "uncertainties associated with the
physical effects and economic valuation steps." The term "partial" is emphasized because only a subset of
the total potential benefits of the CAA could be represented in economic-terms due to limitations in
analytical resources, available data and models, and the state of the science.12 Of paramount concern to the
EPA Project Team was the paucity of concentration-response functions needed to translate air quality
changes into measures,of ecological effect In addition, significant scientific evidence exists linking air
pollution to a number of adverse human health effects which could not be effectively quantified and/or
monetized.15
The CAA requires EPA to consult with an outside panel of experts -referred to statutorily as the
Advisory Council on Clean Air Act Compliance Analysis (ACCACA)- in developing the section 812
analyses. In addition, EPA is required to consult with the Department of Labor and the Department of
Commerce. _* " » . • '
The ACCACA^as organized in 1991 under the auspices and procedures of EPA's Science Advisory
Board (SAB). Organizing the review committee under the SAB ensured that review of the section 812
studies would be conducted by highly qualified experts in an objective, rigorous, and publicly open
. S For example, changes in forced expiratory volume in one second (FEV,) as a result of exposure to ozone were quantified but could not be
expressed in terms of economic value. , -
^ " Although considerable uncertainties surround the direct cost, macroeconomic modeling, emissions modeling,, and air quality modeling
steps, the ranges of aggregate costs and benefits presented in this analysis do not reflect these uncertainties. While the uncertainties in these
components were assessed qualitatively, and in some cases quantitatively, resource limitations precluded the multiple macroeconomic model,
emissfcms model, and air quality model runs which would have been required to propagate these uncertainties through the entire analytical
sequence. As a result, complete quantitative measures of the aggregate uncertainty in the cost and benefit estimates could not be derived.
However, the ranges presented do reflect quantitative measures of the uncertainties in the two most uncertain analytical steps: physical effects
estimation and economic valuation.
12 It should be noted that there is some uncertainty associated with the estimates of economic costs as well and that some omitted components
of adverse economic consequences of pollution control programs may be significant For example, some economists argue that the economic costs
of the CAA reported herein may be significantly underestimated to the extent potential adverse effects of regulation on technological innovation are
not captured. Nevertheless, it is dear that the geographic, population, and categorical coverage of monetary cost effects is significantly greater than
coverage of monetized benefits in this analysis.. , ' '
i -
13 For example, white there is strong evidence of a link between exposure to carbon monoxide and reduced time of onset of angina attack,
there are no valuation functions'available to estimate the economic loss associated with this effect
-------
Chapter 1: Introduction
manner. The SAB ACCACA has met many times during the development of the retrospective study to
review methodologies and interim results. While the full ACCACA retains overall review responsibility
for the section 812 studies, some specific issues concerning physical effects and air quality modeling have
been referred to subcommittees comprised of both ACCACA members and members of other SAB
committees. The ACCACA Physical Effects Subcommittee has met several times and has provided its
own review findings to the full ACCACA. Similarly, the ACCACA Air Quality Subcommittee, comprised
of members and consultants of the SAB Clean Air Scientific Advisory Committee (CASAC), has held ,
several teleconference meetings to review methodology proposals and modeling results. ;i
With respect to the interagency review process, the EPA has expanded the list of consulted agencies to
include the Department of Energy, the Council on Environmental Quality, the National Acid Precipitation
Assessment Program, The Office of Management and Budgei and the Council of Economic Advisors.
While several meetings have been held with the Interagency Review~Cfroup during the course of the
retrospective study, the EPA Project Team now provides opportiuutjrfor integrated interagency, expert
panel, and public consultation through the public review meetings conducted by the SAB ACCACA.
Report Organization
The remainder of the main text of this report summarizes the key methodologies and findings of
retrospective study. The direct cost estimation and macroeconomic modeling steps are presented in
Chapter 2. The emissions modeling iiSummarized in Chapter 3. Chapter 4 presents the air quality
modeling methodology and sample results. Chapter 5 describes the approaches used and principal results
obtained through the physical effects estimation process. Economic valuation methodologies are described
in Chapter 6. Chapter 7 presents the aggregated results of the cost and benefit estimates and describes and
evaluates important uncertainties in the results.
Additional details regarding the methodologies and results are presented in the appendices and in the
referenced supporting documents. Appendix A covers the direct cost and macroeconomic modeling.
Appendix B provides additional detail on the sector-specific emissions modeling effort. Details of the air
quality models used and results obtained are presented or referenced in' Appendix C. The effects of the
CAA on human health and visibility; aquatic, wetland, and forest ecosystems; and agriculture are presented
in Appendices D, E, and F, respectively. Appendix G presents details of the lead (Pb) benefits analysis.
Air toxics reduction benefits are discussed in Appendix H. The methods and assumptions used to value
quantified effects of the CAA in economic terms are described in Appendix I. Appendix J describes some
areas of research which may increase comprehensiveness and reduce uncertainties in effect estimates for
future assessments, and describes plans for future Section 812 analyses.
8
-------
2
Cost and Macroeconomic-
The costs of complying with Clean Air Act (CAA) requirements through the 1970 to 1990 period
affected patterns of industrial production, capital investment, productivity, consumption, employment, and
overall economic growth. The purpose of the analyses summarized in "this chapter was to estimate those
direct costs and the magnitude and significance of resulting changes to theCoverall economy. This was
accomplished by comparing economic indicators under two alternative scenarios: a control scenario
serving as the historical benchmark, including the historical CAA as implemented; and a no-control
scenario which assumes historical CAA programs did not exist. The^stimafed economic consequences of
the Mstbrical CAA were taken as me difference between these two scenarios.
Data used as inputs to the cost analysis can be classified into two somewhat overlapping categories
based on the information source: survey-based information (generally gathered by the Census Bureau)
and information derived from various EPA analyses. For the most part, cost estimates for stationary air
pollution sources (e.g., factory smokestacks) are based on surveys of private businesses that attempt to
elicit information on annual pollution control outlays by those businesses. Estimates of pollution control
costs for mobile sources (erg., automobiles) are largely based on EPA analyses, rather than on direct
observation and measurement of compliance expenditures. For example, to determine one component of
the cost of reducing lead emissions from mobile sources, the Project Team used an oil refinery
production cost model to calculate the incremental cost required to produce unleaded (or less-leaded, as
appropriate) rather than leaded gasoline, while maintaining the octane level produced by leaded
gasoline,.
H - * -
As is the case with many policy analyses, a significant uncertainty arises in the cost analysis as a
consequence of constructing a hypothetical scenario. With this retrospective analysis covering almost
twenty years, difficulties arise in projecting alternative technological development paths. In some cases,
fhe analytical assumptions used to project the alternative scenario are not immediately apparent. For
example, the surveys covering stationary source compliance expenditures require respondents to report
pollution abatement expenditures — implicitly asking them to determine by how much the company's costs
would decline if there were no CAA compliance requirements. While a response might be relatively
straightforward in the few years following passage of the CAA, a meaningful response becomes more
difficult after many years of technical change and investment in less-polluting plant and equipment make it
difficult to determine the degree to which total costs would differ under a "no CAA" scenario. In cases
such as this, assumptions concerning the alternative hypothetical scenario are made by thousands of
individual survey respondents. Where cost data are derived from EPA-analyses, the hypothetical scenario
assumptions are, at least in theory, more apparent. For example, when determining the incremental cost
caused by pollution-control requirements, one needs to make assumptions (at least implicitly) about what
an auto would look like absent pollution control requirements. In either case, the need to project
-------
Chapter 2: Cost and Macroeconomic Effects
hypothetical technology change for two decades introduces uncertainty into the assessment results, and this
uncertainty may be difficult to quantify..
The remainder of this chapter summarizes the basic methods and results of,the direct compliance cost
and macroeconomic analyses. Further detail regarding the modeling methods and assumptions employed,
as well as additional analytical results, are presented in Appendix A. - ^
Direct Compliance Costs
Table 4. Estimated Annual CAA Compliance
Costs (SbiffionsV. ' 'l~!<
Compliance with the CAA imposed direct costs on
businesses, consumers, and governmental units; and
triggered other expenditures such as governmental
regulation and monitoring costs and expenditures for
research and development by both government and
industry. Although expenditures unadjusted for
inflation—that is, expenditures denominated in
"current dollars"—increased steadily from $7 billion to
$19 billion per year over the 1973 to 1990 period,14
annual CAA compliance expenditures adjusted for
inflation were relatively stable, averaging near $25
billion (in 1990 dollars) during the 1376s and close to
$20 billion during most of the 1980s "(see Table 4)".
Aggregate compliance expenditures were somewhat
less than & of one percent, of total Domestic output
during that period, with the percentage falling from 2/3
of one percent of total output in 1975 to •V&'pf one
percent in 1990.
Although useful for many purposes, a summary of
direct annual expenditures may not the best cost
measure to use when comparing costs to benefits.
Capital expenditures are investments, generating a
stream of benefits and opportunity cost15 over the life
of the investment The appropriate accounting
technique to use for capital expenditures in a cost/benefit analysis is to annualize the expenditure. This
technique, analogous to calculating the monthly payment associated with a home mortgage, involves
spreading the cost of the capital equipment over the useful lif e of the equipment using a discount rate to
account for the time value of money.
: Expenditures
Annualized Costs
i S
$1990 at:
IJ
Year Scurrent $1990 3% B'%
1973
1974
1975
1976
1977
, 1978
1379
1980
1981
1982
1983
1984
1985
" 1986
1 1987
1988
, '1989
1990
7.2
8.5
10.6
11.2
11.9
12.0
14.4
, 163
,17.0 ,
ie.o *
15.5
17.3
19.1 /
17.8
18.2^
18.2
19.0
19.0
19.6
2i.4
24.4
24.1
24.1 v
2l6
24>8
25.7
24.4
;2Y'6
20v.l ,
21.6
22.9
i 3 ! f
20.8
20.6
19.8
19.8
19.0
11.0
13.2
133
14,1
15.3
15.0
173
19.7
19.6
18.6
19.1
20.1
22.5
21.1
22.1'
22.0'
22.9
23.6
if
11.0
13:4
13.6
14^4
15.9
15.8
183
20.8
20.9'
20ll
20.7
21.9
24.4
'23.2
24,2
243
253
26.1
.-' '"
TSL ,
" 11.1 -
13;7 '
14.0!"
15.1 *
',16.6
1'6.7
193 '
22.0 }
2?'3'^-'
2l,7
22.5
23.S
26.5
,25.4 •;
26.6
26^7"
27.8 \f
28.7^ "*
'**>' ?<$
" Due to dat« limitations, the cost analysis for this CAA retrospective starts in 1973, missing costs incurred in 1970-72.. This limitation is
not IDcely to be significant, however, because relatively little in the way of compliance with the "new" provisions of the 1970 CAA was required in
the first two years following passage.
a In this context, "opportunity cost" is defined as the value of alternative investments or other uses of funds foregone as a result of the
invesunent. '
10
-------
Chapter 2: Cost and Macroeconomic Effects
For this cost/benefit analysis, "annualized" costs reported for any given year are equal to O&M
expenditures — including R&D and other similarly recurring expenditures'— plus amortized capital costs
(i.e., depreciation plus interest costs associated with the existing capital stock) for that year.: Stationary
source air pollution control capital costs were amortized over 20 years; mobile source air pollution control
costs were amortized over 10 years.16 All capital expenditures were annualized using afive percent,
inflation-adjusted rate of interest. Additionally, annualized costs were calculated-using discount rates of
three and seven percent to determine the sensitivity of the cost results to changes in the discount rate.
Table 4 summarizes costs annualized at three, five, and sevenpercent, as well as annual expenditures. «*?
Total expenditures over the 1973-1990 period, discounted to 1990 using a five percent (net of
inflation) discount rate, amount to 628 billion dollars (in 199Q~doUars).~ Discounting the annualized cost
, stream to 1990 (with both annualization and discounting procectares using a five percent rate) gives total
costs of 523 billion dollars (in 1990 dollars). Aggregate annualized costs are less than expenditures
because the annualization procedure spreads some of the capital cost beyond 1990."
/
indirect Effects of the CAA
Through changing production costs, CAA implementation induced changes in consumer good prices,
and thus in the size and composition of economic output. The Project Team used a general equilibrium
macroeconomic model to assess the extent of such second-order effects. This type of model is useful
because it can capture the feedback effects of an action. In_the~Section 812 macroeconomic modeling
exercise, the feedback effects arising from expenditure changes were captured, but the analogous effects
arising from improvements in human health wefFnofcapfured by the model. For example, the
macroeconomic model results do not reflect theindirect economic effects of worker productivity
improvements and medical expenditure savings caused by the CAA. Consequently, the macroeconomic
modeling exercise provides limited and incomplete information on the type and potential scale of indirect
economic effects. * . •
The effects estimated by the macroeconomic model can be grouped into two broad classes: sectoral
impacts (i.e*, changes in the composition of economic output), and aggregate effects (i.e., changes in the
degree of output or of some measure of human welfare). The predicted sectoral effects were used as inputs
to file emissions models as discussed in Chapter 2. In general, the estimated second-order macroeconomic
effects were small relative'to the size of the U.S. economy. See Appendix A for more detail on data
sources, analytical methods, and results for the macroeconomic modeling performed for this assessment.
16 Although complete data are available only for the period 1973-1990, Cost of Clean includes capital expenditures for 1972. Those capital
expenditure data have been used here. Therefore, amortized costs arising from 1972 capital investments are included in the 1973-1990 annualized
costs, even though 1972 costs are not otherwise included in the analysis.' Conversely, some capital expenditures incurred in the 1973-1990 period
are not reflected in the 1973-1990 annualized costs - those costs are spread through the following two decades, thus falling outside of the scope of
this study (e.g., only one year of depreciation and interest expense is included for 1989 capital expenditures). Similarly, benefits arising from
emission reductions realized after 1990 as a result of capital investments made during the 1970 to 1990 period of this analysis are not included in
the estimates of benefits included in this report
17 This adjustment is required because many 1970 to 1990 investments in control equipment continue to yield benefits beyond 1990,
Annualization of costs beyond 1990 ensures that the costs and benefits of any particular investment are properly scaled and matched over the
lifetime of the investment '•
' ' . 11 ' ' . '' ' -
-------
Chapter 2: Cost and Macroeconomic Effects
Sectoral Impacts
The CAA had variable compliance impacts across economic sectors. The greatest effects were on the
largest energy producers and consumers, particularly those sectors which relied most heavily on
consumption of fossil fuels (or energy generated from fossil fuels). In addition, production costs increased
more for capital-intensive industries than for less capital-intensive industries under the control scenario due
to a projected increase in interest rates. The interest rate increase, which resulted in an increase hi the dolt
of capital, occurred under the control scenario because CAA-mandated investment hi pollution abatement
reduced the level of resources available for other uses, including capital formation.
Generally, the estimated difference in cost impacts under the control and no-control scenarios for a
particular economic sector was a function of the relative energy-intensity and capital-intensity of that
sector. Increased production costs in energy-.and capital-intensive sectors under the control scenario were
reflected in higher consumer prices, which resulted in reductions in the quantity of consumer purchases of
goods and services produced by those sectors. This reduction in consumer demand under the control
scenario led, ultimately, to reductions in output and employment in those sectors. The sectors most
affected by the CAA were motor vehicles, petroleum refining, and electricity generation. The electricity
generation sector, for example, -incurred a two to four percent increase hi consumer prices by 1990,
resulting in a three to five and a half percent reduction in output. Many other manufacturing sectors saw
an output effect in the one percent range.
Some other sectors, however, were projected to increase output under the control scenario. Apart from
the pollution control equipment mdustry, which was not separately identified and captured in the
macrc^conbmicmcidelmgperlbnned for this study, two example sectors for which output was higher and
prices were lower under me conttOl scenario are food and furniture. These two sectors showed production
cost and consumer price reductions pf one to two percent relative to other industries under the control
scenario, resulting in output and employment increases of similar magnitudes. .
Aggregate Effects
As noted above, the control and no-control scenarios yield different estimated mixes of investment. In
particular, the control scenario was associated with more pollution control capital expenditure and less
consumer commodity capital expenditure. As a result, the growth pattern of the economy under the control
scenario differed from the no-control scenario. Under the control scenario, the macroeconomic model
projected a rate of long-run GNP growth about one twentieth of one percent per year lower than under the
no-control scenario. Aggregating these slower growth effects of the control scenario over the entire 1970
to 1990 period of this study results, by 1990, hi a level of GNP one percent (or approximately $55 billion)
lower than that projected under the no-control scenario.
Although small relative to the economy as a whole, the estimated changes hi GNP imply that the
potential impact of the CAA on the economy by 1990 was greater than that implied by expenditures ($19
billion in 1990) or annualized costs ($26 billion hi 1990, annualized at five percent). Discounting the
stream of 1973-1990 GNP effects to 1990 gives an aggregate impact on production of 1,005 billion dollars
(in 1990 dollars discounted at five percent). Of that total, $569 billion represent reductions hi household
i 12
-------
Chapter 2: Cost and Macroeconomic Effects
consumption, and another $200 billion represent government consumption, for an aggregate effect on U.S.
consumption of goods and services equal to 769 billion dollars. Both the aggregate GNP effects and
aggregate consumption effects exceed total 1973-1990 expenditures ($628 billion) and annualized costs
($523 billion, with all dollar quantities in $1990, discounted at five percent).
Changes in GNP (or, even, changes in the national product account category "consumption") do not
necessarily provide a good indication of changes in social welfarefSibcial welfare is riot improved, for
example, by major oil tanker spills even though measured GNP is increasedJby the "production" associated
with clean-up activities. Nevertheless, the effects of the CAA on long-term economic growfii would be
expected to have had some effect on economic welfare. One of the characteristics of the macroeconomic
model used by the Project Team is its ability to estimate a measure of social welfare change which is
superior to GNP changes. This social welfare measure estimates tne monetary compensation which would
be required to offset the losses in consumption (broadly defined) associated with a given policy change.
The model reports a range of results, with the range sensitive~to assumptions regarding how cost impacts
are distributed through society. For the CAA, the model reports an aggregate welfare effect of 493 billion
to 621 billion dollars (in 1990 dollars), depending on the distributional assumptions used. This range does
not differ greatly from the range of results represenfel% 1973-1990 expenditures, compliance costs, and
consumption changes.
Uncertainties and Sensitivities In the Cost and!
Macroeconomic
The cost and macroeconomic analyses for the presenftissessment relied upon survey responses, EPA
analyses, ^nd a macroeconomic simulation model. Although the Project Team believes that the results of the
cost and macroeconomic analyses are reasonably reliable, it recognizes that every analytical step is subject to
uncertainty. Asuoted at the beginning of this chapter, explicit and implicit assumptions regarding
hypothetical technology development paths are crucial to framing the question of the cost impact of the CAA.
In addition, there is no waylo verify the accuracy of the survey results used; alternative, plausible cost
analyses exist that arrive at results mat differ from some of the results derived from EPA analyses; and it is not
clear how the use of a general equilibrium macroeconomic model affects the accuracy of macroeconomic
projections in a macroeconomy characterized by disequilibrium. For many factors engendering uncertainty,
the degree or even the direction of bias is unknown. In several areas, nevertheless, uncertainties and/or
sensitivities can be identified that may bias the results of the analysis.
j.£3-r£ » •
f ' ' ^ri?'^''* '
Productivity and Technical Change
An important component of the macroeconomic model used by the Project Team is its treatment of
technical change and productivity growth. Three factors associated with productivity and technical change
have been identified which may bias the results of the macroeconomic simulation: (1) the long-run effects
of reducing the "stock" of technology, (2) the possible "chilling" effect of regulations on innovation and
technical change, and (3) the role of endogenous productivity growth within the macroeconomic model.
The macroeconomic model projected a decrease in the growth of GNP as a result of CAA compliance.
Decreased growth was due not only to decreased capital investment, but also to decreased factor
13
-------
Chapter 2: Cost and Macroeconomic Effects
productivity. The annual decrement in productivity can be thought of as a reduction of the stock of
available technology. That reduction in stock could be expected to affect macroeconomic activity after
1990, as well as during the 1973-1990 period studied by the Project Team. Thus, to the extent that this
effect exists, the Project Team has underestimated the macroeconomic impact of the CAAFBy disregarding
the effect of 1973-1990 productivity change decrements on post-1990 GNP.
Some economists contend that regulations have a "chilling" effect on technological innovation and, .
hence, on productivity growth. The macroeconomic model allowed policy-induced productivity change
throught the mechanism of price changes and resultant factorshare changes." To the extent that additional
policy-induced effects on productivity growth exist, the ProjectTeam has underestimated the impact of the
CAA on productivity growth during the 1973-1990 period, and, thus, has underestimated macroeconomic
impacts during the 1973-1990 period and beyond.
The macroeconomic model allowed productivity growth to vary with changes in prices generated by
the model. This use of "endogenous" productivity growth is not universal in the economic growth
literature — that is, many similar macroeconomic models do not employ analogous forms of productivity
growth. The Project Team tested the sensitivity of ifie model results to the use of endogenous productivity
growth. If the model is run without endogenous productivity growth, then the predicted macroeconomic
impacts (GNP, personal consumption, etc.) of the CAA are reduced by approximately 20 percent. That is,
to the extent that use of endogenous productivity growth in the macroeconomic model is an inaccurate
simulation technique, then the Project T^irn has overestimated tie macroeconomic impact of the CAA.
Discount Rates
There is a broad range of opinion in the economics profession regarding the appropriate discount rate
to use in analyses such as the current assessment. Some economists believe that the appropriate rate is one
that approximates the social rate of time preference — that is, the rate of return at which individuals are
willing to defer consumption to the future; 'K three percent rate would approximate the social rate of time
preference (all rates used here are "real''; i.e., net of price inflation impacts). Others believe that a rate that
approximates the opportunity cost of capital (e.g., seven percent or greater) should be used.18 A third
school of thought holds that sorne combination of the social rate of time preference and the opportunity
cost of capital is appropriate, with the combination effected either by use of an intermediate rate or by use
of a multiple-step procedure employing the social rate of time preference as the "discount rate," but still
accounting for the opportunity cost of capital.
11 Some would argue that use of the opportunity cost of capital approach would be inappropriate in the current assessment if the results of the
macroecooomic modeling (such as GNP) were used as the definition of "cost," since the macro model already accounts for the opportunity cost of
capital. The appropriate rate would then be the social rate of time preference.
14
-------
Chapter 2: Cost and Macroeconomic Effects
Hie Project Team elected to use an intermediate
rate (five percent), but recognizes that analytical
results aggregated across the study period are
sensitive to the discount rate used. Consequently, all
cost measures are presented at three and seven
percent, as well as the base case five percent. Table
5, 36 summarizes major cost and macroeconomic
impact measures expressed in constant 1990 dollars,
and discounted to 1990 at rates of three, five, and
seven percent. •>••;'
\'
-Table 5*"-O>rapliance Cost, GNP,-anet Consumption ;
5 '
'- 7%^"- •
520 "
Expenditures • ' :
> Annualized Costs
GfiE\" ,' ' ' >\" I 880 ' 1005^ 115^7
JIousehoId'CcHisuinpfion' 500 <569'' .' 653" -
Exclusion of Health Benefits from the Macroeoonomic
Model
The macroeconomic modeling exercise was designed to'capture the second-order macroeconomic
effects arising from CAA compliance expenditures. Those predicted second-order effects are among the
factors used to drive the emissions estimates'and, ultimately, the benefits modeled for this assessment. The
benefits of the CAA, however, would also be expected" to induce second-order macroeconomic effects. For
example, increased longevity and decreased incidence of non-fatal heart attacks and strokes would be
expected to improve macroeconomic performance measures. The structure of the overall analysis,
however, necessitated that these impacts be excluded from th¥ macroeconomic simulation.
The fir|U|dar CAA beneficial effects have been included in the benefits analysis for this study,
including measures that approximate production changes (e.g., income loss due to illness, or lost or
restricted wbrkTdays; income loss due fo impaired cognitive ability; and income loss due to reduced worker
production in certain economic sectors). These measures are analogous to compliance expenditures in the
cost analysis. The second-order benefits impacts, which would result from price changes induced by.
CAA-related benefits,~have'not been estimated. It is likely that the estimated adverse second-order
macroeconomic impacts would Have been reduced had the impact of CAA benefits been included in the
macroeconomic modeling exercise; however, the magnitude of this potential upward bias in the estimate of
adverse macroeconomic impact was not quantitatively assessed.
15
-------
16
-------
3
This chapter presents estimates of emissions reductions duelo the Clean Air Act (CAA) for six criteria
air pollutants. Reductions are calculated by estimating, on a sector-by-sector basis, the differences in
emissions between the control and no-control scenarios. While the relevant years in this analysis are 1970
through 1990, full reporting of emissions was only made for the 1975 to 1990 period since 1970 emission
levels are, by assumption, identical for the two scenarios. The criteria pollutants for which emissions are
reported in this analysis are: total suspended particulates (TSP),19 carbon monoxide (CO), volatile organic
compounds (VOC), sulfur dioxide (SOa), nitrogen oxides (NO,),-and Leadf(Pb).
The purpose of the present study is to estimate the differences in economic and environmental
conditions between a scenario reflecting implementation of historical CAA controls and a scenario which
assumes that no additional CAA-related control programs were introduced after 1970. Because of the
focus on differences in -rather than absolute levels of- emissions between the scenarios, the various
sector-specific emission models were used to estimate both the control and no-control scenario emission
inventories. _,This approach ensures that differences between the scenarios are not distorted by differences
between modeled and actual historical emission estimates.20
Despite the use of models to estimate control scenario emission inventories, the models used were
configured and/or calibrated using historical emissions estimates. The control scenario utility emissions
estimates, for example, were based on the ICF CEUM model which was calibrated using historical
emissions inventory data.21 In other cases, such as the EPA Emissions Trends Report (Trends)
methodology22 used to estimate industrial process emissions, historical data were used as the basis for
control scenario emissions with little or no subsequent modification. Nevertheless, differences in model
~ " In 1987,~EPA replaced the earlier TSP standard with a standard for paniculate matter of 10 microns or smaller (PM10).
20 By necessity, emission models must be used to estimate the hypothetical no-CAA scenario. If actual historical emissions data were adopted
for the control scenario, differences between the monitoring data and/or models used to develop historical emission inventories and the models used
to develop no-control scenario emission estimates would bias the estimates of the differences between the scenarios.
21 See ICF Resources, Inc., "Results of Retrospective Electric Utility Clean Air Act Analysis - 1980,1985 and 1990," September 30,1992,
Appendix C . . - -
22 EPA,1994a: U.S. Environmental Protection Agency, "National Air Pollutant Emission Trends, 1900-1993," EPA-4S4/R-94-027 Office of
Air Quality Planning and Standards, Research Triangle Park, NC, October 1994.
•. " • . " . 17 • • . " •
-------
Chapter 3: Emissions
selection, model configuration, and macroeconomic input data23 result hi unavoidable, but in this case
justifiable, differences between national total historical emission estimates and national total control
scenario emission estimates for each pollutant. Comparisons between no-control, control, and official EPA
Trends Report historical emissions inventories are presented in Appendix B.24
To estimate no-control scenario emissions, sector-specific historical emissions~are adjusted based on
changes in the following two factors: (1) growth by sector predicted to occurunder the no-control
scenario; and (2) the exclusion of controls attributable to specific provisions of the CAA.
" % *" *
Vt / , „ ~
To adjust emissions for economic changes under the no-control scenario, activity leveVthat affect
emissions from each sector were identified. These activity levels include, for example, fuel use^industrial
activity, and vehicle miles traveled (VMT). The Jorgenson-Wflcoxen (J/W) general equilibrium model
was used to estimate changes in general economic conditions, as well as sector-specific economic
outcomes used as inputs to the individual sector emission models.25
The specific outputs from the J/W model used in this analysis are the percentage changes in gross
national product (GNP), personal consumption, and output for various economic sectors under the control and
no-control scenario for the years 1975,1980,1985, and 1990.26 The sectors for which the results of the J/W
model are used include: industrial processes, electric utilities, highway vehicles, industrial boilers, and the
commercial/residential sector. For the off-highway sector^ economicgrowfli was not taken into account as
there was no direct correspondence between J/W sectors and the off-highway vehicle source category activity.
In addition to adjusting for economic activity changes, any CAA-related control efficiencies that were
applied to calculate control scenario emissions were removed for the no-control scenario. In most
instances, emissions were recalculated based on 1970 control levels.
Uncertainty associated with several key modeling inputs and processes may contribute to potential
errors in the emission estimates jiresented herein. Although the potential errors are likely to contribute in
only a minor way to overall uncertainty in the estimated monetary benefits of the Clean Air Act, the most
significant emission modeiingvuncertainties are described at the end of this chapter.
:**. The Jorgenson/Wflooxen macroeconomic model outputs were used to configure both the control and no-control scenario emission model
White this satisfies the primary objective of avoiding "across model" bias between the scenarios, the macroeconomic conditions associated
with the control scenario would not be expected to match actual historical economic events and conditions. To the extent actual historical economic
conditions are used to estimate official historical emission inventories, conformity between these historical emissions estimates and control scenario
emission estimates would be further reduced.
V la general, these comparisons show dose correspondence between control scenario and Trends estimates with the largest differences
occurring for VOC and CO emissions. The Trends report VOC estimates are generally higher than the control scenario estimates due primarily to
the inclusion of Waste Disposal and Recycling as a VOC source in the Trends report. This inconsistency is of no consequence since Waste
Disposal and Recycling sources were essentially uncontrolled by the historical CAA and therefore do not appear as a difference between the control
and no-control scenarios. The higher CO emission estimates in the Trends Report are primarily associated with higher off-highway vehicle
emissions estimates. Again, since off-highway emissions do not change between the control and no-control scenario in the present analysis, this
inconsistency is of no consequence.
55 For example, the change in distribution of households by income class predicted by the J/W model was used as input to the transportation
sector model system. Changes in household income resulted in changes in vehicle ownership and usage patterns which, in turn, influence VMT and
emissions. (SeePechan,1995,p.43).
54 For details regarding the data linkages between the J/W model and the various emission sector models, see Pechan (1995).
18
-------
Chapter3:_ Emissions
The approaches used to calculate emissions for each sector vary based on the complexity of estimating
emissions in the absence of CAA controls, taking economic activity levels and CAA~regulations into
account. For the off-highway vehicle and industrial process sectors, a relatively simple methodology was
developed. The approaches used for the highway vehicles, electric utilities, industriaTBoflers, and
commercial/residential sectors were more complex because the J/W model does not address till of the
determinants of economic activity in these sectors that mightiave changed in the absence of regulation.
The approaches by sector used to estimate emissions for the two scenarios are summarized in Table 6, and
19
-------
Chapter 3: Emissions
are described in more detail in Appendix B.
Summary of Results
fcij!
ij^^ \ < «.
"'' 5. Summary of Sector-Specific Emission Modeling Approaches.
Sector
Modeling Approach
On-Highway Vehicles
Modeled using ANL's TEEMS; adjusted automobile emission estimates by
changes in personal trawl and economic activity in the without CAJ\
Track VMT was obtained firom theTedeial Highway Admiriistration
MOBILESa emission factors Were used to calculate emissions, '" ,t
>"*',*
Lead emission changes from gasoline were estimated by Abt Associates based
on historical gasoline sales and the lead content of leaded gasoline in each
target year. , . £ <_>
Off-Highway Vehicles
ELI analysis based on Trends methods. Recalculated historical emissions
using 1970 control efficiencies from Trends. No adjustment was made to f
activity levels in the without the CAA case. *,>"*' ' >
Electric Utilities
i f ^ v H
ICFs Coal and Electric Utility Model (CEUM) used to assess SO^NO^ and
TSP emission changes. Electricity sales levels were adjusted with results of the
J/W model.
The Argonne Utility Simulation Model (ARGU'S) provided CO andVoC
results. Changes in activity levels were adjusted with results of the JJ/W^. model.
Lead emissions were calculated based on energy'consumption data and Trends
emission factors and control efficiencies.' ' - >
Industrial Combustion
ANL industrial boiler analysis for SOj, NO,, and'TSP using the Industrial^ (
Combustion Emissions (ICE) model. ! > •"
•> * * i (t
VOC and CO emissions from industrial boilers were calculated based on
Trends methods; recalculated using 1970 control efficiencies. , , .,
Lead emissions calculated for boilers and processes based on Trends fuel
consumption data, emission factors, and 1970 control efficiencies. : ;
Industrial Processes
ELI analyzed industrial process emissions based on Trends methods. Adjusted
historical emissions with J/W sectoral changes in output, and 1970 control
efficiencies from Trends.
Lead emissions calculated for industrial processes and processes based on
Trends fuel consumption data, emission factors, and 1970 control efficiencies.
Commercial 7 Residential
ANL's Commercial and Residential Simulation System (CRESS) model yas
used.. , \ f i>/ ,
20
-------
Chapter 3: Emissions
Figure 5 compares the total estimated sulfur dioxide emission from all sectors under the control and
no-control scenarios over the period from 1975 to 1990. Figures 6,7,8,9, and 10 provide similar
comparisons for NO^ VOCs, CO,TSP, fr"
and Lead (Pb) respectively.
Additional tables presented in
Appendix B provide further breakdown
of the emissions estimates by individual
sector. The essential results are
characterized below. For most sectors,
emission levels under the control
scenario were substantially lower than
leyels projected under the no-control
scenario. .
The CAA controls that affected SO2
emitting sources had the greatest
proportional effect on industrial process
emissions, which were 60 percent lower
in 1990 than they would have been
under the no-control scenario. SO2
emissions from electric utilities and:
industrial boilers were each nearly 40
percent lower in. 1990 as a result of the
controls. In terms of absolute tons of
emission reductions, controls on electric
utilities account for over 10 million of
the total 16 million ton difference
between the 1990 control and no-control
scenario SOX emission "estimates.
CAA regulation of the highway
vehicles sector led to the greatest
percent reductions in VOG and NOX.
Control scenario emissions of these
pollutants in 1990^61:6 66 percent and
47 percent lower, respectively, than the
levels estimated under the no-control
scenario. In absolute terms, highway
vehicle VOC controls account for over
15 million of the roughly 17 million ton
difference in control and no-control
scenario emissions.
Differences between control and
no-control scenario CO emissions ate
Figure 5. Control and No-control Scenario Total Spx Emission
Estimates. %
40
o 30
Il20
1 10
u
I-Control
I ••• No-Control
1975
1980 1985
Year
1990
Figure 6. fGbntrol and No-control Scenario Total NO, Emission
Estimates.
ta
40
30
= 20
x
10
• Control
I -•• No-Control
1975
1980 . 1985
Year
1990
Figure 7. Control and No-control Scenario Total VOC Emission
Estimates.
40
| 30
flao
^ ^
i
10
I Control
I ••• No-Control
1975
1980 1985
Year
1990
-------
Chapter 3: Emissions
also most significant for highway
vehicles. In percentage terms, highway
vehicle CD emissions were 56 percent
lower in 1990 under the control scenario
than under the no-control scenario.
Industrial process CO emission
estimates under the control scenario
were about half the levels projected
under the no-control scenario. Of the
roughly 89 million ton difference in CO
emissions between the two scenarios, 84
million tons are attributable to highway
vehicle controls and the rest is
associated with reductions from
industrial process emissions.
For TSP, the highest level of
reductions on a percentage basis was
achieved in the electric utilities sector.
TSP emissions from electric utilities
were 93 percent lower in 1990 under the
control scenario than projected under
the no-control scenario. TSP emissions
from industrial processes were also;
significantly lower on a percentage basis
under the control scenario, with the
differential reaching 76 percent by
1990.
This is not an unexpected result as
air pollution control regulations in .the
1970's focused on solving the visible
paniculate problems from large fuel
combustors. In terms of absolute tons,
electric utilities account for over 5
million of the 16 million ton difference
between the two scenarios and industrial
processes account for almost 10 million
tons.
The vast majority of the difference
in lead emissions under the two
scenarios is attributable to reductions in
burning of leaded gasoline. By 1990,
reductions in highway vehicle emissions
account for 221 thousand of the total
234 thousand ton difference in lead
Figure 8. Control and No-control Scenario Total CO Emission
Estimates.
200
-
iso
50
•B. Control
.4. No-Control
1975
1980 1985
Year
1990
Figure 9. Control and'No-control Scenario Total TSP Emission
Estimates.
40
I 3°
•B
° n
I i
1 10
1975
1980 1985
Year
1990
Figure 10. Control and No-control Scenario Total Pb Emission
Estimates.
I
200
ISO
100
.
i »
m
•B Control I
+• No-Control
1975
1980 1985
Year
1990
22
-------
Chapter 3: Emissions
emissions. As shown in more detail in Appendix B, airborne lead emissions from all sectors were virtually
eliminated by 1990.
As described in the following chapter and in Appendix C, these emissions inventories were used as
inputs to a series of air quality models. These air quality models were used to estimate air quality
conditions under the control and no-control scenarios.
Uncertainty in the Emissions Estimates
The emissions inventories developed for the control and no-control scenarios reflect at least two major
sources of uncertainty. First, potential errors in the macroeconomic scenarios used to configure the sector-
specific emissions model contribute to uncertainties in the emissions model outputs. Second, the emissions
models themselves rely on emission factors, source allocation, source location, and other parameters which
.may be erroneous. ,
An important specific source of potential error manifest in the present study relates to hypothetical
emission rates from various sources under the no^control scenario. Emission rates from motor vehicles, for
example, would have been expected to change during the 1970 to 1990^>eriod due to technological
changes not directly related to implementation of the dean Air Act (ejg., advent of electronic fuel
injection, or EFT). However, the lack of emissions data from vehicles with EFI but without catalytic
converters compelled the Project Team to use 1970 emission factors throughout the 1970 to 1990 period
for the no-control scenario. Although this creates a potential bias in the emissions inventories, the
potential errors from this and other uncertainties in the emissions inventories are considered unlikely to
contribute significantly to overall lincertainty in the monetary estimates of Clean Air Act benefits. This
conclusion is based on the dempnstrably greater influence on the monetary benefit estimates of
uncertaintiesjn other analytical components (e.g., concentration-response functions). A list of the most
significant potential errors in the emissions modeling, and their significance relative to overall uncertainty
m the monetary benefit estimate, are presented in Table 7.
23
-------
Chapter 3: Emissions
•Table?. Uncertainties Associated with Emissions Modeling.
Potential Source of Error
Direction of Potential
Bias in Estimate of
Emission Reduction
Benefits
Significance Relative to Key
Uncertainties in Overall ,
Monetary Benefit Estimate*
I s V it 3s! ' I' » ',
Use of 1970 motor vehicle emission factors
for no-control scenario without adjustment for
advent of Electronic Fuel Injection (EFI) and
Electronic Ignition (El).
Overestimate,
Unknown, but likely to be
minor due to overwhelming
significance of catalysis in
determining emission rates.
Use of ARGUS for utility CO and VOC rather
thanCEUM.
Unknown.
Negligible.
Use of historical fuel consumption to estimate
1975 SO;,, NO,, TSP utility emissions.
Unknown.
Negligible.
Adoption of assumption that utility unit
inventories remain fixed between the control
and no-control scenarios.
Overestimate.
,
Unknown, but likely to be small
since the CAA had virtually no
effect on costs of new coal-fired
plants built prior to 1975 and' *
these plants comprise a large
majority of total coal-fired y
capacity operating in the 1970
to 1990 period. (See ICF
CEUM Report, p. 7). L
24
-------
4
Air Quality
Air quality modeling is the crucial analytical step which links emissions to changes in atmospheric
concentrations of pollutants which affect human health and the environmental. It is also one of the more
complex and resource-intensive steps, and contributes significantly to overall uncertainty in the bottom-line
estimate of net benefits of air pollution control programs. The assumptions required to estimate
hypothetical no-control scenario air quality conditions are particularly significant-sources of uncertainty in
the estimates of air quality change, especially for those pollutants which are not linearly related to changes
in associated emissions. Specific uncertainties are described in detail at the end of this chapter.
The key challenges faced by air quality modelers attempting to translate emission inventories into air
quality measures involve modeling of pollutant dispersion and atmospheric transport, and modeling of
atmospheric chemistry and pollutant transformation. These challenges are particularly acute for those
pollutants which, rather than being directly emitted, are formed through secondary formation processes.
Ozone is the paramount example since it is formed in the atmosphere through complex, nonlinear chemical
interactions of precursor pollutants^ particularly volatile organic compounds (VOCs) and nitrogen oxides
(NO*). In addition, atmosphericlransport and transformation of gaseous sulfur dioxide and nitrogen
oxides to paniculate sulfates and nitrates, respectively, contributes significantly to ambient concentrations
of fine particulate matter. In addition to managing the complex atmospheric chemistry relevant for some
pollutants, air quality modelers also must deal with uncertainties associated with variable meteorology and
the spatial and temporal distribution of emissions. "
Given its comprehensive nature, the present analysis entails all of the aforementioned challenges, and
involves additional complications as well. For many pollutants which cause a variety of human health and
environmental effects, the concentration-response functions which have been developed to estimate those
effects require, as inputs; different air quality indicators. For example, adverse human health effects of
particulate matter are primarily associated with the respirable particle fraction;27 whereas household soiling
is a function of total suspended particulates, especially coarse particles. It is not enough, therefore, to
simply provide a single measure of particulate matter air quality. Even for pollutants for which particle
size and other characteristics are not an issue, different air quality indicators are needed which reflect
different periods of cumulative exposure (i.e., "averaging periods"). For example, 3-month growing
season averages are needed to estimate effects of ozone on yields of some agricultural crops, whereas
27 Particles with an aerometric diameter of less than or equal to 10 microns.
25
-------
Chapter 4: Air Quality
adverse human health effect estimates require ozone concentration profiles based on a variety of short-term
averaging periods.28 ' _
Fortunately, in responding to the need for scientifically valid and reliable estimation otair quality
changes, air quality modelers and researchers have developed a number of highly sophisticated
atmospheric dispersion and transformation models. These models have been employed for years
supporting the development of overall federal clean air programs, national assessment studies, State
Implementation Plans (SIPs), and individual air toxic source risk assessments. Some of these models,
however, require massive amounts of computing power. For example, completing the 160 runs of the
Regional Acid Deposition Model (RADM) required for the present studyrequired approximately 1,080
hours of CPU time on a Cray-YMP supercomputer at EPA's Bay City"Supercomputing Center.
•v " X -^
Given the resource-intensity of many state-of-the-art models, tire Project Team was forced to make
difficult choices regarding where to invest the limited resources available for air quality modeling. With a
mandate to analyze all of the key pollutants affected by historical Clean Air Act programs, to estimate all
of the significant endpoints associated with those pollutants, and to do so for a 20 year period covering the
entire continental U.S., it was necessary to use simplified approaches for most of the pollutants to be
analyzed. In several cases related to primary emissions —particularly sulfur dioxide (SO2), nitrogen oxides
(NO,), and carbon monoxide (CO)— simple "roll-up model" strategies were adopted based on the
expectation that changes in emissions of these pollutants would be highly correlated with subsequent
changes in air quality.29 Significant pollutants involving secondary atmospheric formation, nonlinear
formation mechanisms, and/or long-range transport were analyzed using the best air quality model which
was affordable given time and resource limitations. These models, discussed in detail in Appendix C,
included the Ozone Isopleth Plotting with Optional Mechanism-IV (OZEPM4) model for urban ozone;
various forms of the above-referenced RADM model for background ozone, acid deposition, sulfate,
nitrate, and visibility effects in theiastem U.S.; and the SJVAQS/AUSPEX Regional Modeling
Adaptation Project (SARMAP) Air Otialjty Model (SAQM) for rural ozone in California agricultural
areas. In addition, a linear scaling approach was developed and implemented to estimate visibility changes
in large southwestern tJ.S. urban areas. -
By adopting simplified approaches for some pollutants, the air quality modeling step adds to the
overall uncertainties and limitations of the present analysis. The limited expanse and density of the U.S.
air quality monitoring network and the limited coverage by available air quality models of major
geographic areas30 further constrain the achievable scope of the present study. Under these circumstances,
it is important to remember the extent and significance of gaps in geographic coverage for key pollutants
when considering the overall results of this analysis. Key uncertainties are summarized at the end of this
chapter in Table 8. More extensive discussion of the caveats and uncertainties associated with the air
quality modeling step is presented in Appendix C. In addition, information regarding the specific air
quality models used, the characteristics of the historical monitoring data used as the basis for the control
3 For example, ozone concentration-response data exists for effects associated with 1-hour, 2J5-hour, and 6.6-hour exposures.
9 U is important to emphasize that the correlation expected is between changes in emissions and changes in air quality. Direct correlations
between the absolute emissions estimates and empirical air quality measurements used in the present analysis may not be strong due to expected
inconsistencies between the geographically local, monitor-proximate emissions densities affecting air quality data.
M For example, the regional oxidant models available for the present study do not cover some key Midwestern states, where human health,
agricultural crop, and other effects from ozone may be significant.
26
-------
Chapter 4: Air Quality
scenario profiles, pollutant-specific modeling strategies and assumptions, references to key supporting
documents, and important caveats and uncertainties are also presented in Appendix C.
General Methodology
The general methodological approach taken in this analysis starts with the assumption that actual
historical air quality will be taken to represent the control scenario. This may seem somewhat inconsistent
with the approach taken in earlier steps of the analysis, which used modeled macroecbnomic conditions as
the basis for estimating macroeconomic effects and emissions. However, the central focus of the overall
analysis is to estimate the difference in cost and benefit outcomes between the control and no-control
scenarios. It is consistent with this central paradigm to use actual historical air quality data as ihe basis for
estimating how air quality might have changed in the absence of ^e Qean Air Act.
The initial step, then, for each of the five non-lead (Pb) criteria pollutants31 was to compile
comprehensive air quality profiles covering the entire analytical period from 1970 to 1990. The source for
these data was EPA's Aerometric Information Retrieval System (AIRS), which is a publicly accessible
database of historical air quality data. The vast number of air quality observations occurring over this
twenty year period from the thousands of monitors in the U.S indicates the need to represent these
observations by statistical distributions. As documented hi detail in the supporting documents covering
SOj, NOx, CO, and ozone,32 both lognormal and gamma distributional forms were tested against actual
data to determine the form which provjdjwTthe bestlit to the historical data.33 Based on these tests, one or
the other statistical distribution was adopted for the air quality profiles developed for each pollutant. In
addition to reducing the air quality data to a manageable form, this approach facilitated transformations of
air quality profiles from one averaging period basis to another.
Once the control scenario profiles based on historical data were developed, no-control scenarios were
derived based on the results of the various air quality modeling efforts. Again, the focus of the overall
analysis is to isolate thejdifference in outcomes between the control and no-control scenarios. The no-
control scenario aii quality profiles were therefore derived by adjusting the control scenario profiles
upward (or downward) based on an appropriate measure of the difference in modeled air quality outcomes.
To illustrate this approach, consider a simplified example where the modeled concentration of Pollutant A
underthe no-control scenario is 0.12 ppm, compared to a modeled concentration under the control scenario
of Ojp) ppm. An appropriate 'measure of the difference between these outcomes, whether it is the 0.02
ppll difference in concentration or the 20 percent percentage differential, is then used to ratchet up the
l-case profile to djjnve the no-control case profile. Generally, the modeled differential is applied
-the entire control case profile to derive the no-control case profile. As described below in the
~"" ^sections-covering participate matter and ozone, however, more refined approaches are used
' take account of differential outcomes for component species (i.e., particulate matter),
and background levels of pollutants.
31 Lead (Pb), the sixth criteria pollutant, is analyzed separately. The ability to correlate emissions directly with blood lead levels obviates the
need for using air quality modeling as an intermediate step toward estimation of exposure. . • . '
32 See SAISO^ NO,, and CO Report (1994) and SAI Ozone Report (1995).
33 The statistical .tests used to determine goodness of fit are described in the SAI reports.
27
-------
Chapter 4: Air Quality
The results of the air quality modeling effort include a vast array of monitor-specific air quality profiles
for fine particulates (PM10), total suspended participates (TSP),34 SO* NOa, NO, CO, and 'ozone,-1 RADM
grid cell-based estimates of sulfur and nitrogen deposition; and estimates of visibility degradation for
eastern U.S. RADM grid cells and southwestern U.S. urban areas. All of these data were transferred to the
effects modelers for use in configuring the human health, welfare, and ecosystem physical effects models.
Given the massive quantity and intermediate nature of the air quality^data, they are not exhaustively
reported herein.35 To provide the reader with some sense ofrthe magnitudeof the difference in modeled air
quality conditions under the control and no-control scenarios, some iUustrative;results for 1990 are
presented in this chapter and in Appendix C. In addition, maps depicting; .abstflppfivels of control and no-
control scenario acid deposition and visibility are presMSdSdavoid potential confusion which might arise
through examination of percent change maps ;! , ...'!,"'"
alone;
36
Figure/11. Fr«squeiScy:bistribution of Estimated Ratios for
1990 Control to^oicontrol Scenario 95th Percentile 1-Hour
Average CO Concentrations, by Monitor.
300
200 -
I
100 -
Carbon Monoxide
Figure 11 provides an illustrative
comparison of 1990 control versus no-control;
scenario CO concentrations, expressed as a
frequency distribution of the ratios of 1990
control to no-control scenario 95ih.percentile
1-hour average concentrations at individual
CO monitors. Consistent with the emission
changes underlying these air quality results,
CO concentrations under the control scenario
tend to be about half those projected under the
no-control scenario, with most individual
monitor ratios ranging from about 0.40 to 0.60
percent, and a few with ratios in the 0.60 to 0.80 range.
In considering these results, it is important to note that CO is essentially a "hot spot" pollutant,
meaning that higher concentrations tend to be observed in localized areas of relatively high emissions.
Examples of such areas include major highways, major intersections, and tunnels. Since CO monitors tend
OJD5 025 0.45 0.65 0.85 1.05 1.25
RatbofCAA«o-CAA 95thPemaitael-HourAvesage
34 PM data are repotted as county-wide values for counties with PM monitors and a sufficient number of monitor observations.
M The actual air quality profiles, however, are available on disk from EPA. See Appendix C for further information.
M Large percentage changes can result from even modest absolute changes when they occur in areas with good initial (e.g., control scenario)
air quality. Considering percentage changes alone might create false impressions regarding absolute changes in air .quality in some areas. For
example, Appendix C discusses in detail two such cases: the Upper Great Lakes and Florida-Southeast Atlantic Coast areas, which show high
percentage caanges in sulfur deposition and visibility.
28
-------
Chapter 4: Air Quality
to be located in order to monitor the high CO concentrations observed in such locations, one might suspect
that using state-wide emissions changes to scale air quality concentration estimates at strategically located
monitors might create some bias in the estimates. However, the vast majority of ambient GO is contributed
from on-highway vehicles. In addition, the vast majority of the change in CO emissions between the
control and no-control scenario occurs due to catalyst controls on highway vehicles. Since:CO hot spots
result primarily from highway vehicles emissions, controlling such vehicles would meairCO
concentrations would be commensurately lowered at CO monitors. While variability in monitor location
relative to actual hot spots and other factors raise legitimate concerns about assuming ambient
concentrations are correlated with emission changes at any given monitor, the Project Team believes that
the results observed provide a reasonable characterization of the aggregate change m'amblentCO
concentrations between the two scenarios. , , ;!
Figure 12. 'Frequency Distribution of Estimated Ratios for
1990 Control to No-control Scenario 95th Percentile 1-Hour
Average;cSO2 Concentrations, by Monitor.
300
200 -
100 •
Sulfur Dioxide
As for CO, no-control scenario SO2
concentrations were derived by scaling control
scenario air quality profiles based on the
difference in emissions predicted under the
two scenarios. Unlike CO, SO2 is
predominantly emitted from industrial and:
utility sources. This means that emissions,
and me changes in emissions predicted under
the two scenarios, will tend to be concentrated
in the vicinity of major point sources. Again,
while state-wide emissions changes are used
to scale SO2 concentrations between the
scenarios, these state-wide emission changes
reflect the controls placed on these individual
point sources. Therefore, the Project Team
again considers the distribution of control to .
no-control ratios to be a reasonable characterization of the aggregate results despite the uncertainties
associated with estimation of changes at individual monitors.
^
Figure 12 provides a histogram of the predicted control to no-control ratios for SO2 which is similar to
the one presented for CO. The results indicate that, on an overall basis, SO2 concentrations were reduced
by about one-third. The histogram also shows a much wider distribution of control to no-control ratios for
individual monitors than was projected for CO. This result reflects the greater state to state variability in
SO2 emission changes projected in this analysis. This greater state to state variability in turn is a function
of the variable responses of SO2 point sources to historical CAA control requirements.37 This source-
specific variability was not observed for CO because controls were applied relatively uniformly on
highway vehicles.
OJ05 025 0.45 0£5 035 1.05 1,25
Ratfo ofCAA Mo-CAA 95th Penentite 1-HourAvazqe
'Figure 12 indicates that six monitors were projected to have higher SOj concentrations for 1990 under the control scenario than under the
no-control scenario. AH six of these monitors are located in Georgia, a state for which higher 1990 utility SO2 emissions are projected in the control
scenario due to increased use of higher-sulfur coal. The projected increase in overall Georgia utility consumption of higher sulfur coal under the
control case is a result of increased competition for the low-sulfur southern Appalachian coal projected to occur under the control scenario.
29
-------
Chapter 4: Air Quality
Nitrogen Dioxide
Results for NO2 are presented in Figure
13. These results are similar to the results
observed for CO, and for a similar reason: the
vast majority of change in NO2 emissions
between the two scenarios is related to control
of highway vehicle emissions. While baseline
emissions of NO2 from stationary sources
may be significant, these sources were subject
to minimal controls during the historical
period of this analysis. On an aggregated
basis, overall NO2 concentrations are
estimated to be roughly one-third lower under
the control scenario than under the no-control
scenario.
Figure 13. Frequency Distribution of Estimated Ratios for
1990 Control to No-control Scenario 95th Percentile 1-Hour
Average NO, Concentrations, by Monitor.
300
OJ3S 0.25 045 0.65 OSS 1.05 1.25
RatbofCAAWo-CAA 95thPetDentflsl-HourAveaags
Figure 14. Distribution of Estimated Ratios for 1990
Control to No-Control Annual Mean TSP Concentrations, by
Monitored County.
Particulate Matter
An indication of the difference in
outcomes for particulate matter between the
two scenarios is provided by Figure 14. This
graph shows the distribution of control to
no-control ratios for annual mean TSP in 1990
for those counties which both had particulate
monitors and a sufficient number of
observations from those monitors.38 While the
distribution of results is relatively wide,
reflecting significant county to county
variability in ambient concentration, on a
national aggregate basis particulate matter
concentrations under the control scenario were just over half the level projected under the no-control
scenario. The significant county to county variability observed in this case reflects point source-specific
controls on particulate matter precursors, especially SO2, and the effects of long-range transport and
transformation. i
0.00 020 OAO 0£0 030
RatbofCAANo-CAA AnnualM eanTSP (mtetvalmifeomt)
Ozone
Urban Ozone
Figure 15 presents a summary of the results of the 1990 OZIPM4 ozone results for all 147 of the
modeled urban areas. In this case, the graph depicts the distribution of ratios of peak ozone concentrations
** Given Ihe relative importance of particular mailer changes to the bollom line estimate of CAA benefits, and the fact thai a substantial
portion of Ihe population lives in unmonitored countifes, a methodology was developed to allow estimation of particulate matter benefits for these
unmonilorcd counties. Tin's methodology was based on the use of regional air quality modeling to inierpolate between monitored counties. It is
summarized in Appendix C and described in detail in the SA1 PM Interpolation Report (1°96). •
30
-------
Chapter 4: Air Quality
estimated for the control and no-control
scenarios. While the vast majority of
simulated peak ozone concentration ratios fall
below 1.00, eight urban areas show lower
simulated peak ozone for the no-control
scenario than for the control scenario. For
these eight urban areas, emissions of
precursors were higher under the no-control
scenario; however, the high proportion of
ambient NOX compared to ambient non-
methane organic compounds (NMOCs) in
these areas results in a decrease in net ozone
production in the vicinity of the monitor when
NOX emissions increase.39
Rural Ozone
Figures 16 and 17 present frequency
distributions for control to no-control ratios of
average ozone-season daytime ozone
concentrations at rural monitors as simulated
by RADM and SAQM, respectively.
Both the RADM and SAQM results
indicate relatively little overall change in rural
ozone concentrations. This is primarily
because reductions in ozone precursor
emissions were concentrated in populated
areas. " ' -
Add Deposition
Figure 18 is a contour map showing the
estimated percent increase in sulfur deposition
under the no-control scenario relative to the
control scenario for 1990. Figure 19 provides
comparable information for nitrogen
These results show that acid deposition
rates increase significantly under the no-
control scenario, particularly in the Atlantic
Coast area and hi the vicinity of states for
which relatively large increases in emissions
Figure 15. Distribution of Estimated Ratios for 1990
Control to No-control OZIPM4 Simulated 1-Hour Peak
Ozone Concentrations, by Urban Area.
30
20
g 10
X
0.00 0.20 0.40 0.60 0.80 1.00 1.20
RatbofCAA »o-CAA PeakOaDne (fctBjyalB flpoit)
Figure 16. Distribution of Estimated Ratios for 1990
Control to No-control RADM Simulated Daytime Average
Ozone Concentrations, by RADM Grid Cell. *
200
ISO
100
50
0.00 020 0.40 0.60 0.80 IjOO 120
RatbofCAA M&CAA OaDne^eaaonDayt&eAveiageOaooe (btexraJm iipoiit)
Figure 17. Distribution of Estimated Ratios for 1990
Control to No-control SAQM Simulated Daytime Average
Ozone Concentrations, by SAQM Monitor.
10
Ot
•g 4
§ 2
ss
0.00 020 OAO 0£0 0.80 1:00 120
RatfaofCA A Ho-CAA OaDne-GeosonDaytineA vemgeOaooe (fctenrajta flpoit)
Over an unbounded geographic area, NO, reductions generally decrease net ozone production.
31
-------
Chapter 4: Air Quality
(i.e., Kentucky, Florida, Michigan, Mississippi,
Connecticut, and Florida).
In the areas associated with large increases in
sulfur dioxide emissions, rates of sulfur deposition
increase to greater than or equal to 40 percent The
high proportional increase in these areas reflects
both the significant increase in acid deposition
precursor emissions in upwind areas and the
relatively low deposition rates observed under the
control scenario.40
Along the Atlantic Coast, 1990 nitrogen
deposition rates increase by greater than or equal to
25 percent under the no-control scenario. This is
primarily due to the significant increase in mobile
source nitrogen oxide emissions along the major •
urban corridors of the eastern seaboard.
Visibility
The difference in modeled 1990 control and
no-control scenario visibility conditions projected
by the RADM/EM for the eastern U.S. is depicted
by the contour map presented in Figure 21. This
figure shows the increase in modeled annual
average visibility degradation,1 ih'DlciYiew41 terms,
for 1990 when moving from the control to the
no-control scenario. Since the DeciVIew metric is
based on perceptible changes in visibility, these
results indicate noticeable deterioration of visibility
in the eastern U.S. under thes no-control scenario.
Visibility changes in 30 southwestern U.S.
urban areas were also estimated using emissions
scaling techniques. This analysis also found
significant, perceptible changes in visibility
between the two scenarios. Details of this analysis,
including the specific outcomes for the 30
individual urban areas, are presented in Appendix
C
Figure 18. RADM-Predicted Percent Increase in Total
Sulfur Deposition (Wet + Dry) Under the No-control
Scenario.
K 0 - 25
j«25 - 30
!090 -35
n35 -40
II > 40
Figure 19. RADM-Predicted Percent Increase in Total
Nitrogen-Deposition (Wet + Dry) Under the No-control
Scenario.
M Even small changes in absolute deposition can yield large percentage changes when initial absolute deposition is low. See Appendix C for
further discussion of this issue.
41 The DeciView Haze Index (dV) is a relatively new visibility indicator aimed at measuring visibility changes in terms of human perception.
It is described in detail in Appendix C . •
32
-------
Chapter 4: Air Quality
Uncertainty prevades the projected changes in
air quality presented in this study. These
uncertainties arise due to potential inaccuracies in
the emissions inventories used as air quality
modeling inputs and due to potential errors in the
structure and parameterization bf the air quality
models themselves. In addition, an important
limitation of the present study is the lack of
available data and/or modeling results for some
pollutants in some regions of the country (e.g., .
visibility changes in western U.S. Class I areas
such as the Grand Canyon). The inability to
provide comprehensive estimates of changes in air
quality due to the Clean Air Act creates a
downward bias in the monetary benefit estimates.
The most important specific sources of
uncertainty are presented in Table 8, and are
described further in Appendix C. While the list of
potential errors presented in Table 8 is not
exhaustive, it incorporates the uncertainties with t
the greatest potential for contributing to error in the monetary benefit estimates. Overall, the uncertianties
in the estimated change in air quality are considered small relative to uncertainties contributed by other
components of the analysis.
Figure 20. RADM-Predicted Increase in Visibility
Degradation, Expressed in DeciViews, for Poor
Visibility Conditions (90th Percentile) Under the No-
control Scenario. ,
33
-------
Chapter 4: Air Quality
.mil? ^»ftti^m'"•p«>'! 's*4'*1 * *. ,ivi 'A"*-',-' »; ^ *•
ffiable 8. Key Uncertainties Associated with Air Quality Modeling.
. .
Potential Source of Error
Direction of
Potential Bias
in Estimate of
Air Quality
Benefits
Significance Relative to Key
Uncertainties in Overall Monetary
Benefit Estimate
Use of QZIPM4 model, which does not
capture long-range and night-time transport of
ozone. Use of a regional oxidant model, such
as U AM-V, would mitigate errors associated
with neglecting transport.
Underestimate.
Significant, but probably not major. , , ,
Overall average ozone response of 15% to
NO, and yOC redactions of
approximately 30% and 45%, respectively.
Even if ozone response doubledto 30%,
estimate of monetized benefits of CAA '
will not change very much. Significant '
benefits of ozone reduction, however,
could not be monetized.
Use of early biogenic emission estimates in
RADM to estimate rural ozone changes in the
eastern 31 states.
Underestimate.
Probably minon Errors are estimated to
be within -15%1o +25% of 'the, .ozone
predictions. " » '^_'V '
Use of proxy pollutants to scale op some
particulate species in some areas. Uncertainty
is created to the extent species of concern are
not perfectly correlated with the proxy
pollutants.
Unknown.
Potentially significant. Given ttie relative
importance of the estimated changes in
fine particle concentrations to the
monetized benefit estimate, any
uncertainty associated with fine particles is
potentially significant, floweyer, the (^
potential error is mitigated to some extent
since proxy pollutant measures are applied
to both scenarios. " •( ^
Use of state-wide average emission reductions
to configure air quality models. In some
cases, control programs may have been
targeted to problem areas, so using state-wide
averages would miss relatively large
reductions in populated areas.
Underestimate.
Probably minor.
Exclusion of visibility benefits irt Class I areas
in the Southwestern U.S.
Underestimate.
" > ' 1
Probably minor. No sensitivity analysis
has been performed; however, monetized
benefits of reduced visibility impairment y
in the Southwest would probably not '
significantly alter the estimate of
monetized benefits. ' ' ' ^
34
-------
Chapter 4: Air Quality
** * ^ Vv< , ^ t- * ' v * * A * £ .
*K^ r
-» ( .< * V1
fcPotentia|'§piiree of Error
Direction of
Potential Bias
in Estimate of
Air Quality
; BenelBts
'. Signiecance Relative to Key
Uncertainties in Overall Monetary
Benefit Estimate , 1
^ < N - ^
Lack of model coverage in western 17 states
for add deposition. f . " t , , Z _
^1" ' "
Underestimate.
Probably minor. No sensitivity analysis *,
has been performed; however,' monetized
benefits of reduced acid deposition in the
17 western states would probably not ,''
significantly alter the estimate of
monetized benefits. ':
»•••>/ , -• « / > ... .
Use of spatially and geographically ::
aggregated emissions data to configure
RADM. l^usKojE^«aila1jlevday-specific
meteorological data results in inaBility to '
account for temperature effects on VOCs and
effect of localized meteorology around major
pdintsources, -\ „
Unknown*
'Potentially significant. Any effect which; , '
.might influence the direction of long-range
transport of fine particulates such as
sulfates and nitrates could significantly
influence the estimates of total monetized .
benefits of theCAA. i
' -
•t s1. * ^ "%>^
Ifee of lab-^SO's nromtoiingdata^o derive
the PMJ0:TSP ratios aeeded to developJ&H
PM profiles. Since early control programs
focused significantly on combustion sources,
using recent yearTM10:TSP ratios may lead to
underestimation of early period controls of the
finepartfcletractiotf. ' ^ / -^
Underestimate,
Potentially significant Any
underestimation of the baseline - ,.
contribution of fine particles to total PSl*
•would lead to an underestimation of fine
particle controls, which are a relatively
significant contributor to the monetized
benefit estimate. ' s »-"
Use of constant cp Acentration for organic
aerosols between the tw6 Scenarios. Holding
organic aerosol concentrations fixed omits the
effect of changes in this constituent of fine
particulate matter^ ,
'' ' ' ! '
Underestimate.
Probably minor, because (a) nitrates were
also held fixed and nitrates and organic
aerosols move in opposite directions so the
exclusion of both mitigates f he effect of
omitting either, (b) sulfates are by far the
dominant species in the eastern U.S., and
(c) larger errors would be introduced by
using emissions scaling to estimate
changes in organic aerosols since most
organic aerosols are caused by biogenic
emissions which don't change between the
scenarios.
Unavailability of ozone models for rural areas
outside the RADM and SAQM domains.
Underestimate,
Probably minor. Misses potential human
health, welfare, and ecological benefits of
reducing rural ozone in agricultural and
other rural areas; however, ozone changes
are likely to be small given limited
precursor reductions in rural areas.
RADM controbno-control ratios are in
fact, relatively small. >
35
-------
36
-------
5
Physical Effects
Human Health and Welfare Effects Modeling
This chapter identifies and, where possible, estimates the principal health and welfare benefits enjoyed
by Americans due to improved air quality resulting from the CAA. Health benefits have resulted from
avoidance of air pollution-related health effects, such as premature mortality, respiratory illness, and heart
disease. Welfare benefits accrued where improved air quality averted damage to measurable resources,
including agricultural production and visibility. The analysis of physical effects required a combination of
three components: air quality, population; and health or welfare effects. As structured in this study, the 3-
step process involved (1) estimating changes in air quality between the control and no-control scenarios,
(2) estimating the human populations and natural resources exposed to these changed air quality
conditions, and (3) applying a series of concentration-response equations which translated changes in air
quality to changes hi physical health and welfare outcomes for the affected populations.
Air Quality
The Project Team first estimated changes in concentrations of criteria air pollutants between the
control scenario, which at this step was based on historical air quality, and the no-control scenario. Air
quality improvements resulting from the Act were evaluated hi terms of both their temporal distribution
from 1970 to 1990 and their spatial distribution across the 48 conterminous United States. Generally, air
pollution monitoring data provided baseline ambient air quality levels for the control scenario. Air quality
modeling was used to generate estimated ambient concentrations for the no-control scenario. A variety of
modeling techniques was applied, depending on the pollutant modeled. These modeling approaches and
results are summarized in Chapter 4 and presented in detail in Appendix C.
Population \
Health and some welfare benefits resulting from air quality improvements were distributed to
individuals in proportion to the reduction hi exposure. Predicting individual exposures, then, was a
necessary step in estimating health effects. Evaluating exposure changes for the present analysis required
not only an understanding of where air quality improved as a result of the CAA, but also how many
individuals were affected by varying levels of air quality improvements. Thus, a critical component of the
.benefits analysis required that the distribution of the U.S. population nationwide be established.
37
-------
Chapter 5: Physical Effects
Three years of U.S. Census data were used to represent the geographical distribution of U.S. residents:
1970,1980, and 1990. Population data was supplied at the census block group level, with approximately
290,000 block groups nationwide. Allocating air quality improvements to the population for the other
target years of this study—1975 and 1985 - necessitated interpolation of the three years of population data.
Linear interpolation was accomplished for each block group in order to maintain the variability in growth
rates .throughout the country.
Health &nd Welf&ro Effects
Benefits attributable to the CAA were measured in terms of the avoided incidence of physical health
effects and measured welfare effects. To quantify such benefits, it was necessary to identify concentration-
response relationships for each effect being considered. As detailed in Appendix D, such relationships
were derived from the published science literature. In the case of health effects, concentration-response
functions combined the air quality improvement and population distribution data with estimates of the
number of fewer individuals that suffer an adverse health effect per unit change in air quality. By
evaluating each concentration-response function for every monitored location throughout the country, and
aggregating the resulting incidence estimates, it was possible to generate national estimates of incidence
under the control and no-control scenarios.
In performing this step of the analysis, the ProjectTeam discovered that it was impossible to estimate
all of the health and welfare benefits which have resulted from the Clean Air Act. While scientific
information was available to support estimation of some effects, many other important health and welfare
effects could not be estimated. Furthermore, even though some physical effects could be quantified, the
state of the science did not support assessment blffie;ejcpn6mic value of all of these effects. Table 10
shows the health effects for which quantitative analysis was prepared, as well as some of the health effects
which could not be quantified in the analysis. Table 11 provides similar information for selected welfare
effects. '
While the 3-step analytical process described above was applied for most pollutants, health effects for
lead were evaluated using a different methodology. Gasoline as a source of lead exposure was addressed
separately from conventional point sources. Instead of using ambient concentrations of lead resulting from
use of leaded gasoline, the concentration-response functions linked changes in lead releases directly to
changes in the population's mean blood lead level. The amount of leaded gasoline used each year was
directly related to mean blood lead levels using a relationship described in the 1985 Lead Regulatory
Impact Analysis (U.S. EPA, 1985). Health effects resulting from exposure to point sources of atmospheric
lead, such as industrial facilities, were considered using the air concentration distributions modeled around
these point sources. Concentration-response functions were then used to estimate changes in blood lead
levels in nearby populations.
Most welfare effects were analyzed using the same basic 3-step process used to analyze health effects,
with one major difference in the concentration-response functions used. Instead of quantifying the
relationship between a given air quality change and the number of cases of a physical outcome, welfare
effects were measured, in terms of the avoided resource losses. An example is the reduction in agricultural
crop losses resulting from lower ambient ozone concentrations under the control scenario. These
agricultural benefits were measured in terms of net economic surplus.
38
-------
Chapter 5: Physical Effects
.Table 1Q» Human Health Ejects of tXiteria Pollutants.
Pollutant " ,
•> *• > f v
Ozone ^' " ~^
<• •" < f
'V>;^v
.^ y** v<" s
," *^
-^.
*, ^ *•*
Particulate Matiei^ '
TSiySulfates .'
«;, „ ^
J V •>' v. X '
* < *" „ y
V
Nitrogenbxides , •';:
> * ^
1 x "< „ •• ' •-
Sulfur Dioxide :'
<
< *, *~ ^
- - S ' ' "
Lead '
\ x ^ *•
•? ^
' V " ' *
/ ^
< ' '
0 i
), * ^
V * ^
Quantified Health Effects
Respiratory symptoms
Minor restricted activity days
Respiratory restricted activity
' days1-
Hospital admissions ,:
Asthma attacks
taianges in pulmonary function
Chronic Sinusitis & Hay Fever „
Acute bronchitis
Hospital admissions
' kcsrta^Sy
Chronic bronchitis
Chest illness
Respiratory symptoms
Minor restricted activity days
AffresWcted"acuvitjr4ays
Days of work loss
Moderate or worse asthma
/ "Status (asthmatics) *>,
Hospital Admissions
Decreased time to onset of
angina
Respiratory illness
>
In. exercising asthmatics:
' Respiratory symptoms
i
V, ?
( ^
Hypertension s
Non-fatal coronary heart
disease
Non-fatal strokes
Mortality ,,
IQ loss effect (on lifetime
earnings)
IQ toss effects on special
education needs
Neonatal mortality due to
decreased gestatibnal age
Unquantiffed Health Effects
Increased airway
responsiveness to stimuli
Centroacinar fibroas
Inflammation in the lung '
i*( ^
Changes in puhnonary function
> ^
f<
/ > ^
Behavioral effects
Increased airway
responsiveness
\
In exercising asthmatics:
Changes in pulmonary function
Combined responses of
respiratory symptoms and K
pulmonary function
- changes
Reproductive effects
Fetal effects from maternal
exposure
Other neurological and
metabolic effects
Other cardiovascular effects
v-
^ *" 'V " *> "•
Other Possible Effects
< s
Immunologic changes :
Chronic respiratory diseases /
Extrapulmonary effects (e.g.,
• changes in structure,
function of other organs) '
'" ~ v " l ' ( '
s
* v / "" *• '
^ > Qjroaictt
» (- other than chronic "„
bronchitis
Inflammation in the Jung
^ j
, < '*• *•*•
s •<
r- ^ ~" . * *"
V 5,
* •/
1 > f
v " S *•
Other cardiovascular effects
Developmental effects ' '
v "s
•\ >.
Decreased pulmonary function t
Inflammation in the lung
Immunological changes *
Respiratory symptoms in non-
asthmatics ~ * ',
Hospital admissions:
*. <,
**• "* ~>*
s
""" f f
Cancer -
"•" > ' ! H
-> " $
* i ^
piratory disease;
Another important welfare effect is the benefit accruing from improvements in visibility under the
control scenario. Again, a slightly different methodological approach was used to evaluate visibility
improvements. Visibility changes were a direct output of the models used to estimate changes in air
quality. The models provided estimates of changes in light extinction, which were then translated
mathematically into various specific measures of perceived visibility change. These visibility change
39 ,
-------
Chapter 5: Physical Effects
, t ', ,Jl,>ttUI«IH« li> , 1 > ' ^,» -<» * ' ' ,\ *! V, '^
lie 11. Selected Welfere Effects of Criteria Pollutants. "
A . ' ' •"• " i '* 1
: Pollntaat
^ Ozoae
F
7"
i: - -
r
Pwtknlate Matter/
TSP/Salfates
" Nitrogen Oxides
1
*•
| Sulfur DloxWe
**•• - f "
T
m
; i ,.
Quantified Welfare Effects
Changes in crop yields {for 7 oops)
Decreased worker productivity
* ^ <•
1 ;
Household soiling
Visibility > ' '
, • • t
Visibility
- J
r >-'
k
Visibility
' '" J
J
T *•' /
Unqualified Wel&re Effects , " <-
••*. t . , ,_ ,M>
Changes lit other crop yields
Materials damage
Effects on forests \ >^
Effects omwldlife ''
- > ' ., • t"lj
Other materials damage
Effects on wildlife
1 " ' , > / i »*• '
Crop losses due to acid deposition
Materials damage due to acid deposition
Effects on fisheries due to acidic deposition
Effects on forests i
'* J" i j *><• • '•>>)?
Crop losses due to acid deposition '
Materials damage due to acid deposition
Effects on fisheries due to acidic deposition
Effects on forests " ° ' t ""
measures were then combined with population data to estimate the economic value of the visibility
changes. Other welfare effects quantified in terms of avoided resource losses include household soiling.
damage by PM10 and decreased worker productivity :dueto;pzone exposure. The results of the welfare
and F. ',.'•••
Because of a lack of available concentration-response functions (or a lack of information concerning
affected populations), ecological effects were^not quantified for this analysis. However, Appendix E
provides discussion.of many of the important ecological benefits which may have accrued due to historical
implementation of theGAA. .
Key Analytical Assumptions
Several important analytical assumptions affect the confidence which can be placed in the results of
the physical effects analysis. The most obvious concerns the choices which have to be made among
competing scientific studies. The Project Team relied on the most recent available, published scientific
literature to ascertain the relationship between air pollution and human health and welfare effects. The
choice of studies, and the uncertainties underlying those studies, also created uncertainties in the results.
For example, to the extent the published literature may collectively overstate the effects of pollution,
EPA's analysis will overstate the effects of the CAA. Such outcomes may occur because scientific research
which fails to find significant relationships is less likely to be published than research with positive results.
On the other hand, history has shown that it is highly likely that scientific understanding of the effects of
air pollution will improve in the future, resulting in discovery of previously unknown effects. Important
examples of this phenomenon are the substantial expected health and welfare benefits of reductions in lead
and ambient participate matter, both of which have been shown in recent studies to impose more severe
effects than scientists previously believed. To the extent the present analysis misses effects of air pollution
40
-------
Chapter 5.'Physical Effects
that have not yet been subject to adequate scientific inquiry, the analysis may understate the effects of the
CAA.
Because these resultant uncertainties were caused by the inadequacy of currently available scientific
information, there is no compelling reason to believe that the results of the present analysis are biased in a
particular direction. Some significant uncertainties, however, may have arisen from interpretation of model
results, underlying data, and supporting scientific studies. These assumptions and uncertainties are
characterized in this report to allow the reader to understand the degree of uncertainty and the potential for
misestimation of results. In addition, the overall results are presented in ranges to reflect the aggregate *
effect of uncertainty in key variables. A quantitative assessment of some of the uncertainties in the present
study is presented in Chapter 7. In addition, the key uncertainties in the physical effects modeling steJ3 of
this analysis are summarized in Table 16 at the end of this chapter. The remainder of this section discusses
a few of the most important modeling procedures and associated assumptions.
Mapping Populations to Monitors
As noted above, the Project Team's method of calculating benefits of air pollution reductions required
a correlation of air quality data changes to exposed populations. For pollutants with monitor-level data
(i.e., SC>2,03, NO^, CO), it was assumed that all individuals were exposed to air quality changes estimated
at the nearest monitor. For PM10, historical air quality data were available at the county level. All
individuals residing in a county were assumed to be exposed to that county's PM10 air quality.42
Many counties did not contain particulate matter air quality monitors or did not have a sufficient
number of monitor observations to provide reliable estimates of air quality. For those counties, the Project
Team conducted additional analyses to estimat&PM10 air quality changes during the study period. For
counties in the eastern 31 states, the grid cell-specific sulfate particle concentrations predicted by the
RADM model were used to provide a scaled interpolation between monitored counties. For counties
outside the RADM domain, an alternative method based on state-wide average concentrations was used.
With this supplemerital"analysis, estimates were developed of the health effects of the CAA on almost the
entire continental U.S. population. Compliance costs incurred in Alaska and Hawaii were included in this
study, but the benefits of historical air pollution reductions were not. In addition, the CAA yielded benefits
to Mexico and Canada which were not captured in this study.
In some counties and in the early years of the study period, particulate matter was monitored as TSP rather than as PM10 In these cases,
PM10 was estimated by applying TSP:PM,0 ratios derived from historical data. This methodology is described in Appendix C. '
41
-------
Chapter 5: Physical Effects
Ak quality monitors are more likely to be ,- 'V* ' * t »*~
found in high pollution areas rather than low- Table 12. Percent of Population (of the Continental US)
pollution areas. Consequently, mapping *»* 5°^*^f°r ' -
km" sensitivity scenario. For most pollutants in •••••••••••••••••••••••••••••••»••••••••••
most years, 25 percent or more of the population
resided more than 50 km from an air quality monitor (or in a county without PM10 monitors). Estimated
health benefits are approximately linear to population covered — that is, if the population modeled for a
pollutant in a given year in the sensitivity analysis is 25 percent smaller than the corresponding population
modeled in the base case, then estimated health benefits are reduced by roughly 25 percent in the
sensitivity case. This sensitivity analysis demonstrates that limiting the benefits analysis to reflect only
those living within 50 km of a monitor or within a PM-monitored county would lead to a substantial
underestimate of the historical benefits of the CAA. Since mese|S(tilative results may have led to severely
misleading comparisons of the costs and benefits of the Act, raelProject Team decided to adopt the full 48-
state population estimate as the central case for this analysis despite the greater uncertainties and potential
biases associated with estimating exposures from distant monitoring sites.
II :
ChoiCQ of Study
As noted above, the Project Team estimated health and welfare effects of improved air quality through
the use of concentrafibn-response (CR) functions derived from the peer-reviewed scientific literature. For
some healmeniapcMnts, however, several functions were available, and significant differences were found
among the CR functions in implied health effects. For example, 19 CR functions were available correlating
excess premature mortality to short-term exposure to particulate matter (PM). It is not unusual for two
equally-reputable studies to differ by a factor of three or four in implied health impact. The difference in
implied health effects across studies can be considered an indication of the degree of scientific uncertainty
associated with measurement of that health effect.
Where more than one acceptable study was available, the Project Team used CR functions from all
relevant studies to infer health effects. That is, the health effect implied by each study is reported (see
Appendix D), and a range of reported results for a particular health endpoint can be interpreted as a
measure of the uncertainty of the estimate.
Variance Within Studies
Even where only one CR function was available for use, the uncertainty associated with application of
that function to estimate physical outcomes can be evaluated quantitatively. Health effects studies provided
"best estimates'' of the relationship between air quality changes and health effects, and a measure of the
statistical uncertainty of the relationship. In this analysis, the Project Team used simulation modeling
42
-------
Chapter 5: Physical Effects
techniques to evaluate the overall uncertainty of the results given uncertainties within individual studies,
across studies examining a given endpoint, and in the economic valuation coefficients applied to each
endpoint. The analysis estimating aggregate quantitative uncertainty is presented in Chapter 7.
Health Effects Modeling Results
This section provides a summary of the differences in health effects estimated under the control and
no-control scenarios. Because the differences in air quality between the two scenarios generally increased
from 1970 to 1990, and the affected population grew larger during that period, the beneficial health effects
of the CAA increased steadily during the 1970 to 1990 period. More detailed results are presented in
Appendix D. . - r
Mortality
The most serious human health impact of air pollution is an increase in incidences of premature
mortality. Increased levels of lead (Pb) in the bloodstream and increases in various measures of ambient
participate matter have each been correlated with elevated mortality rates: For presentation purposes in this
analysis, the expected reduction in mortality rates due to CAA-induced air quality improvements have been
expressed as avoided "excess premature mortalities per year." For Pb, separate analyses were conducted
for men, women, and children, and then aggregated to present a total effect for Pb (see Appendix G for
detailed information on methods, sources, and results of the Pb mortality analysis). For PM10, two types of
studies were used as sources of concentration/response functions. Table 13 presents estimated avoided
excess premature mortalities for 1990 only, with the mean estimate and 90 percent confidence interval. See
Appendix D for more detail on results implied by individual epidemiological studies, and on the temporal
pattern of impacts.43 ~ _
Short-Term Exposure Studies
A substantial body of published health science literature recognizes a correlation between elevated PM
concentrations (either PM10, or a related measure such as TSP) and increased mortality rates. Many of the
studies evaluate changes in mortality rates several days after a period of elevated PM concentrations. In
gen'eral, strong correlations have been found. These "short-term exposure" or "episodic" studies are unable
to address two important issues: The degree to which the observed excess mortalities are "premature," and
the degree to which mortality rates are correlated with long-term exposure to elevated PM concentrations
(that is, exposures over many years rather than a few days).
.Because the episodic mortality studies evaluate the mortality rate impact a only few days after a high-
pollution event, it is likely that many of the "excess mortality" cases represented individuals who were
already suffering impaired health, and for whom the high-pollution event represented an exacerbation of an
already serious condition. Based on the episodic studies only, however, it is unknown how many of the
victims would have otherwise lived only a few more days or weeks, or how many would have recovered to
enjoy many years of a healthy life in the absence of the high-pollution event For the purpose of cost-
43 Earlier years ate estimated to have had fewer excess premature mortalities.
, 43
-------
Chapter 5: Physical Effects
benefit analysis, it can be important to determine whether a pollution event reduces the average lifespan by
several days or by many years. Although the episodic mortality studies do not provide an estimate of the
expected life years lost (nor do they address the health status of victims), some have evaluated the age of
the excess premature mortality cases, and have estimated that 80 to 85 percent of the victims are age 65 or
older.
In addition to causing short-term health problems, air pollution (measured by elevated PM10
concentrations) can cause longer-term health problems that may lead to premature mortality. Such long-
term changes in susceptibility to premature mortality in the future will be missed by efforts to correlate
premature mortalities with near-term episodes of elevated pollution concentrations. Consequently, excess
premature mortality estimates based on the results of the "episodic" mortality studies will underestimate
the effect of long-term elevated pollution concentrations on mortality rates.
Long-Term Exposure Studies
Several health effects studies have correlated long-term exposure to criteria air pollutants (measured by
PM10 concentrations, or a related measure such as TSP concentration) with annual mortality rates. These
studies are able to avoid some of the weaknesses of the episodic studies. In particular, by investigating
changes in annual (rather than daily) mortality rates, the long-term studies do not predict most cases of
excess premature mortality where mortality is deferred for only a few days; also, the long-term studies are
able to discern changes in mortality rates due to long-term exposure to elevated air pollution
concentrations. Other characteristics, However, make this class of studies imperfectly suited for the present
analysis. Most importantly, the concentration response functions inferred from the long-term exposure .
studies may overstate health effects of elevated air pollution concentrations if pollution exposures from
before the study period affect health status during the study period.
Avoided Excess Premature Mortality Estimates
Table 13. Criteria Pollutants Health Benefits - Distributions of 1990 Avoided
Prematore Mortalities (thousands of cases reduced) for 48 State Population
The number of premature mortalities
avoided in 1990 (only) estimated by the
health benefits model is presented in
Table 13. The model reports a range of
results for each health endpoint. Here, the
fifth percentile, mean, and ninety-fifth ,
percentile estimates are used to
characterize the distribution. One study
(Pope et aL, 1995) was used as the source
for a concentration-response function for ^ a '<''"„ * ' * ,
long-term exposure. Concentration- ^ is ,'j 1 ,' ,'.,.,%
response data inferred from nineteen mmmmammmammmm^mmm^^^^^mmmaai^^mmmm
studies were used to estimate short-term
exposure effects. The long-term and short-term results are not additive - they should be viewed as
representing a plausible range of avoided premature mortalities. The estimates of premature mortalities
avoided due to reduced lead exposure are additive with the PM-related short-term and long-term exposure
estimates.
1 '' 1
Endpoint
Mortality (long-term exposure)
Mortality (short-term exposure)
Mortality (Lead exposure)
!. ^
Pollutant
1
'PM-10
PM-10
Lead
Annual Cases Avoided
{thousands)
5th - 95th
%ile Mean %Ue
145 „ 238 341
55 58 6|
8 24 38
44
-------
Chapter 5: Physical Effects
Non-Fatal Health impacts
The health benefits model
reports non-fatal health effects
estimates similarly to estimates of
premature mortalities: as a range
of estimates for each quantified
health endpoint, with the range
dependent on the quantified
uncertainties in the underlying
concentration-response
functions. The range of results
for 1990 only is characterized in
Table 14 with fifth percentile,
mean, and ninety-fifth percentile
estimates. All estimates are
expressed as thousands of new
cases avoided in 1990. "Lost IQ
Points" represent the aggregate
number of points (in thousands)
across the population affected by
lead concentrations in 1990. All
"Hospital Admissions" estimates
are in thousands of admissions,
regardless of thePlength of time
spent in the hospital. "Shortness
of breath" is expressed as
thousands of days: that is, one
"case" represents one child
experiencing shortness of breath
for one day. Likewise,
"Restricted Activity Days" and
"Yfork Loss Days" are expressed
in person-days.
'• "Table 14. Criteria Pollutants Health Benefits - Distributions of 1990 Non-Fatal Avoided v
Incidence (thousands of cases reduced) for 48 State Population
:" '"•-., '; :\t~v A
Endpoint "( '" ' ,'\ s : *• •
t >• a s ?~* s S " ••,.•* ^ ^ ^ _,
CJwofljo Djo&cbitx& "* '*•'*'*<
Non-Fatal Lead-induced Ailments ,
Lost IQ Points -_ , '*' ,
• , Hypertension
„ Coronary Heart Disease , , ,
' Atherothrombotic brain infarction
; Initial cerebrovascular accident
Hospital Admissions
All Respiratory , ,
v '- PB+ Pneumonia , ,
' fechemic Heart Disease
Congestive Heart Failure
Other Respiratory-Related Ailments
Children .••',' - ,
Shortness of breath, days
Acute Bronchitis _
, Upper & Lower Resp. Symptoms
'. Adults " "", " '
^/-AnyofWAcufeSymrrtoms
AsthmaAttacks
, Increase in Respiratory Illness
* Any Symptom , ' ' ,
Restricted Activity and Work Loss Days '
VniAD ' s
Work Loss Days fWH))
'- foBnfant(s)
s PM^6,S ' '' . '
'/3Uad'' ,. ',
Lead
lead-
Lead' -
'lead '"f
^ s *''
-^PM-10&03%
'PJif-lQ&O3
'PM-10
>pjir-io&cox
'! '
•," ~-
PM-10
PM-10
PM-10
.f- ', "~\-
PM-10&O3
'PM-1O&O3;
NO2
ISOZ
s
SPM-10&03-
PM-10 1
Annual Cases Avoided ;
fthoosands) '•>"•'
:Jt-'^-'^'
.; \4M: V>1, ^*
"7,600 " io,4ro '13,100
9,70jD% 12,800 15,600
,, '-5' s -36 ^ "105
' 0- „? .IS'
o '; ,9 , 26
" ! " f \'t > -.
;> ' 74'" , 88 "- 104:
'56 „ "63- ' - -71
^8 19 '31;
', "',28 "'.» " "5~0
' • ' f.< '- '•
', „ - „' ' '. "";'
15,000 <8&'fXXt 129,000
,0V '8,700; 21,600
' ,,6,200.. 9^00 12^00,
> - ,' , ' "- * :
25,000 IS? 000 252,000
'-860 1,710' \2^80
4^M)d 9,900, 14^200
, 26 i67 759
^ *" ^ ™ f
8,030 8,940 ' 9,9^0
.. 1390 /l,filO* 1,^30"
The following additional welfare benefits were quantified directly in economic terms:
tiniisehntiil nvflfncr damage, visibility, decreased workerproductivity, and agricultural
I in terms of net surplus). ' ^ - -, . "" ,
45
-------
Chapter 5: Physical Effects
Of her Physical Effects
Human health impacts of criteria pollutants dominate quantitative analyses of the effects of the CAA,
in part because the scientific bases for quantifying air quality and physical effect relationships are most
advanced for health effects. The CAA yielded other benefits, however, which are important even though
they were sometimes difficult or impossible to quantify fully given currently available scientific and
applied economic information.
Ecological Effects
Beyond the intrinsic value of preserving natural aquatic and terrestrial ecosystems and the life they
support, protection of ecosystems from the adverse effects of air pollution can yield significant benefits to
human welfare. The historical reductions in air pollution achieved under the CAA probably led to
significant improvements in the health of ecosystems and the myriad ecological services they provide. For
example, improvements in water quality stemming from a reduction in acid deposition-related air
pollutants such as SO, and NOX likely yielded significant benefits to human welfare through enhancements
in commercial and recreational fishing, wildlife viewing, maintenance of biodiversity, nutrient cycling, and
maintenance and improvement of drinking water quality. Increaiseld growth and productivity of U.S. forests
were likely enhanced due to reduced concentrations of ambient ozone. Protection of forest resources results
in benefits associated with increased"timber production, improved outdoor recreation (e.g., hunting,
camping), and ecological services such as nutrient cycling and carbon sequestration. These improvements ,
in ecological conditions have not been quantified in iiJuVassessment. For a fuller discussion of the possible
ecological effects of the CAA, see Appendix E.
Quantified Agricultural Effects
Quantification of the effects of the CAA on agriculture was limited to effects on seven crops. Changes
in crop yields between rne two scenarios were less than one percent in 1990 for barley, corn, soybeans, and
sorghum. Changes in peanut and cotton yields were estimated in the four to seven percent range, and
waiter wheat yield changes were estimated at one-half of one percent to eight percent. These ranges reflect
usage of alternate exposure-response functions. Exposure-response relationships and cultivar mix reflect
historical patterns and do not account for possible substitution of more ozone-resistant cultivars in the no-
coMroI scenario. Tfius, these are likely overestimates of the actual effect of the CAA on crop yields. It is
unclear whether that likely overestimate is greater than or less than the underestimate of benefits due to
exclusion of all other crops from the quantitative analysis.
Effects of Air Toxics
In addition to control of criteria pollutants, the Clean Air Act resulted in control of some air toxics —
defined as non-criteria pollutants which can cause adverse effects to human health and the environment.
Control of these pollutants resulted both from incidental control due to criteria pollutant programs and
46
-------
Chapter 5: Physical Effects
specific controls targeted at air toxics through the National Emission Standards for Hazardous Air
Pollutants (NESHAPs) under Section 112 of the Act.
Air toxics are capable of producing a wide variety of effects. Table 15 presents the range of potential
human health and ecological effects which can occur due to air toxics exposure. For several years, the
primary focus of risk assessments and control programs designed to reduce air toxics has been cancer.
According to present EPA criteria, there are over 100 known or suspected carcinogens. EPA's 1990
Cancer Risk study indicated that as many as 1,000 to 3,000 cancers annually may Be attributable to the air
toxics for which assessments were available (virtually all of this estimate came from assessments of about a
dozen well-studied pollutants).44
In addition to cancer, these pollutants can cause a wide variety of health effects, ranging from
respiratory problems to reproductive and developmental effects. There has been considerably less work
done to assess the magnitude of non-cancer effects from air toxics, but one survey study has shown that
some pollutants are present in the atmosphere at reference levels that have caused adverse effects in
animals.45
Emissions of air toxics can also cause adverse health effects via non-inhalation exposure routes.
Persistent bioaccumulating pollutants, such as mercury and dioxins, can be deposited into water or soil and
subsequently taken up by living organisms. The pollutants can biomagnify through the food chain and
exist in high concentrations when consumed by humans in foods such as fish or beef. The resulting
exposures can cause adverse effects in humans, and can also disrupt ecosystems by affecting top food
chain species.
Finally, there are a host of dfcher potential ecological and welfare effects associated with air toxics, for
which very little exists in'the way of quantitative analysis. Toxic effects of these pollutants have the
potential to disrupt both terrestrial and aquatic ecosystems and contribute to adverse welfare effects such as
fish consumption advisories in the Great Lakes.46 ,
Unfortunately, the effects of air toxics emissions reductions could not be quantified for the present
study. Unlike criteria pollutants, there was relatively little monitoring data for available air toxics, and that
which exists covered only a handful of pollutants. Emissions inventories were very limited and
inconsistent, and air quak'ty modeling has only been done for a few source categories. In addition, the
scientific literature on the effects of air toxics was generally much weaker than that available for criteria
pollutants.
44 U.S. EPA, Cancer Risk &om Outdoor Exposure to Air Toxics. EPA-450/l-90-004fc Prepared by EPA/OAR/OAQPS.
45 U.S. EPA, "Toxic Air Pollutants and Noncancer Risks: Screening Studies," External Review Draft, September, 1990.
46 U.S. EPA, Office of Air Quality Planning and Standards. "Deposition of Air Pollutants to the Great Waters, First Report to Congress,"
May 1994. EPA-453/R-93-055.
i- " • . . •
47
-------
Chapter 5: Physical Effects
Limitations in the underlying data and analyses of air toxics led the Project Team to exclude the
available quantitative results from the primary analysis of CAA costs and benefits. The estimates of cancer
incidence benefits of CAA air toxics control which were developed, but ultimately rejected:; are presented
in Appendix H. Also found in Appendix H is a list of research needs identified by the Project Team which,
if met, would enable at least a partial assessment of air toxics benefits in future section 812 studies.
t nl _____ ,,,,,„ 1 *
and Welfare Effects of Hazardous' Air Pollutants." '
.f * M
v t,( i* I
Effect Category
Human Health
Human Welfare
Ecological
Other Welfare
Quantified Effects
Cancer Mortality
- nonutility stationary
source '
- mobile source
r
(
Unquantified Effects
1 .'
Cancer Mortality
-utility source
'•area source, „ "r
Noncancer effects >
-neurological , •-,
-respiratory
-reproductive !f
- hematopoietic
- developmental
- immunological
- organ toxicity
Decreased income and
recreation opportunities
due to fish advisories
Odors ,
Effects on wildlife
Effects on plants
Ecosystem effects
Loss of biological
diversity ,
i i
Visibility
Building Deterioration
j
Other Possible Effects
^
•, f"
. -" Vi' ,
N / " ; "'i
f
", < ' -,' ' ^
* ir ^
i •. f U ^
* f t i1-
1 \ «,
%! ' ,
i
>. >•* v i
/ * * •* -i jC
1 ' , 4*
Decreased income ,
resulting from decreased
physical performance
i "" , a
Effects on global climate
1 - •< ; ' , '
. > *
f » ^ i *
* t
( ^ i(
Loss of biological diversity
'(' <<"<,\?''
t , *
i <.* i
48
-------
Chapter 5: Physical Effects
1 Table 16. Uncertainties Associated with Physical Effects Modeling,
Potential Source of Error
.-
Extrapolation of health effects to
populations distant from monitors
(or monitored counties in the case
ofP,Iii}. ''
* ° •> ••„
Choice of CR function (i.e., -
"across-siudy" uncertainties) n
Uncertainty associated with CR >
functions derived from each
individual study^ie?,, "within
study" uncertainty) " '
Exclusion of potential substitution ,.
of ozone-resistant cultivars in
agriculture analysis. ,
Exclusion of other agricultural
effects (crops, pollutants) >"
Exclusion of effects on wetland and
aquatic ecosystems and forests.
No quantification of materials
damage , , -
Direction of Potential
Bias In Physical Effects
Estimate ,
Probable overestimate. ,
* ~ f ~+ ' "
*• \
Unknown.
Unknown.
Overestimate.
Underestimate.
Underestimate.
Underestimate
Significance Relative to Key
Uncertainties in Overall Monetary -
.Benefit Estimate '"
Probably minor. In addition, this
adjustment avoids the underestimation
which would result by estimating effects
for only those people' living near monitors.
Potential overestimate may result to the
extent air quality in areas distant from
monitors is significantly better than in
monitored areas. This disparity should be
quite minor for regional pollutants, such as
ozone and fine particulates.
Significant: The differences in implied
physical outcomes estimated by different
underlying studies are large. * '
" 1
Probably minor. '' '
Insignificant, given small relative .
contribution of quantified agricultural
effects to overall quantified benefit
estimate.
Unknown, possibly significant.
Unknown, possibly significant.
Unknown, possibly significant.
49
-------
50
-------
6
Economic Valuation
Estimating the reduced incidence of physical effects represents a valuable measure of health benefits
for individual endpoints; however, to compare or aggregate benefits across eridpoints, the benefits must be
monetized. Assigning a monetary value to avoided incidences of an individual effect permits a summation,
in terms of dollars, of monetized benefits realized as a result.of the CAA, and allows that summation to be
compared to the cost of the CAA.
For the present analysis of health and welfare benefits, valuation estimates were obtained from the
economic literature, and are reported in dollars per case reduced for health effects and dollars per unit of
avoided damage for welfare effects.47 Similar to estimates of physical effects provided by health studies,
each of the monetary values of benefits applied in this analysis is reported in terms of mean value and a
probability distribution around the mean estimate. The statistical form of the probability distribution used
for the valuation measures varies by endpoint. For example, while the estimate of the dollar value of an
avoided premature mortality isde^cribed by the Weibull distribution, the estimate for the value of a
reduced case of acute bronchitis Is assumed to be uniformly distributed between a minimum and maximum
value. --
Willingness-to-pay Estimates
*• • '
In benefit-cost analysis, the dollar value of an environmental effect on a person is the dollar amount
that is needed to compensate the person to make him or her just as well off as he or she would have been
absent the effect Theoretically, the dollar amount required to compensate a person for exposure to an
adverse effect should be about the same as the dollar amount a person is willing to pay to avoid the effect
Thus, economists speak of "willingness-to-pay" (WTP) as the appropriate measure of the value of avoiding
an adverse effect For example, the value of an avoided respiratory symptom would be a person's WTP to
avoid that symptom.48
47 The literature reviews and valuation estimate development process is described in detail in Appendix I and in the referenced supporting
reports. •
48 The "compensation principle? is used implicitly for cost calculations as well. In the case of the CAA, compliance cost represents the -
compensation required to elicit voluntary provision of labor and capital services, supplies, etc., to meet CAA requirements.
' ' • 51
-------
Chapter 6: Economic Valuation
For most goods, WTP can be observed by examining actual market transactions. For example, if a
gallon of bottled drinking water sells for one dollar, it can be observed that at least some persons are
willing to pay one dollar for such water. For goods that are not exchanged in the market, such as most
environmental "goods," valuation is not so straightforward. Nevertheless, value may be inferred from
observed behavior, such as through estimation of the WTP for mortality risk reductions based on observed
sales and prices of safety devices such as smoke detectors. Alternatively, surveys may be used in an
attempt to elicit directly WTP for an environmental improvement.
Table 17. Health and Welfare Meets UaH Valuation (1990 dollars).
Wherever possible, this
analysis uses estimates of the WTP
of the U.S. population to avoid an
environmental effect as the value of
avoiding that effect. In some cases,
such estimates are not available,
and the cost of mitigating or
avoiding the effect is used as a
tough estimate of the value of
avoiding the effect. For example,
if an effect results in
hospitalization, the avoided
medical costs were considered as a
possible estimate of the value of
avoiding the effect Finally, where
even the "avoided cost" estimate Is
not available, the analysis relies on
other available methods to provide
a rough approximation of WTP^
As noted above, this analysis uses a
range of values for most
environmental effects, or
endpoints. Table 17 provides a
summary of the mean unit value
estimates used in the analysis. The
full range of values can be found in
Appendix I.
Mortality
Some forms of air pollution
increase the probability that
individuals will die prematurely. ' r ' >\ ,, C ' " ,>
The concentration-response ••^•••^^•••••••••••••^^^••••••••^•^•••^^^•^•i
functions for mortality used in this
analysis express this increase in mortality risk as cases of "excess premature mortality" per time period
(e.g., per year). It is not possible to compensate the victims; thus, it is not possible to "value" the lives of
individual victims in a benefit-cost sense. One can, however, determine the WTP of individuals to accept
Endooint
Mortality
Chronic Bronchitis
IQ Changes '
Lost IQ Points
!
Hypertension
Hospital Admissions
Coronary Heart Disease
Atherothrombotic brain
infarction
Initial cerebrovascular accident
* Ischemic Heart Disease
Congestive Heart Failure
Respiratory
Respiratory Illness and Symptoms
Acute Bronchitis
Asthma Attacks
Acute Respiratory Symptoms
^ *. > ^ n r
Upper Respiratory Illness
Lower Respiratory Illness
Shortness of Breath
!
Work Loss Days
Restricted Activity Days
Welfare Benefits
Visibility
1
f ! t '
Household Soiling
i i.
Decreased Worker Productivity
Agriculture {Net Surplus)
Pollutant
PM10&Pb
"PM10 ^ 'ff.
Pb
,(
1 ! fi
Pb
!
Pb
Pb
PM10
f M
Mo
PM,o&03
ru10,o3tnoy
SQ2
PMW
*Ho
PM0
PMW ^ '
PM,0&0,
>
deciview
East
<, >
, West
PM,o
i ! ^
1 5
o,
03
Valuation (mean est.1 .
$4,800,000 percase
$587^500 percase1
*<•'•• ' \ i / > '
H $3,400 percase ' '
> $4X800 percase'
: $681 percase
, ,1 ? . y •>, „ i j
' ''• M ": *- -,", j
, $10,300 percase
$10,100 percase =
^ f t
! j f ) i K ^ ^ " i
^$7,600^ percase (
$lo,300 percase
' $8^00 percase
$7^400 percase
•i * »,
) ; » ^ J i^ !
$45 percase ;
$32 percase ^l-
$24 percase
i f i
•c 1' ' 'i 1 '
$19 percase
, l $12 percase
h< ( $S per day
° i $83 per day
$38 per day
• '> ,'! 'i ' c '<
^ $149 per % visual
$117 ranee cage.
i1" ' ' < '
' ' * J '"' ,v-
•< I f ^
$3 per household
^ , iperPMw
-------
Chapter 6: Economic Valuation
relatively small reductions in mortality risk (conversely, one can determine the compensation required for
individuals to accept small increases in mortality risk). Since air pollution affects mortality rates by
changing the mortality probability (per time period) for many individuals, it is standard practice in cost-
benefit analyses to value this effect of air pollution by examining the WTP for small reductions in mortality
risk. For expository purposes, this valuation is expressed as "dollars per mortality avoided," even though
the actual valuation is based on small changes in mortality risk.49
The mortality risk valuation estimate '
used in this study is based on an analysis of
26 policy-relevant value-of-life studies (see
Table 19). Five of the 26 studies are
contingent valuation (CV) studies, which
directly solicit WTP information from
subjects; the rest are wage-risk studies,
which base WTP estimates on estimates of
the additional compensation demanded in
the labor market for riskier jobs. The
Project Team used the best estimate from
each of the 26 studies to construct a
distribution of mortality risk valuation
estimates for the Section 812 study. A
Weibull distribution, with a mean of $4.8
million and standard deviation of $3.24 million,
provided the best fit to the 26 estimates.
There is considerable uncertainty associated
with this approach, however, which is
discussed in detail later in this chapter.
Table 18. Summary of Mortality Valuatioa Estimates*
millions of $199Q>
Survey-Based Values
Willingness-to pay for environmental
improvement is often elicited through survey
methods (such as the "contingent valuation"
method). Use of such methods in this
context is controversial 'within the
economics profession. In general,
economists prefer to infer WTP from
observed behavior. There are times when
such inferences are impossible, however, ^
and some type of survey technique may be
the only means of eliciting WTP.
Economists' beliefs regarding the reliability of such survey-based data cover a broad spectrum, from
unqualified acceptances of the results of properly-conducted surveys to outright rejections of all survey-
based valuations.
+ Stady ' '
if 1
Kneisher and Leeth (1991) (OS)
Smith and Gilbert (1984)
DiUingham(1985)
Butler (1983)
Mfller and Curia (1991)
Moras and Viscusi (1988a) ,
Viscusi, Magat, and Huber (1991b)
Ge«ax«taL(1985)
Mann and Psaenaropouks (1982)
Kneisner and Leeth (1991) (Australia)
Gerking, de Haan, and Schulze (1988)
Cousineau, Lacroix, and Girard (1988)
Jones-Lee (1989)
Diliingham (1985)
Viscusi<1978,1979y
R.S. Smith (1976)
V.K Smith (1976)
OJson<1981)
VJscasi(I981>
R.S. Smith. (1974)
Moore and Viscusi (1988a) '
Kneisnerand Leeth (1991) (Japan)
Herzog and Schtottman {1987)
Leigh and Fokon (1984)
Leigh (1987)
Gaten(1988)
l>peof
Estimate
> <•
Labor Market
Labor Market
Labor Market
Labor Market
Cont Value
Labor Market
Valuation
(millions
" 1990$L_
-0.6 '
O7S
' 0.9
UL
- 12
" -2LS\
Cont, Value \\ 23
Coat-Value ,
Labor Market ;
Labor Market
Cont Value '
Labor Market ,
Cont Value
Labor Market
Labor Market
Labor Market
Labor Market
3.S
. 2.8"
33
3.4
3.6 '
3^
'3.9
4.1 -
'-4.6
,4,7
Labor Market f 5.2"
Labor Market
Labor Market
Labor Market
Labor Market
Labor Market
Labor Market ,
Labor Market
6.5
, - 7^2
7.3
, " 7-6
94
^7
10.4
Labor Market . f .133
SOURCE Viscusi, 1992 ' " <- • «
49 For example, if the WTP for a one percent reduction in mortality risk is $50,000, then the "benefit per life saved" would be [$50,000 x 100
=] $5 million.
53
-------
Chapter 6: Economic Valuation
In this analysis, unit valuations which rely exclusively on the contingent valuation method are chronic
bronchitis, respiratory-related ailments, minor restricted activity days, and visibility. As indicated above,
the value derived for excess premature mortality stems from 26 studies, of which five use the contingent
valuation method. These five studies are within the range of the remaining 21 labor market studies. All
five report mortality valuations lower than the central estimate used in this analysis. Excluding the
contingent valuation studies from the mortality valuation estimate would yield a central estimate
approximately ten percent higher than the 4.8 million dollar value reported above. The endpoints with unit
valuations based exclusively on contingent valuation account for approximately 30 percent,of the present
value of total monetized benefits. Most of the CV-based benefits are attributable to avoided cases of
chronic bronchitis.
V. * i -.
Chronic Bronchitis
The WTP estimate used for avoided chronic bronchitis is based on studies that elicited from
respondents a willingness to pay for chronic bronchitis in terms of a willingness to trade the risk of chronic
bronchitis for the risk of a fatal automobile accident (a risk-risk tradeoff).50 The valuation of a change in
the risk of a fatal automobile accident is then appliecltoinfer a valuation for a case of chronic bronchitis
avoided. Using this method, four unit values havebeen suggesfeld:51 (1) a mean value of $883,000; (2) a
median value of $457,000 which captures the skewness of the response distribution; (3) $210,000, based
on the mean value, with an adjustment for the severity of the chronic bronchitis case; and (4) a value of
$800,000 derived from the mean risk-risk response, but adjusting for the skewness of automobile mortality
valuation by using the median value for automobile mortality, tor this analysis, the central estimate of the
value of avoiding a case of chronic bronchitis is taken to be the mean of the four suggested estimates,
which is $587,500. ' ,
The uncertainty surrounding this estimate is also derived from the range of estimates summarized
above. It is assumed that individuals are willing to pay one of the four valuation estimates to avoid chronic
bronchitis. Without further information, it-is not possible to weight the proportion of individuals that
would select each alternative value. Therefore, four discrete values, each with equal probability of
selection, describe the uncertainty distribution associated with valuing this endpoint.
Respiratory-Related Ailments
In general, the valuations assigned to the respiratory-related ailments listed in Table 17 represent a
combination of willingness to pay estimates for individual symptoms which comprise each ailment. For
example, a willingness to pay estimate to avoid the combination of specific upper respiratory symptoms
defined in the concentration-response relationship measured by Pope et al. (1991) is not available.
However, while that study defined upper respiratory symptoms as one suite of ailments (runny or stuffy
nose; wet cough; and burning, aching, or red eyes), the valuation literature reported individual WTP
estimates for three closely matching symptoms (head/sinus congestion, cough, and eye irritation). The
available WTP estimates were therefore used as a surrogate to the values for the precise symptoms defined
in the concentration-response study.
30 Vtsojsi et aL (1991) and Krapnfck and Cropper (1992).
31 Review of Clean Mr Act Section 812 Retrospective Study of Costs and Benefits, Report of the Advisory Council on Clean Air Compliance
Analysis (EPA-SAB-ACCACA-9frO03), June 1996. ,
54
-------
Chapter 6: Economic Valuation
To capture the uncertainty associated with the valuation of respiratory-related ailments, this analysis
incorporated a range of values reflecting the fact that an ailment, as defined in the concentration-response
relationship, could be comprised of just one symptom or several. At the high end of the range, the
valuation represents an aggregate of WTP estimates for several individual symptoms. The low end
represents the value of avoiding a single mild symptom.
-. t
Minor Restricted Activity Days
An individual suffering from a single severe or a combination of pollution-related symptoms may
experience a Minor Restricted Activity Day (MRAD). Krupnick and Kopp (1988) argue that mild
symptoms will not be sufficient to result in a MRAD, so that WTP to avoid a MRAD should exceed WTP
to avoid any single mild symptom. On the other hand, WTP to avoid a MRAD should not exceed the
WTP to avoid a work; loss day (which results when the individual experiences more severe symptoms). No
studies are reported to have estimated WTP to avoid a day of minor restricted activity. Instead, this analysis
uses an estimate derived from WTP estimates for avoiding combinations of symptoms which may result in
a day.of minor restricted activity ($38 per day). The uncertainty range associated with this value extends
from the highest value for a single symptom to the value for a work loss day. "Furthermore, the distribution
acknowledges that the actual value is likely to be closer to the central estimate than either extreme.
Visibility
The value of avoided visibility impairment is derived from various studies of the WTP to improve
visibility. These studies estimated WTP values in different cities in the U.S. The values used to monetize
the measured reductions in visibility were found to vary between the Eastern and Western United States
($149 and $117 per percent change in visual range, respectively).
Avoided Cost Estimates
For some health effects, WTP estimates are not available, and the Project Team instead used "costs
avoided"~as a substitute for WTP. Avoided costs were used to value the following endpoints:
hypertension, hospital admissions, and household soiling.
Hypertension and Hospital Admissions
^/ . -
Avoided medical costs and lost work time were used to value hypertension (high blood pressure) and
hospital admissions (this includes hospital admissions for respiratory ailments as well as heart disease,
heart attacks, and strokes). This analysis considers medical costs to be the sum of physician charges,
medication costs, and hospitalization costs. Lost work time is valued at the rate estimated for a work loss
day (see discussion below). Based on this approach, the value per year of avoiding a case of hypertension
is taken to equal the sum of medical costs per year plus work loss costs per year; the resulting value is $681
per case per year. For health effects which lead to hospitalization, the analysis uses the sum of medical
costs (per case) and work loss costs ($6,131 to $10,308 depending on the ailment for which hospitalization
is required).
Presumably, willingness-to-pay to avoid the effects (and treatment) of hypertension would reflect the
value of pain and suffering, and the value placed on dietary changes, etc. Likewise, the value of a health
"..--._ • . - . ' ' I - "
. . • 55
-------
Chapter 6: Economic Valuation
effect that would require hospitalization or doctor's care would include pain and suffering and lost leisure
time, in addition to medical costs and lost work time. Consequently, the valuations for these endpoints
used in this analysis likely represent lower-bound estimates of the true social values for avoiding such
health effects. "
Household Soiling
This analysis values benefits for this welfare effect by considering the avoided costs of cleaning houses
due to particulate matter soiling. The Project Team's estimate reflects the household's annual cost of
cleaning per fig/&? particulate matter ($2.52). Considered in this valuation are issues «uch as ttie nature of
the particulate matter, and the proportion of households likely to do the cleaning themselves. Since the
avoided costs of cleaning used herein do not reflect the loss of leisure time (and perhaps work time)
incurred by those who do their own cleaning, the valuation function likely underestimates true WTP to
avoid additional soiling.
Of her Valuation Estimates
Changes in Children's IQ
One of the major effects of lead exposure is permanently impaired cognitive development in children.
No ready estimates of society's WTP for improved; cognitive ability are currently available. Two effects of
IQ decrements can be monetized, however: reductions in expected lifetime income, and increases in
societal expenditures for compensatory education. These two effects almost certainly understate the WTP
to avoid impaired cognitive development in children., and probably should be considered lower bound
estimates. In the absence of better estimates, however, the Project Team has assumed that the two
monetized effects represent a useful approximation of WTP.
The effect of IQ on expected lifetime income comprises a direct and an indirect effect. The direct
effect is drawn from studies fliatjestimate, all else being equal, the effect of IQ on income. The indirect
effect is through education: there is a correlation between IQ and years of education, and between
education and expected future income. The two effects together imply a change hi expected lifetime
income (discounted to the present at 5 percent) of $3,400 per IQ point change. The calculated effect
overstates net income change because it does not account for the cost of incremental education. If one
were to include expected incremental education cost in the calculation, the indirect effect would be
lessened somewhat Given that the effect of IQ on lifetime earnings is almost certainly an underestimate of
the WTP to avoid impaired cognitive development in children, the Project Team did not reduce the
estimate to account for expected incremental education cost.
In this analysis, it is assumed that part-time compensatory education is required for all children with IQ
less than 70. The Project Team assumed that the WTP to avoid cases of children with IQ less than 70 can
be approximated by the cost ($41,800 per child) of part-time special education in regular classrooms from
grades one through twelve (as opposed to independent special education programs), discounted to the
present at 5 percent. See Appendix G for more detail on valuation methods and data sources for IQ effects
and other lead-related health impacts.
56
-------
Chapter 6: Economic Valuation
Work Loss Days and Worker Productivity
For this analysis, it was assumed that the median daily 1990 wage income of 83 dollars was a
reasonable approximation of WTP to avoid a day of lost work. Although a work loss day may or may not
affect the income of the worker, depending on the terms of employment, it doe's affect economic output
and is thus a cost to society. Conversely, avoiding the work loss day is a benefit.
* ,*••
A decline in worker productivity has been measured in outdoor workers exposed to ozone. Reduced"
productivity is measured in terms of the total daily income of the average worker engaged in strenuous
outdoor labor, estimated at 73 dollars per day (U.S. Bureau of the Census, 1990). -~
Agricultural Benefits
. . *•- ,*
Similar to the other welfare effects, the agricultural benefits analysis estimated benefits in dollars per
unit of avoided damage, based on estimated changes in crop yields predicted by an agricultural sector
model. This model incorporated agricultural price, farm poKcy, and other datafor each year. Based on
expected yields, the model estimated the production levels for each crop, and the economic benefits to
consumers, and to producers, associated with these production levels. To the extent that alternative
exposure-response relationships were available, a range of potential benefits was calculated.
Valuation Uncertainties
The Project Team attempted to handle most valuation uncertainties explicitly and quantitatively by
expressing values as distributions (see Appendix I for a complete description of distributions employed),
using a Monte-Carlo simulation technique to apply the valuations to physical effects (see Chapter 7) with
the mean of each valuation distribution equal to the "best estimate" valuation. This approach does not, of
course, guarantee that all uncertainties have been adequately characterized, nor that the valuation estimates
are unbiased. It is entirely possible that the actual WTP to avoid an air pollution-related impact is outside
of the range of estimates used in this analysis. Nevertheless, the Project Team believes that the
distributions employed are reasonable approximations of the ranges of uncertainty, and that there is no
compelling reason to believe that the mean values emloyed are systematically biased (except for the IQ-
related and avoided cost-based values, both of which probably underestimate WTP).
*" "-"It /?,!M~- " ' ' "
^> '' ., '
- ~ 'One particularly important area of uncertainty is valuation of mortality risk reduction. As noted below
(see Chapter 7), changes in mortality risk are a very important component of aggregate benefits, and
mortality risk valuation is an extremely large component of the quantified uncertainty. Consequently, any
uncertainty concerning mortality risk valuation beyond that addressed by the quantitative uncertainty
assessment (i.e., that related to the Weibull distribution with a mean value of $4.8 million) deserves note.
One issue merits special attention: uncertainties and possible biases related to the "benefits transfer" from
the 26 valuation source studies to valuation of reductions in PM-related mortality rates.
Mortality Risk Benefits Transfer *
Although each of the mortality risk valuation source studies (see Table 19) estimated the average WTP
for a given reduction in mortality risk, the degree of reduction in risk being valued varied across studies
and is not necessarily the same as the degree of mortality risk reduction estimated in this analysis. The
• ' • • -' ' 57 '-'.'.•''•
-------
Chapter 6: Economic Valuation
transferability of estimates of the value of a statistical life from the 26 studies to the Section 812 benefit
analysis rests on the assumption that, within a reasonable range, WTP for reductions in mortality risk is
linear in risk reduction. For example, suppose a study estimates that the average WTP for a. reduction in
mortality risk of 1/100,000 is 50 dollars, but that the actual mortality risk reduction resulting from a given
pollutant reduction is 1/10,000. If WIP for reductions in mortality risk is linear in risk reduction, then a
WTP of 50 dollars for a reduction of 1/100,000 implies a WTP of 500 dollars for a risk reduction of
1/10,000 (which is ten times the risk reduction valued in the study). Under the assumption of linearity, the
estimate of the value of a statistical life does not depend on the particular amount of risk reduction being-
valued.
Although the particular amount of mortality risk reduction being valued in a study may not affect the
transferability of the WTP estimate from the study to the benefit analysis, the characteristics of the study
subjects and the nature of the mortality risk being valued in the study could be important. Certain
characteristics of both the population affected and the mortality risk facing that population are believed to
affect the average WTP to reduce risk. The appropriateness of the mean of the WTP estimates from the 26
studies for valuing the mortality-related benefits of reductions in pollutant concentrations therefore
depends not only on the quality of the studies (i.e., how well they measure what they are trying to
measure), but also on (1) the extent to which the subjects in the studies are similar to the population
affected by changes in air pollution and (2) the extent to which the risks being valued are similar.
The substantial majority of the 26 studies relied upon are wage-risk (or labor market) studies.
Compared with the subjects in these wage-risk studies, the population most affected by air pollution-
related mortality risk changes is likely tb be, on average, older and probably more risk averse. Some
evidence suggests that approximately 85 percent of those identified in short-term ("episodic") studies who
die prematurely from PM-related causes are over 65i52 The average age of subjects in wage-risk studies, in
contrast, would be well under 65. '• • '
The direction of bias resulting from the age difference is unclear. It could be argued that, because an
older person has fewer expected years teft to lose, his or her WTP to reduce mortality risk would be less
than that of a younger person. This hypothesis is supported by one empirical study, Jones-Lee et al.
(1985), which found WTP to avoid mortality risk at age 65 to be about 90 percent of what it is at age 40.
On the other hand, there is reason to believe that those over 65 are, in general, more risk averse than the
general population, while'workers in wage-risk studies are likely to be less risk averse than the general
population. Although the list of 26 studies used here excludes studies that consider only much-higher-
than-average occupational risks, there is nevertheless likely to be some selection bias in the remaining
studies — that is, these studies are likely to be based on samples of workers who are, on average, more risk-
loving than the general population. In contrast, older people as a group exhibit more risk averse behavior.
There is substantial evidence that the income elasticity of WTP for health risk reductions is positive
(although there is uncertainty about the exact value of this elasticity). Individuals with higher incomes (or
greater wealth) should, then, be willing to pay more to reduce risk, all else equal, than individuals with
lower incomes or wealth. One could therefore argue that because the elderly are relatively wealthy, the
affected population is also wealthier, on average, than are the wage-risk study subjects, who tend to be
middle-aged (on average) blue-collar workers. It is possible, however, that among the elderly it is largely
the poor elderly who are most vulnerable to air pollution-related mortality risk (e.g., because of generally
See Schwartz and Dockeiy (1992), Ostro et aL (1995), and Chestnut (1995).
58
-------
Chapter 6: Economic Valuation
poorer health care). If this is the case, the average wealth of those affected by a reduction in pollutant
concentrations relative to that of subjects in wage-risk studies is uncertain.
Although there may be several ways in which job-related mortality risks differ from air pollution-related
mortality risks, the most important difference may be that job-related risks are incurred voluntarily whereas air
pollution-related risks are incurred involuntarily. There is some evidence (see, for example, Violette and
Chestnut, 1983) that people will pay more to reduce involuntarily incurred risks than risks incurred
voluntarily. If this is the case, WTP estimates based on wage-risk studies may be downward biased estimates
of WTP to reduce involuntarily incurred air pollution-related mortality risks. ' "
,J • ' f „
Finally, another important difference related to the nature of the risk may be that 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^erspnal 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.
Table 19,, EsfimatingMortalityitfstfBaseS on Wage-Risk
Studies: Potential Sources and Likely Direction of Bias.
The potential sources of bias in an
estimate of WTP to reduce the risk of air
pollution-related mortality based on wage-
risk studies are summarized in Table 19.
The need to adjust wage-risk-based WTP
estimates downward because of the likely
upward biasffitroduced by the age
'•'•'- ^^w^^-^" •-• *£^_ j" °
discrepancy has received significant attention
(Cfoesmuyi995;iEc, 1992). ITtheage
difference were the"only difference between
the population affected by air quality changes
and the subjects in the wage-risk studies,
there might be some justification for trying to
adjust the valuation for an excess premature
mortality downward. Even in this case,
' however, .the degree of the adjustment would
be unclear. There is good reason to suspect, ^ *
however, that there are biases in both •^^^^••••^•••^••••'i, ' "^
directions, as shown in the table above.
Because in each case the extent of the bias is unknown, the overall direction of bias in the valuation
estimate is similarly unknown.
Factor
Age
Degree of Risk Aversion
Income
Risk Perception:
Voluntary vs. Involuntary
Catastrophic vs "
Protracted Death
Ijkdy Direction of Bias in WtF Estimate
Uncertain, perhaps upward
Downward "' '
Downward, if the elderly affected are a ,
random sample of the elderly;
Unclear, if the elderly affected are poor.
Downward '
t •* *• \
Uncertain, perhaps downward
, t ~-^
59
-------
60
-------
7
Uncertainty
This chapter presents a summary of the monetary benefits of the CAA from 1970 to 1990, compares
these with the corresponding costs, and explores some of the major sources of uncertainty in the benefits
estimates. Monetized economic benefits for the 1970 to 1990 period were derived by applying the unit
valuations discussed in Chapter 6 to the stream of physical effects estimated by the method documented in
Chapter 5. One goal was to retain the uncertainly associated Tvith the health and welfare effects estimates
and the uncertainty associated with the unit valuations.
Two sources of variability describe the uncertainty in the concentration-response (CR) relationships.
The first is the statistical uncertainty associated with the CR relationships reported in the health and
welfare effects literature. The studies quantifying relationships between air pollution levels and an adverse
health or welfare effect typically report a statistical disfinfeiation along with a mean value of the
concentration-response estimate. The distribution captures the inherent uncertainty in the reported estimate.
The second source of uncertainty lies in the choice of studies. Different published results reported in the
scientific literature typically do not report identical findings; in some instances the differences are
substantial. These differences reflect variability of the concentration-response relationship across studies
and are captured in the overall uncertainty by the method to combine physical effects with unit valuations
below. Finally, the uncertainty in the unit valuations is captured by the range of values derived for each
quantified endpoint. For example, the Weibull distribution surrounding the $4.8 million mean estimate of
the value of an excess premature mortality reflects the uncertainty associated with this health effect.
The Project team adopted a Monte Carlo approach to evaluate the combined uncertainties of the
quantified health and welfare effects estimates and the unit values associated with each effect. This
approach permitted the uncertainty associated with each component of the analysis to be retained through
the process of monetizing and aggregating benefits. The first step in the Monte Carlo technique was to
select randomly a study from the set of available studies for a particular health effect. Each study provides
an estimate of the CR function and its standard deviation.' In the second step, a normal statistical
distribution was derived based on the mean and standard deviation of the coefficient associated with the
selected study. A specific value for the coefficient was then randomly selected from the derived normal
distribution. This two-step procedure was repeated to produce a large number of estimates of the
concentration-response relationship for the given endpoint Because the value of the coefficient varied
from iteration to iteration due to the two random selection procedures described above, the predicted health
effects varied as well. The Monte Carlo method thus produced a distribution of predicted health effect
outcomes due to the change in air quality between the control and no-control scenarios.
61
-------
Chapter 7: Aggregate Results and Uncertainty
Concurrently, the Monte Carlo technique was used to estimate the range of monetary benefits
associated with the avoided health and welfare effects. The monetary values assigned to each health or
welfare endpoint were also measured with considerable uncertainty, which was represented by the
distribution of monetary values derived for each endpoint. Each iteration of the Monte Carlo assessment
picked a specific monetary value of the endpoint from the assigned distribution of values. The random
draws of the health or welfare effect incidence and the monetary value of that effect were multiplied to
yield an estimate of monetary benefits for changes in the given effect for that Monte Carlo iteration.
Repeating the process many times generated a distribution of estimated monetary benefits by endpoint..
Combining the results for the individual endpoints using the Monte Carlo procedure yielded a range of
total estimated monetary benefits for each target year (1975,1980,1985 and 1990). This technique
enabled a representation of uncertainty in current scientific and economic opinion in these benefits
estimates. - . •
Aggregate Monetary Benefits
Table 20 presents monetary benefits for each quantified and monetized health and welfare endpoint (or
group of endpoints), aggregated from 1970 to 1990. The mean estimate resulting from the Monte Carlo
simulation is presented, along with the measured credible range (upper and lower fifth percentiles of the
distribution). Aggregating the stream of monetary benefits across years involved compounding the stream
of monetary benefits estimated for each year to the 1990 present value (using a 5 percent discount rate).
Since Monte Carlo modeling was accomplished only for the four target years, estimates for intervening
years were interpolated based on the trend of benefits estimated by combining only the central estimates of
the concentration-response relationship for each endpoint with the central estimate of the unit valuation.
While the monetary benefits estimated using these "point" estimate values did not attempt to capture
uncertainty, they did provide an approximation of the temporal trend of expected benefits associated with
the year by year changes in air quality. The mean values and the credible range for each endpoint in the
intervening years were calculated based on the ratios of the 5th percentile, mean, and 95th percentiles to .
the point estimate in the target years. These ratios (which were fairly consistent) were linearly interpolated
and applied to the point estimate of benefits in the intervening years. For example, point estimates of
monetary benefits were generated for 1976 through 1979 based on interpolation of the air quality estimates
for those years. The Monte Carlo results were used to calculate ratios of the 5th percentile, mean, and 95th
percentile to the point estimate for the target years 1975 and 1980. Linearly interpolating these ratios and •
applying them to the point estimates of benefits in the intervening years, yielded a stream of benefits for all
years described by the mean values and credible ranges. Finally, the monetary benefits for each year were
adjusted to their equivalent 1990 value and summed.
62
-------
Chapter 7: Aggregate Results and Uncertainty
' Table 20, Present Value ofj970 to 1990 Monetized Benefits by Endpoint Categoiy for 48 State Population and"
of $1990, discounted to 1990 at 5 percent) " v ,
, ' ' Eridpoint " " Pollutantfs)
Mortality . . 'l v > - , PM-10\
Mortality , ' ' . 3 „» ~ Pb
j -> •" *• ^ ^-*
Cbronic Bronchitis ',/ ', "TM-10
[Q (Lost IQ Pts. -f '- TOTAL nBillibns of 1990-value "dollars^
Present Value
Sth%ile
$1^.991
$125
, $2,143'
$299
^$77
'$18
'$31
x
$7
"'$55
'$11
$10,500
Mean
$13^42
( $1,550
$7,156'
$466
* $99
$24
$76
A
». $76
$72
$23'
95th %ue
> $30,968
' , $4,096
$1-2,613
$656
$120
$33
$149
$196
$91
$35
$230001 $40600
Table 20 offers a comparison of benefits by endpoint. When aggregating benefits across endpoints care
must be taken to avoid double-counting. In this analysis, double-counting was avoided by assuming that '
any endpoints which measured overlapping effects weretreated as alternatives (i.e., they were not treated
as additive). -To generate benefit estimates for the endpoint categories shown, the Project Team had to
combine certain endpoints to avoid double-counting. In particular, the mortality estimate associated with
PM10 represents^ pooling of the estimates derived from the short-term time series studies and the long-
term cohort study, applying equal weights to each study type. The potential for double-counting also arises
in the Hospital Admissions endpoint. This endpoint includes hospital admissions for all respiratory
ailments, which is treated as an alternative to the benefits estimated for COPD and Pneumonia admissions
combined. Likewise, since their definitions of symptoms overlap, acute bronchitis is treated as an
alternative to the combination of upper and lower respiratory symptoms. All these effects are combined
into the Respiratory-Related Symptoms, Restricted Activity, & Decreased Productivity endpoint category
in Table 115. Finally, the potential for double-counting of the effects of restricted activity was addressed
by combining benefits estimates of "restricted activity days" of various degrees of severity into a single
Benefit category.
TTable 20 also reports the estimated total national monetized benefits attributed in this analysis to the
CAA'from 1970 to 1990. The benefits, valued in 1990 dollars, range from $10.5 to $40.6 trillion with a
central estimate of $23 trillion. This analysis depended on the Monte Carlo technique to aggregate
monetary benefits across endpoints. For each of several thousand iterations, a random draw of the
monetized benefits for each endpoint was selected from the distributions summarized in Table 22 and the
individual endpoint estimates were then summed. This resulted in the distribution of total national
monetized benefits reported above.
63
-------
Chapter 7: Aggregate Results and Uncertainty
The temporal pattern of benefits during the
1970 to 1990 period is related to the difference in
emissions between the control and no-control
scenarios, and magnified by population growth,
during that period. As illustrated by Figure 21,
quantified annual benefits increased steadily
during the study period, with the greatest
increases occurring during the late 1970s. The
mean estimate of quantified annual benefits
grew from 364 billion dollars (expressed as
inflation-adjusted 1990 dollars) to 957 billion
dollars in 1980,1,204 billion dollars in 1985,
and 1,318 billion dollars in 1990.
Figure 21. Monte Carlo Simulation Model Results for
Target Years (in billions of 1990 dollars)
$2,500-
f $2,000-
i
|$1,500-
.$1,000-
| $500-
$o-
P|f
Hj45*%
1
•M
S^i
y -•"
•4 MAS
4 Mwn
PWW
i
i
4 08oi%
4 HMH
1
I
i
^ 96UI?
4 M«M
"'H- ril mi>%!-
4 «h% •— •* aw*
1975 1980 1985 1990
Table 21 presents summary
quantitative results for the retrospective
assessment Annual results are presented
for four individual years, with all dollar
figures expressed as inflation-adjusted,
1990 dollars. The final column sums the
stream of,costs and benefits from 1970
to 1990,'discounted (i.e., compounded)
to 1990 a&peicerilL' "Monetized
benefits" are the mean estimates reported
by the Monte Carlo benefits model. "Net
Benefits" are monetized benefits less
annualized costs for each year.
V . ,, , ,
Table 21. Monetized Annual Benefits and Cblsts, 1970-1990 (in
"billions of 1990-value dollars). '*,'''
Vlonetized Benefits
Annualized Costs (5%)
"fet Benefits
^ftnefit/nnsf ratio
1975
364
-14
1
350
%?/l
1980
957
x 21
'
936
dKft
1985
1.204
2?,
T ?*7O
48/1
,1990
1,318
26
x
1^92
-------
Chapter 7: Aggregate Results and Uncertainty
not estimated, the Project Team refrained from choosing as the primary cost measure one that included
second-order impacts, and instead employed "annualized costs."
Quantified Uncertainty and Sensitivities
Some degree of uncertainty is associated with every stage of the section 812^malysis, from data inputs
for compliance cost estimation to economic valuation of physical effects. For som&f actors, both the degree
of uncertainty and the direction of any potential bias are unknown; for some other factors, no employable
quantitative estimates are available, even though one would expect a non-zero result (for example,
quantitative estimates are unavailable for some human health effects, some human welfare effects/and all
ecological effects). Some important quantified factors, nevertheless, lend themselves to quantified
uncertainty assessments and/or sensitivity analyses. Some of those factors are addressed in this section.
Quantified Uncertainty in the Benefits Analysis
As previously indicated, a Monte Carlo simulation method was used to-aggregate the benefit estimates
generated by this assessment. The two primary sources of quantified uncertainty captured by the benefits
model are the variability in CR relationship (both in terms of statistical variance reported for each source
study, and of differences in CR relationships found in alternative studies), and the distribution of plausible
economic valuations53 of the quantified endpoints. The combination of variability in CR relationships and
in valuation estimates (along with chafiges in other point estimated variables such as population and air
quality) resulted hi a distribution of benefit estimates for each year of the period of analysis. In this
assessment, the benefits distribution is described as the mean value and a ninety percent probability
interval around the mean.
Figure 22. Distribution of 1990 Monetized Benefits of CAA (in
billions of 1990 dollars)
5th percentite = $600
$1,300
SSth percontile » $2,300
Figure 22 depicts the distribution
of monetary benefits for 1990 (similar
distributions were generated for other
years in the analysis period). The solid
vertical bars in the figure represent the
relative frequency of a given result in
the 1990 Monte Carlo analysis. The
largest bar, located above the
'VfllOOO'', indicates thaHnore Monte
Carlo iterations generated monetized
benefits of $900 billion to $1 trillion
than in any other $100 billion range
bin. This tallest bar, however, does not
represent the expected value of the
estimate for total monetized benefit for
1990; rather, the mean value of the
entire distribution, calculated to be
$1.3 trillion, provides this value. The
ninety percent confidence interval, a summary description of the spread of a distribution, is also noted in
the figure.
Illll.
ISthpercenWe
I Mean
v ' v v Av
9Sth pereenSW
Total Monetary Benefits ($ Billion*)
53 Valuation and uncertainty distributions are described in Appendix L
65
-------
Chapter 7: Aggregate Results and Uncertainty
& • «- > - ' * J ••, ' "«'/•' > ;'
Table 22. Quantified Uncertainty Ranges for Monetized Annual
^Benefits and Benefit/Cos* TRJatios, 1970-1990 (in billions of 1990-
value dollars). " ''
Annual benefit distributions were
calculated or inferred for the 1971-
1990 period. Table 22 summarizes the
calculated benefit ranges for 1975,
1980,1985, and 1990, and notes the
benefit/cost ratios implied by the
benefit ranges. The fifth percentile
present value computation'is the sum
of the 1990 present values of the
annual fifth percentile estimates. The
distribution of benefits changes little
(except in scale) from year-to-year:
The mean estimate is somewhat greater
than twice the fifth percentile estimate,
and the ninety-fifth percentile estimate
is somewhat less than twice the mean •••^^•^••••••^^^^•••••^^^^••••^••••••i
estimate. The distribution shape
changes little across years because the variables which induce the distribution spread (i,e., CR
function variability and economic valuation variability) are unchanged from year-to-year. Some
variability is induced by changes in relative pollutant concentrations over time, which then change the
relative impact of individual CR functions.
,
Monetized Benefits
5th percentile
Mean estimate
95th percentile
tanualized Costs (5%)
Benefit/Cost ratio
5th percentile
Mean estimate
1975
161
364- ,
65* ;
14"
J
, 12/1
26/1
4-7/1
1980
"431
957
1,710^
21'
\
21/1
46/1
1985
^ J % *
,558
1,204
2,075"
25
'f-S
'22/1
48/r
stt/i
1990
623
^318
2317
'26
^J Sr
^ ^ *I
24/f
51/1
«Q/1
PV
10^00
23,000
40,600
523
20/1
'44/1
PV=1990 present value reflecting compounding of costs and benefits
4* -tf/VIM *„_. tftf\t\ _« tf „ n--J.4^.y ^* lf! •** •* St "
*om 1971 to 1990 at S
Sources of Quantified Uncertainty'
Figure 23. Uncertainty Ranges Deriving From Individual Uncertainty
Factors
9 $«.
I $40
SK
• SIS
I110
Maan
I i
The estimated uncertainty
ranges for each endpoint category
summarized in Table 20 reflect the
measured uncertainty associated
with both avoided incidence and
economic valuationfThe Project
Team conducted a sensitivity;
analysis to determine the variables
with the greatest contribution to
the quantified uncertainty range.
The results of the uncertainty
analysis are illustrated in Figure
23. ' _ ., •";".• '
In this uncertainty analysis, all
the inputs to the Monte Carlo
uncertainty analysis are held
constant (at their mean values),
allowing only one variable — for ^^^^^^^u^^^^^^^^^^^^^^^^^^^^o^^^^^^m
example, the economic valuation
of mortality — to vary across the range of that variable's uncertainty. The sensitivity analysis then isolates
how this single source of variability contributes to the variation hi estimated total benefits. The first
uncertainty bar represents the credible range associated with the total monetary benefits of the Clean Air
Act, as reported above. This captures the multiple uncertainties in the quantified benefits estimation. The
66
-------
Chapter 7: Aggregate Results and Uncertainty
rest of the uncertainty bars represent the quantified uncertainty ranges generated by single variables. The
most important contributor to aggregate quantified uncertainty is mortality valuation and incidence,
followed by chronic bronchitis valuation and incidence.
Potential Sensitivities in the Cost-Benefit Analysis
Some uncertainties in the present analysis lend themselves to explicit quantitative analysis. In this
section, two analytical variables considered ex ante to have potentially significant influences on the
outcome of this analysis are evaluated. In addition, the potential for bias due to excluded second-order cost
and benefit impacts is discussed. *
-Table 23. Effect of Alternative Discount Rates OB Resent s
Value of Total Monetized' Benefits/Ctosts for 1970 far -
1990 (m trillions of 1990 dollars).
~ - • ' Discount rate "
-
vfean Estimated Benefits
tantalized Costs .
•let Benefits
Jenefit/Cost ratio
*
3%
19.9
0.4
' 19^
48^1
*•
"$%~
"23.0
0.5
22^
44/1
jf
7% *
26.7
W
26.0
41/1
V*
*
Discount Rates
Typically, the choice of discount rate has
an important effect on the results of a multiyear
benefit cost analysis. In this assessment, the
discount rate affects only three factors: IQ-
related benefits estimates (especially estimates of
changes in discounted lifetime income),
annualized costs (i.e., amortized capital
expenditures), and compounding of all costs and
benefits to 1990. Table 23 summarizes the effect
of alternative discount rates on the "best
estimate" results of this analysis. Because ,
monetized benefits exceed costsfor all years in
the analysis period, net benefits increase as the
discount rate increases. Because the annual B/C
ratio increases as one moves frdm 1970 toward 1990 (see Table 21 above), benefit cost ratios decline as
the discount rate increases (because earlier periods are given greater weight). Overall, the cost/benefit
results of the assessment appear to be generally insensitive to the choice of discount rate.
PM-Related Mortality Valuation
~t TO ,- • , '
The economic benefits of avoided premature mortality associated with reduced concentrations of non-
lead criteria pollutants (using PM10 as a surrogate measure) represent the largest source of benefits
quantified in this assessment, and the PM-mortality incidence and valuation estimates are the largest
sources of quantified uncertainty in the Monte Carlo uncertainty analysis. The Project Team conducted two
additional an|Jyjf|||io test the sensitivity of the assessment results to changes in the mortality valuation
method. Both Bnsitivity analyses are discussed in greater detail in Appendix I.
Eighty to eighty-five percent of the cases of premature mortality estimated by some "short-term
exposure" mortality studies involve persons 65 years of age or older. The age profile of the subjects in the
studies used as sources for the mortality valuation estimates is quite different, with an average subject's
age of 40 years or so. If WTP to avoid changes in mortality risk is correlated with age, then an adjustment
to the unit value used in this assessment would be appropriate (at least for mortality estimates based on the
short-term exposure epidemiological studies). Some have suggested that the WTP for reduced mortality
risk of persons over the age of 65 is approximately 75 percent of the WTP of the remainder of the
population. As discussed in Chapter 6 (and in greater detail in Appendix I), EPA believes that the state of
the science has not yet advanced sufficiently to provide broadly convincing evidence in support of age-
67
-------
Chapter 7: Aggregate Results and Uncertainty
specific adjustments to the mortality valuation estimates used as the central case results of this analysis.
Nevertheless, for illustrative purposes, the Project Team adjusted the mean mortality estimation using the
above method, which results in an adjusted benefits estimate that is approximately 79 percent of the central
case estimate (see Table 24).
• * * < , '
Table 24. Results of Mortality Benefits Sensitivity
-Analyses, Monetized Benefits for 197p,to 1990 (in
"trillions of 1990 dollars, discounted at 5'percent)il ,.
Benefit Estimation Method
Central analysis method
;$4.8M/case)
19% adjustment method
JYL: single relative risk
.YL: age-specific relative risk
PMT " Tot.
12.4 22.9
9.51 mo
Note; The first two methods employ averages of the
results of long-term exposure and short-term exposure
studies for mortality incidences. The LYL method
employs only the results of the long-term exposure
study. V; * ,.* ;' I * '"
A second alternative analysis involves changing
the approach taken for both the physical effects and
valuation estimates. Instead of expressing changes in
mortality risk as changes in the expected number of
premature mortalities per year, this method would
transform changes in mortality risk to changes in
expected life-span. Thus, increased air pollution levels
would result in "life years lost" (LYL) rather than
"excess premature mortalities." Similarly, the
estimates of WTP to avoid mortality risk would be
converted to estimates of WTP to avoid a lost life-year.
However, both the epidemiological science and
economic valuation science are inadequately advanced
to allow confident employment of the LYL mejicxj.
Nevertheless, to test the potential magnitude of the
effect that the LYL approach might have on
assessment results, the Project Team attempted to infer
LYL from the long-term PM exposure epidemiplippcal
study, and applied a crude estimate of 5WTP to avoid a
LYL (see Appendix I for greater detail on analytical method).,
'''*'"' ' " ' i . . •'*! • '
Table 24 summarizes the resjdlte of the two sensitivity analyses. Estimated 1970 to 1990 benefits from
PM-related mortality alone and tptkl assessment benefits are reported, in 1990 dollars discounted to 1990
at 5 percent. Two separate LYL resultsi arereported; the difference between the two is the product of
alternative assumptions used to Mer LYLfrbm the results of the epidemiological study. Note that the LYL
results are based on the single long-term PM exposure study used for this assessment, while the central
case and "79 percent adjusted" results are based on premature mortality incidence results averaged across
the long-term exposure study andithe pooled short-term exposure studies.
Second-Order Macroeconomic Impacts
As just mentioned in the section on benefits and costs, the cost measure used for this assessment does
not include second-order adverse macroeconomic impacts arising from compliance expenditures, even
though the magnitude of those impacts were estimated by the Project Team (see Chapter 2 and Appendix
A). It is unclear, however, whether consideration of indirect impacts would increase or decrease overall
cost impacts if one were to include second-order impacts arising from estimated benefits. For illustration,
Table 25 presents 1970 to 1990 compliance costs and some benefit measures which, one Would expect,
should induce second-order impacts of similar magnitudes. The direction of possible second-order impacts
is unclear.
68
-------
Chapter 7: Aggregate Results and Uncertainty
Compliance Costs, Some Avoided Costs, and
Productivity Improvements. 1970-1990 Cm billions of &990):
~ < "- , , - * J "
\nnuatized Compliance Costs (5%)
Avoided Cost estimates0 " ' > "
Q-refetedefifecfe^productiviQ') < .
yj avoided cost and productivity effects , * -•
PV* %*>
'523
275- - 53&
A '$86 " ' 89%
- 741 14^
a, Py=l§»0 present value teflecflng ccanpoufldtog of costs: aaTbeneefe mil
J97I to 1990 at596. J \ <*•'_/ -/
b.%= percentage of annualized cost estimate. , " '-,''''
c. Includes hypertension, hospital admissions, respiratory-related symptoms,
RA0S, decreased worker productivity, and soiling damage (from Table
20). Does not include avoided costs associated witfc other , „„
health endpoints (e.g., chronic bronchitis) ",
Several of the health and welfare
benefits estimated in this assessment are
expressed as "avoided costs." In the
aggregate, businesses and households
would view these avoided costs (if
unavoided) as being essentially identical in
effect to compliance expenditures, tax
treatment of capital expenditures aside (for
example, a business should be financially
indifferent between incurring 100 dollars
of additional pollution abatement
compliance cost versus 100 dollars of
additional health insurance cost). In this
assessment, benefits estimated as "avoided
cost" are about Vi as large as compliance
costs over the 1970 to 1990 period. On —
this evidence alone, one would expect to find that second-order impacts from benefits mitigate, but would
not completely offset, second-order impacts arising from compliance expenditures.
A second category of avoided cost was calculated. Some IQ-related benefits (here estimated as
increased expected lifetime income) are avoided societal costs insofar as they represent avoided decreases
in future production (here referred to asja "productivity" effect). Here, these benefits are presented
separately from the "avoided cost" estimates for two reasons. First, they differ conceptually from the other
avoided cost categories, and it is not so clear (when considering impact on second-order effects) that they
are analogous to compliance cost. Second, the IQ~ efiec^alculations differ from the other benefit
calculations, since they are lifetime benefits discounted to the time of exposure. That is, unlike other
benefit categories and the annualized cost estimates, some of the estimated IQ-related benefits actually
accrue aft«arl99p.t)nly some of the IQ-related benefits, then, would accrue during the 1970 to 1990 period
and would then induce second-order macroeconomic effects which would offset the compliance cost-
induced effects. Calculated IQ-related benefits are roughly 90 percent of compliance cost estimates; some
portion of the total would offset compliance cost-induced second-order macroeconomic effects. It is
unclear whether this result, combined with the avoided cost impact discussed above, would be enough to
completely offset the second-order impacts of compliance costs.
Of course, the other benefit categories would also be expected to induce second-order effects (e.g.,
improved longevity andjpcreased chronic bronchitis incidence would affect labor supply). Furthermore,
although not counted litre as "avoided cost," the benefit estimates for all other health effects include some
amount of "avoided cost." In some cases (chronic bronchitis, for example), one would expect the "avoided
cost" component of the benefits estimate to be substantial relative to aggregate compliance cost.
Finally, environmental regulation can, in some instances, spur rather than impair technological
innovation. Technological improvements motivated by regulatory requirements, such as more efficient
stack scrubbers and low emission diesel engines, can yield beneficial secondary economic effects through
energy savings, increased export earnings, recovery of valuable materials (e.g., solvents), and reduced
waste disposal costs.
It is unclear at this time given current evidence, however, whether the net effect on the cost estimates
of including second-order macroeconomic impacts would be positive or negative. In any case, even if only
the adverse second-order economic effects were included, the resulting increase in the cost estimates would
69
-------
Chapter 7: Aggregate Results and Uncertainty
not significantly affect the bottom line outcome that the monetizable benefits of 1970 to 1990 clean air
programs exceeded costs by a substantial margin.
70
-------
Macroeconomic Modeling
The purpose of this appendix is to describe in detail the estimation of direct compliance costs
associated with the CAA and the effect of those expenditures on U.S. economic conditions from 1970 to
1990. The first section of this appendix describes the dynamic, general equilibrium macroeconomic model
used to examine economy-wide effects. Two broad categories of models were considered for use in. the
assessment: Macroeconomic forecasting models (e.g., the Data Resources Inc. model of the U.S.
economy), and general equilibrium models (e.g., Hazilla and Kopp [1990], and Jorgenson and Wilcoxen
[1990a]). The project team selected the Jorgenson-Wilcoxen(J/W) general equilibrium model of the
United States for this analysis (Jorgenson and Wilcoxen [1990a]) There are two main reasons for
choosing a dynamic general equilibrium approach: To capture both the direct and indirect economic effects
of environmental regulation, and to capture the long-run dynamics of the adjustment of the economy. The
general equilibrium framework enabled the project team to assess shifts in economic activity between
industries, including changes in distributions of labor, capital, and other production factors within the
economy,' and changes in the distribution of goods and services.
' •' '• ftf
The second section describes the data sources for direct compliance expenditures and presents :
estimates of historical ak pollution control expenditures. These estimates are derived primarily from
EPA's 1990 report entitled "EnvironmentalInvestments: The Cost of a Clean Environment"54 (hereafter
referred to as Cost of Clean). Specific adjustments to the Cost of Clean stationary source and mobile
source O&M data needed to adapt these data for use in the present study are also described. These
adjusted expenditure estimates represent the compliance cost data used as inputs to the J/W model to
determine macroeconomic effects. .
The final section presents a summary of the direct expenditure data, presents direct costs in a form that
can be compared to the benefits estimates found elsewhere in the study, and discusses indirect effects
arising from compliance expenditures estimated by the macroeconomic model. The indirect effects
reported by the model are sectoral impacts and changes in aggregate measures of economic activity such as
household consumption and gross national product. These indirect effects are second-order impacts of
compliance expenditures ~ a parallel modeling exercise to estimate second-order economic impacts arising
from the benefits of compliance (e.g., increased output as a result of improved longetivity or fewer
workdays lost as a result of non-fatal heart attacks) has not been attempted.
Macroeconomic Modeling
EPA analyses of the costs of environmental regulations typically quantify the direct costs of pollution
abatement equipment and related operating and maintenance expenses. However, this approach does not
fully account for all of the broader economic consequences of reallocating resources to the production and
use of pollution abatement equipment. A general equilibrium, macroeconomic model could, in theory,
capture the complex interactions between sectors in the economy and assess the full economic cost of air
54 Environmental Investments: The Cost of a Clean Environment, Report of the Administrator of the Environmental Protection Agency to the
Congress of the United States, EPA-230-11-90-083, November 1990.
71
-------
Appendix A: Cost and Macroeconomic Modeling
pollution control. This would be particularly useful for assessing regulations that may produce significant
interaction effects between markets. Another advantage of a general equilibrium, macroeconomic
framework is that it is internally consistent. The consistency of sectoral forecasts with realistic projections
of U.S. economic growth is ensured since they are estimated within the context of a single model.55 This
contrasts with typical EPA analyses that compile cost estimates from disparate sectoral and partial
equilibrium models.
The economic effects of the CAA may be over- or underestimated, if general equilibrium effects are
ignored, to the extent that sectors not directly regulated are affected. For example, it is well known that the
CAA imposed significant direct costs on the energy industry. Economic sectors not directly regulated will
nonetheless be affected by changes in energy prices. However, an examination of the broader effects of
the CAA on the entire economy might reveal that the CAA also led to more rapid technological
development and market penetration of environmentally "clean" renewable sources of energy (e.g.,
photovoltaics). These effects would partially offset adverse effects on the energy industry, and lead to a
different estimate of the total economic cost to society of the CAA.
The significance of general equilibrium effects in the context of any particular analysis is an empirical
question. Kokoski and Smith (1987) used a computable general equilibrium model to demonstrate mat
partial-equilibrium welfare measures can offer reasonable approximations of the true welfare changes for
large exogenous changes. In contrast, the results of Jorgenson and Wilcoxen (1990a) and Hazilla and
Kopp (1990) suggest that total pollution abatement in the U.S. has been a major claimant on productive
resources, and the effect on long-run economic growth may be significant. Again, such conclusions must
be considered in light of the limitations of general equilibrium models.
Choice of Macroeconomic Model
The adequacy of any model or modeling approach must be judged in light of the policy questions
being asked. One goal of the present study is to assess the effects of clean air regulations on
macroeconomic activity. Tvyo broad categories of macroeconomic models were considered for use in the
assessment: short run, Keynesianmodels and long-run, general equilibrium models.
Recognizing that structural differences exist between the models, one needs to focus in on the
particular questions that should be answered with any particular model. The Congressional Budget Office
(1990)noted:
"Both the [Data Resources Incorporated] DRI and the IPCAEO models show relatively limited ,
possibilities for increasing energy efficiency and substituting other goods for energy in the short run...
Both models focus primarily on short-term responses to higher energy prices, and neither is very good
atjexamining how the structure of the economy could change in response to changing energy prices.
The [Jorgenson-Wilcoxen] model completes this part of the picture..."56
55 In the present study, both benefits and costs are driven off of the same macroeconomic projections from the Jorgenson/Wilcoxen model, to
ensure that the estimates are based on a consistent set of economic assumptions.
54 The Congressional Budget Office report (1990) refers to an older (1981) version of the Jorgenson model, not the current (1988) version.
The approach to long-run dynamics differs between the two models. The newer Jorgenson-Wilcoxen model contains both the capital accumulation
equation and the capital asset pricing equation. The 1981 version of the model contained only the capital accumulation equation.
72
-------
Appendix A: Cost and Macroeconomic Modeling
One strategy for assessing the macroeconomic effects of the CAA would be to use a DRI-type model
in conjunction with the Jorgenson-Wilcoxen model to assess both the long-term effects and the short-run
transitions, in much the same way that the Congressional Budget Office used these models to assess the
effects of carbon taxes. However, because of significant difficulties in trying to implement the DRI model
in a meaningful way, the project team chose to focus on the long-run effects of the CAA." Structural
changes (e.g., changes in employment in the coal sector due to the CAA) can be identified with the
Jorgenson-Wilcoxen model. , -
Overview of the Jorgenson-Wilcoxen Model
The discussion below focuses on those
characteristics of the Jorgenson-Wilcoxen model that
have important implications for its use in the
assessment of environmental regulations (see Table
26). The J/W model is a detailed dynamic general
• equilibrium model of the U.S. economy designed for
medium run analysis of regulatory and tax policy
(Jorgenson and Wilcoxen [1990a]). It provides
projections of key macroeconomic variables, such as
GNP and aggregate consumption, as well as energy
flows between economic sectors. As a result, the
model is particularly useful for examining how the
structure of the economy could change in response to
changes in resource prices. For the purpose of this
study, it has five key features: a detailed treatment of
production and consumption, parameters estimated '
econpmetrically form historical data, an endogenous
model of technical change, a rigorous representation
of saving and investment, and free mobility of labor
and capital between industries.
Table 26. Key Distinguishing Characteristics of the
JorgensotfWilcoxen Model/
Dynamic, general equilibrium,
macroeconomic model of the U.S,
econbmy. . , , t
• Econometrically estimated using historic •
Free mobility ofa single type 'of capital
and labor between industries. " "<
Detailed treatment of production and
consumption.
Rigorous representation of savings and
investment. " * ' }
Endogenous model of technical change.
Does not capture unemployment,
underemployment, of the costs bl moving
capital from one industry to another. , '
The first two features, industry and consumer detail and econometric estimation, allow the model to
capture the effects of the CAA at each point in time for given levels of technology and the size of the
eoStiomy's capital stock. A detailed treatment of production and consumption is important because the
j||||pal effects of the Clean Air Act fell most heavily on a handful of industries. The J/W model divides
I^S^sS. production into 35 industries which allows the primary economic effects of the CAA to be
fcapturecL Econometric estimation is equally important because it ensures that the behavior of .households
and firms in the model is consistent with the historical record.
The model's second two features —its representations of technical change and capital accumulation-
complement the model's intratemporal features by providing specific information on how the Act affected
technical change and the accumulation of capital. Many analyses of environmental regulations overlook or
ignore intertemporal effects but these effects can be very important. Jorgenson and Wilcoxen (1990a)
suggests that the largest cost of all U.S. environmental regulations together was that the regulations
reduced the rate of capital accumulation.
•V. , '
The model's last feature, free mobility of a single type of capital and a single type of labor, is important
because it limits the model's ability to measure the short run costs of changes in policy. J/W is a full-
73
-------
Appendix A: Cost and Macroeconomic Modeling
employment model that describes the long-run
dynamics of transitions from one equilibrium to
another. Capital and labor are both assumed to be
freely mobile between sectors (that is, they can be
moved from one industry to another at zero cost) and
to be fully used at all times. Over the medium to
long run, this is a reasonable assumption, but in the
short run it is too optimistic. In particular, the model
will understate the short run costs of a change in
policy because it does not capture unemployment,
underemployment, or the costs of moving capital
from one industry to another. A single rate of return
on capital exists that efficiently allocates the capital
in each period among sectors. Similarly, a single
equilibrium wage rate allocates labor throughout the
economy.
Structure of the Jorgenson-Wilcoxen Model
The J/W model assesses a broad array of
economic effects of environmental regulations.
Direct costs are captured as increased expenditures
on factors of production —capital, labor, energy and
materials— that the various industries must make to
comply with the regulations, as well as additional
out-of-pocket expenditures that consumers must
make. Indirect costs are captured as general
equilibrium effects that occur throughout the
economy as tiie prices of factors of prpductibn
change (e.g., energy prices). Also, the rate of
technological change can respond to changes in the
prices of factors of production. Causing changes in
productivity (Jorgenson and Fraumeni, 1981).
The model is divided into four major sectors: the
business, household, government, and rest-of-the-
world sectors. The business sector is further
subdivided into 35 ministries (see Table 27).57 Each sector
produces a primary product, and some produce
secondary products. These outputs serve as inputs to
the production processes of the other industries, are
used for investment, satisfy final demands by the
household and government sectors, and are exported. "•
The model also allows for imports from the rest of
the world.
Table 27. Definitions of Industries Within the J/W
Model; ' ' "** V <' '
Industry .
Number Description
> " , i ' *
1 Agriculture, forestry, and „
1 fisheries, <1~ "' , ,
2 MetaloajJing „ ,
3 , Coai mining V ' .»
* 4 Crude petroleum and natural gas
5 Nonmetallic mineral mining
6 Construction ,
7 Food and kindred products
8 , Tobacco manufacturers
9 * > Textile mill: products
10 Apparel and other textile
i' " ? products ' "' '
11 , Lumber and wood products
12 Furniture and fixtures , ,
13 Paper and allied products .
14 Printing and publishing
15 Chemicals and allied products
- 16 Petroleum refining
17 Rubber and plastic products
18 Leather and leather products
, 19 Stone, clay, and glass products
, 20 , Primary metals k" \
21 Fabricated metal products
V 22 Machinery, except electrical
23 Electrical machinery <
24 Motor vehicles , ,'
25 Other transportation equipment
26 Instruments
27 Miscellaneous manufacturing
' 28 Transportation and warehousing
' 29 • Communication
30 Electric utilities
, 31 Gas utilities ' " '.
'32 Trade , , , -' • ;,
33 Finance, insurance, and real
estate ^ , 1 ' ,t
34 Other services
35 Government enterprises
a The 35 industries roughly correspond to a two-digit SIC code classification scheme.
74
-------
Appendix A: Cost and Macroeconomic Modeling
The Business Sector
The model of producer behavior allocates the value of output of each industry among the inputs of the
35 commodity groups, capital services, labor services, and noncompeting imports. Output supply and
factor demands of each sector are modeled as the results of choices made by wealth maximizing, price
taking firms which are subject to technological constraints. Firms have perfect foresight of all future prices
and interest rates. Production technologies are represented by econometrically estimated cost functions
that fully capture factor substitution possibilities and industry-level biased technological change.
•a—
Capital and energy are specified separately in the factor demand functions of each industry. The
ability of the model to estimate the degree of substitutability between factor inputs facilitates the
assessment of the effect of environmental regulations. A higlE degree of substitutability between inputs
implies that the cost of environmental regulation is low, while f low degree of substitutability implies high
costs of environmental regulation. Also, different types of regulations lead to different responses on the
part of producers. Some regulations require the use of specific types of equipment. Others regulations
restrict the use of particular factor inputs; for example, through restrictions on the combustion of certain
types of fuels. Both of these effects can change the rate" of productivity growth in an industry through
changes in factor prices. ~~
The Household Sector
In the model of consumer behavior, consumer choices between labor and leisure and between
consumption and saving are determined. A system of individual, demographically defined household
demand functions are also econometrically estimated. Household consumption is modeled as a three stage
optimization process. In the first stage households allocate lifetime wealth to full consumption in current
and future time periods to maximize intertemporal utility. Lifetime wealth includes financial wealth,
discounted labor income, and the imputed value of leisure. Households have perfect foresight of future
prices and interest rates. In the second stage, for each time period full consumption is allocated between
goods and services and leisure to maximize intratemporal utility. This yields an allocation of household's
time endowment Between the labor ma||e|(giving rise to labor supply and labor income) and leisure time
and demands for goods and services. In the third stage, personal consumption expenditures are allocated
among capital, labor, noncompeting imports and the outputs of the 35 production sectors to maximize a
subutility function for goods consumption. As with the business sector, substitution possibilities exist in
consumption decisions. The model's flexibility enables it to capture the substitution of nonpolluting
products for polluting ones that may be induced by environmental regulations. Towards this end,
purchases of energy and capital services by households are specified separately within the consumer
demand functions for individual commodities.
"~~ It is important to be clear regarding the notions of labor supply and demand within the J/W model, and
what is meant by "employment" throughout this report. Labor demands and supplies are represented as
quality-adjusted hours denominated in constant dollars. The labor market clears hi each period; the
quantity of labor services offered by households is absorbed fully by the economy's producing sectors.
However, inferences regarding the number of persons employed require information on labor quality and
work-hours per person over time and across simulations. Neither of these are explicitly modeled.
The Government Sector
The behavior of government is constrained by exogenously specified budget deficits. Government tax
revenues are determined by exogenously specified tax rates applied to appropriate transactions in the
. • •-•'•'.- 75
-------
Appendix A: Cost and Macroeconomic Modeling
business and household sectors. Levels of economic activity in these sectors are endogenously determined.
Capital income from government enterprises (determined endogenously), and nontax receipts (given
exogenously), are added to tax revenues to obtain total government revenues. Government expenditures
adjust to satisfy the exogenous budget deficit constraint.
The Rest-of-the-Woiid Sector
The current account balance is exogenous, limiting the usefulness of the model to assess trade
competitiveness effects. Imports are treated as imperfect substitutes for similar domestic commodities and
compete on price. Export demands are functions of foreign incomes and ratios of commodity prices in
U.S. currency to the exchange rate. Import prices, foreign incomes, and tariff policies are exogenously
specified. Foreign prices of U.S. exports are determined endogenously by domestic prices and the
exchange rate. The exchange rate adjusts to satisfy the exogenous constraint on net exports.
Environmental Regulation, Investment, and Capital Formation
Environmental regulations have several important effects on capital formation. At the most obvious
level, regulations often require investment in specific pieces of pollution abatement equipment. If the
economy's pool of savings were essentially fixed, the need to invest in abatement equipment would reduce,
or crowd out, investment in other kinds of capital on a dollar for dollar basis. On the other hand, if the
supply of savings were very elastic then abatement investments might not crowd out other investment at
all. In the J/W model, both the current account and government Budget deficits are fixed exogenously so
any change in the supply of funds for domestic investment must come from a change in domestic savings.
Because households choose consumption, and hence savings, to maximize a lifetime utility function,
domestic savings will be somewhat elastic. Tfius, abatement investment will crowd out other investment,
although not on a dollar for dollar basis.
••v ,-;; ." - .'(,."*. . '
The J/W assumption that the current account does not change as a result of environmental regulation is
probably unrealistic, but it is not at all clear that this biases the crowding out effects in any particular
direction. By itself, the need to invest in abatement capital would tend to raise U.S. interest rates and draw
in foreign savings. To the extent this occurred, crowding out would be reduced. At the same time,
however, regulation reduces theprofitability of domestic firms. This effect would tend to lower the return
on domestic assets, leading to a reduced supply of foreign savings which would exacerbate crowding out.
Which effect dominates is an empirical question beyond the scope of this study.
. In additional to crowding out ordinary investment, environmental regulation also has a more subtle
effect on the rate of capital formation. Regulations raise the prices of intermediate goods used to produce
new capital- This leads to a reduction in the number of capital goods which can be purchased with a given
pool of savings. This is not crowding out in the usual sense of the term, but it is an important means by
which regulation reduces capital formation.58
** Wilcoxcn (1988) suggests that environmental regulation may actually lead to a "crowding in" phenomenon. Wilcoxen examined the effects
of regulation at the firm level, and introduced costs into the model related to the installation of capital. He found that when firms shut down their
plants to install environmental capital, they take account of the adjustment costs and often concurrently replace other older capital equipment This
effect, however, is not captured in the current version of the Jorgenson-Wilcoxen model.
76
-------
Appendix A: Cost 'and Macroeconomic Modeling
The General Equilibrium
The J/W framework contains intertemporal and intratemporal models (Jorgenson and Wilcoxen
[1990c]). In any particular time period, all markets clear. This market clearing process occurs in response
to any changes in the levels of variables that are specified exogenously to the model. The interactions
among sectors determine, for each period, aggregate domestic output, capital accumulation, employment,
the composition of output, the allocation of output across different household .types, and other variables.
The model also produces an intertemporal equilibrium pajii from the initial conditions at the start of
the simulation to the stationary state. (A stationary solution for the model is obtained by merging the
intertemporal and intratemporal models.) The dynamics of th| J/W model have two elements: An
accumulation equation for capital, and a capital asset pricing iijuatioii: Changes in exogenous variables
cause several adjustments to occur within the model. First, the single stock of capital is efficiently
allocated among all sectors, including the household sector. Capital is assumed to be perfectly malleable
and mobile among sectors, so that the price of capital services in each sector is proportional to a single
1 capital service price for the economy as a whole. The value of capital services is equal to capital income.
The supply of capital available in each period is the result of past investment, i^e., capital at the end of each
period is a function of investment during the period and capital at the beginning of the period. This capital
accumulation equation is backward-looking and captures the effect of investments in all past periods on the
capital available in the current period.
The capital asset pricing equation specifies the price of capital services in terms of the price of
investment goods at the beginning and end of each period, the rate of return to capital for the economy as a
whole, the rate of depreciation, and variables describing the tax structure for income from capital. The
current price of investment godds incorporates an assumption of perfect foresight or rational expectations.
Under this assumption, the price of investment goods'in every period is based on expectations of future
capital service prices and discount rates that are fulfilled by the solution of the model. This equation for
the investment goods price in each time period is forward-looking.59
"*'•"' „ ",' ' ' ~
One way to characterize the J/W model —or any other neoclassical growth model— is that the short-
run supply of capital is perfectly inelastic^ since it is completely determined by past investment. However,
the supply of capital is perfectly elastic in the long run. The capital stock adjusts to the time endowment,
while the rate of return depends only on the intertemporal preferences of the household sector.
, A predetermined amount of technical progress also takes place that serves to lower the cost of sectoral
production. Finally, the quality of labor is enhanced, giving rise to higher productivity and lower costs of
'production.
' Given all of these changes, the model solves for a new price vector and attains a new general
equilibrium. Across all time periods, the model solves for the time paths of the capital stock, household
consumption, and prices. The outcomes represent a general equilibrium in all time periods and in all
markets covered by the J/W model.
59 The price of capital assets is also equal to the cost of production, so that changes in the iate of capital accumulation result in an increase in
the cost of producing investment goods. This has to be equilibrated with the discounted value of future rentals in order to produce an intertemporal
equilibrium. The rising cost of producing investment is a cost of adjusting to a new intertemporal equilibrium path.
77
-------
Appendix A: Cost andMacroeconomic Modeling
Configuration of the No-control Scenario
One of the difficulties in describing the no-control scenario is ascertaining how much environmental
regulation would have been initiated by state and local governments in the absence of a federal program. It
may reasonably be argued that many state and local governments would have initiated their own control
programs in the absence of a federal role. This view is further supported by the fact that many states and
localities have, in fact, issued rules and ordinances which are significantly more stringent and
encompassing than federal muiimnm requirements. However, it may also be argued that the federal CAA
has motivated a substantial number of stringent state and local control programs. ',
A e
Specifying the range and stringency of state and local programs that would have occurred in the
absence of the federal CAA would be almost entirely speculative. For example, factors which would
complicate developing assumptions about stringency and scope of unilateral state and local programs
include: (I) the significance of federal funding to support state and local program development; (ii) the
influence of more severe air pollution episodes which might be expected in the absence of federally-
mandated controls; (iii) the potential emergence of pollution havens, as well as anti-pollution havens,
motivated by local political and economic conditions; (iv) the influence of federally-sponsored research on
the development of pollution effects information and control technologies; and (v) the need to make
specific assumptions about individual state and local control levels for individual pollutants to allow
estimation of incremental reductions attributable to federal control programs.
Another complication associated with the no-control scenario is the treatment of air pollution control
requirements among the major trading partners of the U.S. Real-world manifestation of a no-control
scenario would imply that public health and environmental goals were not deemed sufficiently compelling
by U.S. policymakers. Under these conditions, major trading partners of the U.S. in Japan, Europe, and
Canada may well reach similar policy conclusions. Simply put, if the U.S. saw no need for air pollution
controls, there is little reason to assume other developed industrial countries would have either. In this
case, some of the estimated economic benefits of reducing or eliminating air pollution controls hi the U.S.
would not materialize because U.S. manufacturers would not necessarily gam a production cost advantage
over foreign competitors. However, like the question of state and local programs in the absence of a
federal program, foreign government policies under a no-control scenario would be highly speculative.
Given the severity of these confounding factors, the only analytically feasible assumptions with respect
to the no-control scenario are that (a) no new control programs would have been initiated after 1970 by the
states or local governments in the absence o^ a federal role, and (b) environmental policies of U.S. trading
partners remain constant regardless of U.S. policy.
Elimination of Compliance Costs in the No-Control
Case
Industries that are affected by environmental regulations can generally respond in three ways: (I) with
process changes (e.g., fuudized bed combustion); (ii) through input substitution (e.g., switching from high
sulfur coal to low sulfur coal); and (iii) end-of-pipe abatement (e.g., the use of electrostatic precipitation to
reduce the emissions of particulates by combustion equipment).60 Clean air regulations have typically led,
Regulation may also affect the rate of investment, and change the rate of capital accumulation.
r
78
-------
Appendix A: Cost and Macroeconomic Modeling
to the latter two responses, especially in the short run. End-of-pipe abatement is usually the method of
choice for existing facilities, since modifying existing production processes can be costly. This approach Is
also encouraged by EPA's setting of standards based on the notion of "best available technology"
(Freeman, 1978). .
All three possible responses may lead to: (I) unanticipated losses to equity owners; (ii) changes in
current output; and (iii) changes in long-run profitability. If firms were initially maximizing profits, then
any of the above three responses will increase its costs. Fixed costs of investment will be capitalized
immediately. This will result in a loss to owners of equity when regulations are introduced. As far as '
firms are concerned, this is just like a lump sum tax on sunk capital. Such effects will not affect growth or
efficiency. However, regulations could also change marginal costs and therefore current output. In
addition, they could change profits (i.e., the earnings of capital), and thus affect investment. Both of these
effects will reduce the measured output of the economy.
On the consumption side, environmental regulations change consumers' expectations of their lifetime
wealth. In the no-control scenario of this assessment, lifetime wealth increases. This causes an increase in
consumption. In fact, with perfect foresight, consumption rises more in earlier time periods. This also
results in a change in savings.
Capital Costs - Stationary Sources
To appropriately model investment in pollution control requires a recognition that the CAA had two
different effects on capital markets. First, CAA regulations led to the retrofitting of existing capital stock
in order to meet environmental standards. In the no-control scenario, these expenditures do not occur.
Instead, the resources that were invested in pollution abatement equipment to retrofit existing sources are
available to go to other competing investments. Thus, at each point in time, these resources might go to
investments in capital in the regulated industry, or may go into investments in other industries, depending
upon relative rates of return on those investments. This will affect the processes of capital formation and
deepening.
Second, the CAA placed restrictions on new sources of emissions. When making investment
decisions, firms take into account the additional cost of pollution abatement equipment Effectively, the
"price" of investment goods is higher because more units of capital are required to produce the same
amount of output. In the no-control scenario, there are no restrictions on new sources and hence no
requirements for pollution control expenditures. Effectively, the "price" of investment goods is lower.
Thus, at each point in timi, investors are faced with a lower price of investment goods. This results in a
different profile for investment over time.
g^ fff „ * -„•"•-• - ."•-', • •
Operating and Maintenance Costs - Stationary Sources
In addition to purchasing pollution abatement equipment, firms incurred costs to run and maintain the
pollution abatement equipment. In the no-control scenario, resources used to pay for these operating and
maintenance (O&M) costs are freed up for other uses. The model assumes that the resources required to
run and maintain pollution control equipment are in the same proportions as the factor inputs used in the
underlying production technology. For example, if 1 unit of labor and 2 units of materials are used to
produce 1 unit of output, then one-third of pollution control O&M costs are allocated to labor and two-
thirds are allocated to materials. These adjustments were introduced at the sector level. O&M
expenditures are exclusive of depreciation charges and offset by any recovered costs.
79
-------
Appendix A: Cost and Macroeconomic Modeling
Capital Costs - Mobile Sources
Capital costs associated with pollution control equipment were represented by changing costs for
motor vehicles (sector 24) and other transportation equipment (sector 26). Prices (unit costs) were reduced
in proportion to the value of the pollution control devices contained in cars, trucks, motorcycles, and
aircraft.
Operating and Maintenance - Mobile Sources
Prices for refined petroleum products (sector 16) were changed to reflect the resource costs associated
with producing unleaded and reduced lead gasoline (fuel price penalty), the change in fuel economy for
vehicles equipped with pollution control devices (fuel economy penalty), and the change in fuel economy
due to the increased fuel density of lower leaded and no lead gasoline (fuel economy credit). Third,
inspection and maintenance costs and a maintenance credit associated with the use of unleaded and lower
leaded (i.e., unleaded and lower leaded gasoline is less corrosive, and therefore results in fewer muffler
replacements, less spark plug corrosion, and less degradation of engine oil) were represented as changes in
prices for other services (sector 34).
Direct Compliance Expenditures Data
Sources of Cost Data
Cost data for this study are derived primarily from the 1990 Cost of Clean report. EPA publishes cost
data in response to requirements of the Clean Air and Clean Water Acts. The following subsections
describe Cost of Clean data in detail, as well as adjustments made to the data and data from other sources.
Corf o/Cfea» Data
EPAis requireti to compile and publish public and private costs resulting from enactment of the Clean
Air Act and the Clean Water-Act The 1990 Cost of Clean report presents estimates of historical pollution
contrpXexpenditures for the years 1972 through 1988 and projected future costs for the years 1989 through
2000. This includes federal, state, and local governments as well as the private sector. Estimates of capital
costs, operation and maintenance (O&M) costs, and total annualized costs for five categories of
environmental media, including air, water, land, chemical, and multi-media, are presented. It should be
noted that these estimates represent direct regulatory implementation and compliance costs rather than
social costs. The Cost of Clean relied on data from two governmental sources, the EPA and the U.S.
Apartment of Commerce (Commerce).
EPA Data
EPA expenditures were estimated from EPA budget justification documents.61 Estimates of capital
and operating costs resulting from new and forthcoming regulations were derived from EPA's Regulatory
41 The nuia source of data for EPA expenditures is the Justification of Appropriation Estimates for Cofnmittee on Appropriations.
80
-------
Appendix A: Cost and Macroeconomic Modeling
Impact Analyses (RIAs). RIAs have been prepared prior to the issuance of all major regulations since
1981. Finally, special analyses conducted by EPA program offices or contractors were used when other
data sources did not provide adequate or reliable data.
Commerce Data
Data collected by Commerce were used extensively in the Cost of Clean for estimates of historical
pollution control expenditures made by government agencies other Chan EPA and by the private sector.
Two Commerce agencies, the Bureau of Economic Analysis (BEA) and the Bureau of the Census
(Census), have collected capital and operating costs for compliance with environmental regulations since
the early 1970's. Commerce is, in fact, the primary source of original survey data for environmental
regulation compliance costs. Commerce publishes a number of documents that report responses to surveys
and comprise most of the current domain of known pollution abatement and control costs in the United
States, including:
>• A series of articles entitled "Pollution Abatement and Control Expenditures" published annually in the
Survey of Current Business by BEA (BEA articles);
>. A series of documents entitled "Pollution Abatement Costs and Expenditures" published annually in
the Current Industrial Reports by Census (PACE reports);"and, >
> A series of documents entitled Government Finances published annually by Census (Government
Finances).
BEA articles contain data derived from a number of sources, including two key agency surveys—the
"Pollution Abatement Costs and Expenditures Survey"r(PACE Survey) and the "Pollution Abatement Plant
and Equipment Survey" (PARE Survey)— which'are conducted annually by Census for BEA. Data have
been reported for 1972 through~1987.62
PACE reports have been published annually since 1973 with the exception of 1987. Figures for 1987
were estimated on the basis of historical shares within total manufacturing. These reports contain
expenditure estimates derived from surveys of about 20,000 manufacturing establishments. Pollution
abatement expenditures for air, water and solid waste are reported by state and Standard Industrial Code
(SIC) at the four-digit level. According to Census, surveys conducted since 1976 have not included
establishments with fewer than 20 employees because early surveys showed that they contributed only
about 2 percent to the pollution estimates while constituting more than 10 percent of the sample size.
•• Each year Census conducts a survey of state, local, and county governments; and survey results are
published in Government Finances. Census asks government units to report revenue and expenditures,
including expenditures for pollution control and abatement
Non-EPA Federal expenditures were estimated from surveys completed by federal agencies detailing
their pollution control expenditures, which are submitted to BEA. Private sector air pollution control
expenditures, as well as state and local government air pollution expenditures, were taken from BEA
articles.
62 The most recent BEA article used as a source for air pollution control costs in the Cost of Clean was "Pollution Abatement and Control
Expenditures, 1984-87" in Survey of Current Business, June 1989.
81
-------
Appendix A: Cost and Macroeconomic Modeling
Stationary Source Cost Data
Capital Expenditures Data
Capital expenditures for stationary air pollution
control are made by factories and electric utilities for i
plant and equipment that abate pollutants through
end-of-line (EOL) techniques or that reduce or
eliminate the generation of pollutants through
changes in production processes (CIPP). For the
purposes of this report EOL and CIPP expenditures
are aggregated.63 Table 28 summarizes capital
expenditures for stationary air pollution control,
categorized as "nonfarm business" or "government
enterprise" expenditures.
Nonfarm business capital expenditures consist of
plant and equipment expenditures made by 1)
manufacturing companies and 2) privately and
cooperatively owned electric utilities and 3) other
nonmanufacruring companies. "Government
enterprise" is, according to BEA, an agency of the
government whose operating costs, to a substantial
extent, are covered by the sale of goods and services.
Here, government enterprise means specifically
government enterprise electric utilities. Government
enterprise capital expenditures are pollution
abatement expenditures made by publicly owned
electric utilities.64
Operation and Maintenance Expenditures Data
, Stationary source O&M expenditures are made
by manufacturing establishments, private and public
electric utilities, and other nonmanuf acturing
businesses to operate air pollution abatement
equipment O&M expenditures for electric utilities
are made up of two parts: 1) expenditures for
operating air pollution equipment and 2) the
additional expenditures associated with switching to
alternative fuels that have lower sulfur content (fuel
Table 28. Estimated Capital and O&M
-Expenditures fef Stationary Source Air Pollution
Control (millions of current dollars).
Nonfarm
Business*
Year
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
toon
Sip/
2,968
3,328
3,914
3,798
3,8lf '
3,977
4,613 >
5,051
5,135
5,086
4,155
4,282
4,141
4,090
4,179
4,267
4,760
41*0
t
Q&M^
\
1,407
1,839
2,195
(2,607
3,163
3,652
4,499 ,
5,420
5,988
5,674
6,149
6,690
6,997
7,116
7,469
7,313
7,743
S.tfftR
, ' re ' •> _.
Government '
Enterprise
v "si
« Caplc
* s
^82,
104
'102
!l56
,197
205
,|85
^loC*
v-Ji^O
, '451
' ?08
> '4%$
416
328
" 512
277
' 243
, 235
yzfi
n f
O&Mrf
i
^
2?
'56
'45
'^8
, 60
72
,'106
148
135
141
' 147
189
^140
130
161
173
J '154
Sources: j " '
a. Non-fann capital expenditures for 1973-87 are from Cost
of Clean, Table B-l, line 2,
b. Non-farm O&M expend!!
xpenditutes for 1973-85 are from Cost
of Clean, Table B-l, line &
a Government enterprise capital expenditures for 1973-87
are from Cost of Clean, Table B-9, line 1.
d. Government enterprise O&M expenditures for 1973-85
are from Cost of Clean, Table B-9, fine 5.
All other reported expenditures are EPA estimates.
0 Survey respondents to the Census annual Pollution Abatement'Surveys report the difference between expenditures for CIPP and what they
would have spent for comparable plant and equipment without pollution abatement features. Disaggregated capital expenditures by private
minBftcturing establishments can be found in annual issues of Census reports.
** BEA calculates these expenditures using numbers obtained from Energy Information Agency (EIA) Form 767 on steam-electric plant air
quality control.
82
-------
Appendix A: Cost and Macroeconomic Modeling
differential). Expenditures to operate air pollution abatement equipment are for the collection and disposal
of flyash, bottom ash, sulfur and sulfur products, and other products from flue gases.65 O&M expenditures
are net of depreciation and payments to governmental units, and are summarized in Table 28. O&M data
were disaggregated to the two digit SIC level for use in the macroeconomic model.
For both capital and O&M expenditures, historical survey data were not available for each year
through 1990 prior to publication of Cost of Clean. For the purpose of the 812 analysis, EPA projected
1988 to 1990 capital expenditures and 1986 to 1990 O&M expenditures. Those projections were used in
the macroeconomic simulation, and have been retained as cost estimates to ensure consistency between the
macroeconomic results and the direct cost estimates. Since completion of the macroeconomic modeling,
however, BEA has published expenditure estimates through 1990. A comparison of more recent BEA
estimates with the EPA projections used in the 812 analysis can be found in the "Uncertainties in the Cost
Analysis" section, below.
Recovered Costs
"Recovered costs" are costs recovered (i.e., revenues realized) by private manufacturing establishments
through abatement activities. According to instructions provided to survey participants by Census,
recovered costs consist of 1) the value of materials or energy reclaimed through abatement activities that
were reused in production and 2) revenue that was obtained from the sale of materials or energy reclaimed
through abatement activities. Estimates of recovered costs were obtained from the PACE reports and are
summarized in Table 29. In this analysis^jecovered'costs were removed from total stationary source air
pollution control O&M costs - that is, net O&M cost in any year would be O&M expenditures (see Table
28) less recovered costs. Recovered cost data were disaggregated to the two digit SIC level for use in the
macroeconomic model.
45 Farber, Kit D. and Gary L. Rutledge, "Pollution Abatement and Control Expenditures: Methods and Sources for Current-Dollar Estimates,"
Unpublished paper, Bureau of Economic Analysis, U.S. Department of Commerce, October 1989.
83
-------
Appendix A: Cost and Macroeconomic Modeling
Source Data
Costs of controlling pollution emissions from
motor vehicles were estimated by calculating the
purchase price and O&M cost premiums associated
with vehicles equipped with pollution abatement
controls over the costs for vehicles not equipped with
such controls. These costs were derived using EPA
analyses, including EPA RIAs, the Cost of Clean, and
other EPA reports.66 This Appendix summarizes the
Section 812 mobile source compliance cost estimates
and provides references to published data sources
where possible. Further information on specific
methods, analytical steps, and assumptions can be
found in McConnell et al. (1995),67 which provides a
detailed description of the Section 812 mobile source
cost estimation exercise (referred to as Cost of Clean
(1993, upublished)) and compares the method and
results to other similar analyses (including Cost of
Clean (1990)).
/
, ^Esfiroatedj Recover^ C3osts for
Stationary Source Air PoUufloii Control (millions
of current dollars). ' - ^
Year
1972
1973
1974
1975 <
1976
1977
1978"
•!
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988,
1989
190ft
] PACE* * J&g
t « k ""
<.
*• 3
> I
<
f v<«
$ 1
750
862
1,000 '
— - 858
822
866
' ' 76f' * , , '
S60
' <
, W03,t ^ ;
r *j
.
imated
248 •
/ x
19^9
"296
389' .
496
'-557 {
.617
*750
!86i v
997
857
'822- "
870" ^
^568 /
867 " '
987 '
,^«»
i '* ' ' I | , k-H '"•' ( , ' J ' -t>
It -iX 'll^ilt ' •* i>
*
-------
Appendix A: Cost and Macroeconomic Modeling
Table 30.' Estimate Capital ant 6peraiion
arid Maintenance Expenditures for Mobile
Coatee Air Pollution Control (rniffiona of ^
current dollars), s - ' , ~ <
Capital Expenditures Data
Capital expenditures for mobile source emission
control are, associated primarily •with pollution abatement
equipment on passenger cars, which comprise the bulk of
all mobile sources of pollution. These capital costs reflect
increasingly stringent regulatory requirements and
improvements in pollution control technologies over time.
Each of the following devices have been used at one time
or another dating back to the Clean Air Act Amendments
of 1965: air pumps, exhaust-gas retirculation valves, high
altitude controls, evaporative emissions controls, and
catalysts. The cost estimates for each component were
computed on a per-vehicle basis by engineering cost
analyses commissioned by EPA. The resulting per-
vehicle capital costs were multiplied by vehicle
production estimates to determine annual capital costs.
Table 30 summarizes mobile source capital costs.
Operation and Maintenance Expenditures Data
Costs for operation and maintenance of emission
abatement devices include the costs of maintaining
pollution control equipment plus the cost of vehicle
inspection/maintenance programs. Operating costs per
vehicle were multiplied by total vehicles in use to
determine annual cost. Mobile source O&M costs are
made up of three factors: 1) fuel price penalty, 2) fuel
economy penalty and 3) inspection and maintenance
program costs as described below. These costs are
mitigated by cost savings in the form of maintenance
economy and fuel density economy. Table 31
summarizes mobile source O&M expenditures and cost
savings by categories, with net O&M costs summarized
above in Table 30. The following sections describe the
components of the mobile source O&M cost estimates.
"! Fuel Price Penalty
>• -2. ~ ' -•,•.-. ' ' ' .
Historically, the price of unleaded fuel has been several cents per gallon higher than the price of leaded
fuel. CAA costs were calculated as the difference between the cost of making unleaded gasoline and
leaded gasoline with lower lead levels and the cost of making only leaded gasoline with a lead content set
at pre-regulatory levels. These cost estimates were developed using a linear programming model of the
refinery industry. Prices of crude oil and other unfinished oils, along with the prices of refinery outputs,
were adjusted annually according to price indices for imported crude oil over the period of analysis. The
relative shares of leaded and unleaded gasoline arid the average lead content hi leaded gasoline also were
• Year
~
1973
1974
1975
1976
1977
1 197*
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
Capital1 -
\ . w"t ^
"276 " _ *
242 '"
V70;^ ,
"1,961
s' . "2,248
2,513* ' <
2>4I s
2,949' ^
' 3,534
* -Si
3,551,
V 4,331
5^679
- 6,387
v 6,886 ' ' ;
6;85i
^'7,206, /.
^7,053
7,299 ,[
p >si,765
/ 2,35l'-
2^282
v 2,0^0V
" X-786 '
\ 908v
^Ij229^
'" "1,790
^389"
v, 555
, ^326''
' ^ 337 '
-1,394
v " -I',j302
; -X575
-1,'636 "
; ^.1,^16
a. Capitalexp.:
^each; Tables C-2Ato C-9A, Mne 10 on eachj'conveited ftom,
$1986 to current dollars. - ' ^ - *
b. O&M exp.: EPA analyses based on sources and
metbodsin: Costs and Benefits of Reducing Lead in
GasottttK Final Regulatory ImpactAnafysis,U.$.
Environmental Protection Aginty, Office of Policy
Analysis,BPA-230-05-85-006,;Febwaryl985?andC
-------
Appendix A: Cost and Macroeconomic Modeling
adjusted adjusted annnually according to the historical
record.
These estimates may tend to understate costs due
to a number of biases inherent in the analysis process.
For example, the refinery model was allowed to
optimize process capacities in each year. This
procedure is likely to understate costs because
regulatory requirements and market developments
cannot be perfectly anticipated over time. This
procedure resulted in estimates that are about ten
percent less than estimates in other EPA reports.68
However, new process technologies that were
developed in the mid-1980s were not reflected in
either the base case or regulatory case runs. It is
reasonable to expect that regulatory requirements
would have encouraged development of technologies
at a faster rate than would have occurred otherwise.
Fuel Economy Penalty
The fuel economy penalty benefit is the cost
associated with the increased/decreased amount of fuel
used by automobiles with air pollution control devices
(all else being equal). An assumption that can be
made is that the addition of devices, such as catalytic
converters, decrease automobile fuelefficiency.69 If
this assumption is true, air pollution control devices
increase the total fuel cost to consumers. An
alternative assumption is that the use of catalytic
converters has increased fuel economy. This increase
has been attributed in large measure to the feedback
mechanism built into three-way catalytic converters.70
Under this assumption, the decrease in total fuel cost
to consumers is considered a benefit of the program.
Table 3i. Q&lft Costs and Credits (millions of
*. ^ 3-XS •» •**« **. i^-s ! '
* curreatdollars). -
Year
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
Fuel
^Price
Penalty
91
244
358
468
A\ i
568
766
1187
1912
2181
2071
1956
2012
3057
2505
2982
3127
3476
3754
f
Fuel
Ecoa*
Penalty
^
-1700
2205
2213
'2106
1956
166$ '
18 f s ^
%%-98
-289
-514
-738 ,
-1527
-1826
-2120
-2386 '
-2542 ''
-273? ,
Jb@5$.
-283*8
-3859,'
-4126
-44J92 *
-4794
-5089
Total
Costs
*f
1765,
*i§5i
2282
2060
,1786'
908
1229
1790
1389
"555;
-155
-326
^337*
-1394
-1302
-1575
4636
'm*
, f.
* Inspection and maintenance costs less fuel density savings and
maintenance savings. ' ' *f ' *V " i
' ' ' i « i f
•. ** *. "
Sources: All results ate presented in Jprgenson etui (1993), pg.
AJ.7. FPP results are based on a petroleum refinery cost model
run for the retrospective analysis. FEP and Net I&M are based
on data and methods from Costs and Benefits of Hectoring Lead
in Gasoline: Fatal Regulatory Impact Analysis, U.S.
Env&onmenlal Protection Agency, Office of Policy Analysis,
• EPA-23005-85-006, February 1985; and Cost of Clean (1990).
Specific analytic procedures are summarized in McConnellef ^
For the purposes of this study, sensitivity analyses were performed using data presented in the Cost of
Clean report These analyses were conducted to evaluate the significance of assumptions about the
relationship between mile per gallon (MPG) values for controlled automobiles and MPG values for
uncontrolled cars. Based on results of these and other analyses, fuel economy was assumed to be equal for
" Costs and Benefits of Reducing Lead in Gasoline: Final Regulatory Impact Analysis, U.S. Environmental Protection Agency, Office of
Policy Analysis, EPA-230-05-85-006, February 1985.
** Memo from Joel Schwartz (EPA/OPPE) to Joe Somers and Jim DeMocker dated December 12,1991, and entitled "Fuel Economy
Benefits." Schwartz states that since this analysis is relative to a no Clean Air Act baseline, not a 1973 baseline, fuel economy benefits are not
relevant. la the absence of regulation, tuning of engines for maximum economy would presumably be optimal in the base case as well.
79 Memo from Joseph H. Somers, EPA Office of Mobile Sources, to Anne Grambsch (EPA/OPPE) and Joel Schwartz (EPA/OPPE) entitled
"Fuel Economy Penalties for Section 812 Report," December 23,1991.
86
-------
Appendix A: Cost and Macroeconomic Modeling
*•
controlled and uncontrolled vehicles from 1976 onward. This may bias the cost estimates although in an
unknown direction.
Inspection and Maintenance Programs
Inspection and maintenance programs are administered by a number of states. Although these
programs are required by the Clean Air Act, the details of administration were left to the discretion of state
or local officials. The primary purpose of inspection and maintenance programs Is to identify cars that
require maintenance —including cars that 1) have had poor maintenance, 2) have been deliberately .„
tampered with or had pollution control devices removed, or 3) have used leaded gasoline when unleaded is
required— and force the owners of those cars to make necessary repairs or adjustments.? Expenditures for
inspection and maintenance were taken from the Cost of Clean.
Beneficial effects of the mobile source control program associated with maintenance and fuel density
were also identified. These cost savings were included in this study as credits to be attributed to the mobile
source control program. Credits were estimated based on an EPA study^where more detailed
explanations may be found.
..''••"
" • -'"£•
Maintenance Credits
Catalytic converters require the use of unleaded fuel, which is less corrosive than leaded gasoline. On
the basis of fleet trials, the use of unleaded or lower leaded gasoline results in fewer muffler replacements,
less spark plug corrosion, and less degradation of engine oil, tfius reducing maintenance costs.
Maintenance credits account for the majority of the direct, (non-health) economic benefits of reducing the
lead concentration in gasoline. Sv; <- ..
>"-£• - ~ •."•"- " . - -.
~~ ~ <
-------
Appendix A: Cost and Macroeconomic Modeling
Unlike the other private
sector expenditure data used
for this analysis, the survey
data used as a source for
private sector R&D
expenditures cannot be
disaggregated into industry-
specific expenditure totals.
Consequently, private sector
R&D expenditures are
omitted from the
macroeconomic modeling
exercise (the macro model is
industry-specific). The R&D
expenditures'are, however,
included in aggregate cost
totals used in the benefit-cost
analysis.
The Cost of Clean and
the series of articles
"Pollution Abatement and
Control Expenditures" in the
Survey of Current Business
(various issues) are the data
sources for ^Other Air
Pollution Control
Expenditures." State and
local expenditures through
1987 are found in Cost of
Clean', 1988-90 expenditures
are from more recent issues
of the Survey of Current
Business (BEA). Federal
government expenditures are
from BEA (various issues).
Private R&D expenditures
were reported in Cost of
Clean. Since publication of
Cost of Chan, however,
BEA has revised its private
sector R&D expenditure
series (BEA, 1994 and 1995).
modeling exercise, the revised
portions of the 812 analysis.
'Table 32. Other Air Pollution Control Expenditures (millions ofi current ""
"dollars). " > '* - , ' * ' ^" '<
t.
E
f.
*
if
t'
i"
f
I
|
F
»
*
t
i
6
I
It
(t
Regulations
Year
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
Abatement
and
Monitoring
Fed.'
47
56
88
105
106
90
103
95
85
87
136
115
98
67
80
65
70
71
State &
Local*
0
0
1
1
,!„„
]
0
0
0
0
0
4
14
12
14
15
10
12
13
State &
Fj«Le L
50
52
66 ,
69
S8Q
,93J
' f
100
122 '
108
93
88
101
-103
106
110
120
130
133
4 115
|31
139,
i35
161
183
200
207
226
230
239
250
250
307
300
320
360
343
' Research
i and Development
c
Private*
* 492
'520
f487
1 "' 562
" 675
"sds
933
' 85t
479l8
76'l
'' ' 691
665
: 775
833^
, 887
934
984
'749
V 1
-
Fed/
126
100
idc5
131
144
146
,105
130
ill
lfc>
13*3
165
^247
*217
200
2i20
23p
231
State &
Local*
6
, 7
'8
6
7
8
* '7
' 5"
«x
y. ^
'$
' 4
3
4
2'
* 1
, , 2
,2
Total
^
^
*
/ 836
'-866
897
1?009
1,174
1,325
1,4^
1,410"
1,^48
1,229
1^97
1,314'
' 1,488
, 1,548
, 1,594
1,670
1,788
1,542
, , , „
Sources: ,' t ft. ,' ,,,;
a. Federal government abatement expenditures: 1973-82, "Pollution Abatement and Control .
Expenditures". Survey of Current Business (BEA) July 1986 Table*9 line 13; 1983-87, BEA June
1989Tabte 7 liiw 13j 1988-90, BEA May 1995 Tabfc 7 line 13. , " J ''' ' \ „
< b. Sute and local abatement expenditures: 1973-87, Costof Clean, TableB-9 line 2; 1988-90,
BEAMayl99STabte7Iinel4. „ , i' {" < ' '" ,' - „, '
Ci Federal government ^tegstoomtorihg" expenditures: 1^73-82, BEA July 1^86, Table 9 line 17;
1983-87, BEA June 1989 Table 6 line 17; 1988-90, BEA May 1995 Table 7 line 17.
d. State and local government "regs/monftoring" expenditures: 1973-87, Cost of Clean, Table B-9
line 3; 1988-90, BEA May 1995 Table 7 fine 18. „ ( , , ' '" -f
e. Private sector R&D expenditures: 1973-86, BEA May 1994 Table 4 (no line #) [total R&0
expenditures in $1987 are converted to current dollars using the GDP price deflator series found,
elsewhere in this Appendix — netting out public sector R&D leaves private sector expenditures];
1987-90,BEAMay 1995Table7line20. 1 "' ! / „ ' ^
* £ Federal government R&D expenditures: 1973-82, BEA. July 1986Tabte 9 line 21; 1983^7,.
BEA June 1989 Table 6 line 21; 1988-90, BEA May 1995, Table 7 line 2J.
g. State and local government R&D expenditures: 1973-87, Cost of Clean, Table B-9 Hne4; 1988-
90,BEAMayl99STable7]nie22. , < <• ' , Js , •
Since private R&D expenditures were not included in the macroeconomic
series can be (and has been) used without causing inconsistency with other
88
-------
Appendix A: Cost and Macroeconomic Modeling
, Table 33. Estimated Annual, CAA Cbrapliance
• Expenditures ($millions).
Direct Expenditures ana Costs
Compliance with the CAA imposed direct costs on businesses, consumers, and governmental units,
and triggered other expenditures such as governmental regulation and monitoring costs and expenditures
for research and development by both government and industry. As shown in Table~33; annual direct
CAA compliance expenditures - including R&D, etc.- over the period from 1973 to 1990 were „
remarkably stable, ranging from $20 billion to - ,^ - * "~
$25 billion in inflation-adjusted 1990 dollars.
This is equal to approximately % of one
percent of total domestic output during that
period, with the percentage falling from % of
one percent of total output in 1973 to % bf one
percent in 1990.
i '
Although useful for many purposes, a
summary of direct annual expenditures is not
the best cost measure to use when comparing
costs to benefits. Capital expenditures are
investments, generating a stream of benefits
(and opportunity cost) over the life of the
investment. The appropriate accounting ,
technique to use^for capital expenditures in a
cost/benefit analysis is to annualize the
expenditure - i.e., spread the cost ove"r the
useful life of theinvestment, applying a ;
discount rate to account for the time value of
money. ~~. ~ ~~ ~
For this cost/benefit analysis, all capital
expenditures have been annualized at three,
five, and seven percent (real) rates of interest.
Therefore, "annualized" jcosts reported for any
given year are equaLtfpD&M expenditures
(plus R<^,etc.y expenditures) plus amortized
capital cc^tl|i.e,j depreciation phis interest
costs»aslolilted with the pre-existing capital
stock) for that year. Stationary source air
pollution control capital costs are amortized
over 20 years; mobile source air pollution
control costs are amortized over 10 years.
Table 34 summarizes annual expenditures and annualized costs. ,
Due to dattf limitations, the cost analysis for this CAA retrospective starts in 1973, missing costs
incurred in 1970-72. Cost of Clean, however, includes capital expenditures for 1972. In this analysis,
amortized costs arising from 1972 capital investments are included in the 1973 to 1990 annualized costs,
-
Year,
,1973
1974
^1975^
1976
1977
1978
1979
1980
1981
"1982
1983
1984'
1985
1986
1987
1988
S1989
1990
1
Capital1
3326
3,674
5^86
5,915
6,256
6,695
7,839
8398
9,120
9445
8,908
10377
10,'856
11,288
11307
tt,7l6
12,048
11,707
O&M ,
and
"Otter**
- 3,838
4,816
5,030
5,238
5;626
5340
6332
7,906
7,863
6,812
6,612
6,955
8,243
6343
6,904
6,462
6,946
7312
s
£
• Total
7,164
,8^490
10,616
1 11,153
11,882
12,035
14371
16,304
16,983
"15,957
15320.
17332
19,099
17,831
18^211
18,'178
18,994.
19,019
GDP
Pr&f
Defl.
413,
44.9
49.2
"52.3^
- 55/
60.3
JS53
71.7
^ 78.9'
' "83.8
87.2
<' 91.0
94.4
96.9
100,0
,103.9
108.5
113.2
average expenditure, 1973-1990 ($1990) = ,
Total
Exp,
f$1990>
\9,S35
2i,405
* 24,425
" 24,139
24,062
22393
24,837
25,741
„ 24367
21,555
20,148
~ 21360
22,903
- 20,831
20,615
19,805
19J817
,19,019
•22,081
a. Capital expenditures are the sum of stationary source-{private
sector plus government enterprise) and mobile source capital
expenditures.
b. The sum of stationary source O&M, mobile source 6&M,
anatement, regulation and monitoring, and R&D expenditures.
89
-------
Appendix A: Cost and Macroeconomic Modeling
even though 1972 costs are not otherwise included in the
analysis. Conversely, only a portion of the (e.g.) 1990
capital expenditures are reflected in the 1990 anmialized
costs — the remainder of the costs are spread through the
following two decades, which fall outside of the scope of
this study (similarly, benefits arising from emission
reductions in, e.g., 1995 caused by 1990 capital
investments are not captured by the benefits analysis).
The stream of costs from 1973 to 1990 can be
expressed as a single cost number by discounting all costs
to a common year. In this analysis, all costs and benefits
are discounted to 1990 (in addition, all costs and benefits
are converted to 1990-value dollars, removing the effects
of price inflation). There is a broad range of opinion in
the economics profession regarding the appropriate
discount rate to use in analyses such as this. Some
economists believe that the appropriate rate is one that
approximates the social rate of time preference - three
percent, for example (all rates used here are "real", i.e.,
net of price inflation impacts). Others believe that a rate
that approximates the opportunity cost of capital (e.g.,
seven percent or greater) should be used; A thkdjichool
of thought holds that some combination of the social rate
of time preference and the opportunity cost otcapiMt-is ,
appropriate, with the combination effected eimiif By iise
of an intermediate rate or by use of a multiple-step
Table 34. Compliance Expenditures and
Amiualized Costs, 1973 to 1$90 ($1990
millions). ' > •" ^
t
,
#.
•*
u
'
*c
J-
>
•>
H
*"
Year
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
toon
*
Expend.
19,635
21,405
24,425
24,139
24,062
22,593
24,837
' 25,741
24,367
21,555
20,148
21,560
22,903
20,831
20,615
19,805
19,817
tomo
"" f
V- 'if i
Annu.
at 3%
10..9S7
13431
13^14
14,123
15,253
14,961
17^09
&&&
19,590
"18,643 '
19,095
20,133
22,516
21,109
22,072
22,012 h
22,916 '
ft SOS
i t ^i
f (
allzedCo!
at 5%
£1,042
13,435
' 13,638
14,611
, 15,904
15J,776
'18^282
20,812 '
20,905
20,125
26,734^
21,909 "
'24^447
23",161
24,23f
24^8f
25,288
9A naa
.+ ' ' -
^ ,
its
at 7%-
11,134
13^655
\
13,988
15,139
16,608
J&653.
19^31
2^046
22^21
21J720
22,498
23,819
26,523
25,364
26^62
26,719
27,836
•7S 7f 1
^
„
,
'
i i *
^ ,> * " T
:erence as
the "discount rate," but still accounts for feeost of capital. The 812 Project Team determined to use a
range of discount rates (three, five, anil seven percent) for the analysis.
Expenditures and annualized costs discounted to 1990 are found on Table 5,35. Expenditures are
discounted at all three rates; annuafized costs are
discounted at the rate corresponding to that used in •••••••^^••••^^•^••••••^•••HB
the annualization procedure (i.e., the "annualized at
3%" cost stream is discounted to 1990 at three
percent). The final row presents the result of an
explicit combination of two rates: Capital costs are
iannualized at seven percent, then the entire cost
stream is discounted to 1990 at three percent.
Table 35, Costs Discounted to 1990 ($1990
millions) ! , !'
i 1 •*
Expenditures 520,475 627,621 760,751
' ' , ' ' , \ •' ' ',
Annualized Cbsts 416^04 522,906 , 65??003
Annualized.at7% 476,32
indirect Effects of the CAA
ill'
In addition to imposing direct compliance costs
on the economy, the CAA induced indirect economic effects, primarily by changing the size and
composition of consumption and investment flows. Although this analysis does not add these indirect
effects to the direct costs and include them in the comparison to benefits, they are important to note. This
90
-------
Appendix A: Cost andMacroeconomic Modeling
Table 36. Differences in Gross National . '
Product Between the Control and No-control
Scenarios. ~,
section summarizes the most important effects of the
CAA, as estimated by the J/W macroeconomic
simulation.
GNP and Personal Consumption
Under the no-control scenario, the level of GNP
increases by one percent in 1990 relative to the control
case (see Table 36). During the period 1973 to 1990, the
percent change in real GNP rises monotonically from 0.26
percent to 1.0 percent. The increase in the level of GNP
is attributable to a rapid accumulation of capital, which is
driven by changes in the price of investment goods. The
capital accumulation effect is augmented by a decline in
energy prices relative to the base case. Lower energy
prices that correspond to a world with no CAA
regulations decreases costs and increases real household-
income, thus increasing consumption.
Removing the pollution control component of new
capital is equivalent to lowering the marginal price of
investment goods. Combining this with the windfall gain
of not having to bring existing capitaBnto compliance
leads to an initial surge in the economy's rate of return,
raising the level of real investment The investment -
effects are summarized in Figure 24, which is found at the "•-,''- /:,
-- A."" ~~Z~r- £f -, " ,— It «,"*•''»
end of this appendix. More rapid (ordinary) capital mmmmmmmmm i ,••
accumulation leads to a decline in the rental price of
capital services which, in turn, stimulates the demand for capital services by producers and consumers.
The capital rentarprice reductions also serve to lower the prices of goods and services and, so, the overall
price level. Obviously, the more capital intensive sectors exhibit larger price reductions.73 The price
effects from investment changes are compounded by the cost reductions associated with releasing
resources from the operation and maintenance of pollution control equipment and by the elimination of
higher prices due to regulations on mobile sources.
' To households, no-control scenario conditions are manifest as an increase in permanent future real
earnings which supports an increase in real consumption in all periods and, generally, an increase in the
demand for leisure (see Table 38). Households marginally reduce their offer of labor services as the
income effects of higher real earnings dominate the substitution effects of lower goods prices. The
increase in consumption is dampened by an increase in the rate of return that produces greater investment
(and personal savings).
<
Year
1973
1974
1975
:* 1976
1977
1978
1979
1980
1981
, 1982
1983
1984
1985
1986
1987
1988
< 1989
1990
Nominal %
Change'
>4>:09
-0.18 ~,
41.10
* -0.00
4).iO
-0.16
-0.16
"- -0.14
41.14
41.19,"
41.19"
41.17 _
41.12
-oi* x
41.15 ~
41J2P
41.21
41.18
Real%
"• Change.
'- '$,26' "
* ;0.27
0.44^
b.4^
x 0.54
,/0.56
" 0.63
^0.69
0.73 -
0.74' ''
,0.78
0.84
- 0.95 '
43.9g
1.0L
15 * M
- 1x00
0.99
' 1'OD -
73 Not surprisingly, at the industry level, the principal beneficiaries in the long run of eliminating the costs associated with air pollution
abatement are the most heavily regulated industries. The largest changes in industry prices and outputs occur in the motor vehicles industry. Other
industries that benefit significantly from the elimination of environmental controls are refined petroleum products, electric utilities, and other
transportation equipment Turning to manufacturing industries, metal mining and the primary metals have the largest gains in output from
elimination of air pollution controls. ' ' -,
91
-------
Appendix A: Cost and Macroeconomic Modeling
Finally, technical change is a very important aspect of the supply-side adjustments under the
no-control scenario. Lower factor prices increase the endogenous rates of technical change in those
industries that are factor-using. Lower rental prices for capital benefit the capital-using sectors, lower
materials prices benefit the materials-using sectors, and lower energy prices benefit the energy-using
sectors. On balance, a significant portion of the increase in economic growth is attributable to accelerated
productivity growth. Under the no-control scenario, economic growth averages 0.05 percentage points
higher over the interval 1973 to 1990. The increased availability of capital accounts for 60 percent of this
increase while faster productivity growth accounts for the remaining 40 percent. Thus,:the principal effect
arising from the costs associated with CAA initiatives is to slow the economy's rates of capital
accumulation and productivity growth.
As with the cost and expenditure data presented , ,
above, it is possible to present the stream of GNP and Table 37, GNP and Consumption "impacts ,
consumption changes as single values by discounting 'Discounted to 1990 ($1990 billions) > „
the streams to a single year. Table 37 summarizes " '" 3% 5% 7%'
the results of the discounting procedure, and also Expenditures ' ' ,520 628 .761
includes discounted expenditure and annualized cost Annualized Costs a 417 523 657
data for reference. Accumulated (and discounted to QJOP * '" 880 V\loQ5 ' U5i
1990) losses to GNP over the 1973 to 1990 period '^Household Consumption' 500 '" 56<> €53
were half again as large as expenditures during the HHamfGoVt Consumption '676 ,769 . ,881
same period, and approximately twice as large as ; p ••"»>"• r ,.
annualized costs. Ixissesm household consumption Source: Expenditures and annuitized costs ftom above;
were approximately as great as annualized costs. macroeconomic impacts ftom Joigensoaetal. (1993),
'•'" ' """ ' '
;••;,- . -,-vi. , Si-* . ' , , < i - f i
Although they have value as descriptors of the
magnitude of changes in economic ^activity, neither '
GNP nor consumption changes are perfect measures of changes in social welfare. A better measure is
Equivalent Variations (EVs), which measure the change in income that is equivalent to the change in
(lifetime) welfare due to removal of the GAA. As part of its macroeconomic exercise, EPA measured the
EVs associated witS '.removal of the C%V Elimination of CAA compliance costs (disregarding benefits)
represents a welfare gain of $493 billion to $621 billion, depending on assumptions used in the analysis.74
This result does not differ greafly from the range of, results represented by expenditures, anualized costs,
and consumption changes.
Prices
Oneprincipalconsequenceof the Clean Air Act is that it changes prices. The largest price reductions
accrue to the most heavily regulated industries which are the large energy producers and consumers (see
Table 39). But these are also the most capital intensive sectors and it is the investment effects that are the
dominant influences in altering the course of the economy. Focusing on energy prices, under the
no-control scenario the price of coal hi 1990 declines by 1.3 percent, refined petroleum declines by 3.03
percent, electricity from electric utilities declines by 2.75 percent, and the price of natural gas from gas
utilities declines by 1.2 percent. The declining price of fossil fuels induces substitution toward fossil fuel
energy sources and toward energy in general. Total Btu consumption also increases.
74 Jotgensonctal.,1993.
92
-------
Appendix A: Cost and Macroeconomic Modeling
Sectoral Effects: Changes in Prices and Output by
Industry
At the commodity level, the effect of the CAA varies
considerably. Figure 25, found at the end of this appendix,
shows the changes in the supply price of the 35 commodities
measured as changes between the no-control case and the
control-case for 1990.
In 1990, the largest change occurs in the price of motor
vehicles (commodity 24), which declines by 3.8 percent in the
no-control case. Other prices showing significant effects are 1
those for refined petroleum products (commodity 16) which
declines by 3.0 percent, and electricity (commodity 30) which
declines 2.7 percent. Eight of the remaining industries have
decreases in prices of 1.0 to 1.4 percent under the no-control
scenario. The rest are largely unaffected by environmental
regulations, exhibiting price decreases between 0.3 and 0.8
percent.
To assess the intertemporal consequences of the CAA,
consider the model's dynamic results and the adjustment of
prices between 1975 and 1990. Initially, in 1975, the biggest
effect is on the price of output from petroleum refining (sector
16), which declines by 4.3 percent. But by 1990, the price of
petroleum refining is about 3,0 percent below control scenario
levels. In contrast, the price of motor vehicles (sector 24) is
about 2.4 percent below baseline levels in 1975, but falls to
about 3.8 percent below baseline levels in 1990.
Table 38. Difference in Personal
Consumption Between the Control and
No-Control Scenarios. <
Year
1973
1974
1975
1976
1977
1978
1979
1980
1981*
1982
1983
1984
1985
1986
1987
1988
1989
1990
Nominal?
Change
" -ftoV
-0.01
- -040
-040
-040*
-0.09
-0.11
-042
-043
" -042
-0.13
> -0.15
-049
-049
^049
-047
' -0.17
-048
$# Bbal,4*,
- Cfaaag^
0.33 ,
0.43 Kv
' oli4
0^9
' 0,63 -
0.68
* 0.71
0.74
0«1
0.85'
' „ 0.86
488 v
J ,0.94
0..98
, 1.03
iM
-* , V*
The price changes affect commodity demands, which in turn determine how industry outputs are
affected. Figure 26, found at the end of this appendix, shows percentage changes in quantities produced
by the 35 industries for th& years 1975,1980,1985 and 1990. As noted earlier, the principal beneficiaries
under the no-control scenario are the most heavily regulated industries: motor vehicles, petroleum refining,
and electric utilities.
In 1990, the motor vehicle sector (sector 24) shows the largest change in output, partly due to the fact
that the demand formotor vehicles is price elastic. Recall that the largest increase in prices also occurred
in the motor vehicles sector. The 3.8 percent reduction hi prices produces an increase in output of 5.3
percent relative to the base case.
Significant output effects are also seen in the petroleum refining sector (sector 16) with a 3.2 percent
increase, in electricity (sector 30) with a 3.0 percent increase, and in other transportation equipment (sector
25) with a 1.6 percent increase. The large gains in output for these industries are mostly due to the decline
in their prices. In manufacturing, the sectors exhibiting the most significant output effects are metal
mining (sector 2) with a 2.0 percent increase, and primary metals (sector 20) with a 1.8 percent increase.
Twenty of the remaining industries exhibit increase in output of less than 0.9 percent after pollution
controls are removed.
93
-------
Appendix A: Cost and Macroeconomic Modeling
Table 39. Percentage ISfierence ia Energy Prices Between'
''the Control and No-control Scenarios. ' r
Year
1973
1974
1975
1976
1977
1978
1979'
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
v
Coal
-0.44
-0.47
-0.42
-0.57
-0.74
-6.86
-0.91
-0.94
-0.97
-0.98
-1.09
-1.12
-1.21
-1.27
-131
-130
-131
-130
Refined
Petroleum
"~w JI-I-L- I-L— .-
-5.99
•4.84
-4.28
- -3.83
-3.43
-3.28
.-2:92 '
' -2.76
-2.50 <
-2.42
-2.35
-2.26
M *J 52O
~*t.O2*
'-3.35
-3.56
-3.61 "
-3.45 "
-3.03
Electric
-Utilities
T •" -->— Ji~i— i ^
-2.11
' -2,53 '
-2.19
-2.12
-2.22 '
-239T
-2.81" '
-2:97
-2.76
1
-2.63
,p-2.58,
-2.49
-2.62 '
-2.69
A**
" -2.75 i
-2,74
i
,•>
Utilities
B — ; — — i
f
-032 *
-0.44
-03]f
-044 '
' -6.59
v-JCt68
xfcn,
' -0.69
I
-0.77
4X85 ,
-0.91
-0.97
,-1.12 *
-1.18
-i-19
f I •} 1
-1J9
'-1.20
While most sectors increase output under
the no-control scenario, a few sectors decline
in size in the absence of air pollution controls.
The most notable of these are food and
kindred products (sector 7) which decline by
0.5 percent, furniture and fixtures (sector 12)
which decline by 0.6 percent, and rubber and
plastic products (sector 17) which decline by
0.3 percent These sectors are among the least
capital intensive, so the fall in the rental price
of capital services has little effect on the prices
of outputs. Buyers of the commodities
produced by these industries face higher
relative prices and substitute other
commodities in both intermediate and final
demand. The rest of the sectors are largely
unaffected by environmental regulations.
Changes in Employment Across Industries
The effect of the CAA on employment
presents a much more complicated picture.
Although Jorgenson-Wilcoxen is a full-
employment model and cannot be used to
simulate unemployment effects, it is useful for mma^i^^mmmmmmmmm^^^mmi^^a^mmmmmm*
gaining insights about changes in the patterns
of employment across industries. Percentage changes hi employment by sector for 1990 are presented hi
Figure 27, which is located at the end of this appendix.
For 1990, the most significant changes hi the level of employment relative to the control scenario
occur in motor vehicles (sector 24) which increases 1.2 percent, other transportation equipment (sector 25)
which increases 0.8 percent electric utilities (sector 30) which increases 0.7 percent, and primary metals
(sector 20) which increases 0.6 percent The level of employment is higher relative to the control case hi
10 other industries.
For a few sectors, the no-control scenario results hi changes hi real wages that cause reductions in'
employment The most notable reductions in employment under the no-control scenario occur in: tobacco
manufacturing (sector 8), which declines 1.2 percent; furniture and fixtures (sector 12), which declines 0.8
percent; rubber and plastic products (sector 17), which declines 0.8 percent; food and kindred products
(sector 7), which declines 0.7 percent, stone; clay and glass products (sector 19), which declines 0.6 percent;
and instruments (sector 26), which declines 0.6 percent. These sectors are generally those in which the level
of output was lower in 1990 relative to the control scenario, since they are among the least capital intensive
and the fall in the rental price of capital services has little effect on the prices of outputs. Buyers of the
commodities produced by these industries face higher relative prices and substitute other commodities in both
intermediate and final demand. It is interesting to note that several of the least capital intensive sectors
experience insignificant employment effects hi the short run (1975) under the no-control scenario, but
increasingly adverse effects over the 20-year period of analysis. These include food and kindred products,
furniture and fixtures, rubber and plastic products, stone, clay and glass products, and instruments.
94
-------
Appendix A: Cost and Macroeconomic Modeling
Examination of the transition of employment in the economy from the initial equilibrium to 1990
reveals that the employment effects of the CAA on motor vehicles, transportation equipment, electric
utilities, and primary metals persist over the entire period of analysis. Employment varies from: an
increase of 1.7 percent in 1975 to 1.2 percent in 1990 in motor vehicles; from an increase ^0.7 in 1975 to
0.8 percent in 1990 in transportation equipment; from an increase of 1.2 percent in 1975lo 0.7 percent in
1990 in electric utilities; and from an increase of 0.8 percent in 1975 to 0.6 percent in 1990.
Figure 24. Percent Difference in Real Investment Between Control and No-control Scenarios.
MM; n» wn int •am wn mi i«o mi
1K2 1H3 1W4
YMT.
1MB 1M» 1WT 1
95
-------
Appendix A: Cost and Macroeconomic Modeling
Figure 25. Percent Difference in Price of Output by Sector Between Control and No-control Scenario for 1990.
10 11 12 II 14 1« 1« IT W.W a> « 3S M X 20 24272«2»303132>3
Sector
Figure 26. Percent DflSei
*»
«%
4%
9K
S M
•
a.
t%
0%
•t*
-a%
rence in Quantity of Output by Sector Between Control and No-control Scenario for 1990.
- ; ..: - :, - '
-
1 1
Illl. ... -H -ll
1 •• LJ_
r | • - -- »
Sactor
>282>272>2>303132333438
96
-------
Appendix A: Cost and Macroeconomic Modeling
'
[Figure 27. Percent Differei
zo%
UK
§ °-°*
a.
-1.0%
•a«
ice in Employment by Sector Between Control and No-control Scenario for 1990.
..1 .11 1. II -.1.
»| ipl ' 1 | I1 |l '••
r- "
12345875 :8 .10 11 12 13 14 15 18 17 IS IB » 21 22 23 M 262827282030 31 3233M36
Sector
M
Uncertainties in the Cost Analysis
Potential Sources of Error in the Cost Data
f * '" '
Because of the importance of the Cost of Clean data for this assessment, the project team investigated
potential sources of error due to the use of industry's self-reported costs of compliance with air pollution
abatement requirements. Concerns about the accuracy of responses include (1) misreporting by firms in
response to federal agency surveys, and (2) omission of important categories of compliance cost from the
data collected or reportedHy these federal agencies.75 Table 40 contains a summary of the results of the
analysis.
75 Memorandum 6om Industrial Economics, Incorporated to Jim DeMocker (EPA/OAR) dated 10/16/91 and entitled "Sources of Error in
Reported Costs of Compliance with Air Pollution Abatement Requirements."
97
-------
Appendix A: Cost and Macroeconomic Modeling
Wiable40. Potentia^Sources of Error and Their Effect on total Costs of Compliance, * ,
V I
tffli
l mil i
1 71
•i(iii(i|i
.4.4
«
rifi
*'!
^
i
i '
Source of Error
Lack of Data at Firm Level
Misallocation of Costs:
Inclusion of OSHA and Other
Regulatory Costs
Exclusion of Solid Waste Disposal
Costs Related to Air Pollution
Abatement
Exclusion of Costs;
Exclusion of Private R&D Expenses
Exclusion of Energy Use by Pollution
Abatement Devices1**
Exclusion of Depreciation Expenses®
Exclusion of Recovered Costs
Omission of Small Firms
NET EFFECT
Effect on Capital Costs '
Under-reported
Percent Unknown'
'
Over-reported
Percent Unknown
—
:_
' r
Under-reported by 1 to 2% >
Under-reported
Effect on O&M Costs
„ > Under-reported,
Percent Unknown „
, ' , '<,. , , ',
•Over-reporteti
Percent Unknown
, /Under-reported
Percent Unknown
• , > <
i
Under-reported by 14 to.
(varies by year)
Under-ieportedby 1 to 3%
(varies by year)
Under-reported by 1 to 2%
(varies by year)
s * ' \ ** * ',
Over-reported by 1% Plus
1 Under-reported by 1 to 2%
Under-reported^
M Escrgy outlays ore part of the data on O&M costs and depreciation expenses ore no/. Accordingly, in the J/W model, energy outlays are
considered along with other operating expenditures in terms of their impacts on unit costs. Depreciation is represented fully in the capital
" accumulation process, as the undepreciated capital stock at the beginning of any period gives rise to the flow of capital services available to
, producers aad consumers. ' „ / „> - *" $ j ' , " £'
*'* ' l^i,' * J ' : ' " ', ' - ,\ ', '<,'«''
Source: Industrial Economics, Incorporated, memorandum to Jim DeMocker, EPA/OAR, "Sources of Error in
Reported Costs of Compliance with Air Pollution Abatement Requirements." October 1(5/1991.
..' Uji.ii1 i"t Jl I M' II I' 1 . >. " ' - ;, , , • - u , *' u- » '
The major finding from this analysis indicates that total O&M costs are likely to be under-reported due
to exclusion of private research and development (R&D) expenditures. Note, however, that although these
costs were excluded from those used for the macroeconomic modeling, they were included in the overall
direct cost estimate of the CAA; see "Other Direct Costs," above. These costs are excluded from the
macromodeling because they cannot be disaggregated by industry and, more importantly, because there is
no information on what was purchased or obtained as a result of these expenditures.
Based on the need indicated by the lEc review, modifications to the BEA data were made to remedy
some of the biases noted above. In particular, recovered costs for stationary source air pollution, e.g. sulfur
98
-------
AppendixA: Cost and Macroeconomic Modeling
removed using scrubbers that is then sold in the chemical market, have been accounted for in the data set
used in the model runs. •
Mobile Source Costs . ;:-:>
For the Section 812 analysis, EPA used the
best available information on the estimated cost of
mobile source air pollution control. Several other
sources of cost estimates exist, however, including
a cost series produced by the Department of
Commerce Bureau of Economic Analysis (BEA),
which is summarized in Table 41. The BEA
estimates differ significantly from EPA estimates,
particularly with respect to estimates of capital
costs and the "fuel price penalty" associated with
the use of unleaded gasoline.
EPA's capital cost estimates are based on
estimates of the cost of equipment required by
mobile source regulations. BEA's estimates are
based on survey data from the Bureau o| Labor
Statistics (BLS) that measures the incre|l|m the
per-automobile cost (relative to the previous
model year) due to pollution contro|ind fuel
economy changes for that model year. The
differenc i
annual
factor of ^
costs to the e
of comrx)nentsej
costs for
is significant: BEA's
^*SS?5*£>«a^WW?iV*sf:.;i&
^JlEPA's by a
I. EPA may underestimate
engineering cost estimates
• design and development
BLS estimates
add me incremental a^ffi^ts to all past costs
tf\ /1^r4x7ic» f/^fol /nin*0't%f_'i7a^Jr\efciV:.^r:>Oii/^l« on
to derive total cninent-year^sts; Such an
apprSach overestimates costs td the extent that it
i account for cost savings due to changes in
aent mixes over time.
Table 41. BEA Estimates of Mobile Source Costs.
•Year
1973
^1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
, 1986
1987
1988
1989
1990
Capital
' Em
, 1,013
1418
2431
2,802
-3,371
3,935"
4,634
5463
7^29
7,663
v 9426
11J900
13,210
14368
13,725
16,157
15,340
14421 -
•
, Net
I&M*
1404 f
1,380
1420
1,420
1 1,289
1,136
,931
726
552,
409 ,
274,
118
J165
(331)
(453)
(631)
<271)
(719)
Fuel
Price
Penalty
* A.
' *
97
-309
- 701
1^09.
1,636-
2,217
^996
, 3418
4,235
4,427
4995
4422
3,672'
3,736
1,972s
1370
Fuel v
Economy
Penalty '
" ,' ^97
'1480*
.,1444
,1363,
1,408
"1J397
1^792'
" - 2^20
",. 2^252
-1,876
lv *. \
1433"
'895'
658
420
183
(55)
* Inspection and maintenance costs less fuel density savings
and maintenance savings.
e Jpiirce pollution control devices required the use of unleaded fuel. Unleaded gasoline is
uce than is leaded gasoline, and generally has a greater retail price, thus imposing a cost
'EPA estimated the "fuel price penalty" by using a petroleum refinery cost model to
determine the expected difference in production cost between leaded and unleaded gasoline. BEA's "fuel
price penalty" was the difference between the retail price of unleaded gasoline and that of leaded gasoline.
A detailed description of the data sources, analytic methods, and assumptions that underlie the EPA
and BEA mobile source cost estimates can be found in McConnell (1995).
99
-------
Appendix A: Cost and Macroeconbmic Modeling
Endogenous Productivity Growth in the Macro Model
For each industry in the simulation, the JW model separates price-induced changes in factor use from
changes resulting strictly from technical change. Thus, simulated productivity growth forjfaeh industry
has two components: (a) an exogenous component that varies over time, and (b) an endogenous component
that varies with policy changes. Some reviewers have noted that, although not incorrect, use of
endogenous productivity growth is uncommon hi the economic growth literature. EPA conducted a
sensitivity run of the J-W model, setting endogenous growth parameters to zero (i.e., removing endogenous
productivity growth from the model).76
Endogenous productivity growth is an important factor in theJW model. For example, for the period
1973 to 1990, removal of the endogenous productivity growth" assumptions reduces household income by
2.9%-3.0% (depending on whether one uses a world with CAA or one without CAA as the baseline). In
comparison, removal of CAA compliance costs results in a (X6<%-0.7% change in household income
(depending on whether one uses, as a baseline, a world with or one without endgenous productivity
growth). That is, use of the endogenous productivity growth assumption has four to five times the impact
of that of CAA compliance costs.
Although very important to the simulated growth of the economy within any policy setting, the
endogenous productivity growth assumption is less important across policy settings. Under the base (i.e.,
"with endogenous productivity growth")_scenario, the aggregatejwelfare effect (measured as EVs, see
above) of CAA compliance costs and indirect effects is estimated to be 493 billion to 621 billion dollars
(in 1990-value dollars). If one removes the endogenous productivity growth assumption, the aggregate
welfare effect declines to the range 391 billion to 494 billion dollars (in 1990-value dollars) [Jorgenson et
al., 1993, pg. 6-15], a reduction of about 20 percent. : ,
Stationary Source Cost Estimate
Revisions'
As noted above on page 8,3, the costs
used for stationary sources^in the macro-
modeling (and retained in this cost analysis)
were projected for several years in the late
1980s. Since that time, BEA hasi released
historical expenditure estimates for those
years based on survey data. A comparison
of the expenditure series can be found hi
Table 42. Apparently, EPA's projections
overestimated stationary source compliance
expenditures by approximately $2 billion
per year for the period 1987 to 1990. Since
expenditures from all sources are estimated
to be $18 billion -$19 billion (current
dollars) per year during 1987 to 1990, this
implies that EPA has overestimated
compliance expenditures by more than ten
',Table 42. Comparison of EPA and BEA Stationary Source
Expenditure Estimates (millions of current dollars)
Private sector Gov't Enterprise
Year capital O&M capital O&M
EPA Estimates
Total
1986
1987
1988
1989
1990
4,090 7,116
4,179 7,469
4,267 7,313
4,760 7,743
4,169 8,688
BEA Estimates
1986 4,090 7,072
1987 3,482 5;843
1988 3,120 6,230
1989 3,266 6,292
1990 4,102 6,799
312
277
243
235
226
312
246
121
229
200
' 146
130
161
173
154
"i82
141
161
152
154
11,658
12,t>55
11,984
, 13,237
11,656
9,712
9,939
41,255
,.,,,-> > r
"Recovered Costs" are not included in this table. r t •
Sources for "BEA Estimates": for 1986, "Pollution Abatement and Control
Expenditures," Survey of Current Business (BEA) Tune 1989, Table 7; for
1987-90, BEA May 1995, Table 8.
For greater detail, see Jorgensoa et al., 1993.
100
-------
AppendixA: Cost and Macroeconomic Modeling
percent during this period. Although a substantial overstatement for those years, the $2 billion per year
overestimate would have little impact (probably less than two percent) on the discounted present value in
1990 dollars of the 1973 to 1990 expenditure stream.
Amortization Period for Stationary Source Plant and Equipment
In developing annualized costs, stationary source capital
expenditues were amortized over a twenty-year period. That
is, it was assumed that plant and equipment would depreciate
over twenty years. It is possible that stationary source plant
and equipment has, on average, a useful lifetime significantly
greater than twenty years. The Project Team tested the
sensitivity of the cost analysis results to changes in stationary
source capital amortization periods.
Table 43 presents total annualized compliance costs
assuming a 40-year amortization period for stationary source
capital expenditures (all other cost components are unchanged
from the base analysis). All costs are in 1990-yalue dollars, ad
three alternative discount rates are used in the annualization
period. Table 44 presents the results discounted to 1990, and
compared to the base case results (i.e., using a twenty-year
amortization period). Doubling the amortization period to 40
years decreases the 1990 present value of the 19J3 to 1990
cost stream by approximately 40 billion dollars. This
represents sfcoange of six percent to nine percent depending
on the discount rate employed.
•"•• f $ -J1"
Table 44, ErfecTqCAmortizatioa Periods on
Annualized Costs Discounted to 4990 (billions
*of$1990) , '' '
Discount rate
20-yr amortization period 417 523 ' 657
<• v ^ *" M
40-yr amortization period ^379* 483 617
Table 43, Annualized Costs Assuming
40-Year Stationary Source CapM •
Amortization Period, 1973 to 1990
($1990 millions).
Year
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1O
-------
Appendix A: Cost and Macroeconomic Modeling
Cosf and Macroeconomics Modeling References
Bureau of Economic Analysis, U.S. Dept. of Commerce, "Pollution Abatement and Control
Expenditures," Survey of Current Business, various issues.
Chase Econometrics Associates, Inc., "The Macroeconomic Impacts of Federal Pollution Control jj;
Programs: 1976 Assessment," Report prepared for the Council on Environmental Quality and the ?r
Environmental Protection Agency, 1976.
Congressional Budget Office, Carbon Charges as a Response to Global Warming: The Effects of Taxing
Fossil Fuels, Washington, DC, U.S. Government Printing Office, August 1990.
Data Resources, Inc., "The Macroeconomic Impacts of Federal Pollution Control Programs: 1978
Assessment," Report prepared for the Environmental Protection Agency and the Council on
Environmental Quality, 1979.
Data Resources, Inc., "The Macroeconomic Impact of Federal Pollution Control Programs: 1981
Assessment," Report prepared for the Environmental Protection Agency, July 17,1981.
Freeman, A.M., "Air and Water Pollution Policy," in P.R. Portney (ed.), Current Issues in Z7.S.
Environmental Policy, Johns Hopkins University Press, Baltimore, 1978.
Hazilla, M., and RJ. Kopp, "Social Cost of Environmental Quality Regulations: A General Equilibrium
Analysis " Journal of Political Economy, Vol. 98, No. 4, August 1990.
Industrial Economics, Incorporated, "Sources of Error in Reported Costs of Compliance with Air Pollution
Abatement Requirements," memorandum to Jim DeMocker, EPA/OAR, October 16,1991.
Jorgenson, Dale W. and Peter J. Wilcoxen, "Environmental Regulation and U.S. Economic Growth,"
RAND Journal of Economics, Vol. 21, No. 2, Summer 1990(a), 314-340.
Jorgenson, Dale W., and Peter J. Wilcoxen, "Energy, the Environment and Economic Growth," in
Handbook of Natural Resource and Energy Economics, Allen V. Kneese and James L. Sweeney, eds.,
^Volume 3, Chapter 2% North-Holland, Amsterdam, forthcoming, 1993.
Jorgenson, Dale W., and Peter J. Wilcoxen, "Intertemporal General Equilibrium Modeling of U.S.
H E^ Journal of Policy Modeling, Vol. 12, No. 4, Winter 1990(c), 715-744.
Jorgenson, Dale W., Richard J. Goettle, Daniel Gaynor, Peter J. Wilcoxen, Daniel T. Slesnick, "The Clean
Air Act and the U.S. Economy," Final report of Results and Findings to the U.S. EPA, 27 August
1993.
Jorgenson, Dale W., arid Barbara M. Fraumeni, "Relative Prices and Technical Change," in E. Berndt and
B. Field, eds., Modeling and Measuring Natural Resource Substitution, MTT Press, Cambridge, MA,
1981.
102
-------
Appendix A: Cost and Macroeconomic Modeling
Jorgenson, Dale W., and Barbara M. Fraumeni, "The Accumulation of Human and Nonhuman Capital,
1948-1984," in R.E. Lipsey and H.S. Tice, eds., The Measurement of Saving, Investment, and Wealth,
University of Chicago Press, Chicago, n, 1989.
Kokoski, Mary F., and V. Kerry Smith, "A General Equilibrium Analysis of Partial-Equilibrium Welfare
Measures: The Case of Climate Change," American Economic Review, Vol. 77, No. 3, June 1987,
331-341.
McConnell, Virginia, Margaret A. Walls, and Winston Harrington, "Evaluating the Costs of Compliance
with Mobile Source Emission Control Requirements: Retrospective Analysis," ResdurcesTfbr the
Future Discussion Paper, 1995.
; x \
U.S. Environmental Protection Agency, Environmental Investments: The Cost of a Clean Environment,
Report to the Congress, Office of Policy, Planning and Evaluation, EPA-230-12-90-084, December
1990.
Verleger, Philip K., Jr., "Clean Air Regulation and the L^. Riots," The Wall Street Journal, Tuesday, May
19,1992,A14.
Wilcoxen, Peter J., The Effects of Environmental Regulation and Energy Prices on U.S. Economic
Performance, Doctoral thesis presented to the Department of Economics at Harvard University,
Cambridge, MA, December 1988. _
103
-------
Appendix A: Cost and Macroeconomic Modeling
104
-------
Appendix Bs Emissions
This appendix provides additional detail regarding the methodologies used to estimate control and
no-control scenario emissions from each of the six principal emission sectors: industrial combustion,
industrial processes, electric utilities, on-highway vehicles, off-highway vehicles, and
commercial/residential sources. This appendix is adapted from the draft report "The Impact of the Clean
Air Act on 1970 to 1990 Emissions;
Section 812 Retrospective Analysis,"
March 1,1995 by Pechan Associates,
which distills the methodologies and
results associated with the emission
modeling efforts by Argonne National
Laboratory (ANL), ICF Resources
Incorporated (ICF), Abt Associates
(Abt), and the Environmental Law
Institute (ELI).
The control scenario emission
results are similar, but not identical, to
official EPA historical emission
estimates provided by the EPA
National Air Pollutant Emission
Trends Reports.77 Comparisons^
between the current estimates and the
Trends data for SOj, NO,, VCFC, and
CO are present in Figures 28,29;
31, and 32, respectively. No
comparison between the particulate
matter (PM) emission estimates
developed for this analysis and the
estimates generated for the EPA
Trends Report are presented,
however, since Trends estimates are
jpresented for only PM^a subset of
theTSP measures developed for this
analysis. More detailed tables
providing emission estimates by
sector aWby target year for TSP,
SOj, NO*, VOC, CO, and Lead are
presented in Tables 61,62, 63,64, 65,
and 66, respectively, at the end of this
appendix.
Figure 28. Comparison of Control, No-control, and Trenifc SO,
Emission Estimates.
40
30
§
.2 »
1120
I
10
-•-Control
•+-No-Control
^•TRENDS
1975 1980 1985
Year
1990
Figure 29. Comparison of Control, No-control, and Trends NOX
Emission Estimates.
40
30
" =§20
a I
10
1975
1980
1985
1990
Year
77 EPA/OAQPS, "National Air Pollutant Emission Trends 1900 -1994," EPA-454/R-95-011, October 1995.
105
-------
Appendix B: Emissions Modeling
Figure 30. Comparison of Control, No-control, and Trends VOC
Emission Estimates. '
Though the EPA Trends and the present study emission profiles are similar to each other, they should
not be expected to match precisely. This is because the emission estimates developed for the present study
are based on modeled macroeconomic and emission sector conditions. Even though the macroeconomic
and sector models themselves are constructed and calibrated using historical data, modeled replications of
historical trends would not be
expected to precisely capture actual
historical events and conditions which
affect emissions. Relying on modeled
historical scenarios is considered
reasonable for the present analysis
since its purpose is to estimate the
differences between conditions with
and without the CAA Comparing
actual historical emissions with
modeled no-control emissions would
lead to an inconsistent basis for
comparisons between scenarios.
Using models for both scenarios
allows potential model biases to
essentially cancel out.
H
40
30
M
10
1975
1980 1985
Year
1990
Figure31. Comparison of Control, No-control, and Trends CO
Emission Estimates.
200
150
§
I lioo
I1
I 50
ta-
in general, however, these
comparisons show close •
correspondence between control -
scenario and Trends estimates with
the largest differences occurring for
VOC and.CO emissions. The Trends
report VOC estimates are generally highejr
than the control scenario estimates
due to the inclusion of Waste
Disposal and Recycling as a VOC
source in the Trends report. This
inconsistency is of no consequence
since Waste Disposal and Recycling
sources were essentially uncontrolled
by the historical CAA and therefore
do not appear as a difference between
the control and no-control scenarios.
The higher CO emission estimates in
the Trends Report are primarily
associated with higher off-highway vehicle emissions estimates. Again, since off-highway emissions do
not change between the control and no-control scenario in the present analysis, this inconsistency is of no
consequence.
The following sections of this appendix summarize the methodologies used to model control and no-
control scenario emissions for each of the six major emission sectors. Additional details can be found in
the supporting documents listed in the References section of this appendix.
1975
1980 1985
Year
1990
106
-------
Appendix B: Emissions Modeling
industrial Boilers ana Processes
••/ ' , •
For the purposes of the retrospective analysis, the industrial sector was divided into two components:
(1) boilers; and (2) industrial processes and process heaters. The factors affecting emissions from these
two source types are different, and, as a result, separate methods were used to calculate control and
no-control scenario emissions in each of the target years. To analyze the change in emissions from
industrial boilers, ANL used the ICE model (Hogan, 1988). This model was developed under the auspices
of NAPAP to forecast State-level fuel choice and emissions from conventional, steam raising, sincliiitrial
boilers. For the retrospective analysis of industrial processes and fuel use emissions from process heaters,
ELI used the EPA Trends methods and the ANL MSCET database (EPA, 1991; Kohout, 19916). The
Trends report contains estimates of national emissions for a variety of industrial sources for the time period
of interest. The MSCET data base provided the spatial distribution used to calculate State-level emissions.
The distinction between industrial boilers and non-boiler industrial processes was necessitated by the
structure of the CAA regulations and by the factors affecting emission levels from these two source types.
Boilers are regulated differently from processes and process heaters. Emissions from industrial processes
are primarily a, function of levels of industrial activity. The emissions from fuel combustion, however, are
a function of energy use and fuel choice as well as industrial activity. Fossil fuel emissions in the absence
_ of the CAA are not proportional to industrial output, since the level of energy use is a decision variable for
the firm in its production process. Therefore, in the ICE model simulations used to estimate no-control
scenario boiler emissions, the level (and type) of energy use were determined first, and then the effects of
emission regulation were taken into account.
Overview of Approach
Industrial Boilers
~ r " _ ' . ..'.-•.-
ICE model inputs include fuel prices, total fossil boiler fuel demand by industry type, and
environmental control costs. The outputs of the ICE model were SOj, NO*, and TSP emissions by State,
industry, and boiler size class. The model runs in 5-year increments and has a current base year of 1985.
~~ The boiler demand input data were required at the State level. The following seven industry types
were included in the ICE model: Standard Industrial Classification (SIC) codes 20, 22, 26,28, 29,33, and
"other manufacturing." ANL's approach assumed that industrial boiler fuel use occurs only in the
manufacturing sector. Fuel price data in each of the target years was required at the Federal Region level.
Prices by grade of coal and petroleum product, such as sulfur content and heating value, were used by the
model. These prices were used to determine the cost of compliance and to determine emissions when the
regulations are not binding. .
Control costs were computed by engineering subroutines in the model. These costs were used by the
ICE model's fuel choice component to determine the effect of CAA-related costs on the market share of a
particular fuel. This fuel choice decision only applies to new industrial boilers, since the cost of existing
emission controls are not in the ICE data base and fuel choice is not re-evaluated for existing boilers.
107
-------
Appendix B: Emissions Modeling
Industrial Processes and In-Process Fuel Combustion
The calculation of historical emissions from industrial processes uses EPA Trends methods to estimate
national emissions for the analysis years, then allocates these emissions to States using the State shares
from the MSCET data base.
' i *'* •
MSCET uses a variety of methods to estimate historical emissions for the various industrial sectors. _,:::.
For industrial process emissions, MSCET is based on historical data on industrial activity to allocate Jjv
emissions based on the State level distribution of the polluting activities. The" State level distribution and
benchmark is based on the 1985 NAPAP Inventory (EPA, 1989). This approach implies that the MSCET
data corresponds directly to the 1985 NAPAP Inventory, and that, for any State, the sum of the emissions
from Source Classification Codes (SCCs) that comprise the MSCET industry sector are equal to the
MSCET data for that State and sector. Data from Trends are used by MSCET to provide information on
changes in the aggregate level of control for years other than the 1985 benchmark. Since no direct
correspondence existed between the Trends data and MSCET, a relationship was developed to link
MSCET sectors to Trends industry categories to industry categories in the J/W model, which was used to
change activity levels for the no-control scenario.
Table 45 shows the relationship between the sector definition used by MSCET, Trends, and the J/W
model. The mapping from MSCET to J/W and Trends is used to provide the changes in aggregate activity
and emission control for the calculation of no-control scenario emissions.
Establishment of Control Scenario Emissions
Energy use and corresponding emissions were broken down between boilers and non-boiler industrial
processes. The latter category includes furnaces, kilns, internal combustion engines (e.g., compressors),
and other non-steam types of process heat The focus of this analysis is on boiler emissions, which were
subject to increasing stringent regulations over the 1970 to 1990 period. (Emissions from some types of
industrial processes were also regulated, but regulation of non-boiler sources was targeted on the emissions
from the industrial process itself, not oh its fuel combustion) For this study, ANL assumed that only boiler
fuel use is affected by emission regulations. The non-steam boiler portion of industrial fuel use is not
directly affected by the CAA. This portion of the emissions may be affected indirectly by changes in
industry activity level and fuel consumption. The emissions from non-boiler industrial processes were
calculated separately by ELI.
Control Scenario Boiler Emissions
Control scenario boiler SO^ NO,, and TSP emissions were calculated by the ICE model. The MSCET
data base provided an estimate of historical emissions for total fossil fuel combustion by industry. Since
MSCET does not identify the two required components of boiler and non-boiler emissions, ANL defined
the residual of the ICE model control scenario and MSCET as the non-boiler or in-process fuel use
emissions. For the relevant study period, MSCET provided a control scenario
108
-------
a
,|
I
-1
i
**
-,
S
IS
I
:I
If!
Jf
£
Ill 3'£ 5';
.1
s -
.9,8
. §
-*
_
S f f 5<
« -g 8- •§
<3
1 -i
B. .2
11
«
s s
o S
ON
O
-------
•a
9
•3
i
I
1
t
5
I
Ck
e
1
I.
il
I i i i i
lliiii
£ £ £ £
li
" 3
o* o. a. in in in
i
i
1
.
I
1 1
-------
estimate of total boiler and non-boiler emissions, which was used to calculate the control scenario State-
level boiler emissions based on a special run of the ICE model.78
In order to use ICE to model the historical emissions path, it was necessary to construct a new ICE
model base year file and new user input file so that the model could begin its calculations from 1975
conditions. Construction of the base year file is discussed first. The construction was completed in two
stages, using two different data sources. The user input file has several elements, including energy prices
and historical boiler fuel use. The description of the construction of the control scenario user input file
follows the discussion of the ICE base year file. The model base-year file provided the energy use in
boilers and corresponding emission control regulations (State Implementation Plans -SDPs- for example)
by severaltcategories. These categories include: „« ~
\, ^_ 4 '
> State;
>• Industry group (one of seven);
»• Fuel type (natural gas, distillate or residual fuel oil, and coal);
> Boiler size class (MMBTU/hr, one of eight categories);
> Utilization rate (one of five categories); and
>• Air quality control region (AQCR).
For the purposes of ANL's analysis, only the first three categories were assumed to vary. In other
words, for each State, industry, and fuel type-combination, the distribution of boiler size, utilization rate,
and AQCR was assumed to be constant. Over time, however, changes in the aggregate composition of
State, industry, and fuel type would cause corresponding changes in the aggregate composition of the other
three characteristics. As mentioned previously, the current base year file was 1985. The retrospective
analysis required a 1975 base year. Because of data limitations, the approach to construct a new base year
was achieved in the following two steps: the construction of a 1980 interim base year file from the 1985
file, and thea the construction of the 1975 file from the interim 1980 file. ,
"^ ^ V —*
Estimates of boiler fossil fuel consumption in 1980 for each State and major fuel type were provided
by Hogan (Hogan, 1988). These estimates are based on the assumption that the industry mix, size,
utilization, and AQCR distribution within a State are constant. Through assuming this relationship, the
1985 ICE base year was scaled to matclithe data for 1980, thus forming the 1980 interim base year data.
To construct the 1975 base year file, the assumption of a constant industry mix for a State and fuel
type was no longer necessary, since detailed data on each industry for 1980 and 1975 were available from
" chased Heat And Ppjfer (PURHAPS) model data files (Werbos, 1983). These PURHAPS data files
; derived from the Anlnual Survey of Manufactures: Fuels and Electric Energy Purchased for Heat and
||wer (DOC, 1991). The available data in these files were for total fuel use not boiler fuel use. To make
of these data, it was necessary to assume that the fraction of fuel used in boilers, for any given State
ani industry, remained constant from 1975 to 1980. To the extent that the fraction of boilers' heat versus
process heat applications is a function of the specific industrial production process, this assumption is
reasonable. .
78 MSCET does not provide State-level estimates of TSP, while ICE does. To estimate total regional TSP from fuel combustion, the Trends
model was employed. These national emissions estimates were allocated to the States based on the State-level shares of TSP from the NAPAP
inventory.
,111
-------
Appendix B: Emissions Modeling
Based on the assumption of constant boiler fuel fraction of total fuel use, the ratio of 1975 to 1980
energy use for each State, industry, and fuel type was applied to the corresponding record of the 1980
interim base year file to produce 1975 base year files.
Control Scenario Industrial Process Emissions
To estimate boiler emissions of sulfur oxides (SOJ, NOX, and VOC from industrial processes, data
from Trends were used. The percentage change in national emissions by Trends category was applied to
the appropriate sector from MSCET to obtain State-level emissions. In some cases there are several .
categories in Trends that match directly with MSCET categories (see Table 45). In these cases, me Trends
sectors were aggregated and the percentage change was computed. It was assumed that the level of control
in each industry sector implied by Trends was uniform across States. The changes in emissions in each
State are not equal to those at the national level, since the industry composition in each State varies.
/
Development of Economic Driver Data for the
* -s-fcggj"*; *• *.--"*" -
Control Scenario - Industrial Boilers anst
The results of the J/W model were the primary source of activity in the ICE model driver data. These
results were also used by ELI to produce the national results for industrial processes from Trends. Both
ICE and Trends use the forecasted change in industrial activity that results under the no-control scenario.
These data were in the form of industry specific changes in energy consumption and industrial output, for
boilers and industrial processes. .
Economic Driver Data for Industrial Boiler Approach
Using the 1975 base year file as a starting point, the ICE model estimated fuel choice and
emissions based on a user input file containing total boiler energy demand and regional energy prices. The
1975, interim 1980, and original 1985 base year files contained the required information on energy
demand for each industry group and State, so the data in these three files were aggregated across fuel type,
and other boiler characteristics (for example, size). These aggregated data provided the energy demand for
three of the target years. Since 1990 State-level data on energy use by industry group were not available at
the time of the study, the NAPAP base case forecast for the ICE model for 1990 was used to provide the
demand data for this year.
The user input file for ICE also requires a price input for each target year. These prices were input by
Federal Region for distillate oil, 4 grades of residual oil (by sulfur content), natural gas, and 11 grades of
coal (by sulfur content and coal rank, i.e., bituminous and sub-bituminous). Prices for 1985 and 1990 were
obtained from the NAPAP base case user input file. The prices for 1975 and 1980 are from U.S.
Department of Energy (DOE) data on State-level industrial energy prices (DOE, 1990). Regional prices of
natural gas, distillate oil, steam coal, and residual oil were constructed by aggregating expenditures across
States within each region and dividing by total British thermal unit (BTU) consumption for the years 1975,
1980, and 1985. Since prices by sulfur content grade are not reported by this DOE source, ANL assumed
that the sulfur premium implied by the 1985 ICE model input file was proportional to the average price.
112
-------
Appendix B: Emissions Modeling
Based on this assumption, the ratio of the regional coal and residual oil price in 1975 and 1980 to the 1985
price was applied to the 1985 price in the ICE model base case file for each grade of fuel. To provide
additional consistency between the NAPAP analysis and ANL's study, the distillate oil and natural gas
prices were benchmarked to the 1985 ICE model prices as well.
One possible inconsistency arises using this procedure. The residual oil and natural gas markets are
closely linked, particularly for industrial customers. These markets, specifically the* gas market, underwent
tremendous changes over the study period. To model the effect of these structural changes on the sulfur
premiums in residual oil would require a detailed oil and gas supply model that was beyond the scope of
this project. Moreover, the CAA regulations themselves create the potential for sulfur premiums. THis
potential effect of the CAA was not captured, though, because of the assumption of proportional fuel sulfur
premiums on residual fuel oil. The relationship between market driven sulfur premiums in the coal market
and the CAA was given additional consideration in this analysis through the use of an explicit coal supply
•model.
The J/W data for industrial energy consumptions was supplied in the form of percentage change in cost
shares. In order to compute the percentage change in the quantity of energy used, ANL used the following
identity:
The percentage change in Iris the percentage change in cost share, minus the change in price,
pluslhe change in value of shipments. These calculations were performed for each energy type
anIKhdustry sector in tip J/W model. The ICE model requires total fuel use, so the fuel specific
percentages were weighted by historical fuel consumption to produce an aggregate change in fuel
consumption to apply to the ICE model input data files.79
ICE also uses energy prices to simulate boiler fuel choices. The control scenario forecasts of
energy prices in ICE were adjusted based on the percentage changes in energy prices, by coal, oil
and natural gas. This implicitly assumes that the, oil and coal fuel sulfur premiums, by region, are
proportional to the average national price. To test this assumption for the coal market, additional
"ICE uses six of the manufacturing industries from the JW model directly. The remaining industries' percentage changes were weighted to
produce the "other" category. ,
113 ' . . /
-------
Appendix B: Emissions Modeling
modeling of the coal prices was performed using the coal market component of the ARGUS
model.
It is possible that in some regions low sulfur coal prices to the industrial sector may be lower
than the national average. This was not found to be the case. For example, hi 1990, delivered
regional industrial coal prices change by less than two-thirds of one percent. "In most cases, the
percentage change was near zero. This result appears to occur because of the highly regional
nature of the coal market. While the artificial demand for low sulfur coal may fall, power plants
near low sulfur coal reserves now find it advantageous to buy this local coal, wnich*faisle^ ffie
price back to an equilibrium level near to that of the control scenario. This is even more likely to
be true of industrial delivered prices, since industrial prices are more affected by transportation
costs than are the utility prices. No additional ICE modeling was performed.
Economic Driver Data for the Industrial Process Approach
The J/W model was also used to account for activity level changes hi the calculation of
industrial process emissions under the no-control scenario. The correspondence between Trends,
MSCET, and the J/W model was used to apply changes hi industrial activity in each target year
to each industrial process.
No-control Scenario Emissions
Industrial Boiler Emissions of SO2, NO,, and TSP
The CA^ imposed different regulations, SIPs, and New Source Performance Standards (NSPS) that
apply to industrial boilers of varying size. The primary effect of CAA regulations on industrial boilers was
simulated by defining the Air Quality Control Region (AQCR), the resulting SIPs, and subsequent NSPS
for boilers. The industrial boiler SIP regulations were included in the ICE base year file discussed in the
previous section. Since the ICE model estimates new boiler emissions for each target year, the boiler
NSPS are input through tne ICE user files. Industrial NSPS were implemented in two phases. The 1971
regulations are imposed for the study years 1975 and 1980. The 1984 NSPS revisions are imposed in the
study years 1985 and 1990. Forthe no-control scenario, ANL set the SIPs and NSPS to a flag that
indicated "no regulation." "
Industrial Boiler Emissions of CO and VOC
" Two of. the criteria pollutants emitted by industrial fuel combustors, CO and VOC, were hot included
as outputs <>f the ICE model. Therefore, CO and VOC emissions were analyzed separately using Trends
methods. Control scenario CO and VOC emissions were taken directly from Trends.
To estimate CO and VOC emissions from industrial combustion for the no-control scenario, fuel use
for industrial manufacturing was adjusted, reflecting fuel consumption changes estimated by the J/W
model. These changes in the level of fuel consumption by industrial combustion were also used in ANL's
ICE boiler model. Changes in industrial combustion fuel use by manufacturing in the no-control scenario
114
-------
Appendix B: Emissions Modeling
are reported in Table 46. These estimates represent an average of several sectors, which were developed
by ANL as part of the modeling process for IC3E.
• '••,.'.' ' • ' A, - '
No-control scenario emissions were computed using 1970 emission factors. Since there were no add-
on controls for industrial combustion VOC and CO emissions, it was not necessary to adjust the no-control
scenario for changes in control efficiency.
Emission estimates were regionalized using State-level emissions data from industrial borers recorded
in MSCET. For the control scenario estimates, VOCs were regionalized using the MSCET State-level
shares for industrial fuel combustion. In the no-control scenario, the State-level shares were held constant.
The control scenario emissions of CO were regionalized using the control scenario NOX emissions from the
ICE model. This approach assumes that CO emissions are consistent with NOX emissions. The no-control
scenario CO emission estimates from industrial combustion sources were regionalized using no-control
NOX emission estimates from industrial combustion sources.
Industrial Process Emissions • ' ' ,
A wide range of controls were imposed on industrial processes. These emission limits are embodied in
the assumptions of control efficiencies in ihef rends model. Data on national no-control scenario
emissions from industrial processes were provided by EPA. These data were combined with MSCET to
produce regional-level results.
and indu
Associates.
Lead Emissions
ead emissions from industrial boilers
^ ^. is were "completed by Abt
nethbds used for calculating kad
emissions from mdustrial processes and industrial
boilers werej|TOfir.~ The^tarting point was the TRI,
which provides air toxics emissions data for
manufacturing facilities with more than 10 employees.
To estimate lead emissions from industrial boilers and
processes, 1990 facility-level lead emissions data were
extracted from the TRI. jThese data were then adjusted
to create estimates of lead emissions from industrial
sources under the ccptrol and no-control scenarios for
each o| the tirget years. For the control scenario, lead
emissions fff3¥9"'75> 1980, and 1985 were obtained by
extracting an emission factor and a control efficiency
for each lead-emitting industrial process in the Trends
data base. These emission factors and control
efficiencies were multiplied by the economic activity
data for each year for each process as reported in
Trends to yield estimated control scenario emissions
by industrial process. Each industrial process was
assigned a code to correspond with energy
Table 46.
No-control Scenarios.
^Yea*
1975
-
1980
' '
">
1985
'••
„
1990
Fuel Type :
Coal ,
Oil
Gas
Coal
Oil
Gas , , ;
Coal
Oil ' , ' „
Gas
Coal
Oil ;
Gas
Fuel Use Changed '
..'"'. : 1.0042^ '%
• * +:63ii > ' '•
, -.00(54 •' ;
, 40061 .-. .
•fjoib? / .."
-.0095
3ffil,
t:o°^9 V
. - • -^ , •„•/'
\0079 ":' "'f
-.0091? - ;,
115
-------
Appendix B: Emissions Modeling
consumption data by industrial process compiled in the National Energy Accounts (NBA) by the Bureau of
Economic Analysis, and emissions were summed over all processes to obtain a total for each target year.
For consistency with the other emission estimates in this analysis, industrial process no-control
, scenario lead emissions were adjusted for changes in industrial output, and for changes in emissions per
unit of output due to control technology applications. Changes in industrial output were accounted for
using results from the J/W model. Lead-emitting industrial processes in the Trends data base were
assigned to a J/W sector. For each sector, the percentage change in economic output was used to adjust the
economic activity data for that process from the Trends data base. These adjusted economic output figures
were used with the 1970 emission factors and control efficiencies to derive the estimated no-control
scenario lead emissions for each industrial process in each target year. The process-level emissions were
then aggregated to the NBA-code level as in the control scenario.
The lead emission estimates from industrial processes, by NBA code, were used to derive percentage
changes in emissions under the control and no-control scenarios by NBA code for application to the TRI
emissions data. Since TRI data are reported by SIC code, NBA codes were "mapped" to the appropriate
SIC codes, and then the percentage change for each NBA code was used to represent the percentage
change for all SIC codes covered by that NBA code.
t, •!
To calculate lead emissions from industrial boilers, Abt Associates developed estimates of lead
emissions from industrial combustion under the CAA for each of the target years. The Trends data base
contains national aggregate industrial fuel consumption data by fuel type. For each fuel type, the fuel
consumption estimate was disaggregated by the share of that fuel used by each NBA industrial category.
The Trends data base also contains emission factors for industrial fuel use, by fuel type, as well as control
efficiencies. Thp lead emissions from industrial combustion for each NBA category were derived by
multiplying the fuel-specific combustion estimate for each NBA category by the emission factor and
control efficiency for that fuel type. The result was emissions of lead by NBA code and by fuel type.
Emissions from all fuel types were then summed by NBA code. The NBA data were used to disaggregate
the industrial fuel consumption figures, based on the assumption that the ICE are the same among all
industries covered by a given NBA code.
To estimate no-control scenario lead emissions, the macroeconomic effect of the CAA and the change
in emissions per unit of output that resulted from specific pollution control mandates of the CAA were
both taken into account. As in the control scenario, the national aggregate industrial fuel consumption
estimate by fuel type was disaggregated by the share of that fuel used by each NBA industrial category.
Tlie fuel use was then adjusted in two ways: some NBA codes were specifically modeled by the ICE
mode| and for me remaining NBA codes, J/W percentage changes in fuel use were applied. These fuel
use estimates were men combined with the 1970 emission factors and control efficiencies for industrial
combustion by fuel type from the Trends data base to obtain no-control scenario combustion-related lead
emissions from industrial boilers by NBA code. These estimates of total lead emissions by NBA codes
were matched to SIC codes, and then to the data in the TRI data base. This approach assumed that an
average emission value was assigned to all reporting TRI facilities in a given SIC code.
116
-------
Appendix B: Emissions Modeling
Off-Highway Vehicles li
The off-highway vehicle sector includes all transportation sources that are not counted as highway
vehicles. Therefore, this sector includes marine vessels, railroads, aircraft, and off-road internal
combustion engines and vehicles. As a whole, off-highway vehicle emissions are'a relatively small
fraction of total national anthropogenic emissions.
Overview of Approach
The process used by ELI to determine the national level of emissions from the off- highway
transportation sector is similar to the procedure outlined above for industrial processes. To estimate the
emissions of criteria air pollutants from these sources under the no-control scenario, the historical activity
levels were held constant, rather than attempting to calculate -a new no-control scenario level of off-
highway vehicle activity. This assumption was necessary since the off-highway activity indicators (amount
of fuel consumed, and landing and take-off epics for aircraft) do not have direct correspondence with a
given J/W category. The national no-contfcilscenarioemissions of criteria air pollutants from these
sources were simply derived by recalculating emissions using 1970 emission factors.
Development of Control Scenario ^
1 ' •*£* ^ •* ^ ~r • ' - '• '.
To estimate control scenario emissions, the analysis relied on Trends methods, using historical activity
indic^tors|%rru^ionfactore; arid control efficiencies Essentially, the estimates of off-highway emissions
under the control scenario represent the historical estimates from the Trends data base.
No-control Scenario Emissions Estimates -
The: .calculation of off-highway emissions for the no-control scenario required the Trends data to be
adjusted to reflect changes in controls and economic activity in each of the target years. Tanking source
activity changes with economic activity for this section is not straightforward. The economic activity data
for"off-highway engines and vehicles are expressed either in terms of amount of fuel consumed, or in terms
of landing and take-off cycles for aircraft. Neither of these off-highway activity indicators has a direct
correspondence with a given J/W sector, making the sort of direct linkage between Trends categories and
•J/W sectoral outputs that was used for industrial processes inappropriate.
^ y • -• . .
In the absence of a link between the economic factors that are determinants of emissions from this
sector "and the available economic activity forecasts, the no-control scenario emissions of criteria air
pollutants from off-highway mobile sources were estimated based on the same historical activity levels
used for the control scenario. Although there were changes in sectoral output and personal income that
might have had an effect on off-highway vehicle usage, these changes were deemed to be small and not
likely to have a major effect on the emissions from this sector.
117
-------
Appendix B: Emissions Modeling
Emission factors for each of the off-highway sources were also held constant at 1970 levels to
calculate no-control scenario emissions for each target year. The national emissions of criteria air
pollutants from these sources were then recalculated using 1970 emission factors.
National and! State-Levei Off-Highway Emission
Estimatest ~ .. • ' ,fff
Table 47 summarizes national-level emission estimates for off-highway sources. The emission
estimates derived from using the methodology discussed above yielded results that seem counter-intuitive.
The emissions from off-highway sources, in particular the emissions from aircraft, are lower in the
no-control scenario than those projected for the control scenario for most pollutants. This is a result of
calculating emissions using 1970 emission factors, since the 1970 emission factors for aircraft are lower
than the aircraft emission factors in later years.
ELI identified several potential sources of uncertainty in the emission estimates for this sector. First,
the assumption that the total level of off-highway vehicle fuel consumption is constant between the two
scenarios may be flawed. Second, the use of 1970* emission factors in the no-control scenario may fail to,
capture significant changes in technology. These technological changes are implicitly captured in the
control scenario and it is possible that these technological changes may also have occurred under a
no-control scenario. • *
118
-------
Appendix B: Emissions Modeling
_Table 4t^ Difference in'Control and'NoKcbntrof Scenario Off-Highway Moj&fle Source Enrfssioni,
y
v^ •
,'TS^
\ s> ^
"s
v- j
/" ?
NO, '
»•*. * (
< f -v
<. •*
^sa
< zr *
*• j i
CO
^ ?
VOCs
>
•V*
5> (
Control Scenario;
No-Control Scenariot
Percentage Increase:
Control Scenario:
-No-Control Sceaarioi „
» , ' - '•'" 3
Percentage Increase;
"
Control Scenario:
">>>':
, No-Control'Scenario: 1
> *
Percentage Increase:
> > ••*
Control Scenario:
No-Control Scenario; „ *
* *- *• w >*s
Percentage Increase: ,
•> ^ V
Control Scenario;
No-Control Scenario:
y S-S *
Percentage Increase:
V ? ,-j
1975
'^ 268.6
260.8 , •
4 -3%^ X
,1^87.6
' 1,974.6
-1%'
4 364.6
363.2 .
0% ,
8^12.8 ,
8,511.0
,° 'o% - s
1,374,9 *,
IJ8SJ9
' 1%,
; ",19W ,
281d
268.8
" -4% ,^
12,176.7
, 2,150^1
^-1% "
j-
" ssr.r
528.6
0%
, 8,101.4
, '8,071.2
Q%
' 1,370.8
1,416.1
3%
. 1985
, 268,7
'261.2^
-3%
2,077^
2V042.7 "
-2%
406.4 ^
403.0 .
-1*'
7^8ll9 .
7,880.2
0%
1334.8
1,388.'6
4%
4.
. 1990;
1 ,,280;9 , -
-26&9 ,]-
" -4%'
2,085.9
- ' 2,058.9'-
~-Sv
v -m s ,.
30£5: ,
"" 386:9V '
. -1*
< *
8,079.0 '
•* -s <
"8,'077.7
S "v
6%
A
^ 1,405.0
1,485.8
' 6%
Note: Emission estimates arc expressed in thousands of short tons. Percentage increase is the differential between
scenarios divided by Ihe Control Scenario projection. K '
One possible response to the biases created by the use of 1970 emission factors for all years in the
" no-control scenario is to test how results might differ if the emission factors used for the control scenario,
which would include technological change, were also used for the no-control scenario. However, using
this treatment of emission factors, the emissions projections from the adopted methodology from non-
highway sources in the no-control scenario would be identical to the emissions projections under the
control scenario. The reason for this is that the economic activity levels were not adjusted for the
calculation of emissions under the no-control scenario.
119
-------
Appendix B: Emissions Modeling
In Older to disaggregate the national data to a State level, the methodology used the MSCET data base,
which is described starting on page 107. Emissions of VOC, SO,, and NOX were regionalized using the
State-level shares from the MSCET methodology. The emissions of TSP were regionalized by using the
State-level shares for SOX reported by MSCET, and the emissions of CO were regionalized using the State-
level shares for NOj, also reported by MSCET. The potential bias that this introduces is likely to be small,
due to the relative homogeneity of off-highway vehicle emission sources. As with regionalization of
industrial process emissions, the State-level shares are held constant between the two scenarios. To the
extent that the distribution of economic activity between States was not constant over the period of the
analysis, holding State-level emission shares constant may bias the results, although the direction and
magnitude of the potential bias is unknown.
On-Highway
This section addresses the highway vehicle portion of the transportation sector. Highway vehicle
emissions depend on fuel type, vehicle type, technology, and extent of travel. Emissions from these
vehicles have been regulated through Federal emission standards and enforced through in-use compliance
programs, such as State-run emission inspection programs. Vehicle activity levels are related to changes in
economic conditions, fuel prices, cost of regulations, and population characteristics. Emissions are a
function of vehicle activity levels and emission rates per unit activity.
TEEMS was employed by ANL to analyze the transportation sector. The modeling system links
several models, disaggregate and aggregate, to produce State-level estimates of criteria pollutants. The
system is subdivided into two modules: an activity/energy module and an emissions module. Each
module contains multiple models. TEEMS has been documented in several reports and papers (Mintz and
Vyas, li&i; Vyas and Saricks, 1986; Saricks, 1985). It has been used for several policy analyses and
assessment studies for DOE and NAPAP. This section presents an overview of the approach used to
conduct the analysis of the transportation sector. Also included hi this section is a summary of the
methodology used by Abt Associates to estimate changes in lead emissions from highway vehicles in each
targetyear. <
Overview of Approach
/' ' i
TEEMS has two modules: an activity/energy module and an emissions module. The activity/energy
module calculates emissions based on: (I) personal travel; (2) goods movement; and (3) other
transportation activity inputs.
Personal Travel
Personal travel activity and resulting fuel consumption were calculated for each target year using
procedures that disaggregate households by demographic and economic attributes. Economic driver data,
developed from U.S. Government data and macroeconomic model(s) of the domestic economy, formed the
basis for household disaggregation. Modeling procedures were employed by ANL to project movement of
households between various attribute classes, and vehicle holdings were projected in terms of the number
and type of vehicles held by each household type. National totals were then developed by aggregating the
vehicle holding estimates for each household type, accounting for the number of households of that type.
120 ,
-------
Appendix B: Emissions Modeling
Travel estimates, in terms of VMT, were calculated using the same approach, and based on the VMT of
each household type. The basis for household transportation activity projection has been empirically
established through analysis of the 1983-84 Nationwide Personal Transportation Survey (NPTS) (FHWA,
1986; Mintz and Vyas, 1991). VMT are projected using this empirical relationship, and estimates of the
elasticity of VMT to vehicle operating cost are then made. Energy consumption was estimated in each
target year using VMT, shares of VMT by vehicle type, and exogenously developed vehicle characteristics.
The following three models and an accounting procedure were employed to develop target year
personal travel activity projections: - - ^
1. The first model projected the target year distribution of households by their attributes. This
model employed an iterative proportional fitting (IFF) technique and projected the number of
households in each cell of the household matrix - each Ibf winch is defined by various
categories within six household attributes. "
2. The second model projected changes in vehicle ownership resulting from changes in income
and cost of vehicle operation. The model applied estimated ownership changes to each target
year household matrix such that the control values within each of the household attributes,
excepting vehicle ownership, remained unchanged.
3. The third model estimated the composition of household vehicle fleet by type (cars and
trucks), size, technology, and fuel.
4. An accounting procedure applied VMT pervehicle~to vehicle ownership in each combination
of household attributes. VMT and energy consumption were accumulated by vehicle type, .
size, and fuel.
BE.~JW 1. * ~ „ •
Each of these models is described separately in the following subsections.
Iterative Proportional Fitting (IFF)
This IFF model modified a control scenario matrix of household counts. A household matrix was
developed from the 1983 SPTS data'and upgraded to the year 1985 using published aggregate data. The
procedure used in constructing the 1985 household matrix has been documented elsewhere (Appendix B
of Mintz and Vyas, 1991): The matrix is defined by six attributes: (1) residential location (central city,
suburb, rural); (2) household income; (3) age of householder; (4) household size; (5) number of drivers;
cles. The household matrix has 3,072 cells, some of which are illogical (such as 1
ogical cells were replaced with zeros.
Household shares within each attribute in each target year were developed exogenously using data
from the Bureau of the Census and selected macroeconomic model runs. The projected total of households
and shares of households in each category of an attribute were supplied to the IFF model. The model
modified the control scenario household matrix to match the specified shares and total number of
households. . . '
The IFF model treated household distribution within each attribute as a set of vectors. These vectors
were scaled to match the specified shares and household total. Following the initial scaling, a gradual
121
-------
Appendix B: Emissions Modeling
scaling technique was used to move in the direction of the target shares. The scaling process was repeated
until closure was achieved for all attribute classes. Since vehicle ownership levels were estimated by the
vehicle ownership model (described in the next section), shares within the sixth household attribute
(number of vehicles held) were not specified, leaving it uncontrolled. This flexibility of an uncontrolled
attribute helped to facilitate the model operation. The number of households in each class of vehicle
ownership within the output matrix represents distribution of households using the control scenario (1985)
relationship of vehicle ownership to other household attributes.
Vehicle Ownership Projection (VOP) " ~ " ' - ,
The VOP model projected the changes in vehicle ownership resulting from changes in the number of
licensed drivers, disposable personal income, and annual fuel cost of vehicle operation. The model is
based on historical household ownership rates. A target per-driver ownership rate was computed using
disposable income and fuel cost This target rate represented desired ownership if income and fuel cost
were the only determinants. A parameter representing ownership responsibilities such as acquisition effort,
disposal effort, parking requirements, and other indirect aspects was applied to adjust this target. The new
ownership rate was used to estimate the number of household vehicles.
c j
Jr ~v
The household matrix created by the IFF model was revised to match the projected household vehicle
ownership. Household shares within the first five attributes remain constant while those within the sixth
attribute (i.e., number of vehicles) were variable. A deviation measure was defined and its value for each
class within the first five attributes was minimized. A set of simultaneous equations was solved using
Lagrangian multipliers. ,
»
Projection of Vehicle fleet Composition
The composition of household vehicles was projected for each household matrix cell using a vehicle
choice model called the Disaggregate Vehicle Stock Allocation Model (DVSAM). Vehicles are defined by
type (auto, light truck), size (small, mid-size, full-size auto; small pickup, small utility/minivan, standard
pickup, large utility/standard van; or any other size classification), fuel (gasoline, diesel, methanol, ethanol,
or compres^d natural gas), andltechnology (stratified charge, direct injection, electric, fuel cell, or
Brayton).
The model computed vehicle composition based on an individual vehicle's utility to households and
household needs. A menu of vehicles classified by the previously mentioned vehicle attributes was
supplied to the model. The menu specified characteristics of each vehicle available to households.
Vehicles were characterized by price, operating cost, seating capacity, curb weight, and horsepower.
These variables formed the basis for computing "utility" (analogous to consumer satisfaction). The
household matrix provided demographic and economic attributes which, when combined with vehicle
usage in miles, define household needs. Vehicle usage (VMT) was computed as a function of income,
number of drivers, and number of vehicles. A logit model was applied to compute vehicle ownership
shares. Several model enhancements facilitated modeling of limited range vehicles, and representation of
supply constraints and/or regulated market penetration.
Activity/Energy Computation
122
-------
Appendix B: Emissions Modeling
An accounting procedure was applied to compute personal travel activity in terms of VMT by vehicle
type. Control scenario VMT per vehicle estimates for each cell in the household matrix were developed
from the 1983 NPTS. These rates were adjusted within the procedure on the basis of changes in average
vehicle operating cost per mile for each cell. The vehicle composition projection model computes
ownership shares and share-weighted change in vehicle operating cost Elasticity values were applied to
this change.
ANL assumed that VMT per vehicle remained nearly unchanged for a household matrix cell over time
(with the exception of the effect of changes in vehicle operating cost). In other words, variation of VMT
across household types is far greater than within household types. VMT per household vehicle remained
stable during the period from 1977 to 1984 (Klinger and Kuzmyak, 1986). Some increaSe^pre observed
in recent years, which were attributed to lower fuel prices and increased household income fDOC, 1991;
FHWA, 1992). (A portion of the increase could be attributed to the method of computing average VMT
per vehicle.) The assumption that VMT per vehicle for each cell remained nearly constant and was elastic
relative to vehicle operating cost is reasonable. As households move from one cell of the matrix to
another, they "acquire" the VMT per vehicle rate of that cell. Thus, this approach accounted for changes in
VMT per vehicle due to increased household affluence7 increased rate of driver licensing, changes in fuel
price, and changes in vehicle technology.
f
Goods Movement
Energy and activity demand resulting from movement of 24 aggregate categories of commodities is
estimated by this subcomponent of the TEEMS activity module. Changes in commodity
demand/production were provided by growth indexes by two-digit SIC generated by a macro model. A
model that projects shifts in mode shares among truck, rail, marine, air, and pipeline modes was used,
foUowedbya-prdcedure to compute ton miles of fravel for each mode, VMT by fuel type for trucks, and
energy consumption by operation type for non-highway modes. The model used 1985 control scenario
data, which were compiled from railroad waybill sample and publications, waterborne commerce
publications, transportation statistics, and other sources. The procedure used in developing the 1985
control scenarioSig^l^ftas been documented in an ANL report (Appendix A of Mintz and Vyas,
"" : • ''^s
. , • .
This goods movemenffli^l^was not used for this retrospective analysis because of funding and time
constraints. A procedure td estimate truck VMT by fuel type was employed in its place. Published
historical VMT values (FHWA, 1988; 1992) were used along with VMT shares by fuel and truck type
from Truck Inventory and Use Surveys (TTUS) (DOC, 1981; 1984; 1990).
*" s • i .
Other Transportation Activities
The activity/energy module also has other models for developing activity and energy use projections
for air, fleet automobiles, and bus modes. Fleet automobile activity estimates from an earlier study (Mintz
and Vyas, 1991) were used while other modes were not analyzed.
Lead Emissions
123
-------
Appendix B: Emissions Modeling
Estimates of lead emissions in the transportation sector were developed by Abt Associates based on
changes in reductions of lead in gasoline. This estimation required the estimates of lead in gasoline
consumed over the period from 1970 to 1990 and the amount of lead content in gasoline that would have
been consumed in the absence of the CAA. These values were calculated using the quantityof bom leaded
and unleaded gasoline sold each year and the lead concentration in leaded gasoline in each target year.
Data on annual gasoline sales were taken from a report by ANL that presented gasoline sales for each State
in each target year. For the control scenario, data on the fraction of gasoline5 sales represented by leaded^,
gasoline were used. For the no-control scenario, all of the gasoline sold was assumed to be leaded. Dati-
on the lead content of gasoline was obtained from ANL for 1975 through 1990. For 1970 through ,1975,
the analysis assumed that the 1974 lead content was used. ~
Estimation of No-control Scenario Emissions
TEEMS emissions projections were carried out by ANL in the following tfiree steps:
1. Development of emission factors;
2. Allocation of highway activity to States; and
3. Development of highway pollutant estimates.
~ -«.« *
The following subsections describe the procedures used for computing highway vehicle emissions.
Development of Emission Factors
EPA's ftloBILESa Mobile Source Emission Factor model was used to provide all of the highway
veMcle emission factors used to estimate 1975 to 1990 emission rates (EPA, 1994b). Documentation of
the MOBILESa/model is found in the tJser's Guide for the MOBILES model.8*
Although the actual emission factors used by ANL are not documented in either the original ANL
TEEMS model report or in the Pechan summary report, the Project Team provided direction that defined
the emission factors to be used. For the control scenario, ANL was directed to use the official EPA
emission factors prevailing at the time for each target year. For example, the official EPA emission factor
being used in 1980 for on-highway vehicle NO, was to be used to estimate 1980 control scenario on-
highway vehicle NOX emissions. For the no-control scenario, the official EPA emission factors used to
estimate emissions in 1970 were to be used throughout the 1970 to 1990 period.
It is important to note that using the 1970 on-highway vehicle emission factors to estimate no-control
scenario emissions ior the entire 1970 to 1990 period may bias scenario emission differentials upward.
This is because it is possible that technological changes to on-highway vehicles unrelated to CAA
compliance strategies may have yielded incidental reductions in emissions. However, EPA Office of
Mobile Sources (EPA/OMS) experts indicate that the two major technological changes in vehicles
EPA/OAR/OMS, "User's Guide to MOBILES," EPA-AA-AQAB-94-01, May 1994; see also 58 FR 29409, May 20,1993.
124
-------
Appendix B: Emissions Modeling
occurring during the period of the analysis -electronic ignition and electronic, fuel injection- would have
yielded negligible emission reductions in the absence of catalytic converters;81
Another: potential bias is introduced by assuming .the CAA had no substantial effect on vehicle
turnover. However, two factors render this potential bias negligible. First and foremost* lihder the no-
control scenario retired .vehicles would be replaced by new but equally uncontrolled vehicles. Second,
no-control scenario vehicle use is greater m terms of VMT'per year. This means no-control scenario
vehicles would reach the end of their service lives earlier, offsettragto some exfent1 the alleged incentive,
to retire vehicles earlier due to costs imposed by CAA control requirements,, „ „ -«, ' •.
''. •' • -•-.'•' . .•'".''•.•'.'•."• *'<""" ^ \
• r . . • ' '. • •' ^M1** ,_ *•
Allocation of Highway Activity to States •,-..'- - ? "
TEEMS1 activity module generated national activity and energyestimates. These activity totals were
allocated to States through a regionalization algorithm that used time series data on historical highway
activity shares by Slate. A trend extrapolation methodology was usedthat^bilizes shifts after,5 years
in the future. For the retrospective analysis, historical highway activity shares" for each target year were
developed using data published by the Federal Highway Administration (PHWA) (FHWA, 1988; 1992).
Development of Highway Pollutant Estimates "" f
.'. Highway emission estimates were calculated in both scenarios for each target year using VMT -
estimates generated by TEEMS and emission factors from MOBILESa. Control scenario activity levels
were adjusted for the no-control scenanoTising economic forecasts and historical data. .
•••••', ' ^ "*. ,, • :'''•'."•''..'•' ••'
Control Scenario Emissions Calculation* A ^ \"
Control scenario data for/the transpprtation sector were compiled from several sources. Household .
', counts and'iliaresi of louseholds b^^ix, attributes were obtained from various editions of the Statistical
Abstracts of the tlnjtedjStates. Hous^oljijncome information wasi obtained from the control scenario
run of the J/W moHel. Fuel prices were^^ obtained, from the Annual Energy Review (PGE, 1992) while
vehicle fueLeconomy and aggregate VMT per vehicle were obtained from Highway Statistics (FHWA,
1988; 1992). Table 48 lists datajsources for the control scenario run. . .
Table 49 shows household shares prepared for the EPF model.' The total number of households
mc|eased from 63.4 million in 1970 to 93.3 million in 1990. A gradual shift from rural to urban was. -'\
observed with movemenfto suburbs,within urban, areas. The effect of economic downturns in 1975 and
1980 was an increase in'share for the lowest income category; more households moved to the highest
resulted in increases in smaller and younger households. The trend in younger households reversed after
1980 as household formation slowed. Average household size dropped from 3.2 in 1970 to 2.67 in 1990.
84 Telephone conversation between Jim DeMocker, EPA/OAR and EPA/QMS/Ann Arbor Labpratory staff (date wikntfwn). Nevertheless,
.tiie Project Team did consider reviewing emission factors .for European automobiles to attempt to estimate no-control scenario emission factors
for 1975 through 1990 reflecting the use of electronic &el injection and electronic ignition but no catalytic converter. However, the'Project
Team concluded that differences in fuel/air mix ratios used in Europe would probably obscure any differences in emission rates attributable to
me use of electronic fuel injection and electronic ignition. . , . "- ' '
"'•-•.;.-''. ' •.. •• / >.." ' .-,• •.. 125 .-. ,' '• • .•• • • --• ."• • •-•
-------
Appendix B: Emissions Modeling
The number of licensed drivers increased throughout the analysis period as more and more young people
.were licensed to drive. 1 , " . , . . . •. •
Data for the VOP model included disposable income per capita, fuel price, overall personal vehicle
fuel economy, and annual usage in terms of VMT. Table 50 shows these data for each year in the '.
analysis period. •'•'.'. ,
1 '•• .'••• ••'•••:••' • ' «. *# > *
. * if ^
Data preparation for the model that projected household vehicle composition was limited to
characterization of existing technology vehicles. Seven vehicle size and type combinations were •
characterized for 1975 and 1980 while one vehicle, minivan/small utility, was addeSfor 1985 and 1990.
Control scenario vehicle characteristics are tabulated in Table 51. TEEMS* activity and^energy , '
computation procedure was executed to produce personal vehicle travel and energy consumption '.•
estimates. •''-••-..'. ",/..-*
f » V,
, . . . . V • • • .
Commercial truck travel was not modeled but, historical data published bj^the FHWA (FHWA,
1987; 1991) were used; FHWA publishes truck travel by three categories:* f)*S-axle, 4-tire trucks; 2)
single unit trucks; and 3) combination trucks. All 2-axle, 4-tire trucks were treated as light-duty trucks. v
VMT by personal light trucks were subtracted from the published totals to arrive at commercial light .
track VMT. Diesel truck VMT shares of total YMT werg^ob^ned i|om TTUS (DOC, 1981; 1984; .
1990). TIUS data were also used to split VMT"by smgle imit and cdmbination trucks. All combmation
trucks were assumed to be the heaviest; class 7 and class 8, whife single unit trucks could be of any size
class - 3 through 8. Gasoline and 'die$ei£ryMT totals were developed for these heavy-duty trucks and
were kept constant for the control an4: no-control scenarios.
"^ •' ." ' " 'is&'t i- '
126
-------
Appendix B: Emissions Modeling
^ "
: ,/, ',,, ' Data Item
hcsBehtid abates, fey Ibat adiib»tes
''ispo$abl& lacome, ^
Statistical Afas&aCis > ,
< , - - ,-, -
YeMele 'fleet '{WP '
"if\^-",
•* w •**
'> " A
127
-------
Appendix B: Emissions Modeling
e "49*
-*'
HomeMd (MSEoii) ^;" ,*r^ ; ,\&A " "
Population (Mi8toj$- ' ' c' ' - •. * 2J&4&
Attribute - ';'] Sf . XV 'Av A. ^ ; ' ' -
' •• •.''•/ ,. ^ ^s 5
jL0C3tJQ3l ' V1-" ,, •.-. xv"' •v*'* •• "
Central City , , V % , ,33.2\,
Suburbs , , "* ':"" \^"'\3&.$ ,
• Raral < ' '- '" - ^ 33.2 ' *
<$13,000 , '\ :•, £ " ?'; V- ,;'"25.9 " ,v
$B,000>$33,OW' ' "- '- ;O f 34.<>
$33,006- $52,500 V ,; * ^ &•& ''
Age of Householder (Yll) , "' - ,
<*33 s \ &J&
35-44 ' - - '^/'Vx t'184 ' «"
45-64 / s " \'\>'% ':-3g.3 %>
^ ^ •*« ** ^
Household Sfee ' ; / '-' "
3-4 ' ' , , ' v aa /v '
. •. „, .i^vsur
> ** ""-v ,VAS s^A-iet
, ^ / %, '^ ' ^J*"- -
Licensed Drivess <« , * 4V % - -
0 "-'"'", §4,
1 „>/ •., 27.S
2 , \; V*4*M ' '
i-7i,r ; 'v'»»'"\ - 8^ - ' 93,3 •:
:^2ts:5 ^ ^ '*&%".>, &T£ >-;&ks* ;
; t ^WM^ip^pi^W '; J - ' -,,'
•&&' : 'i98o , --' im , im :
^ -% ' , " , » - '' ' '. ' ' :
\- 32.Q 31.9 -- ' 31J5 ', '31.4
5fc/J J\ s •^J'Y jft "* "* 'ftO 1 ^O "^i
«5U^'1 -.':
-- ^^L<|s^'s^ -31-.1 -„ 3Q-,3^-'\ 3$3 ""-
*pjj% «j ' ^ 1 "3 rfl ""** 4 A ^ ' 4 ^ *«t
,. " "\, f" ~< f v - f "f , f - ,-/\ \ ' !
^Z9^X ^"~ 31,1 */ "••••• "2J9L3 ^ "27 i4 " •
tw '' ' .• **"• *"% > ' '''V'
s X^x?* ^ s "* *!*£*** ' "• ^ , ^23+7 <24-!ip , " :
wfA-*^ ;
^3.0 \ ;f ^2 ' . 33,'5 " ' 3,218 "
' ? , ' ' ''' f-'' ' '
, , 27.3, , , 27.0 26.2 - , f 26^
' I,
^M* Jl nlJ t ^
128
-------
Appendix B: Emissions Modeling
Table $6. jEimo^caadTeMclelf^^ , ,; " „,,
,.
s
f f
'
A
f'-
••;"
V
•; ^
% 'v
",
s
*
'"-'',
'
-
^
V. ,
Y-ear-0'- '
f
-Mo- \
1972; ";
1S73 ~
*19?4 '
1975' '
1976 -r
1977
1978, t
1979' s '
198Q - "
,1981
1982 s
1983 '
-1984'
1985 '
1986'*'*
1987, ,
1988
lW '""
•f ff f f s
199V
Disposable Inc ;
/ ,
\ ^
" {-7^990 ;
v\ - -8,43^ /'
" ,s,27f *" ;
8,349 - '; '
/n ,8'^1\
8,742 ^ \ - ,
, ; 9,070^ ,
",,- ' 9M- V,;
/ 4»E'/ ,
;- ^ - ^ , 9,093 "' <
% ' 9,050 '' '
J/ 9,239 , -
9,«91 ^ -"
1 '-' "9^881 '-
' * N ' ^
^10,174^ % ""
10,564
^ , -" " 13.4,^, , 's , ' 9,563-
/-1.03; ,,s ' ^13^ ',- '„ % 'S^E» "
' '.'' V
I.O2 , 13-5 - „ " 9^833 '
1^1\ / , - 13:8^- - /'9#Mi J
0"^? ' 14 O"" ^ ''V10 i^S1 ' ^
ff \ *)\ N *• \-& A. ^ •• G at?'7
i-
;"•."
-^
,-
,
,
"•^
^
,
,
,
129
-------
Appendix B: Emissions Modeling
- '.
pjigfjj^^
praMe '51, !'!]Con^>l Set
i'T;!l!
'C v "i '!j
fill
'J"i' !!fl"i i'iii,'.'1'
piiffii
^||I
J&ISlMliB.hif
Vehicle lype
and Size
(Seats)
AaiomobHe
SmaH(2-4)
" I1'!' ;
Compact (4)
Mid-size (5)
Laige(6)
Light truck
Std. truck
Compact
StaVVan/Stdl
Otflity (11-15)
'l • , ' T ' ,".'
Minivin/Small
Utility (7-8)
Uto!!1!**--**1 f«V
.i i ,
nario Personal Characteristics,*
H|[4sr *-nr j; ^wi /•
1975
Curb Engine
Weight Power
(IB) (hp)
2,770 91
3,625 115
f
4,140 128
4,900 155
4,530 141
3,745 108
5,010 145
Economy
(inpjsT
17,2
14^
13^
12.2
!
*
11,2
14.2
9.9
>f C t<. , >' V >• - ''. •
» ^ -'> " /'< ^ ,,.^^
1980
Weight Power Econoniy
OB) flip) fmptD
, X
2^35" ^83 19,6*
^ *
_ ,3^35'-" 105 16.9
si V "
3,730 116 " 15.1
4,840 ' 153 x 13JJ '
i
-j ><' , '•*' "
4,455 '1^ 1X6 „
3,580 99' 1519
H
4,975 144 'll.4
t I
** •*! " ^ ^
* /V f *i ,^
....
^
K
*
X
?t?i s;a«5|frf ts%" , „',
t'j::!.i«b''J!:::;aiK;-i3l;:.:: 'iV; t r . ' "
1*
VehideType
and She
(Seats)
Automobile
Small (2-4)
Compact (4)
MW-sizc(5)
Laige(6)
U^it truck
SkL track
Compact '
Std,V*n/Std,
Utilify (11-15)
Utility^)11
W^t
2,225
2,775
3,180
3,975
4460
3,495
4,920
4,125
1985
Engine
Power
(hp)
75
90
108
^
135
132
9O
142
I
101
Economy
•j.
22.7
19.3
16.8
14.6
13.1
17.2
12.4
T ^
16.7
' H
Curb
Weight
OB)
'»
2,135
2^95
3,050
' ' 3',705
r
4,000
3^60
' 4,765
\ ^ ** j *>
_,
3,910
1990
f s ^
Power
(hp)
{
«i
75
i ^
108
it 1* if
^130
128
90
138
f
, ^
v ^i
IDS
•, " s
Fud
Economy
'! i '
"1,1 i .
24.9 ;(
22.0'
'j95
TI i i
174
i t !
, 14.1
18.9
12.9
r ^
y- < ^
*
1&2
s < ( !
MJ..*;,. irNote; "Average for all vehicles of each type and size.
130
i i s
» l
-------
Appendix B: Emissions Modeling
No-control Scenario Emissions • , "
The control scenario data were modified to reflect no-control scenario emissions using economic
changes predicted by the J/W model, EPA, and ANL. The J/W model predicted a slight loss of
employment and drop in GNP in terms of nominal dollars. However, the lower rate of inflation coincided
with a real GNP rise. ANL's information from the model did not include any indexes for converting
/,s3|y££^S!'*'' — g$&
nominal income to real income. ANL assumed real income changes to be similar to those of real GNP Jnd
modified household shares by income classes accordingly. Tie model also predicted a slight drop in
refined petroleum price beginning in 1973. The predicted djj||was theJargest (5.35 percent) m|L973,
reached the lowest level (2.16 percent) in 1984, then increased to a second peak (3.44 percent) in 1988,
' •' ^-^'^^^*£S";?i5, • ^^K^f"
and dropped again from 1989 to 1990. Since these changes wer^ Mcbhsistent with historical patterns of
leaded and unleaded gasoline price change, ANL developed an e|^at| of changes in fuel price resulting
from the cost of removal of lead from gasoline and other inMstmcnire costs involved with distributing a
new grade of fuel. Subsequently, EPA provided a set of fuel costs for iise in die analysis. Both ANL and
EPA fuel prices followed a similar pattern, although their tnagnitudes differed. The no-control scenario
was analyzed with EPA fuel prices. ANL also estabi^^^^^ffiipnship with cost of regulation/emission
control technology, and the effect of costs on vehicle pric^^^^^ec^oi^ directly from the EPA
publication Cost of Clean Environment (EPA,,1990). TfieSe^elK^^^reused in the analysis.
The IFF model was executed for target years 1975,1980,1985, and 1990 using a set of revised
household shares by income class. Table
52 shows the revised shares. Comparing
Table 52 no-control scenario shares with
those in Table 49 for the control scenario,
there seems to be a slight shift away from
travel by the lowest income group and
toward the middle income groups.
3Table 52. ^Distribution of Households by Income Class lor
ifo-control Scenario. „ i * " '
The vehicle ownership projection -
model was executed for Che above four
target years using the data listed in Table
53. Changes in fleet characteristics are
summarized in Table 54.
•Household Shares (%),
Attribute
Income (1990 $)*
,^$13,000 "-"
$13,000-33,000
$33,000-52,000
*&yX**>
1975
26.3
373
22.8
13.6
1980
'
26.2
37.6
22.6
13.6
1985
-
253
38.4
22.0
,14.3
by Year
1990
* ,
24*7,
38.4
22.6
< 14.3
* Note: , ^ * Approximated to 1990 dollars.
131
-------
Appendix B: Emissions Modeling
1 " MW'!'"1 ji>"'';w~(H>>;i»"1" x ? > '", ,,,>i>!^, <-, - - , / ' *
3. Economic and Vehicle Usage Data for VeblcIe'Ownejship Projection - ,
« li Jj.i«UM«fc i «.«J. , ° , , * *« ., 4
tro! Scenario. _ , i(
11
1
If f
n
4f f
K
1
f
It
» |l
1
I
l|
S
ii
B
ii
fWf
*B
•1
A!
iU
i
'•A
Year
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
Disposable
Income per
Capita (84 $)
7,597
7,769
7,990
8,463
8,297
8,406
8,600
8,795
9,126
9,216
9,114
9,158
9,116
9312
,9,775
9,976
10,244
10,282
10,676
10,827
11,019
Fuel Price
(84$)/GalIon
0.91
0.88
0.83
0.84
1.06
1.02
1.01
1.01
0.96
1.19
i
1.51
1.53
136
125
1.18
1.06
0.84
0,86
0.83
0.88
0.97
*> ?
Miles/
Gallon
13.5
' 13.5
13.4 * f
' &3
" f i
13.4 .
.&* '
j
13.5 v
i
13.8
14.0
f *
14.4
ly
155
16.0
16.8
17.2
17.9
18.3'
18.4,
'-> i > ^ >
19.4 ,
ft * f
20.1
l
k
20J
21.0 '"
>
VMT/VeWcle
"10,143
^ 10,247 ' ,
"V > fi
1^,353
10,189 '*
14 ,'"','
' 9,569^
VV,736^
> Jd f i
,9,854
v » " if
9,963
, I0'l7f, rf
, 9,557 /'
1 x *^
, 9,234 ^
9,234
9,447
t irfi * *! <. >v
9>50
, 9,582
9,607 ' .
9,738
>*! 1 ^ £f * *
t 10,201 ^
^214 ' ^
9,902
1 ' , ''•
9,849 '
t |Soi5 The effect of leduOkms to vehicle price and vehKleoperatmg cost, and mcreases in fuel ecoM
^l^ horsepower were reflected in the menu of the vehicle choice model (DVSAM). Vehicle weight an&
,; seating capacity were kept unchanged from the with CAA tun. Table TV-7 shows ihe changes in various
ivehidcaiributes. , „, , , , ', , ,. ' f * •>
132
-------
Appendix B: Emissions Modeling
, -s -T » «• %- „ ^ >
-Table; 54C* Peiceaf Changes lo Key VeMcie Characteristics Between the
- " ' '. ' uffs '
Vehick ' ? - Price ttTws - HP "-
, - "
Small Auto^ "" ,-235 ; OJ)1 059
Compact Auto -2,35^ Cuttf* 059 ^
W&dsS»Auto' ^-2\35 ' &01 O^S9
Large Atrip "j-235< HB.01 059
Small Truck -1.30 0,01 059 *
^tdT|Hck ^1,30 ,p.0t 0::59V
std valuta ' -r.3o, vaoi/4 059 ;
MVn/Sm, ' „ V ,4
Utilitv* 1 , J , " . N
^ ^ s. \
Price
-2.76
, ,-2.76
^-2.76
-2.76'
-2.71
-2.71
-2.71'
1980
«pg
- 0.22
0.22*. '
™0.22
0.22
4^ ;
0.2! ;
0.22
™ ** ^
* ,.
HP
i-8i:
,i«k
1,81
1.81
1.81
1,81
1,81,
v
i Jr
\
* \
5
-
v>
^
Vehicle
Small Auto '"
Compact Auto'
Midsize Auto
Large Auto
Small Truck
Std Truck
'StdVau/Util ,
MVri/Sm
IMlhV
+
Price
-3^5
-3^5
-3.25
-3.25
-2.53
-253
-253
-253
1985
mpg
'0.62
0.62
0^2
0.62.
0.62
0.62
0,62
0.62^
"•
HP
2.20
(120
2.20
'2.25
2.20 v
' 2.20
' 12.20
W,
11
- Price
' -2.94
-2.94
-2.94
'-2.9^
-258
-258
-258
-258
1990
mpg
6.95
0.95
d.95
0.95
" 0.95 .
0.95
0.95
0.95
'
HP
2.77;
2.77
2.77
2.77.
2.77
2.77
2.77
2.77
Motet * Average change for each vehide size and type combination.
133
-------
Appendix B: Emissions Modeling
Utilities
The electric utility industry retrospective analysis was prepared using two different utility simulation
models. ICF utilized its CEUM to estimate control and no-control scenario emissions for SO^ PM, and
NO, in each of the target years. ANL's ARGUS model was used to estimate'electric utility CO and VOC
emissions for the same period. This mix of modeling approaches was used because, while CEUM was
determined to be a better tool for examining fuel shifts that were affected bytfie CAA than ARGUS, the
CEUM model was not initially set-up to evaluate CO or VOCemissions. Although CEUM can be (and
eventually was) configured to provide emission estimates for pollutants other than SO^ NO,, and PM,
ARGUS was already configured to provide VOC and CO emissions. However, it should also be noted that
VOC and CO emissions from utilities are quite low, as efficieritfuel combustion reduces both pollutants.
Thus, for this sector, the presence or absence of the CAA would not produce any different VOC or CO
control techniques. VOC and CO emission rates for this sector differ primarily based on the fuel and
boiler type. Therefore, a simpler modeling approach was judged to be acceptable and appropriate for these
two pollutants. This chapter presents the methodology used to estimate utility emissions under the control
and no-control scenario using the CEUM and ARGUS models. The method used by Abt Associates to
estimate lead emissions from utilities is also presented.
Overview of Approach
i' '~~
The CEUM model uses industry capacity data and specific unit-by-unit characteristics, operating costs
data, electricity demand estimates under the control and no-control scenario, and historical fuel prices to
estimate SOj, TSP, and NOX emissions for 1981), 1985, and 1990. Changes in electric utility emissions,
costs, and regional coal production were developed using ICF's CEUM with a calibration to historical
electricity generation, fuel use, and emissions. The ARGUS model, which was used by ANL to estimate
utility VOC and CO emissions, is driven by operating costs, industry capacity and generation data, demand
for coal, and unit-level operating characteristics. The J/W model is used to incorporate predicted changes
in electricity demand under the no-control scenario. Finally, Abt Associates relied upon energy use data,
the Trends data base, and the Interim 1990 Inventory to calculate utility lead emissions based on coal
consumption. The approaches used by each of these three contractors are discussed individually in the
following sections.
Establishment of Control Scenario Emissions
;••! ,"•;. ,t :; . . • •
A common feature of the approaches taken by ICF and ANL was to identify conditions that are inputs
to me CEUM and ARGUS models, respectively, in the control scenario. Later in the analysis, these
variables were revised to reflect no-control scenario conditions. The next section discusses the specific
assumptions used in the CEUM analysis.
Key Assumptions in the Development of the ICF Analysis
At EPA's direction, ICF made several assumptions in conducting this analysis for purposes of
consistency with other ongoing EPA efforts assessing the effects of the CAA. These include the
134
-------
AppendixB: Emissions Modeling
macroeconomic assumptions regarding the effects of the CAA on economic growth^ or more specifically,
electricity demand, developed from other EPA commissioned efforts. Each is described briefly below.
Pollution Control Equipment Costs ,
Only limited actual data were available for this analysis on the historical capital and operating costs of
pollution control equipment. Accordingly, for this analysis, the actual capital and operating costs of
scrubbers were estimated using EPA scrubber cost assumptions adjusted to reflect actual data from a ~~~~
survey of scrubbed power plants with scrubbers installed during the 1970ssand early 1980s. For those:
power plants with actual survey data, actual capital costs were used. For other pre-1985 scrubbers, ICF
relied on the average costs from the survey data. For paniculate control equipment (primarily electrostatic
precipitators, or ESPs), costs were estimated based on limited actual data, and a 1980 Electric Power
Research Institute (EPRI) study of ESP and baghouse costs. Based on this information, ESPs were
estimated to cost an average of $50 per kilowatt (in 1991$). The development of more detailed data on
actual power plant pollution control costs was beyond the scope of ICF's analysis. ICF concluded that such
an effort would not significantly change the national or regional cost estimates developed by its approach.
Electricity Demand and Fuel Prices
' Consistent with other EPA ongoing analyses, ICF assumed that the CAA resulted in a reduction in
electricity demand of 3.27 percent in 1980, 2.77 percent in 1985rand 2.97 percent in 1990. Also
consistent with these studies, ICF assumed that natural gas prices and oil prices would not be affected by
the CAA. Coal prices were estimated to change in line with increases and decreases in demand for specific
coal supplies (and consistent with ICF?s detailed modeling of coal supply and demand). The average prices
of all residu|l pils consumed were also estimated to change due to a greater use of more expensive lower
sulfur resiSullils,under the CAA.
Coal, Nucjjsdf>Jfydro, and Oil/Gas Capacity
At EPA'S|directidn, ICFs"apjproach,was based on the assumption that no changes in the amount of
nuclear, coilf lydro, or*oiJ|/gas»stream or combined cycle capacity would be built or in place in 1980,1985,
or 1990. Given that the driving factors associated with the actual decisions to build new baseload capacity
were not based solely on economics but entailed financial, regulatory, and political factors as well, the
actual effect of the CAA on these build decisions is very uncertain. To the extent that more coal-fired
power plants would be bupt and fewer oil/gas-fired power plants constructed, the actual emission
reductions assodated with the CAA would be greater than those estimated by ICF, while the estimated
cpstsof the CAA jpuSS be greater (because fewer, lower-cost, coal-fired power plants would be on line
under tp£C^^^^»wever, the CAA had virtually no effect on the costs of constructing new coal-fired
power playplit came on line prior to about 1975 and a relatively moderate cost effect on coal-fired
power plants that came on line through the early 1980s (since these power plants were not required to
install scrubbers). Since a large majority of coal-fired power plant capacity came on line prior to 1975,
ICF concluded that the effect of the CAA on the amount of total coal-fired capacity was not expected to be
very large.
Natural Gas Consumption
135
-------
Appendix B: Emissions Modeling
The analysis assumed that the amount of natural gas consumed under the no-control scenario could not
exceed the actual amount of consumption in 1980,1985, and 1990. In part, because of natural gas price
regulation and the oil price shocks of the 1970s, natural gas was often unavailable to electric utilities in the
early 1980s. Since the CAA is relatively unrelated to the questions of supply availability and price
regulation of natural gas, ICF assumed that no additional gas supplies would be available if the CAA had
never been adopted. It is possible, however, that in the absence of the CAA, industrial and commercial ,
users of natural gas would have used more oil or coal. To the extent that this would have occurred, there?
would have been more natural gas supplies available to the electric utility sector. This increase in sujpfy
would have resulted in an increase in the estimated costs of the CAA, and a corresponding decreasejin the
estimated emission reductions. ICF concluded, however, tharthis effect would not be very significant.
State and Local Environmental Regulations
At EPA's direction, ICF assumed that there would be no State and local emission limits or other
emission control requirements under the no-control scenario. Accordingly, ICF assumed that there would
be no 862, NO^ or TSP emission limits under the no-control scenario and ttipSU scrubbers, NOX controls,
and ESPs/baghouses (at coal-fired power plants) were installed as a result cf The CAA. (The more limited
amount of participate control equipment installed at oil-fired plants was assumed to have been installed
prior to the passage of the CAA.) In the case of particulate controlequipment, some ESPs and other
equipment were installed at coal plants prior to the 1970 CAA. To the extent that this is the case, the
estimates of the costs of meeting the CAA have been overstated. ICF concluded, however, that the amount
of such capacity was not substantial.
RetirementAge
The analysis assumed that unit retirement age was constant between the control and no-controls
scenarios. Adoption of this assumption might bias the emission reduction estimates upward to the extent
turnover rates of pl^r (and presumably higher-emitting) units may be slower under the control scenarios,
because more significant CAA control requirements focused on new units. However the vast majority of
existing coal and oil capacity was built after 1950 and it is generally acknowledged that a relatively short
technical plant lifetime would be about 40 years. As such, even if the no-control scenarios resulted in no
life-extension activity, there would be virtually no effect over the 1970 to 1990 timeframe of the analysis.82
ICF 1975 Control Scenario Emissions
The 1975 emissions under both scenarios were calculated differently than emissions in 1980,1985,
and 1990. In calculating or estimating 1975 SO2 emissions for the control scenario (i.e., "actual" 1975),
the weighted average emission rates at the State level, in the year 1975 were estimated, based on plant
level average sulfur content of fuel deliveries from Federal Energy Regulatory Commission (FERC) Form
423 and assumed AP-42 sulfur retention in ash. These weighted average emission rates were then applied
to actual State-level electric utility fuel consumption in the year 1975 (DOE, 1991). In the case of NOX
emissions, first, an estimate of Statewide NOX emissions in the year 1975 was derived based on the use of
°rhis conclusion is supported by R. Nelson, T. Tietenberg and M. Donihue, 'Differential Environmental Regulation: The Effects on Electric
utility Capital Turnover and Emissions,'' Review of Economics and Statistics, Vol. 75, no. 2 (May 1993), pp. 368-373, which indicates that a delay
in utility retirement due to new source regulation exists but that it has only a small effect on emissions.
136
-------
Appendix B: Emissions Modeling
the same NOX emission rates, by fuel type, as developed for the 1980 no-control scenario modeling runs.
These emission rates were specific to the fuel type (coal, oil, or natural gas). These Statewide NOX
emission rates or factors were then applied to actual fuel consumed by electric utilities in the year 1975, in
order to obtain estimated "actual" 1975 emissions. As before, the fuel consumption at a Stale level was
derived from the State Energy Data Report (DOE, 1991). ICF calculated the weighted'average heat
content (BTU/lb) by State from the 1975 FERC Form 423 data and used these figures with the TSP
emission factors (Ibs/ton) to derive emission rates by State (Ibs/MMBTU). These emission rates were then
applied to 1975 fuel consumption estimates obtained from the State Energy Data Report. For the control
scenario 1975 estimates, ICF used the 1975 factors.
/- *
For the remaining target years, ICF used the results of CEUMjuns that provided fuel consumption
figures in 1980,1985, and 1990, respectively. Emissions were then calculated using the appropriate
emission factors for each year.
ARGUS Modeling Assumptions
The portion of the electric utility sector analysis conducted by ANL with the ARGUS model is
described in this subsection. ARGUS contains four major components: BUILD, DISPATCH, the
Emissions and Cost Model, and the Coal Supply and Transportation Model (CSTM). An overview of
ARGUS can be found in Veselka etal (1990). Only the rJISPATCHPand CSTM modules were used for
the present analysis. A brief description of the ARGUS components used in this analysis is found in the
following subsections. „
DISPATCH Module
The DISPATCH module contains a probabilistic production-cost model called the Investigation of
Costs and Reliability in Utility Systems (ICARUS). This module calculates reliability and cost information
for a utility system! ICARUS represents detailed, unit-by-unit operating characteristics such as fuel cost,
forced outage rate, scheduled maintenance, heat rate, and fixed and variable operating and maintenance
(O&M) costs. These components are used to efficiently compute system reliability (such as loss-of-load
probability and unserved energy) and production costs.
^y ~ . *'•••- ,
The input data required by ICARUS include monthly load duration curves, annual peak demands, and,
for both new and existing units; unit sizes, capital costs, fixed and variable O&M costs, fuel types and
costs, heat rates, scheduled maintenance, and equivalent forced outage rates. The output from ICARUS
Includes annual summaries of capacity, generation, cost, and reliability for the entire generating system.
? -JCSTM Module
The CSTM module determines the least-cost combination, on a per BTU basis, of coal supply sources
and transportation routes for each demand source. First, it estimates coal market prices based on regional
demands for coal from all economic sectors. To generate market prices, CSTM estimates regional coal
production patterns and coal transportation routes. The CSTM input data are grouped into three major
categories: demand, supply, and transportation. CSTM uses supply curves from the Resource Allocation
and Mine Costing (RAMC) Model (DOE, 1982). Every region has a separate curve for one or more of the
60 different coal types that may be produced in that region. CSTM modifies the original RAMC supply
curve by dividing the single RAMC curve into two curves, one representing deep mines and the other
137
-------
Appendix B: Emissions Modeling
representing surface mines, but still uses the same ranges for heating values and mine prices that define the
supply curves in RAMC. Prices fluctuate as a result of different raining methods, size of mining
operations, reserve characteristics, and depletion effects.
The transportation data defines the network that connects 32 coal supply origins with 48 demand
centers. Transportation cost is affected by distance, terrain, congestion, variable fuel costs, cost escalators
for fuels and facility upgrades, and competition. CSTM first computes the production cost for each coal
supply region and coal type. It then matches supply sources with transportation routes to find the lowest
delivered costs. ' • ^
""»
Coal demand for a particular region is based on the amount, geographic region, economic sector, and
range of coal types. There are 44 domestic demand regions. CSTM allows demand to be met by one, or a
combination of, different supply regions.
The ARGUS input data for existing units are based on the Argonne Power Plant Inventory (APPI).
APPI is a data base of operating and planned generating units in the United States that was current through
1988 at the time of ANL's analysis. This data base is updated annually based on information in the
regional North American Electric Reliability Council (NERC) reports, reports from the Energy Information
Administration (EIA), and other sources. Unit operating characteristics (fixed O&M, variable O&M, heat
rate, forced outage rate, and scheduled maintenance) are based on regional data as defined in the EPRI
report on regional systems and other historic data (EPRI, 1981).
ANL used the 1988 inventory to generate a 1990 inventory. The 1990 inventory was then used to
generate a separate unit inventory for the target years 1975,1980 and 1985. The target year inventories
were generafed by removing uiiits^hose on-line year was greater than the target year, from their respective
inventory. The regional capacity totals in these preliminary inventories were tabulated by major fuel
category (nuclear, coal, oil and gas steam) and compared to the regional historic NERC totals. This review
identified capacity differences, espeaaliy.in 1975 and 1980 inventories. The original plan was to add
phantom units to match the regional historic; totals. However, based on the need for State-level emissions,
it was decided that a more thorough review of the unit inventories was required.
ANL's detailed review included an examination of the nuclear and coal units greater than 100
megawatt equivalent (MWe) in each target year. Missing units, with the appropriate unit size and State
code, were added so that the regional totals were comparable. The availability of coal units was based on
the on-line year of the unit as reported in the EIA report Inventory of Power Plants in the United States
(DOE, 1986). The coal units were also checked against the EIA Cost and Quality Report (EIA, 1985) to
verify the existence of flue gas desulfurization (FGD) systems in each of the target years. The nuclear unit
inventorieswere verified with the EIA report An Analysis of Nuclear Power Plant Operating Costs (DOE,
1988). The review also included oil and gas steam units greater than 100 MWe. The total capacity of the
oil and gas steam units were compared because many units switched primary fuel from oil to gas during the
relevant time period. The oil and gas units were compared to historic'inventories based on information
provided by Applied Economic Research. In addition to thermal generation, the hydro and exchange
energy was reviewed. For each target year, the hydro generation and firm purchase and sale capacity data
was adjusted to reflect the historic levels. These two components, hydro and firm purchase and sales, are
accounted for first in the loading order. If these variables are overestimated, there will be less generation
from coal units. Likewise, if they are underestimated, there will be too much coal generation. The hydro
138
-------
Appendix B: Emissions Modeling
and firm purchases and sales can vary significantly from year to year because of weather conditions and
other variables. Therefore, it was important that they be accurately represented.
No-control Scenario Emissions
In order to calculate utility emissions under the no-control scenario, inputs to both the CEUM and
ARGUS models were adjusted to reflect no-control scenario conditions. The changes made to each
model's base year input files are discussed separately in the following sections.
ICF Estimates of SO2, TSP, and NO, Emissions in the No-control Scenario
— --K~ •** ^--v. •* *&"„ .*•£*
As described on page 136, ICF utilized a different methodology to calculate 1975 emission estimates.
Rather than relying on the use of detailed modeling runs, ICF based the 1975 emission estimation on
historic fuel consumption and sulfur content data in 1975. This subsection first outlines the process used
to calculate no-control scenario emissions in 1975 and then presents the methods used for the remaining
target years.
1975 Utility SO* NO* and TSP Emissions
To develop State-level no-control scenario utility SO2 emissions, ICF developed no-control scenario .
SO2 emission rates. A reasonable surrogate for these emission rafes is SO2 rates just prior to the
implementation of the SIPs under the CAA. ICF developed 1972 rates (based on the earliest year available
for FERC Form 423) and compared these with 1975 rates. In each State, the greater of 1972 or 1975 rates
was used in the calculation of SO2 emissions in the absence of the CAA. To develop State-level no-control
scenario SO2 emissions, no-control scenario fuel consumption data were needed. ICF assumed that the
demand for electricity in 1975 would be 2.73 percent higher than the actual energy sales in 1975. This
assumption is identical to the no-control scenario electricity demand projections derived from the J/W
projections. For the purpose of this analysis, it was further assumed that this increment in demand would
have been met in 1975 from the oil and coal-fired plants in each State. The increase in consumption of
these fuels was assumed to be in the same proportion as their share in the 1975 total energy mix for
electricity generation in that State. It was assumed that the generation of nuclear, gas-fired, and other
electricity generation would not change. A sensitivity case without an assumed electricity demand change
was also calculated. (The sensitivity analysis results are presented later in this appendix, starting on page
|or NOX emissions under the no-control scenario, it was also assumed that the 1975 electricity sales
thave been 2.73 percent higher than was the case in 1975. No-control scenario TSP emissions in
[on national emission rate numbers from EPA that were converted to pounds per million
f erage energy content of fuels in each State. No-control scenario TSP emissions were
calculated based on 1970 emission factors (Braine, Kohli, and Kim, 1993).
1980,1985, and 1990 Utility Emissions
For 1980,1985, and 1990, ICF calculated no-control scenario emissions based on fuel consumption
figures from the CEUM runs, and 1970 emission factors from EPA.
139
-------
Appendix B: Emissions Modeling
Electric utility SO2 emission estimates are approximately 10 million tons (or about 38 percent) lower
by 1990 under the control scenario than under the no-control scenario. Most of this estimated difference
results from the imposition of emission limits at existing power plants through the SIPs under the 1970
CAA. Most of these SIPs were effective by 1980 (with some not fully effective until 1985). Most of the
additional reductions that occurred during the 1980s were the result of the electric utility NSPS, which
required the installation of 70 to 90 percent SO2 removal control equipment.
By contrast, electric utility NOX emission estimates under the control scenario are only about 1.2
million tons, or 14 percent, lower than under the no-control scenario by 1990. This occurs because, under
the implementation of the 1970 CAA, only a few existing power plants were subject to NOr emission
limits. Virtually all of the estimated reductions are the resulFof NOX &SPS, which generally required
moderate reductions at power plants relative to uncontrolled levels. In addition, electricity demand is
estimated to be about 3 percent lower under the control scenario. This decrease reduces the utilization of
existing power plants and also contributes to lower NOX emissions (and other pollutants as well).
Electric utility annualized costs (levelized capital, fuel, and O&M) are estimated to be $0.2 billion
lower in 1980, $1.5 billion higher in 1985, and $1.9 bflfion higher in 1990 under the control scenario.
Note, however, that this reflects the effects of two offsetting factors: (1) the higher utility compliance costs
associated with using lower sulfur fuels, and the increased O&M and capital costs associated with
scrubbers and particulate control equipment; and (2) lower utility generating costs (fuel, operating and
capital costs) associated with lower electricity demand requirements. In 1980, the increase in fuel costs
due to higher generation requirements (under the no-control scenario), was larger than the decrease in
capital and O&M costs and thus yielded a cost increase over the control case.
However, lower electricity demand for the utility sector would translate into higher costs in other
sectors (as electricity substitutes are used). This effect was captured to some extent by the original J/W
macioeconomic modeling conducted for the present analysis.
Average levelized U.S. electricity rate estimates are approximately 3 percent higher under the control
scenario during the 1980s. Note that year by year, electric utility revenue requirements and capital
expenditures (not estimated by ICF) would be estimated to have increased by a greater percentage
particularly in the 1970s and early 1980s as incremental capital expenditures for scrubbers and ESPs were
brought into the rate base. ,
Significant shifts in regional coal production are estimated to have occurred between the control and
no-control scenarios. High sulfur coal producing regions such as Northern Appalachia and the
Midwest/Central West are estimated to have lower production under the control scenario, while lower
sulfur coal producing regions such as Central and Southern Appalachia are estimated to have higher coal
production.83
** At EPA^s direction, ICFs analysis did not estimate the effect of shifts in non-utility coal consumption on regional coal production, nor did it
consider the possibility that fewer new coal powerplants might have been built due to the CAA as discussed earlier. Both of these factors could
result in a greater estimated change in total U.S. coal production than estimated herein although the difference is not likely to be very significant.
-------
Appendix B: Emissions Modeling
ARGUS No-control Scenario
Regional fuel prices, for the thermal units, were based on
historic information from the EIA Form 423 data for the year
1977,1980 and 1985. The 1977 data was used for 1975.
Fixed and variable O&M costs were adjusted from the 1988
level, and all cost data were converted to 1985 dollars.
The load data were based on regional historic MERC data
for each of the target years. The shapes of the monthly load
duration curves 'are the result of modifications based on the
data in the EPRI report on regional systems (EPRI, 1981). The
shapes were modified to match the projected 1988 monthly
load factors for the NERC regions. These load shapes were
held constant for all years.
Table 55. W Estimates of Percentage
Increases in National Electricity "
Generation Under Nb-controi Scenario,
Year
Increase
1975
1985
1990
The actual peak-loads were selected from historic ~ •^^^••i ' ""•'''" •• •
information and used with the existing load duration curves.
The system was dispatched so that the calculated generation could be compared with historic data.
Discrepancies were resolved by adjusting the peak load so that the annual generation was on target This
procedure was repeated for each of the target years.
The electric utilities were expected to have an increase in generation as identified by the J/W data.
Table 55 identifies the increase in national level generation by year. The national level increase in
generation j|l| applied to each power pool.
In adlffi||i|p||^ changes,~coal units with FGD equipment were modified. These units had their
FGD eqiupment removed along with a 3 percent decrease in heat rate, a 2 percentage point decrease in
forced outage rate|ancl a 50 percent decrease in then* fixed and variable O&M costs. These changes were
incorporated inj|fj|l ARGUSjnodel for each of the target years. Model runs were then conducted to -
arrive at estimates of VOC and CO emissions in the no-control scenario.
imation of Lead Emissions from Utilities
L order to estimate Pb emissions from electric utilities in each of the target years, data from three
: sources were used. Energy use data for the control and no-control scenarios were obtained from
i coal use estimates prepared for the Section 812 analysis by ICF (Braine and Kim, 1993). The
provided emission factors and control efficiencies, and the Interim 1990 Inventory
identified utility characteristics. The ICF data bases provided the amount of coal consumed for both the
control and no-control scenarios in each of the target years. A correspondence between the Interim
Inventory and the ICF data base was achieved through the plant name variable. Using emission factors for
lead and control efficiencies for electric utilities, estimates of lead emissions per plant per year were
calculated. These factors were obtained from the Trends data base. It was assumed that pollution control
on coal-burning power plants under the no-control scenario would be the same as the pollution control
level in 1970. Therefore, the control efficiency from 1970 is used as the basis for the no-control case.
141
-------
Appendix B: Emissions Modeling
CEUM Sensitivity Case
In addition to comparing actual (control scenario) historical costs and emissions with the higher
electricity demand under the no-control scenario, ICF also evaluated emissions in a sensitivity case without
the CAA or no-control scenario with the same electricity demand (versus the no-control scenarios with
higher demand). The purpose of this sensitivity analysis was to isolate the incremental electric utility .
compliance costs and reductions in emissions associated with the CAA from the lower resulting generation
costs and emissions due to lower estimated electricity demand under the CAA. The incremental effects of
the CAA when compared with this case indicate:
»• Estimated reductions in emissions due to the CAA are somewhat lower if measured against the
sensitivity case without the CAA with the same electricity demand than the emissions without the
CAA with lower demand. This occurs because lower electricity, demand under the no-control scenario
sensitivity results in lower utilization of existing coal and oil plants which, in turn, results in lower
emissions. As noted above, in some sense, the changes in emissions represent the effects of electric
utility compliance actions under the CAA, absent the effect of lower resultant demand for electricity.
> When measured against the sensitivity case without the CAA (with the same electricity demand),
electric utility annualized costs are estimated to have increased by about $5 to $6 billion during the .
1980 to 1990 period. This reflects the following cost factors: (1) higher annualized capital costs
associated primarily with scrubbers and ESPs installed by electric utilities to comply with the CAA;
(2) higher O&M costs associated with the additional air pollution control equipment; and (3) higher
fuel costs associated with using lower sulfur coal and oil in order to meet the emission limit
requirements of the CAA,
Commercial/Residential
The Commercial and Residential Simulation System (CRESS) model was developed by ANL as part
of the Emissions and Control Costs Integrated Model Set and used in the NAPAP assessment (Methods for
Modeling Future Emissions and Control Costs, State of Science and Technology, Report 26) (McDonald
antpouth, 1984). CRESS is designed to project emissions for five pollutants: SOX, NOX, VOC, TSP, and
CO. The CRESS output is aggregated into residential and commercial subsectors related to both economic
activity and fuel use. The introductory material provided in this appendix about CRESS describes the base
year as being 1985. It appears in this way because CRESS was originally developed to operate using the
1985 NAPAP' Einission Inventory as its base year data set. For the five pollutants reported by CRESS,
emission estimates are provided for the following sectors:
>• Commercial/institutional
coal, including point and area categories of anthracite and bituminous boilers;
liquid fuel, including boiler and space heating uses of residual, distillate, LPG, and other fuels;
natural gas boilers, space heaters, and internal combustion engines;
wood used in boilers and space heaters; and
other mixed or unclassified fuel use.
142
-------
Appendix B: Emissions Modeling
> Residential
• coal, including area sources of anthracite and bituminous;
• liquid fuel, composed of distillate and residual oil;
• natural gas; and
• wood.
> > Miscellaneous _ " , -
• waste disposal, incineration, and open burning; and
• other, including forest fires, managed and agricultural burning, structural fires, cutback asphalt
paving, and internal combustion engine testing.
In addition, VOC emissions are projected for these source categories: ~~~
>• Service stations and gasoline marketing;
*• Dry-cleaning point and area sources; and .
*• Other solvents, including architectural surface coating, auto-body refinishing, and
consumer/commercial solvent use. - '
This section describes the use of CRESS to estimate control and no-control scenario emissions from
the commercial/residential sector. » €'
Control Scenario Emission^ '
, ~ .^ m
i-1^, ~~* "- - '
For meTSfAPAFassessment, 1985 GRESS output corresponded to the 1985 NAPAP Inventory (EPA,
1989), which served as the benchmark for any projections. The design of CRESS is such that emissions by
NAPAP SCC are input for each State, then projected to future years by scaling them to economic data such
as energy demand. In estimating emissions, differences in emission controls associated with new,
143
-------
Appendix B: Emissions Modeling
replacement, and existing equipment are taken into account where such differences are considered
significant. The basic modeling approach is shown in the following equation:
where:
Q s emissions in year t or the base year, year 0
, * ^*2~. *"**v^r£
E = emission factor for the source category b in the base year, orfor a subcategory j subject to controls in year t
(this takes into account changes in emission rates that may occur as a result of emission regulations or
technology changes) ~
D- driver data indicating activity levels in the base and future years
n Sf ' '
f= fractionof total activity in year t differentially affected by emission controls
The calculations are carried out in two subroutines, one for SO^ NOX, TSP and CO, and One for VOC.
Typically SO^ NOX, TSP, and CO emissions are projected by multiplying the 1985 NAPAP SCC data
or base year data by the ratio of the driver data (activity level) value in the projection year to its value in the
base year. Because there are few controls on SOX or NOX emissions from the sources covered by CRESS,
projected emissions for most sectors are proportional to the expected activity levels. Thus,
IP ™ S- ^
I t | I I I 1 : -, 1,1 ( f i ^ , I t !»,<;••,» f .>, /f >•,
-i i ».i i j • r (i, •• j r «- ,, -••','„'4 • Ms-'i
There are a few source; types, such as commercial/institutional boilers, for which emission controls are
mandated. These are modeled by multiplying the 1985 emission data by the ratio of the controlled
emission factor to the base-year emission factor. Emission factors for each source type are weighted by the
proportion of base year activity in each subsector to which controls are expected to apply.
where:
the fraction of base-year activity accounted for by existing source b, replacement source r, or new
source n hi year t
144
-------
Appendix B: Emissions Modeling.
The effective emission factor (E ) for the sector is calculated by weighing the portions of sectoral
emissions subject to NSPS controls and those likely to continue at existing levels. An appropriate Internal
Revenue Service-based rate at which new equipment replaces existing sources is applied to each sector in
the model. This is done to estimate how emissions might change as older sources are retired and replaced
by new sources that emit at lower rates.
The SOj^NO/TSP/CO subroutine varies hi new and replacejment'enussion-spurce fractions subject tct
NSPS controls. These fractions are applied to the emission-source replacement rates, in addition, ratios
for new source emission factors are varied by State. However, emission ratios for any pollutant/source
type combination do not vary over the projection period.
The VOC estimation methodology is similar, but allows variation in emission factors over time.
Emission ratios are calculated from files of replacement and existing source emission factors weighted by
the replacement rate for each sector and new source factors by State. These are input for each 5-year
projection interval. For most source categories, VOC controls are not eavlsioned, and the 1985 NAPAP
emissions for the category are simply scaled proportionally to changes in the driver (activity level) data.
\ ,
For sources to which controls apply, a variation on the following equation is employed:
In equation 7, the emission factors for new and existing sources are effectively weighted by the
proportidn-of total Activity hi year t to which controls apply.
-» s ,
In using CRESS-for the CAA retrospective analysis, the base year was 1975. CRESS requires
emissions information by State and NAPAP source category as input Since detailed information on
emission levels for 1975 by Ni&PAP source category were not available, the data were developed from a
combination of sources. TJSe~pf6oedure for calculating 1975 emissions based on the 1985 NAPAP
invenpry is described below.'The emissions module uses these initial values hi conjunction with activity
Ites to project control and no-control scenario emissions.
ions Data -
ace the starting point for the analysis was 1975, emissions data by State and SCC for SO^ NOX,
__,=^_P,juid CO were required. Available emissions information for this year was not at the Jevel of
detal%eeded by CRESS. The 1985 NAPAP Inventory, which contains the necessary level of detail, in
conjunction with information from EPA's National Air Pollutant Emission Estimates, 1940-1990 (Trends)
and ANL's MSCET, was used to construct an emissions inventory for 1975. The model then uses these
„ emissions as a benchmark for the analysis.
The method for constructing the 1975 emissions data base was consistent for all pollutants; however,
two different sources of emissions data were necessary in order to obtain time series information on all
pollutants. MSCET contains monthly State-level emission estimates from 1975 to 1985 by emission
•''''. " . 145. ' . ' " ' •
-------
Appendix B: Emissions Modeling
source group for SOz, NO,, and VOC. Therefore, MSCET information was used for SOz, NO,, and VOC,
while Trends data were used for TSP and CO. Emission source groups from MSCET were matched with
1985 NAPAP Inventory SCCs. The MSCET methodology is benchmarked to the 1985 NAPAP Inventory
and uses time series information from Trends in conjunction with activity information to estimate State-
level emissions for SOz, NO,, and VOC. Although the level of detail contained in the NAPAP Inventory
could not be preserved because of the aggregation needed to match with MSCET emissions sources,
MSCET provided the State-level spatial detail required by CRESlff-
„ ~ VJ,V
Once the 1985 emissions by SCC and State from the 1985 NAPAP Inventory were matched with
emission source groups and States from the MSCET data base, an estimate of 1975 emissions was
computed by multiplying the 1985 NAPAP Inventory emissions valueby the ratio of 1975*MSCET
emissions to 1985 MSCET emissions. Ratios were computed and applied for each combination of State,
pollutant, and MSCET emission source group.
This method of constructing an emissions inventory for 1975 utilizes the State estimates from MSCET,
thus capturing the spatial shifts that occurred over the analysis period. It is assumed that NAPAP provides
the most reliable point and area source information in terms of the level of 1985 emissions (which is also
the assumption of the MSCET methodology). Note that if there were a 1-to-l correspondence between
MSCET and NAPAP, this method would be equivalent to using the MSCET methodology directly for
constructing 1975 emission levels.
A similar method was used for TSP and CO, but since these pollutants are not included in MSCET,
the Trends ratio of 1975 to 1985 emissions for these two pollutants was used. Thus, for TSP and CO, all
States were assumed to have experienced the same change in emissions as indicated by the national
figures.
» i ii ^ , <•
It should be noted that in addition to the loss in spatial detail, the Trends source groups generally
spanned several NAPAP source categories. The strength in the Trends information is the consistency of
emissions estimates over time. It is considered to be the most reliable data for tracking changes in
emissions over the time period of the analysis, and was therefore chosen for developing 1975 estimates for
TSPandCO. "
The 15 source categories reported in Trends were matched with those in the 1985 NAPAP Inventory.
The ratios of 1975 emissipHs to 1985 emissions by source category that were applied to the 1985 NAPAP
emissions data are shown in Table 56. The 1975 emissions data estimated from the above procedure
served as the benchmark and initial value for the CRESS emissions module for both scenarios.
146
-------
Appendix B: Emissions Modeling
Trends Source Category ,
co*
211*. ./.;flH
' ^o,9i;
^''"'4:43'
1,17, ,,1.00
111*. /.vU76
CAA regulation of commercial/residential
emissions was limited and largely confined to
fuel combustion sources (SO^ NOX, TSP),
gasoline marketing (VOC), dry cleaning
(VOC), and surface coating (VOC). NSPS
regulations of small (over 29 MW capacity)
fuel combustors were promulgated in 1984
and 1986. For purposes of emissions
calculations, the stipulated NSPS for SQj,
NO,, and TSP were incorporated into the
control scenario for 1985 and 1990. Emission
rates for source categories subject to VOC
regulation were similarly adjusted.
Energy Data
Nearly 75 percent of the source categories
in CRESS use energy consumption by State
and sector as the driver for the emissions
calculation. State-level energy consumption
statistics are published by EIA in State Energy
Data Report, Consumption Estimates, I960-
1989, and are electronically available as part
of the State Energy Data System (SEDS)
(DOE, 1991). The SEDS data bas^contains
annual energy cbrisumption estimates by
sector for the various end-use sectors:
residential, commercial, industrial and
transportation, and electric utilities.
• _&:f ,._ _~
Seypi fuel-type categories are used in
CRESS: coal, distillate oil, residual oil,
, liquid petroleum gas, wood, and
ricity. The model assumes zero
ption of residual fuel oil in the
_*i&<-'iv'
[ sector and zero consumption of
•-••*•'?-*?* *
I in the commercial sector. Energy
consumption for each fuel-type was expressed
in BTUs for purposes of model calculations. t: ''- .' '"'''.' ' '- ' - "^ '' "'
With the exception of wood consumption, all
of the energy consumption statistics used in
CRESS were obtained from SEDS.
Residential wood consumption estimates were derived from two data sources. State-level residential
sector wood consumption estimates for 1975 and 1980 were obtained from Estimates of U.S. Wood Energy
Consumption from 1949 to 1981 (EIA, 1982). State-level wood consumption, however, was not available
Commercial/Institutional Fuel
Combustion: "•',';
:_CoaI '„'„,'-, / '^
Natural-Gas :„ " -, v • ;
Fuel da/ . r ^ ' J...
, Othe/ • I
Residential Fuel Combustion:
"Coat s v ' "
x Naturalfias ' '
Fuel'Oil-. - ^ r ,s- ,
%--Woo4 ^ ' - ;_ ,
Miscellaneous: Forest'Fires
Solid Waste BfaposalC
Incineration , - (
Open Burning - ,
Miscellaneous Other Burning •
Industrial Processes: Paving
Asphalt Paving and Roofing
MiscellaneousOther
2.1
1.33
0.56
0.56
Note: "These values are the ratios of 1985 Trowfc emissions to
1975 ?VemisemIssfcins for each source category. Forexampfe, the
comraercial/ institutional fuel combustion: coal emission ratio of
2.11 is computed as the ratio of the 1975 TSP emissions of 40
,147
-------
Appendix B: Emissions Modeling
for 1985 and 1990, therefore, regional information from an alternative publication, Estimates of U.S.
Biofuels Consumption 1990 (EIA, 1990), was used to derive State-level residential wood use figures.
Regional 1985 and 1990 wood consumption was distributed among States using 1981 State shares. All
wood consumption figures were converted to BTLTs using an average value of 17.2 million BTU per short
ton.
Economic/Demographic Data
Emissions from slightly more than 25 percent of the CRESS source categories follow State-level
economic and demographic activity variables. The demographic variables used by CRESS include State-
level population, rural population, and forest acreage. State population is the activity indicator for six
emissions source categories for SO^ NO,, TSP, and CO, and^Es VOC source categories. State population
data were assembled from the SEDS data base. Rural population, which is the indicator of residential open
burning activity, is computed as a fraction of total State population. Forest wildfires and managed open
burning activity are related to 1977 State-level forest acreage. The demographic information is assumed to
be invariant to CAA regulations and thus is the same in the control and no-control scenario scenarios.
Car stock (or vehicle population), the driver variable for the auto body refinishing, is approximated by
State motor vehicle registrations. Highway Statistics, an annual publication by the FHWA, was the source
for data on State motor vehicle registrations. The three source categories connected with gasoline
marketing are driven by State-level gasoline sales in gallons. Stall gasoline consumption was obtained
from the SEDS data base. Housing starts and 10 percent of the existing housing stock were combined to
form the activity indicator for architectural surface coating emissions. Housing data compiled by the U.S.
Bureau of the Census were available in the Statistical Abstract of the United States (DOC, 1975; 1977;
1982; 1983; 1987; 1993). Regional-level data for 1975 was allocated to the States based on the 1980 State
distribution.
No-control Scenario Emissions
Adjustments to control scenario emissions in each of the target years to reflect conditions under the
no-control scenario were achieved through emission factors, energy input data, and economic/demographic
data. The adjustments made to each of these variables to generate no-control scenario emissions are
discussed individually in the following subsections.
Emissions Data
CAA regulation of the commercial/residential sector was minimal. For regulated source categories,
emission factors were revised to reflect pre-regulation emission rates. Six commercial/residential source
categories were regulated for VOC emissions: Service Stations Stage I Emissions, Service Stations Stage
II Emissions, Dry Cleaning (perchloroethylene), Gasoline Marketed, Dry Cleaning (solvent), and Cutback
Asphalt Paving. Commercial-Institutional boilers were regulated for SO2 and TSP and internal combustion
sources were regulated for NOX emissions. All NSPS were removed for these sources to estimate
no-control scenario emissions levels.
148
-------
Appendix B: Emissions Modeling
Energy Data , . .
State-level energy demand for the residential and commercial sectors for the no-control scenario was
estimated from the J/W model forecast. Final energy demand estimates for the household sector were
calculated by an EPA contractor for the purposes of the no-control scenario analysis. State allocation of
the national-level estimates was based on historic State shares, i.e., this assumes that there is no change in
the distribution of energy demand across States as a result of removing regulations. In addition, the J/W
model estimates an aggregate -"_-". T „
refined petroleum category and
does not distinguish among
liquid petroleum gas, distillate
oil, and residual oil. The
relative shares among these three
categories of petroleum products
remained constant between the
control and no-control scenarios.
The information on percentage
change in energy demand by
fuel type as provided by the J/W
model is listed in Table 57.
Table 57. Percentage Change in Real Energy Demand by Households
from Control io No-control Scenario. " ~ , ' > > '
Year
1975"
1980
1985
1990,
Coal Refined
' Petroleum
1>.48
1.50
^ 3
4.76 t
3.84\
*J D/\^
•jfy\j
433
Electric Natural ,.
s~ v ' Gas*
3.62
426
„ 3.88
4.18
„ , '2l42
, , * 2.12
-2.41-
> _, 2.77
The differential for
commercial sector final energy demand was calculated from the combination of four intermediate product
flow categories from the J/W forecast. The National Income and Product Accounts (NTJPA) for the
commercial sector correspond to J/W SIC categories 32 through 35: .
(32) " ~3^l&esale and Retail Trade;
(33) Finance, Insurance, arid Real Estate;
(34) Omef Services; and
(35) Government Services.
Percentage change information from the J/W forecast for energy cost shares, value of output, and
energy prices was used to cMculate the differential in commercial sector energy demand for the no-control
scenario. The energy cost share is defined as the cost of energy input divided by the value of the output.
M tinier to calculate the percentage change in commercial sector energy demand, the change in energy
e was subtracted from the percentage change in energy cost, and added to the change in the value of
output. Each of thes| variables was available from the J/W model results. This.calculation was performed
F the folar energy types, and each of the four NIP A categories. The change in commercial sector
einarid was obtained by taking the weighted average of the four NIP A categories. Since data on
relative energy demand for NIP A categories were not readily available, square footage was used as a proxy
for calculating the weights. These data were taken from the Nonresidential Buildings Energy
Consumption Survey, Commercial Buildings Consumption and Expenditure 1986 (EIA, 1989). The
resulting estimate for commercial sector changes in energy demand is provided in Table 58.
The national-level change in commercial sector energy demand was allocated to the States using
historic shares. Implicit is the assumption that removal of CAA regulations does not alter the State
distribution of energy use.
149
-------
Appendix B: Emissions Modeling
Economic/Demographic Data Table'58. Percentage Change in Commercial Energy Demancffrom
*Contr<>I to l^p-e:ontrol Scenario. " ' '^ ' ° ' I
!*»»,
Aw tt
State population was
assumed not to vary as a result
of CAA regulations, thus only
the economic variables were
revised for the no-control
scenario. No-control scenario
housing starts and car stock were
derived from J/W forecast
information on construction and
motor vehicles. The differential ^ ,^_
for categories 6 (construction) ^"">m^^^^^fm
and 24 (motor vehicles and
equipment) was applied to control scenario values to
obtain no-control scenario levels. The percentage;
change from the J/W forecast is given in Table 59.
State-level gasoline sales is one of the activities
forecasted by the transportation sector model. The
percentage change in gasoline sales calculated by the
TEEMS model was used in the no-control scenario as
a CRESS model input.
Year
1975
1980
1985
1990
Coal
-0.13
031
0.48
0.39
Refilled
Petroleum
336 s
1.90
1.98
2.26
Electric
130,'
2.06
1.72
1.74
Natural
Gas :
^ f f-
-0^82
* ; ,-0.22
Table 59. JW Percent Differen|fel in Economic
Variables Used in CARESS, { " -/ '
J. J>
£
'^
Year
1975
1980
1985
1990
% <
* ; A
ijl *<.H
J
Construction
0.70
0.14
0.41
0.29
'• ' J '"•, , ",<•
.Motor
Vehicles
f ;,
' ^9," t
-- 'U^ ' '" ;
/\ t ' *
150
-------
Appendix B: Emissions Modeling
ear (In thousands of short
* v ,, '<-.,*, * •>
•c * •"*' - •" s t ,/:,
<" » „ - ^'v^~
Sector ^ '„•>-, . -J97S
Transportation: -'' *• - *' f* %•
Highway Vehicles *~ 50tf v
~^iHfcnwa| Vehicles, „ ( ^270-
Stationary Sources: • , ,J
.' HecttidUwTities' ~- - J#20
Industrial Processes " ~* r t $620
Industrial Boilers - "„ ' 74Gt '
Commercial/Residential . 2,020
TOTAL*X • " j> * v , . /. 11,070
With the CAA
»^8tf
/ > h „
<•
- ' 760 -
280
880
3,650
480
^510
8^50,
1985
-,*_ ,
t.
770
270' ,
',
' - 4JKK
^,040 ,
250
2&8Q
7,460
v J
Without the CAA
1990.'
*<, "*
820
280
( 430
3#80
' '240
2^50
,7,390
, 1975.
v^ 5, *"
f
770
_260!
'
3,460^
v 11,120'
780
2,020
18,410
1980
f, t A
£
910
270
4,480^
12,000
550
2^20
20,730
1985
> >.
1,030
26»
„
5,180
11,710,
^360
2,700'
2U50
' Difference
,1990, Emissions
., ' , " r.
^
1480
\ '270"
" ^
5,860,
', 127960'
400^ v
"2»5oO
,23,230^
s'
'^30%)
5%
' (93%)
X76%)
<41%) ^
P^%)
(68%)
*#'*'•*. <.- ^ ^ ^. ^ -*-' ^
Notes: Theesiimates of emission levels with and without the CAA were developed specifically for this Section 812 analysis using models
„':.'•'''•• designed to simulate conditions in the absence of the CAA. These numbers should not be interpreted as actual historical emission
' ~ ~ ~ "
f>, „ « . ^x
Tafele 61. SO, Emissions Under the Control an
Transportation:
Highway Vehicles- ""
Off-Highway Vehicles ~
Stationary Sources: , ' .
Electric Utflities .
f Industrial Processes^' ^ ->
fodustrirf Boilers
Commercial/Residential
TOTAL* ' . ' » t ,
A 1975
!
380
- 378
* >
18,670
, ,4230
, 3,440
1,000
- '28380 > ~^
1980
, "
" 450"
530 ;
'
17,480
3,420
'3",18&
800
25,860
1985
,.
500
,410
.,
16J050
2,730
2,660
590
2^950
1990
-
_ '570
390
K
16^10
2,460
2,sa>v
*690
23,440
1975
,
380
, 360
^
' 20,690
5,560
3,910
1,000
31,900
" f
Without the CAA
1980
450
530
25,620
5^940
' 4,110
810
37,460
1985
'500
400
25,140 -
5,630, l
4,020
610
36,310
1990
>.
560
i
26,730
6,130
4,610
fib
39440
DUTereace
In 1990
Eil5Sk)fM
' * ^
1%
*' 1% *
« i ,
(38%)
^,
(60%)
(39%) s
(3%)
C ,(40%)
The estimates of emission levels with and without the CAA were developed specifically for tins Section 812 analysts using models
designed to simulate conditions in the absence of the CAA. These numbers should not be interpreted as actual historical emission
estimates.'' f "< ^ »
*"-'**• ~ ' * t ' s J
•Totals may oTfier slightly &om sums due to rounding.
151
-------
Appendix B: Emissions Modeling
ins Under the Control and No-control Scenarios by Target Year (in thousands of short
:,,;/,.._-, ,..,n
^^j^i-if.
I & i
\, .,,*! '
i ^ ty
';," : , -, -. ::::: With the CAA
Sector "
Transportation:
Highway Ve&icfcs
Off-Highway Vehicles
Stationary Sources:
r EkctricUtaities
loduitritl Processes
IndiutriilBoikrs
Comnrera'il'Residentii]
TOTAL*
1975
8,640
1,990
5^40
750
4,090
1,060
22,060
1980
9,340
2,180
6,450
760
3,680
960
23,370
1985
8,610
2,080
6»66JJ,
690
3^40
880
22,460
1990
8,140
2,090
7,060
3,710
930.
22,640
1975
9,020
1,980
5,740
760
4,120
1,060
22,680
1 \ !
V- '* ,,, '^- ' ,\
Without the CAA
1980
11,060
2,150
7,^
830
3^60
970
25,830
1985
13,160
2,040
7,780
'790
3,680
890*
28,35'0
i
\
1990
1S390 '
2,060"
8^00"
1,090
* "dso
3L680
DUTennce
hi 1990
Eniasioitt
(47%)"
* tW
(29%)
IlBllliii!S8^ 1™* TLMiL si^nl t JidSi'S'Jiil^^ \t ^-ASM ^ •> "^ ^ v V ty *" f ~*
aj||l||B^ 3r ff ^ T Irfp 1 & Mi 111 ftp *" P's s"l*r*' **» " * *.•* M ' /& ^ 4
Notes^ Jtee^tumtes of emission jteveteu^t^M^^ 812 analysis using models
•;_=;• .^ designed to simulale conditions in the absence of the CAA. These numbers should' not be interpreted as actual historical emission
, t ,
slighily from sums due to rounding.
ible 63. VOC Emissions Under the Control and No-control Scenarios by Target Year (In thousands of short :
i>.rf, *.i|l'l^;!J,1;il^l1'iJ,!il'^l!Hl,l!j|lll,^I:ll!'lll!|^,'!lP|l^,l ' T s s E f J u ^ •** -*^
I
i
i
i
i
IP,
i.
'f
V, ' '. '.. ," . ,"'••': . "'' With the CAA
Sector
Traasportttioo:
H^hwayVehlcJes
Off-Higkwjy Vehkks
SUtFontry Sources:
Electric Utilities.
udttstrial Processes
IndujtrulBoflers
Commerdal/Residentia!
TOTAL-
1975
12^20
1380
20
5^10
ISO
4,980
24,660
1980
10,770
1,370
30
6,780
' 150
5,480
24,580
1985
9,470
1^40
30
6,230
" fed
5,820
23,030
1990
,,
7,740
1,410
40
5,630
150
5,870
20,840
1975
14,620
1,390
20
6,130
150
4,980
27,290
WithouttheCAA
1980
!'
16,460
1,42&
30
7,930
150
5,700
31,680
1985
•••"< i
ii
19,800
W96
^
30
7,2901"
140
5 ^
6,080
34,730
i s" '
1996
-
23,010
f j f
\
f f
40
6,810
' 150
6,130*
37,630
Difference
InI99ft
V
' t 51
tf ) t
t' , f *
I ^
_ W
0%'
f
(45%)
Notes The estimates of emission levels with andwiihout the CAA were developed specifically for this Section 812 analysis using models
designed to simulate conditions in the absence of the CAA. These numbers should not be interpreted as actual historical emission
• estimates. " ' ' ' ' . > ^ , „
,!.' , I' i'.! 'Li, '' , I ' ' , ,' ' ! - " ' ?' ~* * S ;
-------
Appendix B: Emissions Modeling
"'•^ ;,"*.„ 4*,-^ ' ' v "- ,. ^ ™ 4 < <• ' '* - - < < > ' " „<• <"
• Table 64. CO Emissions Under the Control and No-control Scenarios by Target Year (in thousands of snort
""' *"** <"* "- •*
With the CAA
DHTenace
~~ Inl99»
Se&wv
198S , 1990
1975 " 1980 , .1985
Ttansportationt,*
Highway Vehides ,:>
OfHlighway Vehicles >*
83,580
72,490' \ 65,430
8,190 „ ,; 7,880 % ^080
90,460' 105,530 131,420 149^80
8,070 7,880
Industrial Etoaers' < ~
Commercial/Sesidehtial
7,580.,
, '720
280, f 290 370
6,990 ' 4,840 " 5,140
710s . 670' ^ ,740
13,130 14,140 13,150
110,880 .109.170 100300 ' „ 32,900
2»> 290
9^40 „ -9,120
'720 710
,10^50 13,170
119,430 136.880. 163^280 181,860
300 * , 380
, 8,860 „ 10,180,
620 v 740'.
r4;200
m
• Notes: TJie estimates of emissronleveU»^
'" ;;-; '' .,,... designed to simulate condilHMis in the absence of the CAA. These numbers should not be interpreted as actual historical emission
*?<"-J *T" : s
;. Lead (Pb) Emissions Under the Control and No-control Scenarios by Target Year (in thousands of
t.-^-'S^ . *. t -s "*• *• *-f •• *
- ,. »
"* "* v If.^
Sector ' -s „
Transportation:
Highway Vehicles '
Stationary Source:
Industrial Processes ,
.Industrial Combustion ~ <
umaies « <"..
fOTAL» " -' >;
'
With the CAA
"1975
180
3
> 4
,1
190
, 1980
> V
86
1
. „ 2
1
90
1985
- -„
,22
1-
t)
0
23
«:1990
^
2
/• •«
0
O
3
, 1975-
><
203
7
5-
-- 2
, 217
Without the CAA
1980
'207
7
"5
3
221
1985
> 214
6
"5*
4
228
; : Difference
v- , te!990
1990 * **Eni!3s!cffi$
-
223 " (99%)'
( X
-5 ^7^)
t fo&y\
4 (95%)
237 (99%)
The estimates of emission levels with and without the CAA were developed specifically for Ibis Section 812 analysis using models
":'' designed to simulate conditions in the absence of the CAA. These numbers should not be interpreted as actual historical emission,
estimates.
-' \ >* ' •. > v /
t; ,f " totals may differ slightly from sums due to rounding, „ '
153
-------
Appendix B: Emissions Modeling
Emissions Modeling References
Abt, 1995: Abt Associates Inc., "The Impact of the Clean Air Act on Lead Pollution: Emissions
Reductions, Health Effects, and Economic Benefits from 1970 to 1990," Final Report, Bethesda, ME?,,
October 1995.
ANL, 1990: Argonne National Laboratory, "Current Emission Trends for Nitrogen Oxides, SuMuuT J;*
Dioxide, and Volatile Organic Chemicals by Month and State: Methodology and Results," Argonne,
IL, August 1990.
ANL, 1992: Argonne National Laboratory, "Retrospective Clean Air Act Analysis: Sectoral Impact on
Emissions from 1975 to 1990," (Draft), Argonne, IL, July 1992.
Braine, Bruce, and P. Kim, "Fuel Consumption and Emission Estimates by State," ICF Resources, Inc.,
Fairfax, VA, memorandum to Jim DeMocker, EPA, April 21,1993.
Braine, Kohli, and Kim, 1993: Braine, Bruce, S. Kohli, and P. Kim, "1975 Emission Estimates with and
without the Clean Air Act," ICF Resources, Inc^Tairfax, VA, memorandum to Jim DeMocker, EPA,
April 15,1993.
DOC, 1975: U.S. Department of Commerce, Bureau of the Census, "Statistical Abstract of the United
States: 1975 (96th Edition)," Washington/DC, September 1975.
DOC, 1977:, U.S. Department of Commerce, Bureau of the Census, "Statistical Abstract of the United
States: 1977 (98thEdition)," Washington, DC, September 1977.
DOC, 1981: U.S. Department of Commerce, Bureau of the Census, "1977 Truck Inventory and Use
Survey," TC-77-T-52, Washington, DC, August 1981.
DOC, 1982: U.S. Department of Commerce, Bureau of the Census, "Statistical Abstract of the United
States: 1982-1983 (IC&d Edition)," Washington, DC, December 1982.
IHDC, 1983: U.S. Department of Commerce, Bureau of the Census, "Statistical Abstract of the United
States: 1984 (lO^th Edition)," Washington, DC, December 1983.
DOC, 1984: U.S. Department of Commerce, Bureau of the Census, "1982 Truck Inventory and Use
Survey," TC-82-T-52, Washington, DC, August 1984.
DOC, 1987: U.S. Department of Commerce, Bureau of the Census, Statistical Abstract of the United
States: 1988 (108th Edition), Washington, DC, December 1987.
DOC, 1990: U.S. Department of Commerce, Bureau of the Census, "1987 Truck Inventory and Use
Survey," TC87-T-52, Washington, DC, August 1990.
154
-------
Appendix B: Emissions Modeling
DOC, 1991: U.S. Department of Commerce, "Annual Survey of Manufactures: Purchased Fuels and
Electric Energy Used for Heat and Power by Industry Group," M87(AS>1,-Washington, DC, 1991.
DOC, 1993: U.S. Department of Commerce, Bureau of the Census, "Statistical Abstract?6flie United
States: 1993 (113th Edition)," Washington, DC, 1993.
DOE, 1982: U.S. Department of Energy, Energy Information Administration, "Documentation of the „„
Resource Allocation and Mine Costing (RAMC) Model," DOE/NBB-0200,1982. jf
DOE, 1986: U.S. Department of Energy, Energy Information Administration, "Inventory of PowefPlants
in the United States 1985," DOE/EIA-0095(85), Washington, DC, August 1986. - r ~
DOE, 1988: U.S. Department of Energy, "An Analysis of Nuclear Power Plant Operating Costs," Energy
Information Administration, DOE/EIA-0511(88), 1988.
DOE, 1990: U.S. Department of Energy, Energy Information Administration, "State Energy Price and
Expenditure Report 1988," DOE/EIA-0376(88)^W^hington, DC, September 1990.
DOE, 1991: U.S. Department of Energy, Energy Information Administration, "State Energy Data Report:
Consumption Estimates - 1960-1989," Dt)E/EIA-0214(89)," Washington, DC, May 1991.
DOE, 1992: U.S. Department of Energy, Energy Information Administration, "Annual Energy Review
1991," DOE/EIA-0384(91), Washington, DC, 1992.
EIA; 1982: Energy Information Administration, "Estimates of U.S. Wood Energy Consumption from
1949 to-1981," U.S. Department of Energy, DOE)EIA-0341, August 1982.
'"~~ f- ' . "
EIA, 1985: Energy Information Administration, "Cost and Quality of Fuels for Electric Utility Plants,"
U.S. DepartoenTof Energy, DOE/ELM)091(85), 1985.
EIA, 1989; Energy Information Administration, "Nonresidential Buildings Energy Consumption Survey:
Commercial Buildings Consumption and Expenditures 1986," U.S. Department of Energy, DOE/EIA-
Jp8(86), May 1989. .*
" ' ' /„ .' ' - • , -
l990: Energy Information Administration, "Estimates of U.S. Biofuels Consumption 1990," U.S.
ent of Energy, DOE/EIA-0548(90), October 1990.
:?U:S. Environmental Protection Agency, "Compilation of Air Pollutant Emission Factors,
. Stationary Point and Area Sources," AP-42, Fourth Edition, GPO No. 055-000-00251-7,
sarch Triangle Park, NC, September 1985.
EPA, 1989: U.S. Environmental Protection Agency, "The 1985 NAPAP Emissions Inventory," EPA-
600/7-89-012a, Research Triangle Park, NC, November 1989.
"^ -
EPA, 1990: U.S. Environmental Protection Agency, "The Cost of a Clean Environment," EPA-230-11-90-
083, November 1990.
155
-------
Appendix B: Emissions Modeling
EPA, 1991: U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards,
"National Air Pollutant Emissions Estimates, 1940-1990," EPA-450/4-91-026, Research Triangle
Park, NC, November 1991.
EPA, 1992: U.S. Environmental Protection Agency, "1990 Toxics Release Inventory," EPA-700-S-92-
002, Washington, DC, 1992.
EPA, 1994a: U.S. Environmental Protection Agency, "National Air Pollutant Emission Trends, 1900-
1993," EPA-454/R-94-027, Office of Air Quality Planning and Standards, Research Triangle Park,
NQ October 1994. "' , " ' '
EPA, 1994b: U.S. Environmental Protection Agency, Office of Mobile Sources, "User's Guide to
MOBILES (Mobile Source Emission Factor Model)," EPA-AA-AQAB-94-01, Ann Arbor, MI, May
1994.
EPRI, 1981: Electric Power Research Institute, "The EPRI Regional Systems," EPRI-P-1950-SR, Palo
Alto, CA, 1981.
FHWA, 1986: Federal Highway Administration, U.S. Department of Transportation, "1983-1984
Nationwide Personal Transportation Survey," Washington, IX!, 1986.
FHWA, 1988: Federal Highway Administration, U.S. Department of Transportation, "Highway Statistics
1987," PB89-127369, Washington, DC, 1988.
FHWA, 1992: Federal Highway Administration, U.S. Department of Transportation, "Highway Statistics
1991," FHWA-PL-92-025,Washington, DC, 1992. '
Gschwandtner, 1989: Gschwandtneir, (aerhard, "Procedures Document for the Development of National
Air Pollutant Emissions Trends Repor£'E.H. Pechan & Associates, Inc., Durham, NC, December
1989. • -.-....,•';':/:.^;,r
Hogan, 1988: Hogan, Tim, "Industrial Combustion Emissions Model (Version 6.0) Users Manual," U.S.
Environmental Protection Agency, EPA-600/8-88-007a, 1988.
ICF, 1992: ICF Resources, Inc., "Results of Retrospective Electric Utility Clean Air Act Analysis - 1980,
1985 and 1990," September 30,1992.
Jorgenson and Wilcoxen, 1989: Jorgenson, D.W., and P. Wilcoxen, "Environmental Regulation and U.S.
Economic Growth," Harvard University Press, Cambridge, MA, 1989.
Klinger and Kuzmyak, 1986: Klinger, D., and J.R. Kuzmyak, "Personal Travel in the United States, Vol.
I: 1983-84 Nationwide Personal Transportation Study," U.S. Department of Transportation, Federal
Highway Administration, Washington, DC, August 1986.
Kohout et al., 1990: Kohout, Ed, "Current Emission Trends for Nitrogen Oxides, Sulfur Dioxide, and
Volatile Organic Compounds by Month and State: Methodology and Results," Argonne National
Laboratory, ANL/EAIS/TM-25, Argonne, IL, 1990.
156
-------
Appendix B: Emissions Modeling
Lockhart, 1992: Lockhart, Jim, "Projecting with and without Clean Air Act Emissions for the Section 812
Retrospective Analysis: A Methodology Based Upon the Projection System used in the Office of Air
Quality Planning and Standards National Air Pollutant Emission Estimate Reports," (Draft Report),
Environmental Law Institute, November 16,1992.
McDonald and South, 1984: McDonald, J.F., and D.W. South, "The Commercial and Residential Energy
Use and Emissions Simulation System (CRESS): Selection Process, Structure, and Capabilities,"
Argonne National Laboratory, ANL/EAIS/TM-12, Argonne, IL, October 1984.
Mintz and Vyas, 1991: Mintz, M.M., and A.D. Vyas: "Forecast of Transportation Energy Demand
through the Year 2010," Argonne National Laboratory, AUL/ESD-9, Argonne, IL, April 1991.
Nelson R., T. Tietenberg and M. Donihue, "Differential Environmental Regulation: The Effects on
Electric Utility Capital Turnover and Emissions," Review of Economics and Statistics, Vol. 75, No. 2,
May 1993.
Pechan Associates, 1995. The Impact of the Clean Air Act on 1970 to Commissions; Section 812
Retrospective Analysis. Draft Report. March 1,1995.
Saricks, 1985: Saricks, C.L., "The Transportation Energy and Emissionis Modeling System (TEEMS):
Selection Process, Structure, and Capabilities," Argonne National Laboratory, ANL/EES-TM-295,
Argonne, IL, November 1985. '
US EPA/OAR/OMS. User's Guide to MOBILES. EPA-AA-AQAB-94-01. May 1994.
1990: VeselkaVT-D,, et al., "Introduction to the Argonne Utility Simulation (ARGUS)
ModeltArgonne.National Laboratory, ANL/EAIS/TM-10, Argonne, IL, March 1990.
* V ~! , "
Vyas and Saricks, 1986: Vyas, A.D., and C.L. Saricks, "Implementation of the Transportation Energy and
Emissions Modeling System (TEEMS) in Forecasting Transportation Source Emissions for the 1986
Assessment," Argonne National Laboratory, ANL/EES-TM-321, Argonne, IL, October 1986.
Werbol, 1983: Werbos, Paul |., "A Statistical Analysis of What Drives Energy Demand: Volume in of
the PURHAPS ModeKDocumentation," U.S. Department of Energy, Energy Information
- Administration, DQEfllA-0420/3, Washington, DC, 1983.
157
-------
158
-------
Appendix Cs Air Quality Modeling
introduction
This appendix describes in greater detail the various methodologies used to translate differences in
control and no-control scenario emission estimates into changes in air quality conditions. Summary
characterizations of the results of the air quality modeling efforts for 1990 are provided here and in the
main text. Further details and discussion of key analytical and modeling issues can be found in a number
of supporting documents. These documents, which provide the analytical basis for the results presented
herein, are:
>• ICF Kaiser/Systems Applications International, "Retrospective Analysis of Ozone Air Quality in the United
States", Final Report, May 1995. (Hereafter referred to as ^'SAI Ozone Report (1995).")
> ICF Kaiser/Systems Applications International, ^Retrospective Analysis ofParticulate Matter Air Quality in the
United States", Draft Report, September 1992. (Hereafter referred to as "SAI PM Report (1992).")
> ICF Kaiser/Systems Applications International, "Retrospective Analysis ofParticulate Matter Air Quality in the
United States", Final Report, April 199»5. (Hereafter referred tdfSS "SAI PM Report (1995).")
» ICF Kaiser/Systems Applications International, "PM Interpolation Methodology for the §812 Retrospective
Analysis"; Memorandum from J. Langstaff to J. DeMocker, March 1996. (Hereafter referred to as "SAI PM
Interpolation'Memo (1996).")
^ ^ ' ' »
* ICF Kaiser/Systems Applications International, "Retrospective Analysis of SO2, NOX and CO Air Quality in the
United States"-, Final "Report, November 1994. (Hereafter referred to as "SAI SO2, NOX and CO Report
(1994).")
-' •* _
» ICFKaiser/Systems Applications International, "Retrospective Analysis of tfie Impact of ike Clean Air Act on.
Urban Visibility in the Southwestern United States", Final Report, October 1994. (Hereafter referred to as "SAI
SW Visibility Report (1994).")
»- Dennis, Robin L., US EPA, ORD/NERL, "Estimation of Regional Air Quality and Deposition Changes Under
^ Alternative 812 Emissions Scenarios Predicted by the Regional Acid Deposition Model, RADM", Draft Report,
-. October 1995. (Hereafter referred to as "Dennis, R. RADM Report (1995).")
•v . , •
The remainder of this appendix describes, for each pollutant or air quality effect of concern, (a) the
basis for development of the control scenario air quality profiles; (b) the air quality modeling approach
used to estimate differences in air quality outcomes for the control and no-control scenario and die •
application of those results to the derivation of the no-control scenario air quality profiles; © the key
, assumptions, caveats, analytical issues, and limitations associated with the modeling approach used; and
(d) a summary characterization of the differences in estimated air quality outcomes for the control and no-
control scenarios.
159
-------
Appendix C: Air Quality Modeling
Carbon Monoxide
Control scenario carbon monoxide profiles
As described in the preceding general methodology section, the starting point for development of
control scenario air quality profiles was BPA's AIRS database. HourlyCO air quality monitoring data
were compiled for all monitors in the 48 contiguous states for the study target years of 1970,1975,1980,
1985, and 1990. Although the CO monitoring network was sparse in 1970, by 1990 506 monitors in 244
counties provided monitoring coverage for 55 percent of the population in the conterminous U.S. Table 66
summarizes the CO monitoring data derived from AIRS. Additional data regarding the EPA Region
location, land use category, location-setting category, and objective category of the monitors providing
these data are described in the SAI SOa, NO,, and CO Report~(1994).
•iji'Xabie66. S
isii Sit .I ,.i(i
:!
ii
y,
1
ml
I
l,'S
1.
I
m
Year
,; !],„ «i.
i970
1975
1980
1985
1990
[l 1, i. Itl , 1 I > ,'i Si - f - ( 1 , , V ' ""'1 v" i
ummaiy of CD Monitoring Data. :,,::., n
tlivViAi'^i, l*<'"< M.'^ u •"> u V. -" - , «. V ,. '':
Number of
Monitors
82
503
522
472
506
Number of
Counties
54
246
250
232
244
Percent
Population
Covered
n/a
n/a
50%
n/a
55%
t
i
Number of
Samples
1
•j 5
n
, , _ ^
D»U Source: SAI SO^ NO, and CO Report (1994). "
The next step in constructing the control scenario air quality profiles was to calculate moving averages,
for a variety of time periods, of the hourly CO data for each monitor. For CO, moving averages of 1, 3,5,
7,8,12, and 24 hours were calculated. Daily maximum concentrations observed at each monitor for each
of these averaging periods were then calculated. Finally, profiles were developed to reflect the average and
maximum concentrations for each of the seven averaging periods. However, profiles were only developed
for a given monitor when at least 10 percent of its theoretically available samples were actually available.
The purpose of applying this cutoff was to avoid inclusion of monitors for which available sample sizes
were too small to provide a reliable indication of historical air quality.
As discussed in the air quality modeling chapter of the main text, development of representative
distributions for these profiles was then necessary to provide a manageable characterization of air quality
conditions. Initially, two-parameter lognormal distributions were fitted to the profiles based on substantial
evidence that such distributions are appropriate for modeling air quality data. However, given the relative
160
-------
Appendix C:Air Quality Modeling
importance of accurately modeling higher percentile observations (i.e., 90th percentile and higher), a three-
parameter modeling approach was used to isolate the effect of observations equal, or very close, to zero. In
this approach one parameter defines the proportion of data below a cutoff close to zero and the remaining
two parameters describe the distribution of data above the cutoff value. Several other studies have already
demonstrated good fit to air quality modeling data with a three-parameter gamma distribution, and both
lognormal and gamma distributions using a three-parameter approach were developed for the present
study. As documented in the SAISO^ NO,, and CO Report (1994), a cutoff of 0.05 ppm was applied and
both the three-parameter lognormal and three-parameter gamma distributions provided a good fit to the f
empirical data. For CO, the gamma distribution provided the best fit. .,
The control scenario air quality profiles are available on diskette.- The filename for the CO Control
Scenario profile database is COCAA.DAT, and adopts the format presented in Table 67.
No-control scenario carbon monoxide profiles
To derive comparably configured profiles representing CO air quality in the no-control scenario,
control scenario profile means and variances were adjusted in proportion to the difference in emissions
estimated under the two scenarios. Specifically, for all control scenario air quality observations predicted
by the three-parameter distributions falling above the "near-zero" cutoff level, comparable no-control
estimates were derived by the following equation:
'• ^ * -' \*
"~X. **•• "*• -fc **•"
where XNCAA = _ air quality measurement for the non-CAA scenario,
= air quality measurement for the CAA scenario,
=- emissions estimated for the non-CAA scenario,
ECAA = emissions estimated for the CAA scenario, and
~W* = backpxmnW^ncenltration., .
The adjustment for background concentration is made to hold ambient background concentrations of
the pollutant constant between the control and no-control scenarios. To the extent background
concentrations are affected by transport of anthropogenic pollutants from upwind sites, and to the extent
upwind emissions may have been controlled under the control scenario, assuming a fixed background
concentration represents a conservative assumption in this analysis. As discussed in the SAI SO^ NOX>
and CO Report (1994), the CO background concentration used for this analysis was 0.2 ppm, which equals
the lowest typical concentration observed in the lower 48 states.
In the SAI SOj, NOW and CO Report (1994) documenting the CO air quality modeling effort,
reference is made to using county-level emission estimates as the basis for deriving the no-control profiles.
Derivation of these county-level results is described in more detail in the appendix on emissions
estimation. It is important to emphasize here, however, that the county-level CO emissions data were
derived for both the control and no-control scenarios by simple population-weighted disaggregation of
161
-------
Appendix C: Air Quality Modeling
state-level emission totals. Although CO emission estimates were needed at the county level to support the
ozone air quality modeling effort, differences in state-level emissions estimates are
162
-------
Appendix C: Air Quality Modeling
*"x * -\ "" •* \ t
!Table°67. Format of Air Quality Profile Databases*
.vi
•*"'••
\
v<
t
<:
<0
£
•*• :
•V
^
V
;•
Columns
: „ "i *2 '
^ „ -4 '-6"
!; ^ ,»s-9~
, ^ii-is
-"la?- 19"
^21 -30*
^ v w
I >. 3&V41
J;43-44'
- *n46-S5f"
'« v » . X
' ,xm~65
,^ "66-75^
^ 76-85
' v ~ 86V 95'
> " ^105
4Q6T^i5
-116-125'
,126-135
136-145
- Format -
Integer ' „„' „, _
Integer :
Integer -v< ,
, fat eger
Idteger- - -
-Iteal v «"" "^
Real" « s " >
Ifirteger
^JReal(F10.3) „
'Realtl7!)^'
*Real'(FlO.§)
Real'(F103)
NReal(F103)
Real' v< ' - " '' i
~ :• Description
Year (70, 75, 80,85,90) " "'*.',-'
t ^ ->* i. <• > •- f
Averaging time (1,3, 5, 7, $12, 24' hours) , ^
iStateFIPS-code "" _ , ,
"* * v- * v' ~ "^
Cdunty-FlPScodei^ o ^ . - —,<~,
f •> ' ' v » -
Monitor number (digitsr6-10 of mohftor'id)
Latitude ^ ^ ^ ^ "-T4
^ v
Longitude *,-'"- ^
LafitudeAongitudeflag" " , v "" -
" - _, ' ' ' ~ *>
Hourly intennittency parameter pb * „
* ^ , i,
Hourly lognorm^I parameter/t*
Hourly lognormal parameter ob , . /^c,.
Hourly'ganima parameter ab
Hourly ganima parameter Pb
Daily max intermittency parameter p6
Daily tnax lognormal parameter A" v •,
Daily max lognormal parameter o*
Daily max gamma parameter ab
Daily max gamma parameter pb ' l^
'Values for flag: 1 = actual latitude/longitud& values
.. . ', :' 2= latkude/lon&mde values from allocated monitor or previous mcmitor
' ' . ' ° location (monitorpararaeter occurrence code 1)
, -9 = latitude/longitude missing (county center substituted)
^ ^ V ^
^ p ^ ^
' bUnils of concentration are ppm for CO and ppb for SO,, NO, and NO.
/Source; ySAI^SO2, NO, and CO Report (1994).
163
-------
Appendix C: Air Quality Modeling
what drive the difference in the control and no-control air quality profiles for CO. In other words, the
ENCM to-^GM ratios used to derive the no-control profiles according the Equation (1) above are essentially
based on state-level emissions estimates for CO.
As for the control scenario air quality profiles, the no-control scenario air quality profiles are available
on diskette. The filename for the CO No-control Scenario profile database is CONCAA.DAT. The same
data format described in Table 67 is adopted.
Figure 32. Frequency Distribution of Estimated Ratios for
1990 Control to No-control Scenario 95th Percentile 1-Hour
Average CO Concentrations, by Monitor.
Summary differences In carbon monoxide air quality
While the control and no-control scenario
air quality profiles are too extensive to present
in their entirety in this report, a summary
indication of the difference in control and
no-control scenario CO concentrations is
useful. Figure 32 provides this summary
characterization. Specifically, the air quality
indicator provided is the 95th percentile
observation of 1990 CO concentrations
averaged over a 1-hour period. The graph
shows the number of monitors for whwh the
ratio of 1990 control to no-control scenario
95th percentile 1-hour average concentrations
falls within a particular ranger The x-axis
values miSie graph represent the midpoint of
each bin. The results indicate that^liy1990,
CO concentrations under a no-control scenario
would have been dramatidilly higher than
control scenario concentrations.
300
200
100
0.06 0.25 0.45 0.65 0.85 1.05 1.25
R£*to Of CAA-No-CAA 95thPercenUle 1-Hour Averegs
Key caveats and uncertainties for carbon monoxide
A number of important uncertainties should be noted regarding the CO air quality estimates used in
this analysis. First and foremost, CO is a highly localized, "hot spot" pollutant. As such, CO monitors are
often located near heavily-used highways and intersections to capture the peak concentrations associated
with mobile sources. Since this analysis relies on state-level aggregate changes in CO emissions from all
sources, the representativeness and accuracy of the predicted CO air quality changes are uncertain. There
is no basis, however, for assuming any systematic bias which would lead to over,- or under-estimation of air
quality conditions due to reliance on state-wide emission estimates.
A second source of uncertainty is the extent to which the three-parameter distributions adequately
characterize air quality indicators of concern. Appendix C of the SAI SOa, NOX, and CO Report (1994)
presents a number of graphs comparing the fitted versus empirical data for one-hour and 12-hour averaging
164
-------
Appendix C: Air Quality Modeling
periods. In the case of CO, the gamma distribution appears to provide a very, reasonable fit, though clearly
some uncertainty remains.
• ' . . : t .'fa.
Finally, as noted in footnote 29, a central premise of this analysis is that changes in (SO emissions
should be well-correlated with changes in CO air quality. Strong correlation between the state-level
emissions estimates used in this analysis and empirical air quality measurements would not be expected
due to inconsistencies between the state-level scale of modeled emissions veMiis^e monitor-level scale of
the air quality data, and between the modeled control scenario emissions inventories and actual historical
air quality measurements. Under these circumstances, it is particularly important to focus on the primary
objective of the current analysis, which is to estimate the difference in air quality outcomes between
scenarios which assume the absence or presence of historical air pollution controls. In the process of
taking differences, some of the uncertainties are expected to cancel out No attempt is made in the overall
analysis to predict historical air quality, or hypothetical air quality in the absence of the Clean Air Act, in
absolute terms.
Sulfur Dioxide
\
Sulfur dioxide (SOj) emissions lead to several air quality effects, including secondary formation of fine
particle sulfates, long range transport and deposition of sulfuric acid, and localized concentrations of
gaseous sulfur dioxide. The first two effects are addressed later in this appendix, under the participate
matter and acid deposition sections. The focus of this section is estimation of changes in local
concentrations of sulfur dioxide. •
The methodology applied to estimation of local sulfur dioxide air quality is essentially identical to the
one applied for carbon monoxide. As such, this section does not repeat the "roll-up" modeling
methodological description presented in the CO section, but instead simply highlights those elements of
the sulfur dioxide modeling which differ from carbon monoxide.
Control scenario sulfur dioxide prof ties
Unlike the CO monitoring network, the number of monitors as well as the population coverage of the
SO2 monitoring network shrank during the 1980's. Table 68 summarizes the SO2 monitoring data used as
jthe basis for development of the control scenario air quality profiles.
5 As for CO, air quality profiles reflecting average values and daily maxima for 1,3,5, 7, 8,12, and 24
hour averages were compiled from AIRS for monitors in the lower 48 states which had at least 10 percent
of their potential samples available. Applying a cutoff of 0.1 ppb to isolate the zero and near-zero
observations, three-parameter lognormal and gamma distributions were fitted to these empirical profiles.
In the case of SO^ the three-parameter lognormal distribution was found to provide the best fit.
The control scenario SO2 air quality profiles are available on diskette, contained in a file named
SO2CAA.DAT. The same data format described in Table 67 is adopted.
165
-------
Appendix C: Air Quality Modeling
?'t*V, > i ," >"" * '- <"•
•Table 68. Summary of SO2 Monitoring Data.
'ift&ffliiA^ - , :. **4 v , ,,...«;, . * . , ,
Year
1970
1975
1980
f . 1985
| 1990
Number of
Monitors
86
847
1,113
926
769
Number of
Counties
56
340
440
401
374
Percent
Population
Covered
n/a
n/a
60%
n/a
50%
! i }
Number of
Samples
399,717
4,280,303
6,565,589
6,602,615
5,810,230
,!•!>•<<« !, * . '
' * ? • V» ' > _ ,
, , , , > < ( ,
/•
-------
Appendix C: Air. Quality Modeling
Figure 33. Frequency Distribution of Estimated Ratios for
1990 Control to No-control Scenario 95th Pjercentile 1-Hour
Average SO2 Concentrations, by Monitor.
Key caveats and!
uncertainties for sulfur
dioxide
The height of stacks used to vent flue
gases from utility and industrial fossil fuel-
fired boilers has a significant effect on the
dispersion of sulfur dioxide and on the
formation and long-range transport of
secondary products such as particulate
sulfates. Under a no-control scenario, it is
conceivable that some sources might have
built taller stacks to allow higher emission
rates without creating extremely high ground-
level concentrations of flue gases. On the
other hand, it is also conceivable that, in the
absence of post-1970 air pollution control
programs, sources might have built shorter stacks to avoid incurring the higher costs associated with
building and maintaining taller stacks. To the extent facilities would have adopted different stack height
configurations under a no-control scenario, both local exposures to sulfur dioxides (and other emissions
from fossil fuel combustion) and long-range transport, deposition, and exposure associated with secondary
formation products may have been different. However, this analysis assumes that both the location of
individual facilities and the height and configuration of emission stacks are constant between the two
scenarios. If, in fact, stack heights were raised under the historical case due to CAA-related concerns,
increases in local SO2 concentrations under the no-control scenario may be overestimated. However, this
same assumption may at the same time lead to underestimation under the no-control scenario of long-range
transport and formation of secondary particulates associated with taller stacks. For stacks built lower under
a no-control scenario, local SO2 exposures would have been higher and long-range effects lower. Finally,
the comments on uncertainties for carbon monoxide on page 165 apply as well to SO2.
OJD5 025 0-45 OJ55 0£6 1JD5 125
RatbofCAA*Jo-CAA 95th Penxntib 1-Hour Average
Nitrogen Oxides
% .~ ^ • '
: - Similarly to sulfur dioxide, emissions of nitrogen oxides (NOX ) -including nitrogen dioxide (NOj) and
nitrous oxide (NO)- lead to several air quality effects. These effects include secondary formation of fine
particle nitrates, formation of ground-level ozone, long range transport and deposition of nitric acid, and
localized concentrations of both NO2 and NO. The first three effects are addressed later in this appendix,
under the particulate matter, ozone, and acid deposition sections. The focus of this section is estimation of
changes in local concentrations of NO2 and NO. .
The methodology applied to estimation of local nitrogen oxides air quality is essentially identical to the
one applied for carbon monoxide and sulfur dioxide. As such, this section does not repeat the "roll-up"
modeling methodological description presented in the CO section, but instead simply highlights, those
elements of the nitrogen oxides modeling which differ from carbon monoxide.
167
-------
Appendix C: Air Quality Modeling
Control scenario nitrogen oxides profiles
•/••$$&"
After peaking around 1980, the number of NO2 and NO monitors, their county coverage, and their
population coverage shrank between 1980 and 1990. Tables 69 and 70 summarize, respectively, the NO2
and NO monitoring data used as the basis for development of the control scenario air quality profiles.
'^Summary of NO2 Monitoring Data.
i,!i!!:':'i,!,lii|i»i|T»if':*llii''!:i:'lii|'i'i''t'»|ii|||i,|,||,F,,,|f'' r.,i m H i i i . C' -,
Si
!
t
I
j
1
f
I'cSlKj&'SF
Year
1970
1975
1980
1985
1990
*JM'J,»{' M
iS;!ior,iii'L':
Number of
Monitors
45
308
379
305
346
Jfj'"'-'l -'n' '
1 ' „ , . . '
Number of
Counties
32
155
205
182
187
,' \f „
Percent
Population
Covered
n/a
n/a
45%
n/a
40%
1- '**><.''
Number of
Samples
275,534 '
1,574,444
1,984,128
i
2,142,606
2,456,922
•t H ,. ', ".' ' >
Mean
Number of
Samples per
Momitor
, 6,123
5,112 I
5,235
7,025
7,101 '
\ ,J{ ^.^J^
t • n
(D«t«SooiceiSAlSOj,NOIandC»Report(1994)
" i hit I III | | I (Hill I HI | Jt] I » I r J M
r
1*11''*!
Table 70. Summary of NO Monitoring Data,
rt
"i!
lj!l
3
Jn!1
\
I-
Year
1970
1975
1980
1985
1990
Number of
Monitors
39
206
224
139
145
Number of
Counties
28
94
124
86
81
Percent
Population
Covered
n/a
n/a
30%
n/a
15%
Number of
Samples
246,262
1,101,051
1,023,834
956,425
999,808
Mean
Number of
Samples per
Monitor
6314
5,345 -
4,571
i- /
6,881
6,895 s '
Datt SOWCK SAlSOj, NO^and CO Report (1994).
168
-------
. Appendbf C: Air Quality Modeling
As for CO and SO,, air quality profiles reflecting average values and maxima for 1, 3, 5, 7, 8,12, and
24 hour NO2 and NO averages were compiled from AIRS for monitors in .the lower 48 states which had at
least 10 percent of their potential samples available,, Applying a cutoff of 0.5 ppb to both NO2 and NO to
isolate the zero and near-zero observations, three-parameter lognormal and gamma distributions were fitted
to these empirical profiles. For NO2 and NO, the three-parameter gamma distribution was found to
provide the best fit.
The control scenario NO2 and NO air quality profiles are available on diskette, contained in files
named N02CAA.DAT and NOCAA.DAT, respectively. The same data format described in Table 67 is
adopted. .
No-control scenario nitrogen oxides profiles
The no-control air quality profiles for NO2 and NO are derived using the same equation -equation 7 on
page 161- that was applied for CO and SO2. As discussed in detail in the SAI SO2, NOX, and CO Report
(1994),85 nitrogen oxides are emitted almost entirely from anthropogenic sources and they do not have long
atmospheric residence times. Therefore, global background concentrations are very low, on the order of
0.1 or 0.2 ppb. For the present analysis, background concentrations of NO2 and NO were assumed to be
zero. . '
The no-control scenario NO, and NO air quality profiles are available on diskette, contained in files
named NO2NCAA.DAT and NONCAA.DAT, respectively. The data format is described in Table 67.
Summary differences in nitrogen oxides air quality
Figure 34 provides a summary indication '
of the differences in control and no-control
scenario air quality for NO2. As for CO and
SO,, the graph shows the distribution of 1990
control to no-control scenario 95th percentile
1-hour average concentration ratios at NO,
monitors. These ratios indicate that, by 1990,
no-control scenario -NO, concentrations were
significantly higher than they were under the
control scenario. The changes for NO are
similar to those for NO,.
Figure 34. Frequency Distribution of Estimated Ratios for
1990 Control, to No-control Scenario 95th Percentile 1-Hour
Average NO2 Concentrations, by Monitor.
300
200 -
O
I
"100 -
0.05 0.25 . 0^5 0.65 i 0.85 1.05 125
Ratb of CAA JJ o-CAA 951h PeroentSe 1-Hour Average
'85SAI SO3, NOX> and CO Report (1994), page 4-9.
169
-------
Appendix C: Air Quality Modeling
Key caveats and! uncertainties for nitrogen oxitSes
A number of caveats and uncertainties specific to modeling NOX should be noted. First, stack height
and stack height control strategies likely to have influenced local concentrations of SO2 may also have
influenced local concentrations of NO2 and NO. (For a fuller discussion of the stack heights issue, refer to
the section "Key caveats and uncertainties for SO2" on page 167.) In addition, the discussion on page 165
of uncertainties resulting from the use of state-level emissions and the cancellation of uncertainties
resulting from analyzing only differences or relative changes also applies to NOX. > .
Acid Deposition
The focus of air quality modeling efforts described above for carbon monoxide, sulfur dioxide, and
nitrogen oxides was to estimate the change in ambient concentrations of those pollutants as a result of
changes in emissions. Particularly since the emissions modeling was driven by modeled macroeconomic
conditions, rather than actual historical economic activity patterns, neither the emissions inventories nor
the resultant air quality conditions developed for this analysis would be expected to match historical
outcomes. The need to focus on relative changes, rather than absolute predictions, becomes even more
acute for estimating air quality outcomes for pollutants subject to long-range transport, chemical
transformation, and atmospheric deposition. The complexity of the relationships between emissions, air
concentrations, and deposition is well-described in the following paragraph from the RADM report
document developed by Robin Dennis of US EPA's National Exposure Research Laboratory in support of
the present analysis:86
"Sulfur, nitrogen, and oxidant species in the atmosphere can be transported hundreds to
thousands of kilometers by meteorological forces. During transport the primary emissions, SO^
NO# and volatile organic emissions (VOC) are oxidized in the air or in cloud-water to form new,
secondary compounds, which are acidic, particularly sulfate and nitric acid, or which add to or
subtract from the ambient levels ofoxidants, such as ozone. The oxidizers, such as the hydroxyl
radical, hydrogen peroxide and. ozone are produced by reactions ofVOC andNO-g. The sulfur
and nitrogen pollutants are deposited to the earth through either wet or dry deposition creating a
load of pollutants to the earth's surface... However, the atmosphere is partly cleansed ofoxidants
through a number of physical processes including deposition (e.g., ozone is removed by wet and
dry deposition). Dry deposition occurs when particles settle out of the air onto the earth or when
gaseous or fine particle species directly impact land, plants, or water or when plant stomata take
up gaseous species, such as SO? In wet deposition, pollutants are removed from the 'atmosphere
by either rain or snow. In addition, fine particles or secondary aerosols formed by the gas- and
aqueous-phase transformation processes scatter or absorb visible light and thus contribute to
impairment of visibility."
** Dennis, R. RADM Report (1995), p. 1.
170
-------
Appendix C: Air Quality Modeling
Figure 35. Location of the High Resolution RADM 20-km Grid Nested
Inside the 80-km RADM Domain.
The complexity and
nonlinearity of the
relationships between
localized emissions of
precursors, such as SO2 and
VOC, and subsequent
regional scale air quality and
deposition effects are so
substantial that the simple
"roll-up" modeling
methodology used for
estimating local ambient
concentrations of SOj, NO^
and CO is inadequate, even
for a broad-scale, aggregate
assessment such as the
present study. For sulfur
deposition, and for a number
of other effects, addressed in
subsequent sections of this
appendix, a regional air
quality model was required.
After careful review of the
capabilities, geographic
coverage, computing
intensity, and resource
requirements associated with
available regional air quality
models, EPA decided to use ~~"~^™^^^^"^^~""" "-
various forms of the Regional
Acid Deposition Model (RADM) to estimate these effects.87 Figure 35 shows the geographic domain of
theRADM.
Control scenario acid deposition profiles
~. The derivation of control scenario emission inventory inputs to the RADM model is succinctly
described in this excerpt from the Dennis, R. RADM Report (1995):
The RADM model requires a very detailed emissions inventory in both time and space. The
emissions fields are also day-specific to account for the temperature effects on the volatile
organics and the wind and temperature effects on the phone rise of the major point sources. At
the time of the 812 Retrospective Study RADM runs, these inventories had been developed for
1985, using the 1985 NAPAP (National Acid Precipitation Assessment Program) inventory, and
87 For a detailed description of the various forms of the RADM and its evaluation history, see the Dennis, R. RADM Report (1995).
• '• '. • ''-.'"'..' . • 171 . . "' ' • "'
-------
Appendix C: Air Quality Modeling
Figure 36. RADM-Predicted 1990 Total Sulfur Deposition
(Wet + Dryj in kg/ba) Under the Control Scenario.
adjusted for point source emissions to 1988 for the Eulerian Model Evaluation Field Study
funded by NAPAP. These RADM emissions inventories had county-level and detailed SCC and
species-level information incorporated into them to provide the 80- and 20-km detail. The 812
Study emissions are principally computed at the state level While the 1985 812 Study emissions
are close to the NAPAP inventory, they do not exactly match, nor do they have the spatial, nor
economic sector, nor species detail within a state needed to run RADM. To connect the 812 ,
Study emissions to the RADM emissions, the following approach was followed: An
industry/commercial-level disaggregation (including mobile sources) was developed for the 812
emissions to allow different sectors in a state to change their emissions across time without
being in lock step and the detailed NAPAP emissions for every 80- and 20-km RADM grid-cell
were grouped by state to the same level of industry fcommercial aggregation for an exact
correspondence. Then it was assumed that the 812 Study 1985 control emissions were
effectively the same as the 1985 NAPAP emissions. Relative changes in emissions between the
8121985 control and any , ,
other scenario (e.g.t 1985
no-control, or 1990 control,
or 1980 no-control, etc.)
were then applied to the
1985 NAPAP state-level
industry/commercial groups
in the appropriate 80- and
20-km grid cells. Thus,
state-level emissions for
each group would retain the
same state-level geographic
pattern in the different ,
scenarios years, but the mix
across groups could change
with time. In this way, the
more detailed emissions
required by RADM were
modeled for each scenario
year using the 812 Study
emissions data sets.
Although the focus of the present
analysis is to estimate the differences
between the control and no-control
scenarios, it is useful to illustrate the
absolute levels of acid deposition associated with the two scenarios. It is particularly important to
demonstrate the initial deposition conditions to preclude possible misinterpretations of the maps showing
percent change in deposition. A relatively high percentage change in a particular region, for example, may
occur when initial deposition is low, even when the change in deposition is also modest. The RADM-
modeled 1990 control scenario wet and dry sulfur deposition pattern is shown in Figure 36. A comparable
map for nitrogen deposition is presented in Figure 37. Maps of the RADM-predicted 1990 No-control
Scenario sulfur and nitrogen deposition are presented in Figures 38 and 39, respectively.
172
-------
Appendix C: Air Quality Modeling
No-control
scenario acid
deposition profiles
Configuration of the RADM
model for the present analysis
—including allocation of emission
inventories to model grid cells, design
of meteorological cases, treatment of
biogenic versus anthropogenic
emissions, and temporal, spatial, and
species allocation of emissions— are
described in detail in the Dennis, R.
RADM Report (1995). The remainder
of this section provides a summary
description of the acid deposition
modeling effort.
For sulfur deposition, the RADM
Engineering Model (RADM/EM),
which focuses on sulfur compounds,;;
was used to derive annual average
total (wet plus dry) deposition of
sulfur in kilograms sulfur per hectare
(kg-S/ha) under both the control and
no-control scenarios. The relative
changes in annual average total sulfur
deposition for each of the 80-km
RADM/EM grid cells for 1975,1980,
1985, and 1990 were then compaed.
Nitrogen deposition was
calculated in a different manner.
Since nitrogen effects are not included
in the computationally fast
RADM/EM, nitrogen deposition had
fd be derived from the full-scale,
15-layer RADM runs. Because of the
cost and computational intensity of
the 15-layer RADM, nitrogen
deposition estimates were only
developed for 1980 and 1990. As for
sulfur deposition, the relative changes
in annual average total (wet plus dry)
nitrogen deposition, expressed as kg-
'igure 37. RADM-Predicted 1990 Total Nitrogen Deposition
j(Wet + Dry; in kg/ha) Under the Control Scenario.
figure 38. RADM-Predicted 1990 Total Sulfur Deposition
/et + Dry; in kg/ha) Under the No-control Scenario.
173
-------
Appendix C:Air Quality Modeling
N/ha, were calculated for each 80-km
grid cell and for each of the two
scenarios. It is important to note that
ammonia depositin contributes
significantly to total nitrogen
deposition. However, the activities of
sources associated with formation and
deposition of ammonia, such as
livestock farming and wildlife, were
essentially unaffected by Clean Air
Act-related control programs during
the 1970 to 1990 period of this
analysis. Therefore, ammonia
deposition is held constant between
the two scenarios.
Summary
dlffer&nc&s in acid
deposition
Figure 40 is a contour map
showing the estimated percent
increase in sulfur deposition under the
no-control scenario relative to the
control scenario for 1990. Figure 41
provides comparable information for
nitrogen deposition. These maps
indicate that by 1990 acid deposition
would have been significantly higher
across the RADM domain under the
no-control scenario.
Examination of the percent
change sulfur deposition map
indicates relatively large percentage
changes in the upper Great Lakes and
the Florida-Southeast Atlantic Coast
areas. This result may appear
somewhat surprising to readers .
familiar with the historical patterns
of acid deposition. However, a
review of the emission data and the
control scenario sulfur deposition
map reveal the reasons for this result.
Figure 39. RADM-Predicted 1990 Total Nitrogen Deposition
(Wet + Dry; in kg/ha) Under the No-control Scenario.
igure 40. RADM-Predicted Percent Increase in Total Sulfur
Deposition (Wet + Dry; in kg/ha) Under the No-control
Scenario.
174
-------
Appendix C; Air Quality Modeling
First, Figure 36 shows that control scenario deposition rates are relatively low. As described above,
even a small absolute increase in deposition leads to a large percentage increase in areas with low initial
rates of deposition. Second, the scenario differences in SOX emission rates for these areas were substantial.
For example, 1990 no-control scenario total SOX emissions for Michigan were approximately 1.8 million
tons but control scenario emissions for the same year were less than 600,000 tons; a reduction of over two-
thirds. Similarly, 1990 no-control scenario emissions for Florida were over 2.3 million tons, compared to
approximately 800,000 tons under the control scenario; also a reduction of about twcnthirds. Almost 1
million tons of the Michigan reduction and approximately 1.3 million tons of the Florida reduction were
associated with utilities. Emission reductions of these magnitudes would be expected to yield significant
reductions in rates of acid deposition.
Key oayeats and uncertainties for add deposition
Regional-scale oxidant and deposition modeling involves substantial uncertainty. This uncertainty
arises from uncertainties in modeling atmospheric chemistry, incomplete meteorological data, normal
seasonal and temporal fluctuations in atmospheric conditions; temporal and spatial variability in
emissions, and many other factors. Uncertainties specific to the RADM model, and this particular
exercise, are discussed in detail in the Dennis, R. RADM Report (1995). It is important, however, to
highlight some of the potential sources of modeling uncertainty unique to this analysis.
The first source of uncertainty specific to this analysis is associated with the spatial and geographic
disaggregation of emissions data. As discussed in the Dennis, R. RADM report, the RADM model
requires emission inventory inputs which are highly disaggregated over both time and space. The ideal
emissions inventory fed into the RADM model-includes day-specific emissions to account for temperature
effects on VOCs and the significance of localized meteorological conditions around major point sources.
Given the broad-scale, comprehensive nature of the present study, such detailed emissions inventories were
not available. However, the industry/commercial-level disaggregation approach developed for the present
analysis would not be expected to introduce any systematic bias, and the contribution of this disaggregation
of emissions would not be expected to contribute significantly to the overall uncertainty of the larger
analysis.
175
-------
Appendix C: Air Quality Modeling
The acid deposition estimates
included in the present analysis are
limited in that only the eastern 31 of
the 48 coterminous states are covered.
Although acid deposition is a problem
primarily for the eastern U.S., acid
deposition does occur in states west of
the RADM domain. The magnitude
of the benefits of reducing acid
deposition in these western states is
likely to be small, however, relative to
the overall benefits of the historical
Clean Air Act
(Figure 41. RADM-Predicted Percent Increase in Total Nitrogen
[Deposition (Wet + Dry; in kg/ha) Under the No-control
(Scenario.
Matter
Developing air quality profiles for
particulate matter is significantly
complicated by the fact that
"particulate matter" is actually an :
aggregation of different pollutants
with varying chemical and
aerodynamic properties. Particulate species include chemically inert substances, such as wind-blown sand,
as well as toxic substances such as acid aerosols; and include coarse particles implicated in household
soiling as well as fine particles which contribute to human respiratory effects. In addition, emissions of
both primary particulate matter and precursors of secondarily-formed participates are generated by a wide
variety of mobile and stationary sources, further complicating specification of particulate air quality
models. Finally, particulate air quality models must take account of potentially significant background
concentrations of atmospheric particles.
Modeling multiple species and emission sources, however, is not the only major challenge related to
particulate matter which is faced in the present study. Over the 1970 to 1990 period being analyzed,
understanding of the relative significance of fine versus coarse particles evolved significantly. Up until the
mid-1980's, particulate air quality data were collected as Total Suspended Participates (ISP). However,
during the 1980's, health scientists concluded that small, respirable particles, particularly those with an
aerodynamic diameter of less than or equal to 10 microns (PM10), were the component of particulate matter
primarily responsible for adverse human health effects. As of 1987, federal health-based ambient air
quality standards for particulate matter were revised to be expressed in terms of PM10 rather than TSP.
Starting in the mid-1980's, therefore, the U.S. began shifting away from TSP monitors toward PM10
monitors. As a result, neither TSP nor PM10 are fully represented by historical air quality data over the
1970 to 1990 period of this analysis. Furthermore, a large number of U.S. counties have no historical PM
monitoring data at all, making it difficult to estimate changes in ambient concentrations of this significant
pollutant for areas containing roughly 30 percent of the U.S. population.
176
-------
Appendix C: Air Quality Modeling
Given the relative significance of participate matter to the bottom-line estimate of net benefits of the
historical Clean Air Act, it was important to develop methodologies to meet each of these challenges. The
methodologies developed and data used are described primarily in the two supporting documents SAIPM
Report (1992) and SAI PM Report (1995).88 To summarize the overall approach, historical TSP data were
broken down into principal component species, including primary participates, sulfates, nitrates, organic
particulates, and background participates. Historical data were used for the control scenario. To derive the
no-control profiles, the four non-background components were scaled up based on corresponding
no-control to control ratios of emissions and/or modeled atmospheric concentrations. Specifically, the
primary participate component was scaled up by the ratio of no-control to control emissions of PM.
Organic constituents were scaled up by the ratio of no-control to control VOC emissions. In the eastern 31
states where RADM sulfate and nitrate data were available, values for SO4 and NO3 from an appropriate
RADM grid cell were assigned to the relevant county and used to scale these components of PM. For the
western states not covered by RADM, sulfates were scaled up by the change in SO2 emissions and nitrates
were scaled up the change in NOX emissions. No-control profiles were then constructed by adding these
scaled components to background concentrations.
To resolve the problem of variable records of TSP and PM10 data, both TSP and PMJO profiles were
generated for the entire 20 year period. Missing eariyyeafdata for PM10 were derived by applying region-
specific, land use category-specific PM10 to TSP ratios to the historical TSP data. Missing recent year TSP
data were derived for those areas where PM10 monitors replaced TSP monitors by applying the reciprocal
of the relevant PM10 to TSP ratio. The methodology is described in detail in the SAI PM Report (1995).
In addition, to increase estimates of air quality on a geographical basis, an interpolation methodology89
was developed to predict air quality for the control scenario in counties without measured data. PM
concentrations were estimated by first estimating the components of PM (i.e., sulfate, nitrate, and organic
participate, and primary participate). The methodology for developing the concentrations of components
within a county-differed depending upon whether the county was within or outside the RADM domain.
-#• &* - .
For those counties within the RADM domain, the RADM modeled concentrations for 1980 and 1990
were used to predict sulfate air quality. Relationships based on linear regressions that related 1980 and
1990 RADM sulfate concentrations to estimated sulfate participate concentrations were calculated for
counties with AIRS data. Sulfate particulate concentrations were then calculated for all counties in the
domain by applying the regression results to the RADM grid cell concentration located over the county
center."Statewide average nitrate, VOC, and primary particulate concentrations were calculated from
measured ambient TSP and PM10 to describe these constituents in counties without data. Control PM
profiles were developed by adding the RADM-estimated sulfate particulate levels with the statewide
average nitrate, VOC, and primary particulate levels, and background.
— .&-* . . .
For counties outside the RADM domain, an alternate procedure was used. Using the primary and
secondary particulate estimates for counties with data, statewide average sulfate, nitrate, VOC, and primary
particulate concentrations were determined. Control PM10 was predicted by adding the statewide averages
of all primary and secondary particulate, and background. Using this method, all counties that did not
88 In addition, SAI memoranda and reports which supplement the results and methodologies used in this analysis are included in the ,
references.
89 The interpolation methodology is described in detail in SAI, 1996. Memo from J. Langstaffto J. DeMocker. PM Interpolation
Methodology for the §812 Retrospective Analysis. March 1996.
177
-------
Appendix C: Air Quality Modeling
have monitors and are in
the same state are assigned
the same PM concentration
4 j s SK <,
r Table 71. Summary of TSP Monitoring Dafau'
profiles. These interpolated
results are clearly less
certain than results based
on actual historical
monitoring data and are
therefore presented
separately.
Control
scenario
particulate
matter
profiles
The number of TSP
monitors peaked in 1977
and declined throughout the
1980's. Table 71
summarizes the daily (i.e.,
24-hour average) TSP
monitoring data used as the
basis for development of
the control scenario air
quality profiles. Most of
the TSP and PM^ monitors
collected samples every six
days (i.e., 61 samples per
year).
f
1'
1
1
f
lit
Year
1970
1975
1980
1985
1990
Number of
Monitors
751
3,467
3,595
2,932
923
,
Number of
Counties
-245
1,146
1,178 *
1*018
410
> f I i.
Number of
Samples
: -/
,56,804
• ' 221,873
> ,234,503 ,
189,344
t
59,184
* ' * *
"" Mean
Number of
Samples per
<, Monitor
76 '
64
S65 ,' " <
65 "
' ,
. 64' ^'
Source: SAtPM Report (1995).
fl ' •> „ i <- " -
*Table 72. Summary of PMW Monitoring Data.
t
f
t
i
Year
1985
1990
Number of
Monitors
303
1,249
Number of
Counties
194
556
* ;
Number of
Samples
22,031 1
98,904
•" ** f
^ Mean
Number of
Samples per
Monitor
73
^9 .
Data Source: SAIPM Report (1995).
:f>aily PMi0 data were also collected for each year between 1983 and 1990. Table 72 summarizes the
daily PM10 monitoring data used for the control scenario air quality profiles.
Further speciation of TSP and PM10 air quality data serves two purposes in the present analysis. First,
speciation of TSP into PM10 and other fractions allows derivation of PM10;TSP ratios. Such ratios can then
be used to estimate historical PM10 for those years and monitors which had TSP data but no PM10 data.
The reciprocal ratio is also applied in this analysis to expand 1985 and 1990 TSP data to cover those areas
which monitored PM10 but not TSP. The second purpose served by speciation of particulate data is, as
described on page 177, to provide a basis for scaling up concentrations of each species to derive no-control
scenario TSP and PM10 profiles. -
To break the TSP and PM10 data down into component species, speciation factors were applied to the
PM fractions with aerodynamic diameters below 2.5 microns (PM^) and from 2.5 to 10 microns (PMj0).
178
-------
Appendix C: Air Quality Modeling
The PMis speciation factors were drawn from a National Acid Precipitation Assessment Program
(NAPAP) report on visibility which reviewed and consolidated speciation data from a number of studies.90
These factors are presented in Table 73. In the table, fine particle concentrations are based on particle
mass measured after equilibrating to a relative humidity of 40 to 50 percent; and organics include fine
organic carbon.
To develop speciation factors for coarser particles (Le., in the PM^ to PM10 range), a review was
performed by SAI of the available literature, including Conner etal.I(1991), Wolff and Korsog (1989),
Lewis and Macias (1980), Wolff et al. (1983), Wolff et al. (1991), and Chow et al. (I994).91 These ST
speciation factors are summarized hi Table 74. Data were top limited to allow differentiationjbetween
urban and rural locations for coarser particles.
90 J. Trijonis, "Visibility: Existing and Historical Conditions-Causes and Effects," NAPAP Report 24,1990.
91 This literature review, and complete citations of the underlying studies, are presented in the SAI PM Report (1995), pp. 4-2 to 4-6 and pp.
R-l to R-2, respectively. ,
179
-------
Appendix C: Air Quality Modeling
e
i^^llW^ffi'/iV.1' 4
; Component
. „ " , ... • :,-",,!,••: ;.). ";*-.::
Fine particle concentration
Ammonium sulfate
Ammonium nitrate
Organics
• Illg^l^,^ „ „, . ^
Fine particle concentration
Ammonium sulfate
Ammonium nitrate
; Organics
• it i . .
RURAL WEST
Fine particle concentration.
Ammonium sulfate
{ Ammonium nitrate
1
I Organics
T p ,- ^ i|u|| ;ni!n jj j u| ,,,, ^njj, ,,, .',;Jij inil|li,,
i . ,1,1' ifTWMtt A itf ts/ironp
r Jl'ii, \f p^fpfli f^ yTJCnjJt • " L
I Fine particle concentration
1 -• • .- ,
Ammonium sulfate
I Ammonium nitrate
I • •••
i Organics
* 1
Units
.*&- :
/tato,
% Fine particles
% Fine particles
% Fine particles
' ' - '" "i
neha.
% Fine particles
% Fine particles
% Fine particles
"^
^g/m,
% Fine particles
% Fine particles
% Fine particles
> * s
..' .
PSfa
% Fine particles
% Fine particles
% Fine particles
Number of
Data Sets
i "„ • ' '"*
19
19
3
5
r^i./£>
3
3
2
2
~- '"".'-'•
25
25
17
25
— ^ -, ^
16
16
14
16
t> A j
Atithmetk
.Mean
"/, ^ 'Jt, '
< '.
18 ,
'S52 „
1
,24
y- *.'**/:
36
"55 ' *
1
24 «
-1" »•,'"-'
5
35
,4
27
x
-j<- #'••„- '»
35
,16
15
42 ' '
Range of
Values
*/* <- -»
6-46 /
41-66
1
o -^4
-~: 'f'C/T ^,
29-43
, 53 -"57 "
1
15 -32/
•± ' , -V-
1-11
15>-56
' 1-17
14-41
«^ ^ ,? j
..'„'• '^'' ,;
13-74
3-35
- 2-37
25-79
^.^'Data'Sourcest'S
,
CondiSons-Causes and Effects, " NAPAP Report 24, 1990.
I If! ^ , , j,
; and J. Trijonis, "Visibility: Existing and Historical
in|i
180
-------
Appendix C: Air Quality Modeling
e 74. 'Coarse ParticleIPN^tof M10) Chemical Composition by U.S. Region.
.•** -.Yi1" -, ' ', ! , < t'
""V ^ ?
Component ; ~ •>
-^fcjis^^-^
Coarse particle concentration
' , Ammonium sulfate
"• ^x f ^
^ „„ Ammonium nitrate
' , -" 'Organics
- Units '
Number
of Data
f Sets
^^?"3rv";Sflf!^
' ^ms ,
% Coarse particles
% Coarse particles
% Coarse particles
1
3
- 1 •>
2
Arithmetic
Mean
Range of
Values
^ ?
i- ,^"~ "-r ^-^^""v^-r^
N
3
' 4
10 ,
" S5'
„' .1^.4* -
< ^ <
7-.13.8 ,
:*ffii&&^!?^
•S <"
, Coarse particle concentration
Ammonium sulfate
Ammonium nitrate <
> Organics- ,
^g/m, '
% Coarse particles
% Coarse particles
% Coarse particles
18
18
' 18
18
24
6
18
14
7.7-56.7
2.1-1639"
233-28,52
"~ 8.41-25.81'
?D«a Sou**; SAIPMRe^ (1995).
The TSP and PM10 control scenario profiles developed based on this methodology are available on
diskette, under the filenames listed in Table 75.
No-control scenario particulate matter profiles
To derive the no-control TSP and PM10 air quality profiles, individual component species were
adjusted to reflect the relative change in emissions or, in the case of sulfates and nitrates in the eastern
US., the relative change m modeled ambient concentration. The following excerpt from the SAIPM
Report (1995) describ|flhe specific algorithm used:92
-For the retrospective analysis, the no-CAA scenario TSP and PM10 air quality was estimated by
means of the following algorithm:
*• Apportion CAA scenario TSP andPM10 to size categories and species;
*• Adjust for background concentrations;
92 SAI PM Repeat (1995), p. 5-1.
181
-------
Appendix C: Air Quality Modeling
Use a linear scaling to
adjust the non-
background portions of
primary particulates,
sulfate, nitrate, and
organic components
based on emissions
ratios ofPM, SO 2, NO,
and VOC, and Regional
Acid Deposition Model
(RADM) annual
aggregation results for
*• Add up the scaled
components to estimate
the no-CAA scenario
TSPandPMjo
concentrations. "
t i> *** S AT '
<
PMlQCMEAlDAT/
PM10CHI2.DAT
PMlOdp&DAT ?,
W ~\ nr i *vv i { } if i ^ %, i * •* > * >•
* ^ble 7fe. PJ& ISo-Control Scenario Air Quality Profile Filenames*
Component
TSP
TSP
TSP
PM,.
PMT.
PM,.
^ ^
Indicator
Annual Mean
2nd Highest Daily
OQth Percentile
Annual Mean
2nd Highest Daily
(X)thPercentiie
<'s ' -
- Filename
TSPCNMEAJDAT
. TSPNCHI.DAT
? i !•
TSPNOX).DAT
PMlOlsTCME-DAT „
* r> '••
PM10NCHI.DAT
1
PM10NC(X).DAT '
>"-"' " ' ^ l"' '" p~ 1* '•" I "ri!' •** ' * < ' t'1 "
, Itote: "(X)" refers to percentpes from 5 to 95, indicating 19 percentD>tased data files
r available for TSP and 19 similar files available for PM10; for example, tne filename for the
; 50th percentfleTSP air quality data profile for the no-control scenario is named
f-TCSPNCSOJJAT, '' " * „, ' ,
> J , » i i , ' '„ ,.
» SAI PM Report (1995), pp. 5-2 to 5-15.
182
-------
Appendix C: Air Quality Modeling
Figure 42. Distribution of Estimated Ratios for 1990
Control to No-Control Annual Mean TSP Concentrations, by
Monitored County.
50 r-
40
30
20
10
0,00 , 0.20 040 - 0.60 0.80 IJOO^
Rate of CAA Mo^AA AnnualM ean TSP (iitervalm ifcoiit)
Summary differences in
particulate matter air
quality
Figure 42 provides one indication of the
estimated change in particulate matter air
quality between the control and no-control
scenarios. Specifically, the graph provides
data on the estimated ratios of 1990 control to
no-control scenario annual mean TSP
concentrations in monitored counties. The X-
axis values represent the mid-point of the ratio
interval bin, and the Y-axis provides the
number of Bounties falling into each bin.
Figure 42 indicates that annual average TSP ,
concentrations would have been substantially higher in monitored counties under the no-control scenario.
'/ ' - , -
Key caveats and uncertainties for particulate matter
There are several important caveats and uncertainties associated with the TSP and PM,0 air quality
profiles developed for this study. Although further reductions in these uncertainties were not possible for
this study given time and resource limitations, the relative importance of particulate matter reduction
contributions towards total benefits of the Clean Air Act highlights the importance of these uncertainties.
A number of uncertainties were introduced in the process of speciating and rolling up individual
components of particulate matter. First, temporal and spatial variability in the size and chemical properties
of particulate emissions'are substantial. These characteristics change from day to day at any given
location. Second, using changes in proxy pollutant emissions, such as using SO2 as a surrogate for SO4 in
the western states, to roll up individual PM components may introduce significant uncertainty. Third, even
assuming a satisfactorily high degree of correlation between target and surrogate pollutants, relying on
predicted changes in emissions at the state level further compounds the uncertainty. Finally, and perhaps
most important, using PM10 to TSP ratios derived from late 1980's monitoring data may lead to significant
underestimation of reductions in fine particulates achieved in earlier years. This is because historical
Clean Air Act programs focused extensively on controlling combustion sources of fine particulates. As a
result, the share of TSP represented by PM1() observed in the late 1980's would be lower due to
implementation of controls on combustion sources. This would lead, in turn, to underestimation of
baseline PM]0 concentrations, as a share of TSP, in the 1970's and early 1980's. If baseline PM10
concentrations in these early years are underestimated, the reductions in PMIO estimated by linear scaling
would also be underestimated.94 ,
' See SA1 PM Report (1995), p. 5-9.
183
-------
Appendix C: Air Quality Modeling
Nonlinear formation processes, long-range atmospheric transport, multiple precursors, complex
atmospheric chemistry, and acute sensitivity to meteorological conditions combine to pose substantial
difficulties in estimating air quality profiles for ozone. Even in the context of an aggregated, national study
such as this, the location-specific factors controlling ozone formation preclude the use of roll-up modeling
based on proxy pollutants or application of state-wide or nation-wide average conditions. Such
simplification would yield virtually meaningless results for ozone.
Ideally, large-scale photochemical grid models —such as the Urban Airshed Model (UAM)— would
be used to develop control and no-control scenario estimates for ozone concentrations in rural and urban
areas. Such models provide better representations of the effects of several important factors influencing air
quality projections such as long-range atmospheric transport of ozone. However, the substantial
computing time and data input requirements for .such models precluded their use for this study.95 Instead,
three separate modeling efforts were conducted to provide urban and rural ozone profiles for those areas of
the lower 48 states in which historical ozone changes attributable to the Clean Air Act may be most
significant.
First, for urban areas the Ozone Isopleth Plotting with Optional Mechanisms-IV (OZ3PM4) model was
run for 147 urban areas. Table 77 Bstsrtheurban areas modeled with OZEPM4. Although it requires
substantially less input data than UAM, the OZEPM4 model provides reasonable evaluations of the relative
reactivity of ozone precursors and ozone formation mechanisms associated with urban air masses.96 Three
to five meteorological episodes were modeled for each of the 147 urban areas; and for each of these, four
model runs were performed to simulate the 1980 and 1990 control and no-control scenarios. The outputs
of these model runs were peak ozone concentrations for each of the target year-scenario combinations. The
differentials between the control and no-control scenario outputs were averaged over meteorological
episodes and then applied to scale up historical air quality at individual monitors to obtain no-control case
profiles. As for the other pollutants, the control scenario profiles were derived by fitting statistical
distributions to actual historical data for individual monitors.
K For a description of the extensive data inputs requited to operate UAM, see SAI Ozone Report (1995), p. 1-1.
*• See SAI Ozone Report (1995), p. 1-1.
184
-------
Appendix C: Air Quality Modeling
7, Ui*anAreiasafoaeledwitfeOZIPM4.
^ " A s
^Aibaay^NY'
Altoona, PA
', Afderson,IN
~
',AsheviMe,NC
&E
-O Austin; T3t *''
Baltimore, MD
Baton Rouge; LA
"
Birmingham,
(Campaign, IL
Charleston, SC
"Ctiarleston,
' Chattanooga; TTsf-GA
Cflkago, IL
Cincinnati, OH
,aevelarid, OH
. ' COfctado SpriaiS, CO
Columbia, SC
- Columbus, GrA-At '
/ Cblun&us^OH,
Cnmberland, ^^
„ Dallas, TX
Davenport, lA-IL
Decatur,IL -
Denver, CQ,
Detroit, MI
1 Eugene, "OR
* 'Evansville, W
Fayettevme, NG
Fltat^MI '
...Fortpollins, CO
Fort Smith,,!AR-OK
FortWayne/IN
Grand Rapids, MI
Gteetey, CO
GreeiiBay/Wl4
* breensboro,~NC
Gjeen|lle,SC
Housfonj'DC
Huntingdon, WV-KY
Indianapolis, IN '
Iowa City, !A
Jackson, MS
Jacksonville, FL N
JapesviIleRoclcCo,WI
Johns1town,PA
Kndxvifle,
-------
Appendix C: Air Quality Modeling
Second, the 15-layer RADM runs for 1980 and 1990 were used to estimate the relative change in rural
ozone distributions for the eastern 31 states. In addition, a limited number of 20-km grid cell high-
resolution RADM runs were conducted to benchmark the 15-layer, 80-km RADM median ozone response
and to estimate high ozone response. The relative changes in modeled median and 90th percentile rural
ozone were then assumed to be proportional to the changes in, respectively, the median and 90th percentile
ozone concentrations. The domain of the high-resolution RADM is shown in Figure 35 and the general
RADM domain is shown in Figure 43. " ,
Finally, the SARMAP Air Quality Model (SAQM) was run for EPA by the California Air Resources
Board (GARB) to gauge the differences in peak ozone concentrations in key California agricultural areas
for 1980 and 1990. No-control profiles were developed for ozone monitors in these areas by -assuming the
relative change in peak ozone concentration also applies to the median of the ozone distribution. The
domain of the SAQM is shown in Figure 43.
Figure 43. RADM and SAQM Modeling Domains, with Rural Ozone Monitor Locations.
186
-------
Appendix C: Air Quality Modeling
Control scenario ozone prof ties
For ozone, air quality profiles were developed from historical AIRS data and calculated for individual
monitors based on 1,2, 6,12, and 24 hour averaging times. Profiles based on the daily maximum
concentrations for these averaging times were also calculated. Given the significance of seasonal and
diurnal ozone formation, twelve separate profiles of hourly ozone distributions were also developed for six
2-month periods and for daytime and nighttime hours. The 2-month periods are January-February, March-
April, and so forth. The diurnal/nocturnal profiles are divided at? A.M. and 7 P.M.1bocal Standard Time.
All of these profiles are based on constructing 1,2,6,12, and 24-hour moving average profiles from the
hourly ozone data from each monitor.97 A two-parameter gamma distribution is then fitted to characterize
each of these air quality profiles.98 The functional form of the gamma distribution, the basis for deriving
the monitor-specific values for mean and variance, and an analysis of the goodness of fit to the data are
presented in the SAI Ozone Report (1995). - :
Table 78 summarizes the ozone monitoring data
used as the basis for the control scenario profiles. The
distribution of these monitors among urban, suburban,
and rural locations is presented in Table C-2 of the SAI
Ozone Report (1995).
Given the substantial number of alternative air
quality profiles for ozone, approximately 20 high-
density disks are required to hold the profiles, even in
compressed data format. Resource limitations therefore
preclude general distribution of the actual profiles. As
discussed in the caveats and uncertainties subsection
below, however, the substantial uncertainties associated
with model results for any given area preclude
application of these profiles in contexts other than
broad-scale, aggregated assessments such as the present
study. The historical ozone monitoring data used as the
basis for this study are, nevertheless, available through
EPA's Aerometric Information Retrieval System
(AIRS).
Table 78. Summary of Ozone Monitoring Data.
Y«ar
1970
, 1975
1980
1985
1990
Number of
.. Monitors
1
467
791' '
719
834
, Number of
Counties ,
*" Tf
240,
415
415'
477
Data Source: SATOzone Report (1995).
No-controll scenario ozone prof lies
The Specific modeling methodologies for the OZIPM4 runs —including emissions processing,
development of initial and boundary conditions, meteorological conditions, simulation start and end times,
organic reactivity, and carbon fractions-— are described in detail in the SAI Ozone Report (1995).
97 For the nighttime profiles, only 1,2,6, and 12-hour averaged concentrations are derived.
98 Normal and fognormal distributions were also developed and tested for goodness of fit; however, the gamma distribution provided a better
representation of the concentration distribution. .See SAI Ozone Report (1995), page 4-2,
187
-------
Appendix C:Air Quality Modeling
Assumptions and modeling procedures not otherwise described in the SAI report were conducted in
accordance with standard EPA guidance."
Similarly, the RADM modeling methodology used to estimate changes in day-tune rural ozone
distributions in the eastern 31 states are described in detail in the Dennis, R. RADM Report (1995). The
referenced report also provides complete citations of the literature associated with development, standard
application procedures, and evaluation of RADM by the National Acid Precipitation Assessment Program
(NAPAP).
- JP
To derive the no-control scenario results for key California agricultural areas, the California Air
Resources Board and US EPA's Region 9 office agreed to conduct three runs of the SAQM. For the 1990
control scenario, the 1990 SARMAP base case scenario adopted for California State Implementation Plan
modeling was adopted.100 Derivation of 1990 no-control and 1980 control and no-control scenarios was
based on adjusting the aggregate mobile, point, and area source VOC and NOX emissions associated with
each of these cases. For example, the 1980 no-control results were derived by, first, multiplying the 1990
SARMAP base case mobile source VOC emissions by the ratio of 1980 no-control scenario to 1990
control scenario mobile source VOC emissions derived for the present study. Similar adjustments were
made for point and area sources, and for NOX. The SAdM was then re-run holding fixed all other
conditions associated with the 1990 SARMAP base case, including meteorology, activity patterns, and
other conditions. The specific emission ratios used to modify the 1990 SARMAP base case are presented
in Table 79. The ratios themselves were derived by adding on-highway and off-highway emissions to
represent the mobile source category; adding utility, industrial process, and industrial combustion
emissions to represent point sources; and using commercial/residential emissions to represent area sources.
The no-control scenarios were then derived by adjusting the peak and median of the control scenario ozone
distribution based on the ratio of SARMAP-predicted peak ozone concentrations under the control and
no-control scenarios.
. ,. r , M.
ble 79. Apportionment of Emissions Inventories for SAQM Runs.
:?w*i'.tt ........ :*m ' » • • •
i-jO'!1!'.:1.!:!!.,;};- .lit ...... Jllrll'll'
I
t
t-
•
t
VOC
NOX
Source
Category
Mobile
Area
Point
Mobile
Area
Point
1980 Control
to 1990 Control Ratio
1344
0.820
1.284
1.042
0.731
1
0.987
1980 No-Control to
1990 Control Ratio
1.955
0.901
1.439
1.148
0.738
1 J
1.339
1990 No-Control to
1990 Control Ratio
i 3.178
1,106
• i.23'2
14 ),
1.677
i
1.058
1 > ' b f
1.159' *'
,,,,,,,,,
Jiiiw! ! I'.,» iLiiyiillimui i .filiBS.
•» US EPA, Office of Air Quality Planning and Standards, "Procedures for Applying City-Specific EKAtA," EPA-450/4-89-012,1989.
"B DocamenUtion of the SARMAP Air Quality Model and the SARMAP 1990 base case can be found in the SAQM references listed at the
end of this appendix.
188
-------
Appendix C: Air Quality Modeling
Figure 44. Distribution,of Estimated Ratios for 1990
Control to No-control RADM-Simulated Daytime Average
Rural Ozone Concentrations, by RADM Grid Cell.
.200
u
'I 150
«
a' 100
SO
0.00 0.20 , 040 0£0 030 1.00 1.20
RatbofCAA No-CAA O rone-Season D ay tin e A veiage O 2One (iiteivalm Spoilt)
The relative results of the control and
no-control scenario runs of the OZIPM4,
RADM, and SAQM models were then used to
derive the no-control case air quality profiles.
For the urban monitors relying on OZIPM4
.-results, only ozone-season daytime
concentrations could be calculated directly
from OZIPM4 results. This is because
OZIPM4 provides only the maximum hourly
ozone concentration. However, to estimate all
the various physical consequences of changes
in ambient ozone concentrations, the current
study requires estimation of the shift in the
entire distribution of ozone concentrations.
Since it is daytime ozone season
concentrations which are most sensitive to,
changes in VOC and NOX emissions, the
predicted shifts in the most important component of the ozone concentration distribution are reasonably
well-founded. The method adopted for this analysis involved applying the no-control to control peak
concentration ratio to all concentrations in the distribution down to a level of 0.04 ppm. The 0.04 ppm
level is considered at the high end of hypothetical ambient ozone concentrations in the absence of all
anthropogenic ozone precursor emissions. A ratio of i.O is used for ozone concentrations at or near zero.
The methodology is described in more detail in the SAI Ozone Report (1995) on page 4-6.
Estimating changes in rural ozone concentrations is required primarily for estimating effects on
agricultural crops, trees, and other vegetation. For this reason, only the differences in daytime, growing
season.ozone concentrations are derived for the present study. As described in detail in the SAI Ozone
Report (1995) on page 4-7, the no-control rural ozone profiles are calculated by, first, taking the ratio of
the average daytime growing season ozone concentrations simulated by RADM or SAQM (whichever is
relevant for that monitor). The ratio of no-control to control scenario average ozone concentration is then
applied to all the hourly concentrations from that monitor. Profiles based on 1, 2, 6,12, and 24-hour
averages are then calculated for the control case; and averages for daytime hours are calculated for the
no-control case.101 Even though the control and no-control scenario off-season profiles are held constant,
profiles for the no-control scenario are developed for all months of the year since the ozone season varies
throughout the country.
1 The no-conlrol scenario nighttime profiles are assumed to be Ihe same as the control scenario profiles.'
189
-------
Appendix C: Air Quality Modeling
Summary differences in ozone mir quality
Figure 46 presents a summary of the results of the 1990 OZIPM4 results for all 147 of the modeled
urban areas. Specifically, the graph depicts a frequency distribution of the ratio of control to no-control
scenario peak ozone. While the vast majority of simulated peak ozone concentration ratios fall below 1.00,
eight urban areas show lower simulated peak ozone for the no-control scenario than for the control
scenario. For these eight urban areas, emissions of precursors were higher under the no-control scenario;
however, the high proportion of ambient NOX compared to ambient non-methane organic compounds
(NMOCs) in these areas results in a decrease in net ozone production when NOX emissions increase.
Figures 44 and 45 present frequency distributions for control to no-control ratios of average ozone-season
daytime ozone concentrations at rural monitors as simulated by RADM and SAQM, respectively.
These figures indicate that, by 1990, no-control scenario ozone concentrations in the modeled areas ,
would have been generally higher in both . ' • *
urban and rural areas. Rural area
concentrations differences are not as great as
urban area differences due to (a) the
differentially greater effect of CAA emission
controls in high population density areas, and
(b) potential differences in the models used for
urban and rural areas.
Ozone reductions in both rural and urban
areas projected in this analysis are not as
proportionally large'as the estimated
reductions in emissions of ozone precursors
for at least four reasons. First, current
knowledge of atmospheric photochemistry
suggests that ozone reductions resulting from
emissions changes will be proportionally
smaller than the emissions reductions.
Second, biogenic emissions of VOCs, an
important ozone precursor, are significant and
are held constant for the control and no-
control scenarios of this analysis. Biogenic
emissions are important because they
contribute roughly half of the total (manmade
plus natural) VOC emissions nationwide. Due
to this abundance of VOC loading and the
inherent nonlinearity of the ozone-precursor
Figure 45. Distribution of Estimated Ratios for 1990
Control to No-control SAQM-Simulated Daytime Average
Ozone Concentrations, by SAQM Monitor.
10
o
£
E
0.00 0.20 0.40 0.60 0.80 1£0 120
Ratb of C A A H o-CA A O 23ne-£ea3an D aytin e A verage O zone (iiteivalm fipo&t)
Figure 46. Distribution of Estimated Ratios for 1990
Control to No-control OZIPM4 Simulated 1-Hour Peak
Ozone Concentrations, by Urban Area.
30
20
I
= 10
0.00 0.20 0.40 0.60 0.80 1.00
Ratb ofc A A « o-CA A Peak O zone (&texvalnt xlpoiit}
190
-------
Appendix C: Air Quality Modeling
response system,102 historical reductions in anthropogenic VOC emissions can yield minimal reductions in
ozone, especially in rural environments. Third, this rural effect also influences urban areas receiving
substantial ozone transported in from surrounding areas. Consequently, the effect of emission controls
placed in urban areas often is reduced since much of the urban area ozone is imported. Thiis, the problem
is truly regionalized given the importance of transport, biogenic emissions and associated urban-rural
interactions; all contributing toward a relatively non-responsive atmospheric system.103 Finally, physical
process characterizations within OZIPM4 are severely limited and incapable of handing transport,
complex flow phenomena, and multi-day pollution events in a physically realistic manner. Consequently,
it is possible that the OZIPM4 method used herein produces negative bias tendencies in control
estimations. Additional discussion of uncertainties in the ozone air quality modeling is presented in the
following section. -
Key caveats and uncertainties for ozone
There are a number of uncertainties in the overall analytical results of the present study contributed by
the ozone air quality modeling in addition to the potential systematic downward bias discussed above.
First, there are substantial uncertainties inherent in any effort to model ozone formation and dispersion.
These uncertainties are compounded in the present study by the need to perform city-specific air quality
modeling using OZEPM4, which is less sophisticated than an Eulerian model such as the Urban Airshed
Model. However, while the absolute oz%|e predictions for any given urban area provided by OZEPM4
may be quite uncertain, the process o^aggregating results for altuimber of dties and meteorological
episodes should significantly reduce this uncertainty.104 Urban areas for which ozone changes may be
overpredicted are offset to some d|£ree by urb|| areas for which the change in ozone concentrations may
be underpredicted. mweighffigl^fe^ignificaiS^Sf this source of uncertainty, it is important to consider
the central purpose of the present study, which is to develop a reasonable estimate of the overall costs and
benefits of all historical Clean Air Act programs. All analyses are based on relative modeled results, and
ratios of the model predictions for the control and no-control scenarios, rather than the absolute
predictions. As a result of this, the effect of any bias in the model predictions is greatly .reduced due to
partial cancellation. < - __
X ' ' *
Additional uncertainty is contributed by other limitations of the models, the supporting data, and the
scope of the present analysis. Relying on linear interpolation between 1970 and modeled 1980 results to
derive results for 1975, an
-------
Appendix C: Air Quality Modeling
throughout the concentration distribution also contributes to uncertainty. Resource and model limitations
also required that night-time ozone concentrations be held constant between the scenarios. This leads to an
underestimation of the night-time component of ozone transport. Finally, changes in rural ozone in areas
not covered by RADM or SAQM could not be estimated. As a result, potentially significant changes in
ambient ozone in other major agricultural areas, such as in the mid-west, could not be developed for this
analysis. The Project Team considered using an emissions scaling (i.e., a roll-back) modeling strategy to
develop crude estimates of the potential change in rural ozone concentrations in monitored areas outside
the RADM and SAQM domains. However, the Project Team concluded that such estimates would be
unreliable due to the nonlinear effect on ozone of precursor emission changes. Furthermore, the Team
concluded that baseline levels of ozone and changes in precursor emissions in these areas are relatively
low. The decision not to spend scarce project resources on estimating ozone changes hi these rural areas is
further supported by the relatively modest change in rural ozone concentrations estimated within the
RADM and SAQM domains.
Visibility
Two separate modeling approaches were used to estimate changes in visibility degradation in the
eastern and southwestern U.S. These are the two regions of the coterminous U.S. for which Clean Air Act
programs were expected to have yielded the most significant reductions hi visibility degradation. Visibility
changes in the eastern 31 states were estimated based on the RADM/EM results for sulfates; and changes
in visibility in 30 southwestern U.S. urban areas were calculated using a linear emissions scaling approach.
Despite the potential significance of Clean Air Act-related visibility changes in southwestern U.S. Class I
areas, such as National Parks, resource limitations precluded implementation of the analysis planned for
these areas. .
'•»!!.:," . ">" ,• f '''$!•' '•>•;;'•.':: ' ;... '
The RADM/EM system includes a post-processor which computes various measures of visibility
degradation associated with changes in sulfate aerosols.105 The basic approach is to allocate the light
extinction budget for the eastern U.S. among various aerosols, including participate sulfates, nitrates, and
organics. The change in light extinction from sulfates is provided directly by RADM, thereby reflecting
the complex formation and transport mechanisms associated with this most significant contributor to light
extinction in the eastern U.S. Nitrates are not estimated directly by RADM. Instead, RADM-estimated
concentrations of nitric acid are used as a surrogate to provide the basis for estimating changes in the
particulate nitrate contribution to light extinction. The organic fractions were held constant between the
two scenarios. Standard Outputs include daylight distribution of light extinction, visual range, and
:DeciViews106 for each of RADMs 80-km grid cells. For the present study, the RADM visibility post-
processor was configured to provide the 90th percentile for light extinction and the 10th percentile for
visual range to represent worst cases; and the 50th percentile for both of these to represent average cases.
More detailed documentation of the RADM/EM system and the assumptions used to configure the
visibility calculations are presented in the Dennis, R. RADM Report (1995).
>M A complete discussion, including appropriate references to other documents, of the RADM and RADM/EM modeling conducted for the
present study is presented in the subsection on acid deposition earlier in this appendix.
*** The DecTView Haze Index (dV) is a relatively new visibility indicator aimed at measuring visibility changes in terms of human perception.
U h described ia detail in the SAISW Visibility Report (1994), pp. 4-2 to 4-3.
192
-------
Appendix C: Air Quality Modeling
To estimate differences in control and no-control scenario visibility in southwestern U.S. urban areas, a
modified linear rollback approach was developed and applied to 30 major urban areas with population
greater than 100,000.107 For each of the 30 urban centers, seasonal average 1990 air quality data was
compiled for key pollutants contributing to visibility degradation in southwestern U.S. coastal and inland
cities, including NO2 and PM10. PM10 was then speciated into its key components using city-specific
annual average PM10 profile data. After adjusting for regional —and for some species city-specific—
background levels, concentrations of individual light-attenuating species were scaled linearly based on
changes in emissions of that pollutant or a proxy pollutant.108 Using the same approach used for the 1993
EPA Report to .Congress on effects of the 1990 Clean Air Act Amendments on visibility in Class I areas,
light extinction coefficients for each
of these species were then multiplied
by their respective concentrations to
derive a city-specific light extinction
budget.109 This process was repeated
for pre-1990 control and all
no-control scenarios by scaling 1990
results by the relative change in
annual county-level emissions of SO,,
NO^andPM. Based on the
city-specific light extinction budget
calculations, measures for total
extinction, visual range, and
, DeciView were calculated for each
scenario and target year.
Figure 47. RADM-Predicted Visibility Degradation, Expressed
in Annual Average DeciView, for Poor Visibility Conditions
90th Percentile) Under the Control Scenario.
Control scenario
visibility
\
Unlike the other air quality
conditions addressed in the present
study, modeled visibility conditions
are used as the basis for the control
scenario rather than actual historical
conditions. However, like the other
air quality benefits of the historical
'Clean Air Act, it is the differences
between modeled visibility outcomes for the control and no-control scenarios which are used to estimate
visibility benefits. Nevertheless, 1990 absolute levels of eastern U.S. visibility predicted by RADM under
the control scenario are presented in Figure 47 to provide a sense of initial visibility conditions.
107 Complete documentation of the linear scaling modeling, speciation methodologies, spatial allocation of emissions, and other data and
assumptions are provided by the SAISW Visibility Report (1994).
108 For example, sulfate (SO4) concentrations were scaled based on changes in sulfur oxide (SOJ emissions.
109 The term "light extinction budget* refers to die apportionment of total light attenuation in ah area to the relevant pollutant species.
193
-------
Appendix C: Air Quality Mofleling
Mo/aelii
For the southwestern urban areas, 1990 control scenario annual average light extinction budget, visual
range, and DeciView conditions are listed in Table 80. These 1990 results are presented to give the reader
a sense of the initial visibility conditions in absolute, albeit approximate, terms.
No-control scenario visibility
^ t, «»
The no-control scenario visibility results for the eastern U.S. area covered by RADM are presented in
' Figure 48. No-control scenario 1990 outcomes for the 30 southwestern U.S. urban areas are presented in
Table 81. '
194
-------
Appendix C: Air Quality Modeling
''Table 8$." 1990C^o/Swnario Visibility Conditions for 30
' Soutnwestertt U:sl
s
s.
-
1
-
'
y \ v*< > ?'"
< t City ^
Los Arigefes, CA
San Bernardino, CA
Riverside, CA
Anaheim, CA :
jft«M?CA , *
San Diego, CA , ,
Santa Barbara, CA ,
Batorsfleld^CA' _ '
Fresno, CA
Modesto, CA
„ Stockloa/CA „ „
San Francisco, CA
Oakland, CA
San Jose, CA
Monterey, CA
1 Sacramento, CA
Redd%,CA "
Reno,NV ^ , '
'LasVegas;NV „ *'
Salt Lake qty,UT
ProvOjtJT
"Fort Collins, CO
GreeKy.CO,
Denver, CO
Colorado Springs, CO
Pueblo, CO
Albuquerque, MM ^
El Paso, TX
Tucson, AZ ,
Phoenix," AZ
^MWA
- . .»w
201.7
'• 2084 ; "
-170.17 s
1133 -
S ,126.9
112.8
.215.1
211.7 „ ,
: ' 1*8.8'
153.1
120.8,
' 1175
154.6
84.7
v /Ifitl
83.2
l4?.4
157.9
1175.
107.8
'80.7
" 84.2
153.4
,833
88.1
> 91J. ,
109.3
85.6
1253
, Visual Range
"Qoa)
15.2,
' 145 „,
S14.4
tf.6 :
265
23.6s
' 26.6
13.9
14.2*
,20.2- '
19.6
' 24.8" '
255-
19.4
35.4
252,
36.1
* 20.3
19.0 „
255
27.8
37.2
' 35.6
19.6
36.O
34.1
3i9
275
35.O
235
%r'
-29.8
,30.0 •
: 30.4 s
,283
243> -
2SA
24>, *
30-7
'305
* Z7JO
,273
245 ,
24.6' ,
27.4
,21.4
24.8 *
21^2
26.9
> 27.6
24.6
23.8
205 ,
213
273 ,
21.2
21.8, ,
22.1
235
215
253
<
••
'
Data Source: SAISW Visibility Report (1994).
195
-------
Appendix C: Air Quality Modeling
«%i^ *! 4 *' tn A > *•*« * j ; V-% "^ »*•„ >f?>i*<*-«
1990 No-controi Scenario Yisibility Conditions for 30 j? -
!
i
I
i
i
..
?'
1
i
""it
J3;
1
If
i
1
1
f
5
1
i
Southwestern U,S, Cities, ' ' 3""' , f j r *
cay
Los Angeles, CA
'San Bernardino, CA
Riverside, CA
Anaheim, CA
Ventura, CA
San Diego, CA
Santa Barbara, CA
BatosfiekLCA
Fresno, CA
Modesto, CA
Stockton, CA
SanFranciscOiCA
Oakland, CA
San Jose, CA
Monterey, CA
Sacramento, CA
Redding, CA
Reno,NV
Las Vegas, NV
:SaltLakeCily,UT
ProwvOT
Fort Coffins, CO
Greefey.CO
Denver, CO
Cofondo Springs, CO
Pueb!o,CO
Albuquerque, NM
El Paso, TX
Tucson, AZ
Phoenix, AZ
3§3.4
337,3 „,„ _
343.2
286^ ' *
194.8
210.1
183,2 / 1
356.4
349,0
240.1
"248.1
197.3 ,
188.6
253,0
141.4
189.2
128.6 '
416.6
643.8
185.8
159.0
191,2
117,0
284.4
175.8
299.9
175.8
2763
272.2
429.5,
t, VisuA '?
Range(km)
9.0
*»",
.&T' "
' 10L5 *"
, 153-', .
' 143'
. 16.4
8.4 ,5
' 8.6
-s \
1 in
, 35.4 ' t
33^1 f?
• '.riLft!*
? sf/?
.' >J» "J
,35,7 ' :
. 554 ""
SL8-. <
^2.| fc
? ^ f
29.4^" „ '
* &S%
26^, ;'
^ "V
294 „,
'253 <•
37J f
41,6 ~i
J29l2 ,1-
27.7 ,
29^ ''
24.e: :i!
33^
28.7 .'
34.0'
'28.7 ' }
33.2 ^ i
33V .*'
37.6
Dtla Source: SAT SW Visibility Report (1994).
"A"\-, !^
r^rt »• f^\\ '
196
-------
Appendix C:Air Quality Modeling
Figure 48. RADM-Predicted Visibility Degradation,
Expressed in Annual Average DeciView, for Poor Visibility
Conditions (90th Percentile) Under the No-control Scenario.
Summary differences
In visibility
Deciview Haze Index
The DeciView Haze Index (dV) has recently been proposed as an indicator of the clarity of the
atmosphere that is more closely related to human perception thanvvisual range (VR) or total extinction (bext)
(Pitchford and Malm, 1994). It is defined by the equation:
-(8)
where:
b^ = total extinction in inverse megameters (Mm"1)
This index has the value of approximately 0 when the extinction coefficient is equal to the scattering
coefficient for particle-free air (Rayleigh scattering) and increases in value by approximately one unit for
each 10 percent increase in bext. Since the apparent change in visibility is related to the percent change in
b,^ (Pitchford et aL, 1990), equal changes in dV correspond to approximately equally perceptible changes
197
-------
Appendix C:Air Quality Modeling
in visibility. Recent research indicates that, for most observers, a "just noticeable .change" in visibility
corresponds to an increase or decrease of about one to two dV units.
Both VR and dV are measures of the value of bex, at one location in the atmosphere. Both are
unaffected by the actual variability of the compositions and illumination of the atmosphere, so neither is
closely linked to the human perception of an actual scene. The isolation of these parameters from site-
specific variations and temporal fluctuations of the atmospheric illumination increases their usefulness for
comparing the effects of air quality on visibility across a range of geographic locations for a range of time
periods. Each parameter attempts to scale the b^ data so that changes in air quality can be used to provide
an indication of changes in the human perception of a scene.
Modeling Results
The differences in modeled 1990
control and no-control scenario visibility
conditions projected by the RADM/EM
for the eastern U.S. are presented in
Figure 49. The map shows the percent
increase in modeled annual average
visibility degradation under poor
conditions for 1990 when moving from
the control to the no-control scenario-
The results indicate perceptible
differences in visibility between the
control and norcontrol scenario .
throughout the RADM domain. The
relatively large increase in visibility
impairment in the Gulf Coast area is a
reflection of the significant increases in
1990 sulfate concentrations associated
with the no-control scenario. (For related
discussion of effects in this region, see
page 172.)
'igure49. RADM-Predicted Increase in Visibility
tegradation, Expressed in Annual Average DeciView, for
'oor Visibility Conditions (90th Percentile) Under the No-
control Scenario.
The differences in modeled 1990
control and no-control scenario visibility
conditions in the 30 southwestern U.S.
urban areas projected by linear rollback
modeling are presented in Table 82.
WhenTeviewing these visibility
degradation differentials for the 30
southwestern U.S. urban areas, it is important to consider that while estimated differences in visual range
were in many cases very large, changes in the DeciView Haze Index (dV) may be relatively small. This is
because the perception of visibility degradation measured by dV may be small when baseline visibility is
high.110 Even so, the results indicate that, by 1990, visibility in southwestern U.S. urban areas would be
noticeably worse under the no-control scenario.
"» See SAISW VisfciKly Report (1994), page 5-3.
198
-------
Appendix C: Air Quality Modeling
, Tifele 82. Summary of Relative Change ia Visual Range
, ana DeciVieWBeh»e6tt 1990 Control anaitecontrol
Scenario Visibility Conditions fat '30 Southwestern ILS.
f
\
J
/
t,
X
^J
4
f
^
£
if
'City
s ^ V *
Los Angeles, CA< \^ ^ > . ft
SanBerna^dlno^CA •,
KiverskfcvCA
Anaheim, CA ,/ , ",''
Venwi^CA,, /' _" ' -
San Diego, CA ,' _,
Santa Baibara, CA •,
,"Bak«sfield,CA
'Fwsm^'CA '
Modesto, CA
's»odrton,iCA. „ '
SanFiancisco, CA
OaHind,CA
San Jose, CA
Monterey, CA
Sacramento, CA
Redding, CA
Reno,NV- ~} ,, >
tasV<^as,NV
Sail lake City, UT ,
ftovo,UT'
Fort Collins^ CO
Gieeley.CO
Demwr»CO
Colorado Springs, CO
3?uebte,CO
Albuquerque>NM
El Paso, TX
Tucson, AZ
Phoenix; AZ
Vfaua^se
', 69
" 67 ,
-. 65 1 ,
68
,72
65
62
'' 66, -
,,65^
61
62
63 ,
61
64' '
67
59
55
1S3-
308
58
- 48
137-
39
„ 85
in
,240
53
153
218,
243
^DWw:
-5"
*4 ^l
<
" -5 ,
" ^'-5,
-5 *
'-5
-5
-5s '"
-5
-5
-5
-5
-5
-5
-5
-5
-4
-10
s
-14
-5
^
-9
-3
•«
-7
-12
-7
-9
-12
-12
Data Soiiice; SM SW YisibDity Report (1994).
199
-------
Appendix C: Air Quality Modeling
Key caveats and! uncertainties for visibility
• ....... n" • "" • "\
v ?
There ate several sources of uncertainty in the RADM and southwestern U.S. linear scaling model
analyses. For RADM, the use of nitric acid as a surrogate for estimating changes in light-attenuating
nitrate particles ignores the interaction effects of nitrates, sulfates, and ammonia. As a result, increases in
nitrates may be overestimated by the model when both sulfates and nitric acid increase. However, the
significance of this potential overestimation is mitigated to some extent by the relative insignificance of
nitrate-related visibility degradation relative to sulfates which prevails in the eastern U.S. _ _
"*" ™ „.">*.> " ~
Several important uncertainties in the southwestern U.S. urban area visibility analysis are described in
detail in the SAI SW Visibility Report (1994). First, the need to use seasonal average conditions leads to
underestimation of extreme visibility impairment episodes associated with high humidity, since particle
growth due to water absorption is highly nonlinear. Second, although" the use of city-specific light
extinction and PM speciation data is significantly better than reliance on regional averages, uncertainties in
city-specific data may contribute to overall uncertainty in the estimates. However, overall uncertainty
associated with these factors will be reduced to some extent since overestimation of visibility degradation
in some cities will be offset by underestimations in other cities. Finally, the linear scaling used to estimate
the pre-1990 control scenarios and the no-control scenarios was based on changes in county-wide or air
basin emissions. Uncertainties associated with apportionment of state-wide emission changes to individual
counties or air basins may contribute significantly to overall uncertainty in the visibility change estimates.
Such apportionment is particularly difficult for SOX emission changes, since emission reductions achieved
by the Clean Air Act tended to be at relatively remote utility and smelter plants. However, sulfates are a
relatively minor source of light attenuation in western urban areas.
An important overall limitation of the visibility analysis conducted for the present study is that only
southwestern urban areas and the eastern 31 states were included. The Clean Air Act may have ,
contributed toward significant reductions in visibility degradation in other areas. For example, Clean Air
Act programs to reduce ambient paniculate matter may have motivated reductions in silvicultural burning
in some northwestern states. Perhaps Hie greatest deficiency in geographic coverage by the present study is
the omissioa of visibility changes in Qass I areas in the west. ,
Air Quality Modeling References
Chang. "SARMAP Air Quality Model (SAQM)." Final report to San Joaquin Valley wide Air Pollution
Study Agency, 1995.
DaMassa, Tanrikulu, and Ranzier. "Photochemical Modeling of August 3-6, 1990, Ozone Episode in
Central California Using the SARMAP Air Quality Model. Part It Sensitivity and Diagnostic
Testing." Preprints, Ninth Joint Conference on the Applications of Air Pollution Meteorology with
Air Waste Management Association. January 28 - February 2, 1996, Atlanta, Georgia.
Dennis, US EPA, ORD/NERL. "Estimation of Regional Air Quality and Deposition Changes Under
Alternative 812 Emissions Scenarios Predicted by the Regional Acid Deposition Model, RADM."
Draft Report. October 1995.
200
-------
Appendix C: Air Quality Modeling
ICF Kaiser/Systems Applications International. "Retrospective Analysis of Ozone Air Quality in the
United States." Final Report. May 1995.
ICF Kaiser/Systems Applications International. "Retrospective Analysis of Particulate Matter Air Quality
in the United States." Draft Report. September 1992.
ICF Kaiser/Systems Applications International. "Retrospective Analysis of Particulate Matter Air Quality
in the United States." Final Report. April 1995. ~v ~ L.
ICF Kaiser/Systems Applications International. "Retrospective Analysis of SO^ NOx-anct~CO Air Quality
in the United States." Final Report. November 1994. ~~ ~
ICF Kaiser/Systems Applications International. "Retrospective Analysis of the Impact of the Clean Air
Act on Urban Visibility in the Southwestern United States." Final Report. October 1994.
ICF Kaiser/Science Applications International. Memo from J. Langstaff to JiDeMbcker. PM
Interpolation Methodology for the §812 Retrospective Analysis. March 199~6.
ICF Resources Incorporated. "Results of Retrospective Electric Utility Clean Air Act Analysis - 1980,
1985, and 1990." September 30,1992/
Seaman and Stauffer. "Development and Design Testing of the*SARMAP Meteorological Model." Final
report to San Joaquin Valley wide Air Pollution Study Agency. 1995.
Seaman, Stauffer, and Lario^Gibbs. "A Multi-Scale Four Dimensional Data Assimilation System Applied
in the^lf^guin Vsffley'DlSihg SARMAP, Part I: Modeling Design and Basic Performance
Chara^^K^pi^Journal of Applied Meteorology. Volume 34. In press. 1995.
''f^'$^%f'&*Z&'i£^^-&Vr., ' vi ' " •
Tanrikulu, DaM^^^^Ranzieri. "Photochemical Modeling of August 3-6,1990 Ozone Episode in
Central CJajd^Safflllag the SARMAP Air Quality Model. Part I: Model Formulation, Description
and Baae^Performance." "Preprints. Ninth Joint Conference on the Application of Air Pollution
Metrology wim Air-Waste Management Association. January 28 - February 2,1996. Atlanta,
.Georgia.
Trijdnis. "Visibility: Existing and Historical Conditions-Causes and Effects." NAPAP Report 24. 1990.
•* „
iJS EPA, Office of Air Quality Planning and Standards. "Procedures for Applying City-Specific EKMA."
, /EPA450/4-89-012. 1989.
201
-------
Appendix G: Lead Benefits Analysis
Since the concentration-response data are particular to specific sex and adult age groups,
additional population data were also required to determine the sizes of affected subpopulations. For 1990
age and sex, the U.S. Census, 1992 was used, with age groups tallied as necessary. For 1980 age and sex,
the U.S. Census, 1982 was used, with age groups also tallied as necessary. The 1970 age and sex
breakdowns were obtained through personal communication with the Census Bureau.250 The age and sex
percentages were interpolated for intervening years.
**>."'" ,sfe>
Pregnant women are often a subpopulation of interest for lead effects. Although pregnant women
themselves may be harmed by exposure to lead, this analysis was concerned with pregnant women because
of possible effects on their fetuses who will be born and evince effects as young children. To estimate the
number of exposed fetuses who were born during the years of interest,251 birth rates for 1970,1980 and
1990 were obtained from the Census Bureau.252 These birth rates were used to interpolate for years
between 1970 and 1980, and for the years between 1980 and 1990.
Results .
0
For both the controlled and uncontrolled scenarios, Table 106 shows estimated lead emissions
from industrial processes, industrial combustion and electric utilities. Table 107 and Table 108 show the
differences in health impacts between the two scenarios (for industrial processes, industrial combustion
and electric utilities only) for the "foiward-looldng" and "backward-looking'' analyses.
538 Kill Kuellmer, Abt Associates, and the Bureau of Census, Population, Age and Sex telephone staff, Match, 1994.
251 Note that we do not record the number of pregnancies, since the valuation only applies if the child is born and lives to exhibit the effect
Neither are we concerned with whether the births are single or multiple births, since each fetus is at risk, whether a pregnant women carries one or
more fetuses. *
30 Personal communication, Karl Kuellmer, Abt Associates and the Bureau of Census, Population, Fertility/Births telephone staff.
336
-------
Appendix G: Lead Benefits Analysis
'
~.Tab|e 106,"' Es&naied Lead Emissions firpni Electric Utilities, Industrial Processes,4nd industrial OtMobustlon; „
"* * v i " i /- <~ *• ^ " * ^ " 'x <•<•.!• *. ^ ^ j ^ '**•„„ % *
4*
•O
f-
'
CootroUtdScourio > • ,
* - '- * ' * C" s -'
UwMtrolkdSceaario ,
>'t J ' * "=" ' - \ -
UicmtroHed Scenario
rt, v „ A ^
Controlled Scenario " • -
IMattriftlCoadttoliM
llBcootrolkdSceiario ' , •>'
, , . * 1970
-" " <> " '•
* ' . /"
' 7:m
w
4S»:
•^ s
1975
; ^
\ 330?
-3^17
7,124
4,354
4,457
1980
', **
' *«
'- 1,032
6fSQ
1,880
4,653
aws'
* ^
_ -^
'fee
^^
'.«•
.«"-
1 ^1990
' >*
f /*
658
^ ' s^os
. , iS7
4^96
>
C- -W ^ „ >, A
edata oo-jelectric utilities do not exist for years prior to 197$.
y^ ;•«• " ^y--*<• J
337
-------
p.
1
1
§
o ^
II
S'g
II
J8H
£<
*1
"I
.
g
00
"
T-
«
"J
d
"5
es
OO
-------
SSS
$
o
o\
•g
<=>
••••*
00
to
%
I
I
•II
-------
Appendix G: Lead Benefits Analysis
Reduction in Health Effects Attributable to
Gasoline Lead Reductions
Estimating Changes In Amount, of Lead In Gasoline
from 1&7O to 1990 f
The relationship between the national mean blood lead level and lead in gasoline is calculated as a
function of the amount of lead in gasoline consumed. Thus, to calculate the health benefits from gasoline
lead reductions, necessary inputs are estimates of lead in gasoline consumed over the period 1970 to 1990
and the amount of lead in gasoline that would have been consumed in the absence of the Clean Air Act.
These values are calculated using the quantity of both leaded and unleaded gasoline sold each year and the
concentration of lead in leaded and unleaded gasoline for each year in the period of interest. For each year,
the relationship is expressed as:
- i , s" *, *
\
where:
LEAD = average lead per day in gasoline sold in a given year (metric tons/day),
SOLD = total quantity of gasoline sold (million gal/yr),
FK&Cff-s fraction of total gasoline sales represented by leaded gasoline (dimensionless),
Pbtaitj "= ' lead content of leaded gasoline (g/gal), and
Pbm,mfff= lead content of unleaded gasoline (g/gal).
Gasoline Sales (SOLD); Data on annual gasoline sales were taken from a report by Argonne
National Laboratories (1993) which presented gasoline sales for each state in five year intervals over the
period 1970-1990. This analysis used linear interpolation to estimate the gasoline sales for years between
the reported years. These data were summed to obtain national sales figures.
Fraction of Sales that were Leaded and Gasoline (FRACp,,): For the controlled scenario, this
analysis used information reported by Kolb and Longo (1991) for the fraction of the gasoline sales
represented by leaded gasoline for the years 1970 through 1988. For 1989 and 1990, data were taken from
DOE (1990 and 1991, respectively). For the uncontrolled scenario, all of the gasoline sold was assumed to
be leaded for all years. •
Lead Content of Gasoline (Pblcldri and Pb^^^: Argonne National Laboratory in Argonne,
Illinois was the source for the data on the lead content of leaded and unleaded gasoline for the period
1974-1990. Argonne compiled these data from historical sales data submitted to EPA, from Clean Air Act
regulations on lead content, and from recent MVMA surveys. For 1970 through 1973, this analysis
assumed the lead content of gasoline to be at the 1974 level. For the uncontrolled scenario, this analysis
used the 1974 lead content in leaded gasoline as the lead content in all gasoline for each year.
340
-------
Appendix G: Lead Benefits Analysis
Estimating the Change in Blood Lead Levels from the
in the Amount of Lead in Gasoline
Several studies have found positive correlations between gasoline lead content and blood lead
levels.253 Data from die National Health and Nutrition Examination Survey (NHANES H) has been used
by other researchers, who determined similar positive correlations between gasoline lead and blood lead
levels.254 ' _~
The current analysis used a direct relationship between consumption of lead in gasoline and blood
lead levels to estimate changes in blood lead levels resulting from Clean Air Act regulation of the lead
content of gasoline. This relationship was based on regression analyses of the reduction of leaded gasoline
presented in the 1985 Regulatory Impact Analysis (RIA).255 Several multiple regressions were performed
in the RIA to relate gasoline usage with individuals' blood lead levels, which were taken from NHANES n.
These regressions of blood lead on gasoline usage controlled for such variables as age, sex, degree of
urbanization, alcohol consumption, smoking, occupational exposure, dietary factors, region of the country,
educational attainment, and income. The regressions suggested that a decrease of 100 metric tons per day
(MTD) of lead used in gasoline is associated with a decrease in mean blood lead concentration of 2.14
/«g/dL for whites and 2.04 /ig/dL for blacks. In both of these regressions, gasoline use was found to be a
highly significant predictor of blood lead (p < 0.0001).256
To determine a single gasoline usage-blood lead slope for the entire population of the U.S., this
analysis used the average of the slopes for blacks and for whites, weighted by the percentage of blacks and
whites in the U.S. during the time period of the'analysis.257 The resulting relationship is 2.13 fig/dl blood
lead per 100 metric tons of lead in gasoline consumed per day. The same relationship was used to model
changes in both children's and adults' blood lead levels. The U.S. EPA (1985) analyzed data from a study
of black children in Chicago 'during the time period 1976 to 1980 and determined a slope of 2.08 /ug/dl per
100 MTD. This slope for children is very similar to the one used in this analysis.
1970-Forward arid i990-Backward Approaches
.As with the industrial processes and boilers analysis, this analysis used two different approaches to
determine mean blood lead levels based on changes in lead concentrations in gasoline. In the 1970- (
forward approach, the calculations began with the estimated blood lead level for 1970. The change in
blood lead level from one year to the next was based upon the change in the amount of lead in gasoline
sold, as discussed above, for both the controlled and uncontrolled scenarios. For example, to calculate the
blood lead level for^fffi, the calculated change in blood lead from 1970 to 1971 was added to the 1970
value. This process was repeated for each succeeding year up to 1990.
_ - . . • • -.••'.
253 U.S. EPA, 1985; Bfflick et al., 1979; Bfflicketal., 1982. , '.
^ Janney, 1982; Annestetal., 1983; Center for Disease Control, 1993; National Center for Health Statistics, 1993.
255 U,S. EPA, 1985.
, ** U,S. EPA, 1985.
257 U.S. Department of Commerce, 1992. Although the percentages of blacks and whites changed slightly over this time period (1970-1990),
the change did not affect the value of the weighted slope.
341
-------
Appendix G: Lead Benefits Analysis
The 1990-backward approach began with a mean blood lead level in 1990 for the controlled
scenario. For the uncontrolled scenario, the starting blood lead was estimated from the 1990 level used in
the controlled scenario, plus an additional blood lead increment resulting from the difference between the
1990 consumption of lead in gasoline in the controlled and uncontrolled scenarios. Again, the difference
in mean blood lead levels from one year to the next was based on the change in gasoline lead for the
corresponding years. For example, the difference in blood lead levels between 1990 and 1989 was
subtracted from the 1990 level to determine the 1989 level. The process was continued for each year back
to 1970. „ 3P
Relating Blood Lead Levels to Population Health
Effects .
The mean blood lead levels calculated using the methods described above were used in the dose-
response functions for various health effects (e.g., hypertension, chronic heart disease, mortality). This
information was then combined with data on the resident population of the 48 conterminous states in each
year to determine the total incidence of these health effects attributable to lead in gasoline in the controlled
and uncontrolled scenarios. A Department of Commerce Publication (1991) was used to obtain the total
population in 1970,1980, and 1983-1990, while a different publication was the source of the 1975
population values.258 Linear interpolation was used to estimate the populations in years for which specific
data were not available.
For certain health effects, it was necessary to know the size of various age groups within the
population. Two different sources were used to estimate the'proportions of the population in the age
groups of interest. A U.S. Census summary (Dept. of Commerce, 1990) was used for information for 1990
for children and adults and for 1980 for adults, and Census Telephone Staff (Dept. of Commerce, 1994)
provided information for 19180 for children and 1970 for children and adults. The populations for the
intervening years were estimated by linear interpolation.
Dept. of Commerce, 1976.
342
-------
Appendix G: Lead Benefits Analysis
Changes in Leaded Gasoline Emissions and Resulting Decreased Blood Lead Levels and Health
Effects
Table 109 shows the estimated quantity of lead burned in gasoline in the five year intervals from
1970 to 1990. Tables 110 and 111 show the difference in health impacts between the two scenarios (for
lead in gasoline only) for the "forward-looking" and "backward-looking9' analyses..
,
Table 109. Lead Burned in Gasoline (in tons).
< * ""* ' * " v *
•>
•*• > > \. 4*
Controlled Scenario
Uncontrolled Scenario ',
J. 1970
' 176,100^
176,100 '
1975
179,;2aV
.202,600
&S& ,
86,400
206,900
1985
22,000
214,400
/»V
, ^2,300 '
,222,900'
343
-------
t
o
S
O
".S
1
8
J3
1
I
p
5?i
w
s:
Sf-
1
•I
s*i
«
I
I
vo
'8
-------
i
*'
Hi
JS/
f
'
•2
4?
•& 55
I
:f
o
ov
VI
*r>
81
I
I
•tff
I
8
8
0\
00
-------
Appendix G: Lead Benefits Analysis
Lead Benefits Analysis References
Abt , 1992. "The Medical Costs of Five Illnesses Related to Exposure to Pollutants." Prepared for:
Nicholas Bouwes, Regulatory Impacts Branch, Economics and Technology Division, Office of
Pollution Prevention and Toxics, U.S. Environmental Projection Agency, Washington, D.C.
^. •* %*
Abt, 1995. The Impact of the Clean Air Act on Lead Pollution: Emissions Reductions, Health Effects, and
Economic Benefits From 1970 to 1990. Draft Januar^l9, 1995. Prepared for Economic Analysis
and Innovations Division, Office of Policy Planning and Evaluation, US EPAl
Annest, J.L., J.L. Pirkle, D. Makuc, J.W. Neese, D.D. Bayse, and M.G. Kovar. 1983. Chronological trend
in blood lead levels between 1976 and 1980. New England Journal of Medicine. 308: 1373-
1377.
Azar, R.D., et al. (1975). An epidemiologic approach to community air lead exposure using personal air
samplers. In: Griffin, T.B. and Knelson, J.H., eds. Lead. Stuttgart, West Germany: Georg
Thieme Publishers; pp.254-290. (Coulston, F. and Korte, F., eds. Environmental quality and
safety: supplement v. 2).
Bellinger, D. Sloman, J., Leviton, A., Rabinowitz, M., Needleman, H.L., and Waternaux, C. 1991. Low-
level lead exposure and children's cognitive function in the preschool years. Pediatrics. Vol 87.
No. 2:219-227.
- •*•
K"t ,
Bellinger, D.C, 1992, Lead Exposure, Intelligence and Academic Achievement. Pediatrics. Vol 90. No
Bttlick, I.H., A.S. Curran, and D^R. Shier. 1979. Analysis of pediatric blood lead levels in New York City
for 1970-1976. Environmental Health Perspectives. 31:183-190.
J[ i-1" " "
L R ,
Billick, I.H., et al. 1982. Predictions of pediatric blood lead levels from gasoline consumption. U.S.
Department of Housing and Urban Development. [Cited in U.S. EPA, 1985.]
...... i • . '
Brunekreef, B.D., et al. (1981). The Amhem lead study: 1. lead uptake by 1- to 3-year-old children living
', ":""' in the vicinity of a secondary lead smelter in Arnhem, the Netherlands. Environ. Res. 25: 441-
' ' "
Center for Disease Control (CDC). 1993. Personal communication between Abt Associates and Jim
• - Pirkle. November 16.
Centers for Disease Control (CDC). 1985. Preventing Lead Poisoning in Young Children. U.S.
Department of Health and Human Services, Public Health Service, Centers for Disease Control,
Atlanta, GA.
346
-------
Appendix G: Lead Benefits Analysis
Centers for Disease Control (CDC). 1991a. Strategic Plan for Elimination of Childhood Lead Poisoning.
U.S. Department of Health and Human Services, Public Health Service, Centers for Disease
Control. February. ,
v
Centers for Disease Control (CDC). 1991b. Preventing Lead Poisoning in Young Children. U.S.
Department of Health and Human Services, Public Health Service, Centers for Disease Control,
Atlanta, GA. October.
^N ' ' "*' , "*
Chamberlain, A.C. (1983). Effect of airborne lead on blood lead. Atmos. Environ. 17:677-692.
Daines, R.H., et al. (1972). Air levels of lead inside and outside of homes. Ind. Med. Surg. 41:26-28.
Dietrich, K.N.* Krafft, K.M., Shukla, R., Bornschein, R.L., Succop, RA. 1987. The Neurobehavioral
Effects of Prenatal and Early Postnatal Lead Exposure. In: Toxic Substances and Mental,
Retardation: Neurobehavioral Toxicology and Teratology, S.R. Schroeder, Ed. American
Association of Mental Deficiency, Washington DC, pp. 71-95 (Monograph No. 8).
•> .f
DOE, 1990. Petroleum Supply Annual, 1989, Volume 1. -DOE publication number EIA-0340(89)/1
DOE, 1991. Petroleum Supply Annual, 1990, Volume 1. DOE publication number EIA-0340(90)/1
! ';' ''.<._ -
Elixhauser, A., R. M. Andrews, and S. Fox. 1993. "Clinical Classifications for Health Policy Research:
Discharge Statistics by Principal Diagnosis and Procedure." Agency for Health Care Policy and
Research, Center for General Health Services Intramural Research, U.S. Department of Health and
Human Services.
__ «- *va8»J • i
— j~ w ^ ^ _t ;%«j|££-;
Environmental Law Institute CELT) 1992, Projecting With and Without Clean Air Act Emissions for the
Section 812 Retrospective Analysis: A Methodology Based upon the Projection System Used hi
the OAOPS "National Air Pollutant Emission Estimates: Reports. [Jorgenson/Wilcoxen Model
Projections], Jim Lockhart.
Fisher, A:, L.G. Chestnut and D.M. Violette (1989). The value of reducing risks of death: a note on new
jjjl evidence. J. of Policy Analysis and Mgmt-, 8(1):88-100.
M., et al. (1973). Concentration levels and particle size distribution of lead in the air of an urban
and an industrial area as a basis for the calculation of population exposure. In: Earth, D., et al.
eds. Environmental health aspects of lead:, proceedings, international symposium; October 1972;
:erdam, The Netherlands. Luxembourg; Commission of the European Communities, pp.
Goldsmith, J.R. (1974). Food chain and health implications of airborne lead. Sacramento, CA: State of
California, Air Resources Board; report no. ARB-R-102-74-36. Available from NTIS,
Springfield, VA PB-248745.
Industrial Economics, Inc. (ffic) (1992). Memorandum to Jim DeMocker, Office of Policy Analysis and
Review, from Robert E. Unsworth and James E. Neumann. "Review of Existing Value of Life
Estimates: Valuation Document." 6 November, 1992.
- . " 347
-------
Appendix G: Lead Benefits Analysis
Industrial Economics, Inc. (DEc) (1993). Memorandum to Jim DeMocker, Office of Policy Analysis and
Review, from Robert E. Unsworth and James E. Neumann. "Revisions to the Proposed Value of
Life: Methodology for the Section 812 Retrospective." 3 May, 1993.
JS:-.
,m>'' .
Janney, A. The relationship between gasoline lead emissions and blood poisoning in Americans. Prepared
for U.S. EPA, Office of Policy Analysis. [Cited in U.S. EPA, 1985.]
Johnson, D.E., et aL (1976). Base line levels of platinum and palladium in human tissue. Research
Triangle Park, N.C.: U.S. EPA, Health Effects Research Laboratory; EPA reportno. EPA-600/1-
76-019. available from: NTIS, Springfield, Va; PB-251885.
Kolb, J. and K. Longo. 1991. Memorandum to Joel Schwartz, U.S. EPA, Washington, D'C, November 5.
Landrigan, PJ. and EX. Baker (1981). Exposure of children to heavy metals from smelter: epidemiology
and toxic consequences. Environ. Res. 25: 204-2241!
Landrigan, P J., et aL (1975). Epidemic lead absorption near aa ore smelter: the role of particulate lead.
N.Engl.J.Med.292i 123-129.
McGee and Gordon. 1976. The Results of the Framingham Study Applied to Four Other U.S.-based
Epidemiologic Studies of Coronary Heart Disease. The Framingham Study: An Epidemiological
Investigation of Cardiovascular Disease. Section 31, April.
National Center for Health Statistics (NCHS). 1993. Facsimile received by Abt Associates from Margaret
McDowell regarding the types of laboratory tests conducted during NHANES I. December 14.
National Center for Health Statistics (NCHS). 1993. Personal communication between Abt Associates
arid NCHS Public Information Specialist. November 3.
National Energy Accounts, Bureau of Economic Analysis.
NHANES IT, National Healthand(Nutrition Examination Survey, 1976-1980.
NHANES, National Health and Nutrition Examination Survey.
in:, .in,;, y'"•' ' , '
Nordman, C.H. (1975). Environmental lead exposure in finland: a study on selected population groups
[dissertation]. Helsinki, Finland: University of Helsinki.
Oliver, T. l91i. Lead Poisoning and the Race. British Medical Journal 1(2628): 1096-1098. [Cited in
USEPA(1990).]
Piomelli et al. 1984. Management of childhood lead poisoning. Pediatrics; 4:105.
Pirkle, J.L., J. Schwartz, J.R. Landis, and W.R. Harlan. 1985. The relationship between blood lead levels
and blood pressure and its cardiovascular risk implications. American Journal of Epidemiology.
121:246-258. .
348
-------
Appendix G: Lead Benefits Analysis
Pirkle, J. L., et al. 1994. "Decline in Blood Lead Levels in the United States, the National Health and
Nutrition Examination Survey (NHANES). JAMA, July 27,1994, v272 n4, p284.
Pooling Project Research Group. 1978. Relationship of blood pressure, serum cholesterol, smoking habit,
relative weight and ECG abnormalities to incidence of major coronary events: final report of the
Pooling Project. Journal of Chronic Disease. Vol. 31.
Rabinowitz, M., Bellinger, D., Leyiton, A., Needleman, H., and Schoenbaum, S. 1987. Pregnancy
hypertension, blood pressure during labor, and blood lead levels. Hypertension. Vol 10., No. 4,
October. ,
Schwartz, J. 1988. The relationship between blood lead and blood pressure in the NHANES II Survey.
Environmental Health Perspectives. Vol. 78:15-22.
Schwartz, J. 1990. Lead, blood pressure, and cardiovascular disease in men and women. Environmental
Health Perspectives, in press.
Schwartz,!. 1992a. Blood lead and blood pressure: a meta-analysis. Presented at the Annual Meeting of
Collegium Ramazzini. November.
Schwartz,!. 1992b. Chapter 13: Lead, Blood Pressure and Cardiovascular Disease. In: Human Lead
Exposure, H. L. Needleman, Ed. CRC Press.
Schwartz,!. 1993. Beyond LOEL's, p values, and vote counting: methods for looking at the shapes and
strengths of associations. Neurotoxicology.VoL 14. No. 2/3. October.
=sr' * * ",. " *'*;
Shurtleff, D. 1974. Some Characteristics Related to the Incidence of Cardiovascular Disease and Death.
• TheFrdmingham Study: An Epidemiological Investigation of Cardiovascular Disease. Section 30,
February.-
Silbergeld, E.K., Schwartz,!., and K. Mahaffey. 1988. Lead and osteoporosis: mobilization of lead from
bone in postmenopausal women. Environmental Research. 47,: 79-94.
jf > • '
Snee| R.D. (1981). Evaluation of studies of the relationship between blood lead and air lead. Int. Arch.
Occup. Environ. Health 48: 219-242.
:, L.B. and L.S. Levin (1975). A survey of air and population lead levels in selected American
communities. In: Griffin, T.B.; Knelson, !.H., eds. Lead. Stuttgart, West Germany: Georg
jeme Publishers; pp. 152-196. (Coulston, F.; Korte, f., eds. Environmental quality and safety:
supplement v. 2).
Tsuchiya, K., et al. (1975). Study of lead concentrations in atmosphere and population in Japan. In:
Griffin, T.B. and Knelson, !.H.. eds. Lead. Stuttgart, West Germany: Georg Thieme Publishers;
pp.95-145. (Coulston, F.; Korte, F., eds/ Environmental quality and safety: supplement v. 2)
U.S. Census (1982) United States Summary, General Population Characteristics, Table 41: Single Years
of Age by Race, Spanish Origin, and Sex: 1980.
' " ' ' ' ' ' ' 349 "• .'..'•'.'
-------
Appendix G: Lead Benefits Analysis
U.S. Census (1992) United States Summary, General Population Characteristics, Table 13: Single Years
by Sex, Race, and Hispanic Origin: 1990.
U.S. Department of Commerce. 1990. U.S. Census, United States Summary, General Population
Characteristics.
U.S. Department of Commerce. 1980. U.S. Census, United States Summary, General Population 4-
Characteristics.
U.S. Department of Commerce. 1993. Personal Communication between Bureau of Census, Population,
Age and Sex Telephone Staff and Karl Kuellmer of Abt Associates on December 8,1993
U.S. Department of Commerce. 1994. Personal Communication between Bureau of Census, Population,
Age and Sex: Telephone Staff, and Karl Kuellmer of Abt Associates on February 7,1994.
U.S. Department of Commerce. 1987 Census of Manufacturers.
U.S. Department of Commerce. 1992. Statistical* AbstracVofthe United States: 112th Edition. Bureau of
the Census. Washington, DC.
U.S. Department of Commerce. 1991. Statistical Abstract of the United States: lllth Edition. Bureau of
the Census. Washington, DC.
U.S. Department of Commerce. 1976. Statistical Abstract of the United States: 95th Edition. Bureau of
the Census. Washington, DC. -
U.S. DOE, 1992, Cost and Quality of Fuels for Electric Utility Plants 1991. DOE/EIA-0191(91) Energy
Information Administration, August 1992.
U.S. Environmental Protection Agency. 1985. Costs and Benefits of Reducing Lead in Gasoline: Final
Regulato.ry Impact Analysis. Prepared by U.S. Environmental Protection Agency, Office of Policy
Analysis, Economic Analysis Division. February.
nglll, '' "*',""' '_i!''
U.Sl Environmental Protection Agency. 1986a. Reducing Lead in Drinking Water: A Benefit Analysis.
; Prepared by U.S. Environmental Protection Agency, Office of Policy Planning and Evaluation,
Draft Final Report. December.
U.S. Environmental Protection Agency. 1986b. Air Quality Criteria for Lead: Volume HI.
Environmental Criteria and Assessment Office, Research Triangle Park, NC. EPA-600/8-
"83/028cF. June.
U.S. Environmental Protection Agency. 1987. Methodology for Valuing Health Risks of Ambient Lead
Exposure. Prepared by Mathtech, Inc. for U.S. Environmental Protection Agency, Office of Air
Quality Planning and Standards, Ambient Standards Branch, Contract No. 68-02-4323.
350
-------
Appendix G: Lead Benefits Analysis
\
U.S. Environmental Protection Agency. 1990. Review of the National Ambient Air Quality Standards for
Lead: Assessment of Scientific and Technical Information. OAQPS Staff Paper, Air Quality
Management Division, Research Triangle Park, N.C. December.
U.S. Environmental Protection Agency. 1994. Guidance Manual for the Integrated Exposure Uptake
Biokinetic Model for Lead in Children, February 1994, Publication Number 9285.7-14-1, EPA
540-R-93-081.
U.S. Environmental Protection Agency (1994). Cost and Benefits of the Smoke Free Environment Act,of
1993 [HR 3434]. Prepared for Congressman HenryWaxman bythe fodooriyfDivisiion, Office
of Air and Radiation. ~_:>"
U.S. Environmental Protection Agency (U.S. EPA). 1986. Air Quality Criteria for Lead: Volume III.
Environmental Criteria and Assessment Office, Research TriangleTark, NC. EPA-600/8-
83/028cF. June.
U.S. EPA (1990). Office of AirQuality Planning and Standards, Airs Facility Subsystem Source
Classification Codes and Emission Factor Listing for Criteria Air Pollutants, Publication Number
EPA-450/4-90-003, US EPA, Research Triangle Park, March 1990.
U.S. EPA (1990). National Air Pollutant Emission Estimates 1940-3.988. U.S. EPA. Office of Air
Quality Planning and Standards, Technical Support Division, National Air Data Branch. Research
Triangle Park. EPA no. EPA-450/4-90-001.
U.S. EPA (1991). National AirQuality and Emissions Trends Report, 1989. U.S. EPA. Office of Air
Quality Planning and Standards. Research Triangle Park. EPA no. EPA-450/4-91-003.
^ ^^ „ . '"• - - * -
U.S. EPA (f991b> The Interim Emissions Inventory. U.S. EPA. Office of Air Quality Planning and
Standards, Technical Support Division, Source Receptor Analysis Branch. Research Triangle
Park.. .
U.S. EPA-(f992). 1990 Toxics Release Inventory. U.S. EPA. Office of Pollution Prevention and Toxics.
'Washington, D.C. 20460. EPA no. EPA-700-S-92-002.
A database. Graphical Exposure Modeling System Database (GEMS).
W.K.(1992). Fatal Tradeoffs: Public and Private Responsibilities for Risk New York: Oxford
rniversity Press.
1 - • • •
tfield. 1986. Assessing the Risks to Young Children of Three Effects Associated with
Elevated Blood Lead Levels. Argonne National Laboratory. December.
351
-------
352
-------
Appendix Hg Air Toxics
Introduction
Air toxics are defined as air pollutants other than those six criterifpollutantsl6^p|fflQ|M'sets
acceptable concentrations in ambient air. The SARA 313 Toxic Release Inventory (TRI), covering 328 of
the approximately 3000 potentially hazardous compounds detected mjir, estimated that approximately 1.2
millions tons of air toxics were released to the atmosphere in 1987 from TJ.S. stationary sources alone.
While the TRI estimate tends to understate emissions of tox^f or a ntimt||r of reasons, it does show that
large quantities of toxics are emitted into the atmosphere annually.
Effects of air toxic emissions are divided it
"noncancer" effects, e.g. a wide variety of s
defects, neurological impairment, or.i
these air toxic emissions contribute to significant adverse effe
ecosystems. In EPA's 1987 Unfinishe&SJJsiness ReMri*9 cane
ories for study and assessment: cancer;
as abnormal development, birth
Iogical effects. Each year,
health, human welfare, and
and noncancer air toxic risk estimates
were considered sufficiently high, relafffilfo risks addressed by other EPA programs, that the air toxics
program area was among the few ra^Fhigh risk|.
Urn
this Assessment
Thee:
often irreversib]
painful and/or
clinicaljfudies,
It to quantify. Adverse health effects of toxics are
or eliminated by reduction in ongoing exposure, and involve particularly
Therefore these effects are not readily studied and quantified in human
|br example, ambient ozone. In addition, epidemiological studies of
onsjfM often confounded by simultaneous exposure of subjects to a variety of
pollutants. Therefore, effjps of these pollutants are often quantified by extrapolating data from animal
studies to human exposurf"and expressed as risk per unit of exposure. Incidence of noncancer effects, for
example, often are difficult to translate into monetized benefits.
- , SiinUarlyj the quantification of ecological effects due to emissions of air toxics is hampered by
lack of sufficient information regarding contribution of sources to exposure, associations between exposure
to mixtures of toxics and various ecological endpoints, and economic valuation for ecological endpoints.
239 U.S. EPA. Office of Policy Planning and Evaluation. Unfinished Business: A Comparative Assessment of Environmental Problems.
February 1987. > ,
353
-------
Appendix H: Air Toxics
•TaWe 112. Health and \Velfare Effects oJt Hazardous Air Pollutants. . / *"'',"**•
III IIP hi 1 illlk I ' I11 III1" lilll III "I1 ill A ] fll J if i 1 l1t | "^l I*5"1 *if * i % U! Jj*"1 ™w fj v * *" **P /"*" * * Y ' * * ^ *siB> J
I iiif Jl 1 L 1 In!' '|l 1 ' imjr ll Jl H bill' 1 l| ilbl ' l K> \i '"Hf ^'4^4* * ' * l^V ''"•'*"" ^1 * "* * ! ^^ ^ i 5 x*" s * ^rf" >*V ! **& j. ,,
Effect Category
Human Health
Human Welfare
Ecological
111 Ti 1 (ll li HI II Inil in
O Urn- Welfare
Quantified Effects
Cancer Mortality
-nonutflity
stationary source
- mobile source
Unquantified Effects
Cancer Mortality
- utility source
-area source
Noncancer effects
-neurological
-respiratory
- reproductive ,
- hematopoetic "
-developmental
- immunological :
- organ toxicity
Decreased income and
recreation
opportunities due to
fish advisories
Odors
Effects on wildlife
Effects on plants
Ecosystem effects '
Loss of biological
diversity 1(
Visibility *
Materials Damage
Other Possible Effects .
" <•'
*.* » i ^ ^ >
Decreased; income resulting
' from decreased '„ ,
physical performance
4 ' ^- t, ^
ft s , l t"C :
, f ill '"til' III
"I i I' mi in i ill" Mini inn i1
,'-' 1 -
The air toxics portion of this study is, of necessity, separate and more qualitative in nature than the
benefit analysis conducted for the criteria air pollutants. Limitations in the quantitative analyses of air
toxic effects led the project Team to decide to exclude the available quantitative results from the primary
analysis of CAA costs and benefits. Table 112 presents the range of potential human health and ecological
effects that can occur due to air toxics exposure. As indicated, this appendix presents alternative
quantitative estimates of benefits of CAA air toxics control for the cancer mortality endpoint for only
nonutility stationary source and mobile source categories. Noncancer effects and ecological effects are
described qualitatively.
354
-------
Appendix H: Air Toxics
History of Air Toxics Standards under th
Clean Air Act of
The 1970 Clean Air Act required the EPA to list a chemical as a hazardous air pollutant if it met
the legislative definition provided:
' • . ' • " - -*" f
"The term 'hazardous air pollutant1 means an air pollutant to which no ambient air
quality standard is applicable and which in the judgment of the Administrator may cause,
or contribute to, an increase in mortality or an increase in serious irreversible, or \
incapacitating reversible, illness. "2eo
/- i"
Once a HAP was listed, the EPA was required to:
"establish any such standard at the level which m his judgment provides an ample margin
of safety to protect the public health ffomsutt hazardous air pollutant."261
In other words the EPA had to first determine that a chemical was a HAP, and then regulate the emissions
of each HAP based solely on human health effects and with an ample margin of safety. This regulatory
mandate proved extremely difficult for EPA to fulfill, for reasons discussed below, and the result was that
only seven HAPs were regulated over a period of 20 years.
Listing chemicals became a difficult task because of debates within and outside of the EPA
surrounding issues of how much data are needed and which methodologies should be used to list a
chemical as a HAP. An even more difficult issue was how to define the Congressional mandate to provide
an "ample margin of safety^For carcinogens, there is generally no threshold of exposure considered to be
without risk. What level of risk, then, is acceptable, and how should it be calculated? The EPA struggled
to provide answers to these questions, aSr>vas challenged in court. The end result was a 1987 ruling by
the D.C. Circuit Court that provided the EP^A with a legal framework with which to determine an "ample
margin of safety."" This framework was interpreted and used by the EPA in its 1989 benzene regulations.
Quantifiable Stationary Source Air TOMMGS
" " _ One might be tempted to presume that the few federal HAP standards set would have achieved
relatively substantial reductions in quantifiable risk. While some standards set under section 112 of the
Clean Air Act appear to have achieved significant reductions in cancer incidence, the coverage,
quantification, and monetization of the full range of potential adverse effects remains severely limited.
This fact serves to highlight the inadequacy of current methods of evaluating HAP control benefits. This
limited ability to estimate the total human health and ecological benefits of HAP reductions is an important
42U.S.C.§1857(aXl).
42U.S.C§1857(b).
355
-------
Appendix H: AirToxics
area for future research. Thus the quantifiable benefits for CAA air toxics control presented here are
limited in scope.
There are three sources of information that provide a picture of potential stationar^source air
toxics benefits of the CAA. EPA's Cancer Risk studies attempted to broadly assess the magnitude and
nature of the air toxics problem by developing quantitative estimates of cancer risks posed by selected air
toxics and their sources. Secondly, risk assessments conducted in conjunction with the promulgation of
National Emissions Standards for Hazardous Air Pollutants (NjSHA^Ps) offer a snapshot of potential
monetized cancer mortality benefits. Finally, the Project Team attempted to estimate historical non-utility
stationary source HAP-related direct inhalation cancer incidence reductions. Results from each of these
studies are presented below.
Analyses of Cancer Risks front Selected Air Toxic
Pollutants
The Agency conducted two efforts to broadly assess the magnitude and nature of the air toxics
problem. The 1985 report entitled, "The Air Toxics Problem in the United States: An Analysis of Cancer
Risks for Selected Pollutants"262 otherwise known as the "Six Month Study," was intended to serve as a
"scoping" study to provide a quick assessment of the air toxics problem utilizing only readily available
data on compound potencies, emissions, and ambient pollutant concentrations. The Agency updated this
analysis of cancer risks in the 1990 report entitled ^Cancer Risk from Outdoor Exposure to Air Toxics"
referred to here as the "1990 Cancer Risk study."263
For the pollutant and source categories examined, the 1990 Cancer Risk study estimated the total
nationwide cancer incidence due to outdoor concentrations of air toxics to range from as many as 1,700 -
2,700 excess cancer cases per year with 14 compounds accounting for approximately 95 percent of the
annual cancer cases. Additionally, point sources contribute 25 percent of annual cases and area sources
contribute 75 percent of annual cases, with mobile sources accounting for 56 percent of the nationwide
total.264
The Six Month study indicates that the criteria air pollutant programs appear to have done more to
reduce air toxics levels during the f 970 to 1990 period than have regulatory actions aimed at specific toxic
compounds promulgated during the same period. Metals and polynuclear compounds usually are emitted
asparticulate matter and most of the volatile organic compounds are ozone precursors. As such, they are
regulated under State Implementation Plan (SIP) and New Source Performance Standard (NSPS) programs
and Title n motor vehicle regulations. A number of reports cited indicate significant reductions in air toxic
emissions attributable to actions taken under SIP, NSPS and mobile source programs. Additionally, EPA
conducted a comparison of ah- quality and emissions data for 1970 with the estimates of cancer incidence
30 U.S.EPA. Office of Air Quality Planning and Standards. TheAir Toxics Problem in the United States: An Analysis of Cancer Risks for
Selected Pollutants. May 1985. EPA-450/1-85-001.
30 U.S.EPA. Office of Air Quality Planning and Standards. Cancer Risk from Outdoor Exposure to Air Toxics. September 1990. EPA-
45Q/l-90-004a.
** The 1990 Cancer Risk study repotted approximately 500 - 900 more cancer cases per year than toe Six Month Study due primarily to the
inchiskn of mote pollutants, better accounting of emissions sources, and, in some cases, increases in unit risk estimates.
356
-------
Appendix H: Air Toxics
for 1980.265 Methods, assumptions and pollutants included were held constant over the period. The
analysis showed a significant decrease in incidence during the decade due to improvements in air quality,
presumably related to general regulatory programs. For the 16 pollutants studied, estimated nationwide
cancer incidence decreased from 3600 in 1970 to 1600 in 1980. The 1990 Cancer Risk Study did not
attempt to update this analysis.
Although it is difficult to make quantitative conclusions from these two studies regarding the
benefits of CAA air toxics control, it is apparent that the pollutant-specific and source category-specific,
NESHAPs were not structured to reduce significant air toxic emissions from area arid mobile sources. In
fact, the 1990 Cancer Risk Study indicates that considerable cancer risk remained prior to passage of the
1990 CAA Amendments: as many as 2,700 excess cancer cases annually. However, some sjudies indicate
that the criteria air pollutant program played a critical role during the 1970 to 1990 period mac^eving air
toxic emission reductions and therefore decreasing cancer risk.
Cancer Risk Estimates from NESHAP Risk
.Assessments
In looking back at the estimated effects of the HAP standards, EPA found that the effects of the
NESHAPs were not quantified completely. These estimates occurred at a time when emission estimation
and risk assessment methodologies for HAPs were first being developed. One consequence-is that because
emissions were not fully characterizedj air toxic exposures could not be completely assessed. Additionally,
most assessments only focused on the specific HAP being listed under the CAA and did not assess the
reduction of other pollutants, which are currently considered HAPs. For example, while the vinyl chloride
standard reduces emissions of ethylene dichloride, these emission reductions were not assessed in the risk
assessmentf In a different context, reductions of HAP may also achieve reductions of VOC and PM. The
benefits of such reductions generally were also not evaluated. In addition, EPA generally did not assess
the potential exposure to high, short-term concentrations of HAP and therefore did not know whether
toxics effects from acute exposures would have been predicted and possibly addressed by the HAP
standards. ~1
In addition, people living near emission sources of concern are often exposed to a mix of
pollutants at once. Some pollutants have been shown to act synergistically together to create a health risk
greater than the risk that would be expected by simply adding the two exposure levels together. More
research is needed to understand the effects of multiple-pollutant exposures. Finally, HAP risks tend to be
distributed unevenly across exposed populations, with particularly high exposures occurring closest to
emission sources. It should be noted that HAP exposure to specific populations may tend to fall
disprop|^pnately?anlong the poor and minorities, who are more likely to live in close proximity to
emitting facilities. •
With the above caveats in mind, Table 113 provides .information about maximum individual risk
taken from the Federal Register notices for the NESHAPs promulgated before the 1990 amendments to the
Clean Air Act. The benefits are calculated by multiplying the estimated annual incidence reduction by
$4.8 million valuation per statistical life (1990, dollars). These benefit estimates provide a snapshot of
potential monetized benefits for the year in which each NESHAP was promulgated. Of course these
265 Hunt, W.F., Faoro, R.B. and Curran, T.C., "Estimation of Cancer Incidence Cases and Rates for Selected Toxic Air Pollutants Using
Ambient Air Pollution Data, 1970 vs. 1980". U.S. EPA. April 1985.
357
-------
Appendix H: Air Toxics
estimates do not include air toxics benefits for other health and ecological benefit categories or air toxics
benefits from co-control of criteria air pollutants. All uncertainties associated with the original estimates
remain.
h
§
1
*i
5
i
i
1,
I
£
Hflmiffij
PolllltMt
„ V, ..,.,.1; .'_
IKBZOK
beBzeae
beaoae
beazeiK
beozeae
arsenic
"' arsenic
asbestos
!vi«yl
chloride
lancer Incidence
ilWpPffiW f 1 ' t i /]>
dSiksfSt «,*ti^
Source
Category
coke by-
product
storage
vessels
waste
operations
transfer
operations
primary
copper
glass mariuf.
demolition
PVC
production;
frt't i] " i *t - • , <• - ' ,,, ,. * , f
Redactions and Monetized Benefits for NESHAPs. , « , V ,\
«« nT i ' s, > ' , " » ft, , **'/-.' i !:
*tiiHi"'j»iM 'it " •, ! i kt j , . | JK '
Year
Promulgated
1985
1984
1982
1986
1987
1986
1986
1973
1975
Pre-Reg
Maxtaium
iHdlvidual
Risk
l^xlff3
7x10^
4^xKr*
2x10*
6xlO*
1.3x10^ to
5xlO*
TxlO^to
SxlOr*
Post-Reg
Mfflyimiinfr
lodividual
Risk
4JSXI&4
2x10^
3x10^
5xlO's
4x10"*
1.2xlO-3
to 3x10-*
1.7xlO-«
to6x!0-«,
Reducttonin
Cancer
Incidence
{per year)
31
\
i
135
0.01 to 0.06
0.55
0.98
0.09
t
0417 to , "
0.0034
100
1 <•!
10JS > „
J> :
!,
Benefits In
SmflBonper
year
(isw«ti
/• f f-
us*
4A
0.05 to
0.3 '
2.6 , '
1
f 9
0.4
0.02 to 0.6
1
480
50.4 ' ,
'
" f M U ,
••> > f
Aft? >l ",
Non-utility Stationary Source
Reductions
Cancer incidence
The Project Team commissioned two studies to estimate reductions in cancer incidence due to pre-
1990 NESHAPs: the PES Study and the ICF Re-analysis. The methodology used for most air pollutant
evaluations involved a "back calculation" for the estimation of incidence reductions. However, the EPA
has elected not to rely on the results of this analysis given critical methodological flaws. Despite the
Project Team's concerns, the methodology and results of the two studies are presented below in the interest
358
-------
Appendix H: Air Toxics
of full disclosure and to guide efforts to develop a more valid and reliable analysis of the health-related
benefits of HAP reductions in the upcoming section 812 Prospective studies.
t*ES Study
Methodology
The first attempt to estimate, for this study, historical non-utility stationary source HAP-related""
direct inhalation cancer incidence reductions was conducted by Pacific Environmental Services (PES).
The basic approach used in the PES study was to adjust the cancer incidence estimates developed for
EPA's 1990 Cancer Risk study to reflect the changes in emissions of, and exposures to, 14 key HAPs:
arsenic, asbestos, benzene, 1,3-butadiene, carbon tetrachloride, chloroform, hexavalent chromium, dioxin,
ethylene dichloride, ethylene dibromide, formaldehyde, gasoline vapors; products of incomplete
combustion (PICs), and vinyl chloride.
The first step was to compile baseline incidence levels, defined as cancer cases per million
population, for each of the 14 pollutants. The point estimates of incidence from the 1990 Cancer Risk
study were used for this purpose. For some source categories, the "best point estimate" from the 1990
Cancer Risk study was used, for others a mid-point was selected.266 These baseline incidence levels were
based on measured ambient concentrations of the pollutant, modeled concentrations, or both.
The second step involved allocating baseline incidence levels to the individual source categories
known to emit the relevant pollutant In some cases, adjustments were made to reflect differences among
the vintages of source category-specific data.267 All baseline incidence estimates were ultimately expressed
relative to a 1985 base year,568 The assumption was then made that source-category incidence rates were
proportional to the level of emissions from that source category.
f ~~ .
Next, levels of control for each source category-specific incidence rate were estimated for each of
the target years of the present analysis (Le., 1970,1975,1980,1985, and 1990).26? Source category-
specific activity level indicators were then established and linked to changes in corresponding activity
indicators provided by the J/W macroeconomic modeling results. Activity levels were estimated for each
source category, for each of the target years, and for each of the two scenarios.
Finally, source category/pollutant combination incidence levels for both the control and no-control
scenarios were developed. These incidence levels were developed based on the baseline incidence levels,
Jhe activity indicators, and the control levels for each year. Both of these latter two factors varied between
266 For sane of the source categories, tee original NESHAP/Air Toxfc Exposure and Risk Information System (NESHAP/ATERIS) estimates
• of incidence were not available, in which case the baseline incidence was obtained from the 1989 National Air Toxics Information Clearinghouse
NATKH) Database Report. (See PES, "Draft Summary of Methodology Used for Cancer from Stationary Sources," memorandum from Ken
Meardon, PES to Vasu Kilaru, US EPA, March 2Z, 1993, p. 2.)
257 For example, six discrete sources for vinyl chloride were identified in the Six-Month Study Update. Point estimate incidences for each of
these source categories came from separate references with databases corresponding to different years. (See PES, "Retrospective Analysis for
Section 812(a) Benefits Study," September 30,1992, p. 8.)
268
' See PES, March 22,1993 memorandum, p, 3.
20 Control level estimates were based on one of the following: control efficiencies for related criteria pollutants defined in the criteria
pollutant analysis, reference documents such as Control Technology Guidelines (CTGs) or Background Information Documents (BIDs), preambles
for related regulations, orEPA experts. (See PES, March 22,1993 memorandum, p. 3.) .
359
-------
Appendix H: Air Toxics
the control and no-control scenarios. The activity levels differed based on the specific levels of related
sector economic activity predicted by the J/W model for the control and no-control scenario. The control
levels prevailing in each of the target years were used for the control scenario, and the 1970 control level
was applied throughout the 1970 to 1990 period for the no-control scenario.270 The formula used for these
calculations was as follows:271
lf, I.
where:
I
A
P
C
ty
by
Findings
cancer incidence for a source category-pollutant combination
activity level for a source category
population
control level for a source category-pollutant combination
target year (1970... 1990)
base year
The PES analysis concluded mat
substantial reductions in HAP-related cancer
cases were achieved during the reference
period of the present study. The vast majority
of these estimated reductions were attributable
to reduced exposures to asbestos, particularly
from manufacturing and fabricating sources.272
In fact, roughly 75 percent of the total
reduction in cancer cases averaged over the
1970 to 1990 period were attributed to
asbestos control.273 Figure 50 summarizes the
PES study overall cancer incidence reductions
and the relative contribution of asbestos-
related reductions over the study period.
,!, '" ''"fin"
Figure 50. FES Estimated Reductions in HAP-Related
Cancer Cases.
DOtherHAPsI
•Asbestos I
1975
1980 1985
Year
1990
The Project Team had several
concerns about the PES results. First and
foremost, the reductions in asbestos-related cancer cases appeared to be substantially higher than expected,
particularly in the earlier target years. Second, the control scenario activity level indicators for several
sources with which Project Team members were familiar did not appear to be even remotely consistent
** More detailed descriptions of the methodology and associated uncertainties are provided in "Retrospective Analysis for Section 812(a)
Benefits Study," a September 30,1992 memorandum fiom Ken Meardon, PES to Vasu Kilaru, US EPA.
571 See PES, Mirch 22,1993 memorandum, p. 4.
372 PES, "Cancer Risk Estimates from Stationary Sources," memorandum from Ken Meardon, PES to Vasu Kilaru, US EPA, March 5,1993.
273 ICF, "Direct Inhalation Incidence Benefits," Draft Report, November 11,1994, p. 10.
360
-------
Appendix H: Air Toxics
with actual historical activity patterns.274 Finally, the level of documentation of the analytical
methodologies, assumptions, and results was insufficient to ascertain the validity and reliability of the
results. Ultimately, the Project Team determined that it was necessary to conduct a formal review and re-
analysis of the cancer incidence reductions associated with non-utility stationary source HAP controls. The
results of the FES analysis remain a relevant part of the record of the present study, however, since they
provided a substantial basis for the subsequent re-analysis by ICF Incorporated.
IGF RG-analysIs
Methodology _ -„
The purposes of the ICF Re-analysis were to examine the methodology and results of the PES
study, particularly to address the aforementioned concerns of the Project Team, and to develop a revised
set of estimates. Due to significant constraints on the resources remaining for HAP analysis in the section
812 study, however, only a few key HAPs could be investigated in depth and many important issues could
not be addressed.275 Furthermore, the effects of two early and potentially important HAP standards -the
Beryllium and Mercury NESHAPs- could not be evaluated Nevertheless, the ICF Re-analysis clarified
some potential sources of uncertainty in the PES results and provided revised cancer incidence reduction
estimates for several HAPs.
s.
A key uncertainty in the PES results was associated with the use of a "back-calculation" technique
to estimate incidence reductions for some HAPs. The back-calculation technique estimates uncontrolled
incidence by dividing residual incidence by the assumed control efficiency. This approach means
uncontrolledincidence, and therefore incidence reductions, are highly sensitive to small changes in
assumed^^^elSciency?76 In some cases, the PES analysis may have used control efficiencies which
were too M^p, resulting in overestimation of uncontrolled incidence and therefore incidence reductions
attributable to & CAA.277 The vinyl chloride incidence reduction estimates appear to be significantly
influenced by the use of Ms back-calculation technique. Another important source of uncertainty
identified by ICF involved the potential overestimation of incidence totals when source apportionment is
based on measured ambient concentrations.278 ICF was unable, however, to perform an extensive
evaluation of the activity level indicators used in the PES study.279
v* . _*° # . , _ •
_^ The first step undertaken in the re-analysis was to conduct a screening test to identify the HAPs
which accounted for the most significant estimated incidence reductions. Based on this screening analysis,
m For extmpte, the activity indicators for Municipal Waste Combustors (MWCs) incorporated in the PES analysis decline dramatically
throughout the 1975 to 1990 period. (See PES, March 5,1993 memorandum to Vasu Kilara, p. 10). In reality, overall MWC capacity and
throughput increased significantly over this period.
275 For example, the Project Team sought to develop and apply a methodology for estimating a central tendency estimate for the total
carcinogenic risk imposed by all the HAPs examined. The intent was to address concerns about potential overestimation of aggregate risk measures
when combining upper bound risk estimates of multiple HAPs. Unfortu »rtrly, resources were insufficient to continue development of this
methodology.
276 An example of this back-calculation technique illustrating the sensitivity to the assumed control efficiency is presented on page 12 of the
draft ICF report
277 See ICF Draft Report, p. 12.
278 See ICF Draft Report, p. 9.
279 See ICF Draft Report, p. 13.
361
-------
Appendix H: Air Toxics
ICF eliminated 1,3-butadiene, carbon tetrachloride, chloroform, gasoline vapors, chromium, formaldehyde,
and PICs from the detailed re-analysis effort.
Detailed reviews were then conducted for the remaining HAPs: vinyl chloride, dioxins, ethylene
dibromide (EDB), ethylene dichloride (EDO), benzene, asbestos, and arsenic. In the re-analysis of these
HAPs, ICF determined whether a forward- or back-calculation technique was used for the relevant source
categories of a given HAP, reviewed the regulatory history of the relevant source categories to re-evaluate
the assumed control efficiencies, and reviewed the upper-bound unit risk factor for each HAP. Revised •
total incidence reduction estimates for each HAP and for each target year were then calculated using the
same basic calculation procedure used by PES. Finally, ICF identified a number of residual deficiencies in
the analysis which could only be addressed through additional research and analysis.280
findings
The ICF Re-analysis largely affirmed the original results obtained by PES; primarily because the
PES analysis itself served as the basis for the re-analysis and only minor adjustments were adopted for
many critical variables. In particular, most Project Team concerns regarding the PES methodology could
not be resolved, including uncertainties associated with activity levels, assumed control efficiencies, and
the unexpectedly high estimated incidence reductions associated with asbestos. In fact, the ICF Re-
analysis produced a revised upper bound estimate for vinyl chloride-related incidence reductions which
were even higher than the asbestos benefits. .
Several sets of results were developed by ICF and presented in either the November 1994 draft
report or in briefing materials prepared for the Science Advisory Board Clean Air Act Compliance
Analysis Council Physical Effects Subcommittee (SABTCAACACPERS) in May 1995. The first set of
results is based on the assumption of 100 percent source compliance with HAP control requirements. An
alternative set of results was developed assuming an 80 percent compliance rate with applicable standards.
Given the linear effect of changes in compliance rates, these results were precisely 20 percent lower than
the first set of estimates. At the ^ay 1^5 CAACACPERS briefing, estimates based on the 100 percent
compliance estimates were presented. For asbestos, the revised incidence reductions were presented and
characterized as upper bound. The asbestos estimates were then combined with upper and lower bound
estimates for vinyl chloride and for "all other compounds.9' Figure 51 presents the total cancer incidence
reductions derived from the ICF Re-analysis, using the asbestos estimates combined with the lower bound
estimates for non-asbestos HAPs. Figure 52 presents a comparable compilation reflecting the upper bound
estimates for all HAPs.
30 Addition! details of the ICF Re-analysis methodology can be found in ICF, "Direct Inhalation Incidence Benefits," Draft Report,
November 11,1994.
362
-------
Appendix H: Air Toxics
The Project Team remains concerned about these incidence reduction estimates, particularly given
the doubts raised by the SAB/CAAGACPERS at the May 1995 presentation of these results. For instance,
several critical assumptions are needed to make this analysis valid when applied to EPA's NESHAPs. The
flaws in these assumptions are described ,
below, . •
Figure 51. ICF Estimated Reductions in Total HAP-Related
Cancer Cases Using Upper Bound Asbestos Incidence and
Lower Bound Non-Asbestos HAP Incidence.
1975 1980 1985 1990
Year
(1) The risk estimates described in the
1990 Cancer Risk study, which served as the
baseline for determining risk reductions, were
accepted without question. There are myriad
uncertainties in these estimates that must be
recognized, as the study was designed only to
generate rough order-of-magnitude estimates
of the extent of the air toxics cancer problem.
(2) The percent control efficiency for
emission reductions, which is calculated in
each standard, would have to apply across
every facility. Typically, the emission
reductions vary between facilities; using a
single average reduction could skew the
results. ,
(3) There is a direct correlation
between the number of tons of emissions
reduced and incidence reduced by a specific
regulation. Given the assumption of a linear,
non-threshold dose-response curve (as is
typically done for cancer), this is theoretically
correct. ,
(4) Finally, the back calculation
approach assumes that there is 100 percent
compliance with the regulation. ,
EPA staff reviewed the "back
calculation" approach for one of the more
controversial aspects of the vinyl chloride
(VQNESHAP. The PES study estimates '. """ - ' ' .—"
benefits at 426 cases reduced in 1990. The -
ICF Re-analysis resulted in an even higher estimate, between 1,000 and 7,000 cases annually. An analysis
by EPA staff indicated that these vinyl chloride risk estimates are highly suspect given historical cancer
incidence data for hepatic angiocarcinoma, a specific cancer that has been linked to vinyl chloride
(Koppikar and Fegley, 1995). The following analysis demonstrates the inadequacies of the assumptions in
the 1993 study.
Figure 52, ICF Estimated Reduction in Total HAP-Related
Cancer Cases Using Upper Bound Incidence for All HAPs.
12
11
1975
1980 1985
Year
1990
363
-------
Appendix H: Air Tomes
(1) In the actual standard, no control technology was required for emissions from oxychlorination
vents at ethylene dichloride (EDQ/VC plants. Applying "back calculation" for these emissions is
inappropriate.
(2) In 1985, there were an estimated 8,000 fabrication plants which processed resins produced by
PVC plants, thus resulting in VC emissions, which were exempt from the VC NESHAP. They emit very
small quantities of VC and back calculation is not appropriate.
.-..;>., ^ *.
(3) The 1993 study uses a baseline estimate of 18 resii|ual cases from the NESHAP/ATERIS data
base. There is no evidence that these cases resulted only from emissions from PVC and EDC/VC plants.
(4) The risk analysis performed for the October 21,197(5 final VC regulation projeicjed an
incidence reduction of 11 cases per year. '
In contrast, the PES study, using the "back calculation" method derived the following annual
incidence reductions:
1980 - 250 cases
1985-360 cases
1990 - 430 cases
The subsequent back calculation conducted in the ICF Re-analysis resulted hi incidence reductions as
much as an order of magnitude higher than these,
Bven considering the slightly different industrial output assumptions imposed by macroeconomic
modeling, such a stark contrast is difficult to explain except for a critically flawed approach. Growth in
activity and population nor other factors explain the difference hi these two estimates. Given that the same
general methodology was usetf for afl of the air toxic pollutant assessments as was used for the VC
NESHAP evaluation,, there is reason to believe that cancer incidence results for the other air toxic
pollutants are also flawed.
Mobile 'Source HAP Exposure Reductions
EPA's Cancer Risk report estimated that approximately 60 percent of the total carcinogenic risk
posed by HAPs was attributed to mobile sources, with stationary sources contributing 15 percent and area
sources contributing the remaining 25 percent281 The relative importance of mobile sources to total HAP
exposure was a significant motivation behind EPA's subsequent effort to examine exposures and risks
from mobile source HAPs.282 Although available analytical resources were severely limited, the Project
Team nevertheless decided it was necessary to perform at least an initial screening analysis to estimate the
differences in mobile source HAP exposures between the control and no-control scenarios configured for
the present study.
"'jjCancer Risk report, Page ES-12.
** See US EPA/OAR/OMS.'lkfotar Vehicle-Related Air Toxics Study," EPA 420-R-93-005. April 1993.
364
-------
Appendix H: Air Toxics
Methodology
The approach used by ICF/SAI in conducting the mobile source HAP analysis closely followed the
approach used in the EPA Motor Vehicle-Related Air Toxics Study (MVATS).283 Recognizing the dearth
of HAP ambient concentration and exposure data, both studies use carbon monoxide (CO) concentrations
as the basis for estimating mobile source HAP concentrations and exposures. An important difference
between the two studies, however, is that the ICF/SAI study adjusted the estimated change in ambient C0
concentrations to take account of background284 and non-mobile source285 CO emissions. The HAP ~
exposure function used in the ICF/SAI analysis is summarized by the following equation:
A T**M x
. , - , . .
where:
E ' -SB. exposure to motor vehicle-emitted HAP
C = annual ambient CO concentration to annual CO exposure concentration conversion factor
A = county-level annual average ambient CO concentration
B = background CO concentration
S = no-control to control scenario CO concentration adjustment factor (equals 1 for the
control scenario)
M = total CO exposure to mobile source (X) exposure conversion factor
VOC a VOC emissions by year, county, and scenario
HAP = VOCspeciation factor by mobile soujfeeHAP
CO ' = CO emissions by year, county, and scenario
•* *
Details of the derivation of each of the variables applied in the above equation are provided in the
ICF/SAI report" However, in essence, me calculation involves the following basic steps.
First, annual average, county-level CO ambient monitoring data are compiled from the EPA
Aerometric Information RetrievalJSystem (AIRS) database. After adjusting for background and non-
mobile source contributions, these annual average ambient CO concentrations are converted to annual
average CO exposure concentrations. As in the EPA MVATS, this conversion is made based on the
Hazardous Air Pollutant Exposure Model - Mobile Sources (HAPEM-MS) population exposure model,
which takes account of time spent in five indoor and outdoor microenvironments: indoors at home, other
indoor, in-vehicle, outdoors near roadway, and other outdoor.286 After adjusting for CO exposures
attributable to non-mobile sources of CO, the CO exposures are converted to exposures for each of the
mobile source HAPs based on available VOC speciation data and the ratio of co-located VOC and CO
283
2H
US
ICF/SAI, "Retrospective Analysis of Inhalation Exposure to Hazardous Air Pollutants from Motor Vehicles," October 1995, p. 4.
Background CO is produced by the oxidation of biogenic hydrocarbons. See ICF/SAI, p. 7.
1 The EPA MVATS attributed all measured CO to motor vehicles, resulting in an overestimation of motor-vehicle contributions to CO
concentration changes. See ICF/SAI, p. 8. The MVATS assumption would also lead to a subsequent overestimation of changes in HAP exposures.
286 See ICF/SAI, p. 3.
365
-------
Appendix H: Air Toxics
emissions.287 These calculations are repeated for the no-control scenario after adjusting for differences in
CO ambient concentrations for each target year and for differences in fuel composition.
a&suits
By 1990, CAA controls resulted
in significant reductions in exposure to
motor vehicle HAPs. Figure 53
summarizes the nationwide annual
average exposure levels, in micrograms
per cubic meter, for each of the five
HAPs analyzed under the control and
no-control scenarios. Additional
detailed results, including breakdown by
urban versus rural environments and
comparisons with the EPA MVATS
estimates, are provided in the ICF/SAI
report.
Analytical resources to carry
forward these exposure estimates to
derive estimates of the changes in motor
vehicle HAP-related adverse effects
attributable to historical CAA programs were not available.
Figure 53, National Annual Average Motor Vehicle HAP
Exposures (wg/m3).
B enzene A cetaHehyde . . D EselPM
Fosn a2iehyde 1,3-B utadsne
Non-Cancer Health Effects
Broad gaps exist in the current state of, knowledge about the quantifiable effects of air toxics
exposure. This is particularly true for a wide range of health effects such as tumors, abnormal
development, birth defects, neurological impairment, or reproductive impairment, etc. For example, the
EPA's Non-Cancer Study2"8 found that ambient concentrations for a substantial number of monitored and
modeled HAPs exceeded one or more health benchmarks.289 However no accepted methodology exists to
quantify the effects of such exceedences. More data on health effects is needed for a broad range of
chemicals.
217 The same HAP emission fractions used in Ihe EPA MVATS were used herein, except for diesel PM which is not proportional to VOC
emissions. Instead, diesel PM emission faclors were developed using year-specific PARTS diesel PM emission factors and VMT estimates for
diescl-powcred vehicles.
a* U.S. Environmental Protection Agency, "Toxic Air Pollutants and Noncancer Risks: Screening Studies," External Review Draft,
September, 1990.
3*> Relevant benchmarks include Acceptable Daily Intake (ADI), the estimate of daily exposure at which adverse health effects are unlikely;
and Lowest Observed Actual Effect Level (LOAEL), which is the lowest exposure level at which significant adverse health effects are observed.
366
-------
Appendix H: Air Toxics
Through the 1970s and 1980s, the adverse effects of toxic pollution on the Great Lakes became
clear and undeniable. Over the same time period, scientists began collecting a convincing body of
evidence that toxic chemicals released to the air can travel long distances and be deposited on land or water
far from the original sources. An example of this evidence is the presence of such contaminants as PCBs,
toxaphene, and other pesticides in fish in Lake Siskiwit, a lake on an island on upper Lake Superior, which
has no waterborne sources of pollution. Toxaphene, a pesticide used primarily in the southeastern U.S.
cotton belt, has been found as far away as the Arctic, with a decreasing air concentration gradient from the
southeast toward the Great Lakes and the north Atlantic regions.
Similarly, a growing body of evidence showed that pollutants that were persistent (do not easily
break down) and bioaccumulating (not significantly eliminated from the body) were magnifying up the
food chain, such that levels in top predator fish contained levels up to millions of times greater than the
harmless levels in the water. As such, those who ate those large fish, such as humans, eagles, mink, and
beluga whales could receive very high exposures to the pollutants. Wildlife were beginning to show
adverse effects in the wild, that could be duplicated in the lab. In the Great Lakes, such chemicals as PCBs,
mercury, dieldrin, hexachlorobenzene, Lindane, lead compounds, cadmium compounds, DDT/DDE, and
others are of significant concern. In other places in the country, similar effects are being experienced,
especially with mercury, which is transported primarily by air, but exposure to which is primarily through
contaminated fish. It was this kind of information about DDT and toxaphene that led to their being banned
in the U.S. under FIFRA.
While ecological and economical sciences are not yet sufficiently advanced to support the kind of
comprehensive, quantitative evaluation of benefits needed for the present study, selected local and regional
scale adverse ecological effects of HAPs, and their adverse consequences for human health and welfare,
can and have been surveyed. In May 1994, the EPA issued its first "Report to Congress on Deposition of
Air Pollutants to the"Great Waters."290 The^Great Waters Report examined the pollutants contributing to
adverse ecological effects, the potential significance of the contribution to pollutant loadings from
deposition of airborne pollutants, and the potential adverse effects associated with these pollutant loadings.
Key HAPs identified in the Great Waters Report include PCBs, mercury, dioxins, and other heavy metals
and toxic organks.
Of particular relevance to the present assessment, the Great Waters Report demonstrated the
significance of transport and transformation of HAPs through food webs, leading to increased toxicity and
biomagnification. A prime example of adverse transport and transformation is mercury. Transformation
from inorganic to methylated forms significantly increases the toxic effects of mercury in ecosystems. A
prime example of biomagnification is PCBs. As noted in the Great Waters Report:
"Pollutantsof concern [such as PCBs] accumulate in body tissues and magnify up the
food web, with each level accumulating the toxics from its diet and passing the burden
along to the animal in the next level of the food web. Top consumers in the food web, '
usually consumers of large fish, may accumulate chemical concentrations many millions ''_
of times greater than the concentrations present in the water...High risk groups...include
290 USEPA/OAR/OAQPS, "Deposition of Air Pollutants to the Great Waters, First Report to Congress," EPA-453/R-93-055i May 1994.
367
-------
Appendix H: Air Toxics
breast-feeding mothers because breast-fed babies continue to accumulate [pollutants]
from their mothers after birth. For example, they can have PCB levels four times higher
than their mothers after six to nine months of breast feeding. "X1
Because of the risk of significant exposure to infants and other high-risk groups, such as "sport
anglers, Native Americans, and the urban poor,"292 a substantial number of fish consumption advisories
have been issued in recent years. Current fish advisories for the Great Lakes alone include widespread
advisories for PCB's, chlordane, mercury and others, cautioning that nursing mothers, pregnant women,
women who anticipate bearing children, female children of any age and male children age 15 and under
not eat certain high-food chain fish species. It should be noted as well that 40 states have issued .mercury
advisories in some freshwater bodies, and nine states have issued mercury advisories for every freshwater
waterbody in the state (these states are Maine, New Hampshire, Vermont, Massachusetts, New York, New
Jersey, Missouri, Michigan, and Florida).
There is little evidence indicating that the CAA had much beneficial effect on air toxic deposition
to water bodies. Since the early NESHAPs were based on direct inhalation, primarily cancer effects close
to a plant, they did not address the issue of cumulative effects of persistent pollutants far from the source. It
was for this reason that Section 112(m) was included in the 1990 CAA Amendments, with requirements to
study and document the atmospheric contribution of water pollutants, the adverse human health and
environmental effects resulting and the sources that should be controlled to prevent adverse effects, and
additionally, to promulgate regulations to prevent adverse effects.
Conclusions — Research Heeds
-• :;0/s, • , " ' ' /•
As has been demonstrated, mere are broad gaps in the current state of knowledge about the
quantifiable effects of air toxics exposure for a wide range of both human health and environmental
effects. The following discussion outlines areas in which further research is needed in order to adequately
quantify the benefits of air toxics control.
Health Effects
• Develop health effects data on pollutants for which limited or no data currently exists. Such
studies should be focused on pollutants with a relatively high probability of exposure and/or
potential adverse health effects.
• Understand mechanism of action of pollutants, for example through pharmacokinetic modeling.
This will allow for a more accurate assessment of the effects of these pollutants on humans.
* Conduct research on factors that affect variations in susceptibility of human populations and
determine the distribution of these factors in the U.S.
• Conduct research to better understand interactive effects of multiple pollutant exposures.
EPA-453/R-93-055, May 1994, p. ix.
EPA-453/R-93-055, May 1994, p. x.
368
-------
Appendix H: Air Toxics
Develop methodologies to derive alternative estimates of human cancer risk from existing upper-bound
methods.
• . Acquire data and develop dose-response relationships for critical noncancer effects such as
developmental, neurotoxic, mutagenic, respiratory and other effects. In particular, design
methodology to quantify effects of exposures above health benchmarks.
• Acquire data and develop methods to estimate effects from acute exposure.
Exposure Assessment
• Expand data collection efforts: pre- and post-control emissions; HAPspeciation; facilities location;
facility parameters (stack heights, distances from stacks to fencelines, etc.)
• Develop more comprehensive exposure models which incorporate activity patterns, indirect
exposures, total body burden, ratios of time spent indoors to outdoors.
• Continue to refine uncertainty analysis methods.
Ecosystem Effects
.••*'*- .
• Reliable estimates/measures of the levels of persistent bioaccumulating toxics in different media
(air, water column, soils and sediments)
• Work to correlate levels of persistent bioaccumulating toxics with exposures, biota
concentrations/accumulation, and adverse effects, especially subtle effects such as wasting,
behavioral effects, and developmental effects.
• Criteria for effects, such as a wildlife correlate to a RfD or dose-response curve. This work should
be done to complement the mass balance efforts now being completed, which will model source
emissions to water column concentrations, then design research to predict effects on living
..resources giffli those predicted levels.
^ <* , . r
• Work to determine the effects of mixtures of persistent bioaccumulating toxic pollutants, and to
determine cause-effect relationships of exposures over long periods of time. '
• Studies to evaluate toxic effects in less well understood terrestrial systems such as: soil
organisms/invertebrates, food web effects, amphibian effects, effects on endangered species and
phytotoxic effects.
369
-------
Appendix H: Air Toxics
Economic Valuation
* Develop valuation estimates for endpoints for which inadequate estimates currently exist. These
valuation estimates must be consistent with the kinds of damages expected.
• Initiate broad-scope economic valuation of air toxics program using survey techniques.
370
-------
Appendix I: Valuation of Human
Health and Welfare Effects of
Criteria Pollutants
For the Section 812 analysis of health benefits, valuation estimates were obtained from the
literature and reported in dollars per case reduced for health effects, and dollars per unit of avoided damage
for welfare effects. Similar to estimates of physical effects provided by health studies, the monetary values
of benefits were reported both in terms of mean values as well as a probability distribution of estimates.
This permitted an evaluation of the uncertainty associated with the point estimates. It is interesting to note
that the distributions of benefit values varied by endpoint. For example, while the estimate of the dollar
value of a reduced mortality was lognormally distributed, the estimate for the value of a reduced case of
acute bronchitis was uniformly distributed between a minimum and maximum valfite. —
For the welfare benefits analysis, the avoided losses were in many cases measured in monetary
terms directly. For agricultural benefits, however, the benefits associated with estimated changes in crop
yields were evaluated with an agricultural sector model, AfcrSIM. This model incorporates agricultural
price, farm policy, and other data for each year. Based on expected yields, the model estimates the
production levels for each crop, the economic benefits to consumers, and the economic benefits to
producers associated with these production levels. To the extefit that alternative exposure-response
relationships were available, a range of potential benefits was calculated.
Methods Used to Value Health Effects
This section describes the derivations of the economic valuations for health and welfare endpoints
considered in the benefits analysis. An introduction to the method for monetizing improvements in health
and welfare is followed by a summary of dollar estimates used to value benefits and detailed descriptions
of the derivation of each estimate. Economic valuations aregiven both in terms of a central (point)
estimate as well as a probability distribution which characterizes the uncertainty about the central estimate.
All dollar values are in 1990 dollars.
Willingness to pay (WTP) is the measure used for the value an individual places on something,
whether it is something that can be purchased in a market or not. WTP is the maximum amount of money
such that the individual would be indifferent between having the good (or service) and having the money.
For both market and non-market goods, WTP reflects individuals' preferences. Because
preferences are likely to vary from one individual to another, WTP for both market and non-market goods
(e.g., health-related improvements in environmental quality) is likely to vary from one individual to
another. In contrast to market goods, however, non-market goods such as environmental quality
improvements are public goods, whose benefits are shared by many individuals. The individuals who
"consume" the environmental quality improvement may have different.WTPs for this non-market good.
The total social value of the good is the sum of the WTPs of all individuals who consume the good.
371
-------
Appendix I: Valuation of Hitman Health and Welfare Effects on Criteria Pollutants
In the case of health improvements related to pollution reduction, it is not certain specifically who
will receive particular benefits of reduced pollution. For example, the analysis may predict 100 days of
cough avoided resulting from a reduction of pollutant concentrations to achieve a given standard, but the
analysis does not estimate which individuals will be spared those days of coughing. The health benefits
conferred on individuals by a reduction in pollution concentrations are, then, actually reductions in the
probabilities of having to endure certain health problems. These benefits (reductions in probabilities) may
not be the same for all individuals (and could be zero for some individuals). Likewise, the WTP for a
given benefit is likely to vary from one individual to another. Iff&eory, the total social value associated
with the decrease in incidence of a given health problem resulting from a given reduction in pollution
concentrations is
,-,"--•—'y j^t in—-y,- --.' ^ , - ', .„ ' ^~'.!",, ?^,o, „,/' '?"*''*,',''**"-'
" .''i
-------
Appendix I: Valuation of 'Human Health and Welfare Effects on Criteria Pollutants
WTP for a particular health benefit is, of course, unlikely to be the same for all individuals. If the
individuals receiving health improvements, however, are a random sample from the population (i.e., if all
individuals have the same chance of receiving these benefits), then the mean WTP is the population
parameter of interest. Predicted benefits are, in this case, just the mean WTP for the benefit times the
'number of individuals predicted to receive the benefit.
The individuals actually receiving health improvements, however, may not be a random sample of
the population. Suppose, for example, that most of those receiving a particular benefit (e.g., a reduction in
the probability of dying in the current year) as a result of a reduction in pollution are the elderly. If WTP
for this particular health risk improvement among the elderly Is substantially different from WTP for the
same health risk improvement among younger individuals, men economic theory would indicate that using
the population mean WTP will give a biased result.
While the estimation of WTP for a market good (i.e., the estimation of a demand schedule) is not a
simple matter, the estimation of WTP for a non-market good, such as a decrease in the risk of having a
particular health problem, is substantially more difficult. Estimation of WTP for decreases in very specific
health risks (e.g., WTP to avoid one day of coughing of WTP to avoid admission to the hospital for
respiratory illness) is further limited by a paucity of information. Derivation of the dollar value estimates
discussed below was often limited by available information. In the case of hospital admissions, for
example, cost-of-illness estimates combined with opportunity cost estimates were used in lieu of WTP
because of the lack of other information regarding willingness to pay to avoid such hospital admissions.
These estimates are likely to understate total WTP to avoid a hospital admission because they do not
include the value of avoiding the pain and suffering associated with the illness for which the individiial
entered the hospital. WTP to avoid a day of specific morbidity endpoints, such as coughing or shortness of
breath, has been estimated by only a small number of studies (two or three studies, for some endpoints;
only one study for other endpoints). The estimates for health endpoints involving these morbidity
endpoints are therefore similarly based on only a few studies. However, it is worth noting that the total
benefit associated with any reduction in pollutant concentrations is determined largely by the benefit
associated with the corresponding reduction in mortality risk because the dollar value associated with
mortality is significantly greater than any other valuation estimate.
-5 - .
Estimates of WTP may be understated for a couple of reasons. First, if exposure to pollution has
any cumulative or lagged effects, then a given reduction in pollution concentrations in one year may confer
benefits not only in that year but in future years as well. Benefits achieved in later years are not included.
Second, the possible effects of altruism are not considered in any of the economic value derivations.
Individuals' WTP for reductions in health risks for others are implicitly assumed to be zero.
The following Exhibit summarizes the derivations of the economic values used in the analysis.
373
-------
I
1
>
2
<§
1
II
s °"i
*n „
"53 H g}
X) C T3
"3 S iv i
a.
•e
o
o >»
VI J3
1
09
1
-------
2
i
§
s
1
I
••§
i
1
a
I
.1
Iga
•a
-------
l
13 ^S
111
2 8.3
TS O S3
iff
I
e
o
i
a «
1
£ 8
.o
I
b
-------
I
t
1
|
K
I
I
I
§
^
vd
S S *»
••in "
2 " S
i s"°
J 1'
2 E tS
oo
I
ig
11
«
?
J5 o oo
•c « oo
i o"*1*
'•o «» ".
"3 n §
S S 1
.1
•o
'It
1£
g*
|
li
.8>>
g
g
"S
a>
MS
III
6 '•§ 5
00
2 £
I
3
MH
S-
I1
«'I a *! I
IffISi
g •— s 5
S
-------
rtainty DIstribu
1
3 O CJ
•§2 1
ail
8
v»
Ce
">
1
to
ea _ "S .S •"
|18-|.|
- H .&^ "g
Assuming that
d as some com
ility, LRI with
ility. The $ val
eights 0.40, 0.4
I + LRI.
Cen
cha
40
20
ti ts-^
"-'I:
sal
JS
13 J
§ -g •§
a a a $ 3 H
c4 "SSI
«*>.
oo
H
fl
W3
£
VO
en
-------
I
4)
2
i
8.
w
I
o
i—i
«»
It
3 O
oa e
II
ii.
- 09 -2
>B.S
609
••S
13
I
•s
1
CO
od
00
-------
o ? w S rv «=
2 1 -g « Sf 3
llflll
i 111 si
per
. Th
b
i
yeld an estimate of $2,995
yield an estimate of $4,067.
ed effects on lifetime rnin
ent, and (2) indirectly
labor force participa
4 y
y
(1994
(1995) y
combine
.decreme
reduced
§
i
&
I
§
1
8
i
^
e
1
s
S.
TH
&I.S
3 B.
available
V
g
o
1
o
T3
1
n
fies
har
lity.
none available
I
H
^H
oo
1 S 'g S ^ .2 -2
« E •§ .S H € S
in the aggregatio
holdper#g/m3P
I •§
a <-•
§
0
o
oo
CO
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Results of Valuation of Health and Welfare Effects
Table 115 presents the results of combining the economic valuations described in this Appendix
with the health and welfare effects results presented in Appendix D. As noted in Appendix D^ there are
alternative estimates for some health and welfare impacts, which form the basis of several alternative
benefit estimates. Each of the health effects estimates also has quantified statistical uncertainty. The
range of estimated health and welfare effects, along with the uncertain economic unit valuations, were " •
combined to estimate a range of possible results. The combining of the health and economic information
used the Monte Carlo method presented in Chapter 7. Table 115 shows the mean estimate results, as well
as the measured credible range (upper and lower five percentiles of the results distribution), of economic
benefits for each of the quantified health and welfare categories.
The results for aggregate monetized benefits were also calculated using a Monte Carlo method.
The results of the Monte Carlo simulations for the economic values for each of the major endpoint
categories are presented in Table 116. Note that for the upper and lower fifth percentiles the sum of the
estimated benefits from the individual endpoints does not equal the estimated total. The Monte Carlo
method used in the analysis assumes that each health and welfare endpoint is independent of the others.
There is a very low probability that the aggregate benefits will «qual the sum of the fifth percentile benefits
from each of the ten endpoints.
Table 117 shows the estimated total benefits ranges for the four modeled target years of this study:
1975,1980,1985, and 1990. The results of the Monte Carlo simulations of the aggregate economic
benefits for these four target years are depicted in Figure 21.
-':S™. .,..,.„- "t
Table 118 examines the impact of limiting the scope of the analysis to locations with more certain
air quality estimates. The main analysis (as shown in Tables 115 through 117) covers almost the entire
population of the~48 States.293 B6wever,1the air quality information is less certain for locations far from a
monitor. Table 118 presents the results of limiting the analysis to people living within 50 km of an ozone,
NOa, SOa, or CO monitor, or in counties with a PM monitor. The availability of monitors changes over
time. Hence'the proportion of the population included in this analysis changes over time as well. Table
118 indicates that approximately a quarter of the total benefits estimated in the main analysis comes from
areas with less certain air quality information.
The results of the "all U.S. population" analysis provides a more accurate depiction of the pattern
of economic benefits across years. The accuracy of the scale of incidence is less certain. These results
provide a better characterization of the total direct benefits of the Clean Air Act in the lower 48 states than
do the "monitored area only" results because the latter completely omits historical air quality
improvements for about 25 percent of the population. However, the "all U.S. population" results rely on
uncertain extrapolations of pollution concentrations, and subsequent exposures, from distant monitoring
sites to provide coverage for the 25 percent or so of the population living far from air quality monitors.
Thus, the main results presented in Tables 115 through 117 include important uncertainties.
293 Except for lead, two to five percent (depending on pollutant) of the population who live in sparsely populated areas are excluded from the
main analysis to maximize computer efficiency. All of the population of the 48 states is included in the lead analysis.
381 ,
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria. Pollutants
1- I ^
w . ' . ', « , - - i .1 i' i '"' , 1 n. . . " ' H • " '1
T«Wc 115. Oiteris PoUutanls Htalth ind Welfare Benefits - Extrapolated to Entire 48 State Population
1 i: I ("Present ^ilue (in 1990 usine S% discount rate) of benefits fiom 1970 - 1990 (in billions of 1990 dollars).
Mortality
•Mortality (Jong-term exposure)
-'Mortality (short-term exposure)
; Mortality (Lead exposure)
| Chronic Bronchitis
OfterLead-faduced Aflments
' LostKJPomJs
- Ki<70
i Hypeiteauou
Coronary Heart Disease
[; AtberoUuombotic brain infarction
Iaiti»J cetebrovascular accident
i Hospital Admissions
' ^AH Respiratory
| *'COPD + Pneumonia
; Bcfcemfc Heart Disease
t ' - •
* Congestive Heart FaBore
j CHberReq)iratoty-ReIated Ailments
' -- Children ;
"'• Shortness of breath, days
{ ***Acufe Bronchitis
*"Upper& Lower Respiratory Symptoms
j AdaUs ' -
! Any of 19 Acute Symptoms
*; -•- AB
Asthma Attacks
Increase in Respiratory Illness
! Any Symptom
Restricted Activity and Work LossDays
MRAD
- Work Loss Days (WLD)
; Hainan Welfare
HousehoW Sofliog Damage
• VkfcQky- Eastern US.
'! Decreased Worker Productive
Agriculture (Net Surplus)
j
PM-10
* PM-10
Lead
PM-10
Lead
Lead
Lead
Lead
Lead
Lead
PM-10 &O3
PM-10 &O3
PM-10
PM-10 & CO
K
PM-10
PM-10
PM-10
PM-10 &.O3
PM-10 & 03
NO2
S02
PM-10 &O3
PM-10
PM-10
particulates
O3
O3
Present Value (billions of 1990$>
5th %0e
t S f *
, $3,040
$820
1 .^ v
$2,145 !
1 f* i
" $275,
$22 "*"
, t$77'
->' '$1
i. >$0 -
$0 "
, f
$8
$8
- $1 '
$3' '
-'
$0
$0
<*1 M
A
$7
' '$0
$1
$0
*
$2
',
, r * ,
$55 '
$5
$11
Mean'
1 $21,933
$5,340
/ $t'SSO »
lt $7,146 t
> ' \ 'i * >• ^
$438 '
'' ' ' $"26-
'J >
* $4V
i ! rf 1 is-^i^
tl
! '" •" ''$1
i. ,-
•V^ '"''i ' '
!-$9
$4'
. " «, ;,:
1 ' V, 1*
1 1 " #" 1 ^
^,$7'
* t - , "^
« ^ |$2\ /(
J ^ {
y:b <"*• J
• **
- K t J'
$1
$2
' * , V
i $e'! ''
1 '$2 '
e r v y < is
I f f i
$75
$72 ,
, " $5" s
$23
95th %Ue
< ,, ' ' V ,
$51,500
' $11^863
; , "«y*6
•* - 1
•k
! ^
' ^ m^1
J '$120
* iV f
-„ ', ^ *?
^ f
/ -^ _,
k l ; V ( ^ 1.
^, ^ V"1
;
. s'''j ^
/ 2$
>\ i t
! '|X !•* * t1
" , ^
'< '.»7
t •• $18
' * ' T '^
i>( i !
* '! ' ''f ti
if ' *l L
t r 4
.' '$2
' " ^
- ' ','^ ' »^
4, I I *
, ;' ' ?-^
43
'" % ji
•i M$!^
, ,>|
* i
< $34
, ,
To avoWo^bte-coonting of benefits, the following endpoints were treated as alternatives: , '
^Mortality beacSt&eslimated by the 19 short-term exposure studies were pooled with the single long-term mortality study,
aligning equal weight to each study type. > '* ,'< 5 " ''f
1, "Hospital admissions foe COPD combined with those for pneumonia are treated as an equally-weighted alternative to hospital
..; , .admhsio^loraareqwatotyillnesses. J ' ' , ' •, ' f\
••The defuutioos of acute broncbitis and upper and tower respiratory illness overlap; both studies count trouble breathing,
dry cough, and wheezing in their estimates. These two studies are treated as alternatives, which reflects the variability of
pollution-induced respiratory effects in children. <- I '
V
j i
382
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
-Table lift. Present Value of 1970 to 199Q Monetized Benefits by Endpolnt Category for 481
- State Population and Total Monetized Benefits (discount rate of~5 percent) " y- "
-^ „* . Endpoint , 1 - , Pollutants)
Mortality » ,- <• .7. " y »PM40, ,
Mortality •> ~ , pb
Chronic Bronchitis „ 5 v s _ PM-10
[Q (Us* IQ Pfc. + Children'^/ IQ<70X Pb
Hypertension ; , Pb ,
Hospital Admissions - PM-10, 03, Pb, £ CO
Respiratory-Related Symptoms, Restricted PM-10, O3, NO2, &
Activity, & Decreased Productivity N ,SO2
Soiling Damage , , , x PM-10
Vfiibiiify , ,!,. %particulates
Agriculture (Net Surplus) , O3
TOTAL{$BilIions'>
, ; 'Present Value ', ^
5th %ile
$l,99i
$125
. -$2443
'$299
577
$18
$31
$7
1 $55
$11
$10,500
Mean
$13,542
' $1^550
„ $7,1561
' $456
$99
$24
J$76
$76
$72
$23 i
$23000
95th %ile
$3Gi,968
^$4,096
$12,613
$656
' M$i
• &
$149
** V
$196
^$9t
$35
•• $40600
Table 117. Monte Carlo Simulation Model Results for Target Years, Plus Present Value in 1990 Terms ofTbta!
Monetized Benefits for Entire* 1970 to 1990 Period (in billions of 1990-value dollars).
Total Benefits Bv Year 'f $Billionsl
Sthpercentile
Mean
95th oercentile
i *
1975
$?61
$364
' .$655
1980
$431
$957
$1,710
1985
$558
$1,204
1 $2,075
1990
$623
$1,318
$2317
Present Value (5%}
$10,500
$23,000
$40600
Pre&nt value-reflects compounding of benefits &om 1971 to 1990. "
••- s ' > V •.';';.' "\
"Uncertainty Estimates'1 aie results of Monte Carioanalysis combining economic and physicaJ effects uncertainty (i.e., using bolh
between-and wlhJn-sJudy variability). - , "
' Full nncettainty analysis done only for years shown. Uncertainty estimates for intermediate years computed based on ratios of 5th to 50th percemHe and 95th to 50th j
' Ratios interpolated between yearsshown and applied to point estimates for intermediate years. ,
383
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Figure 54. Monte Carlo Simulation Model Results for Target Years (in billions of 1990 dollars)
$2,500 j
$2,000 -
?
i$1,500 -
5
^$1,000 -
*B
c
2 $500 -
CS
*3
£
•
$o J
mm^^m
IMMHMH
'
. ?4 95th%
— ~i^ 95th%
•4 95th%
^ Mean
MPHM
"
*
MMMMM
^ Mean
.
' ~
^~*
^
- :
^-
„ J
'
4 Mean
^ 5th%
"0*
-rl
^
Z-\
~:~
1
^.^
**
MMMM
|4 95th%
WMean
4 5th%
•4 5th%
^^ . ,
4 5th%
1975 1980 1985 1990
384
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
* '» ' ~ ' < „ •< 4' > > I ' ,'<<>" - *
Table 118. Comparison of 1990 (Single Year) Monetized Benefits by Endpoint for 48 State Population and
Momtored Areas (in miinbns of 1990 dollars), - , , '
> •• ' , ' *
/ , ' J. ' 1
• ,,. Endpoint
Mortality " \ - V ^ _ ?.
Mortality -"" /'
Chronic Bronchitis
Q (Lost |Q Points * Children with iij < 70)
lypertension - , ~ -'
Jospital Admissions ,•,•''- ;
lespkatory-RelateXI Symptoms, Restricted
'- , Activity, & Decreased Productivity
JoilingDamage "',/ '"' "/ *
^isnjBity' ^ ; '", \ -,,
\griculture (Net Surplus) - ,
Polluiant(s)
PM-10
Pb
* PM-10
Pb
Pb
PM-10, 03,Pb,& CO
PM-10, 03,NO2, &SO2
, PM-10
- particulates
' ' ' - ' - -" TOTAL ^Millions')
Mean Estimate of Monetized Benefits
fmiUions of 1990-value dollars)
48 State Pop. Monitored Areas*
$737,408 ' x , $465,905
$131,752 ' $131,752
$398,042 * $259^870
$37,639 $37,639
'$8,695 f - $8,695
$1,566 , $1,154
$4/00 $3,ffi29
If
$4,137S ' J $2,718
$4^79 $4,479
$^91 , $991
$1.317,993 > sr $905 257
* Monitored: area/are ^ose within 50 km of an O3, NO2, SO2, or CO monitor or a PM-monitored county. The "48 >
State Population" modeling estimate captures Benefits for populations in unmonitored areas. Air pollution'
concentrations in these areas are ^signed' based on concentrations measured at the closest monitor, for O3, NO2, -
SO2, and CO. PM-10 concentrations in unmonitored counties are derived from those in monitored counties.
Sensitivity Analyses
The uncertainty ranges for the results on the present value of the aggregate measured monetary benefits
reported in Table 115 reflect two important sources of measured uncertainty:
• uncertainty about the avoided incidence of health and welfare effects deriving from the concentration-
response functions, including both selection of scientific studies and statistical uncertainty from the
original studies; and
• uncertainty about the economic value of each quantified health and welfare effect.
These aggregate uncertainty results incorporate many decisions about analytical procedures and specific
assumptions discussed in the Appendices to this report.
385
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
In order to provide a more complete understanding of the economic benefit results in Table 115,
sensitivity analyses examines several additional important aspects of the analysis. First, this section explores the
effect of selecting alternative discount rates on the aggregate present value benefits estimation. Second, this
section examines the sources of the measured aggregate uncertainty, identifying which of the measured
uncertainty components of incidence and valuation for individual health effects categories drive the overall
uncertainty results. Third, this section examines several issues involving the estimated economic benefits of
mortality.
The main analysis reflected in present value results shown in Table 115 uses a five percent discount rate.
The discount rate primarily enters the calculations when compounding the economic benefits estimates from
individual years between 1970 and 1990 to estimate the present value of the benefits in 1990. The discount rate
also directly enters in the calculation of the economic value of an IQ point.294 There is considerable controversy
in the economics and policy literature about the appropriate discount rate to use in different settings. Major
alternatives recommended by various authors include a discount rate based on the social discount rate (typical
estimates are in the 2 to 3 percent range), and a discount rate based on the risk-free rate of return on capital
(typically in the 7 to 10 percent range). Table 119 presents the aggregate uncertainty results using three different
discount rates: 3%, 5% and 7%. While the aggregate benefits estimates are sensitive to the discount rate,
selecting one of these alternative discount rates affects tfae^ggej*ate_b^^
jlj'abie 'lisJ. "Effect of j&ernative Discount Rates on Present Value ofTotal Monetized Benefits for 1970 to 199Q , ^ ^
Present Value in 1990 of Total
Benefits ($Trillions of 1990 Dollars')
5th perccntile
Mean
95th percentile
ll *P Mil 11 1 Ml
3%
$9.1
$19.9
$35.1
f
5%'
$103
"$23.0
$40.6
' i •*„ i -
" ' 7%
* $12.2
$26.7
$47.3
, i
1' -
>
rRtseotvttoereflectscompounding of benefits from 1971 to 1990. ' ' f' ,,»,,.•
in i , ' " i , * -> 14-
"* * ' ' I . > * *"&;
« j, , <- ' 'i1 - ' , "V ' ,.--H" .
The estimated uncertainty ranges in Table 115 reflect the measured uncertainty associated with both
avoided incidence and economic valuation. A better understanding of the relative influence of individual
uncertain variables on the overall uncertainty in the analysis can be gained by isolating the individual effects of
important variables on the range of estimated benefits. This can be accomplished by holding all the inputs to the
Monte Carlo uncertainty analysis constant (at their mean values), and allowing only one variable — for example,
the economic valuation of mortality — to vary across the range of that variable's uncertainty. The sensitivity
analysis then isolates how this single source of variability contributes to the variation in estimated total benefits.
The results are summarized in Figure 56. The nine individual uncertainty factors that contribute the most to the
overall uncertainty are shown in Figure 56, ordered by the relative significance of their contribution to overall
:M Tbc estimated economic value of lost IQ points due to lead exposure is based on the present value of the impact on lifetime earnings. A
discount rate is required to calculate that present value. The impact on income primarily occurs during adulthood, which is 20 to 70 years after the
initial lead exposure. This significant lag results in the discount rate having a significant impact on the estimated economic benefits of the IQ loss.
386
-------
Appendix!: Valuation of Human Health and Welfare Effects of Criteria Pollutants
uncertainty. Each of the additional sources of quantified uncertainty in the overall analysis not shown contribute
a smaller amount of uncertainty to the estimates of monetized benefits than the sources that are shown.
Figure 55. Uncertainty Ranges Deriving From Individual Uncertai^
$40
$35
| $30 -j
I $25 -I
$20 •
$15 •
$10 -
$5
$0
f
I
>-
_ 95th %ile
Mean
I 1 '
5th %ile
>> c
if
CD
Because of the multiple uncertainties in the benefits estimation, the total estimated present value of the
monetary benefits of the 197S5tb 1990 Clean Air Act range from a low of about $10 trillion to a high of about
$40 trillion (in 1990 dollars, discounted at five percent). Most of the uncertainty in the total estimated benefit
levels comes from uncertainty in the estimate of the economic valuation of mortality, followed by the uncertainty
in the incidence of mortality from PM (as a surrogate for all non-lead air pollution). The incidence of lead-
induced mortality also has a significant influence on the overall uncertainty. The importance of mortality is not
surprising, because the benefits associated with reduced mortality are such a large share of overall monetized
benefits.
The uncertainty in both the incidence and valuation of chronic bronchitis are the two other significant
factors driving the overall uncertainty range. The modeled uncertainty in the other remaining health and welfare
endpoints in the analysis contribute relatively small amounts to the overall uncertainty in the estimate of total
monetary benefits of the Clean Air Act. Most of these other endpoints account for a relatively small proportion
of the overall benefits estimates, making it unlikely mat they could contribute significantly to the overall
uncertainty. Estimates of either the mean values or standard errors of these variables are generally very small
relative to estimated total monetary benefits. .
387
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Because the economic benefits associated with premature mortality are the largest source of monetized
benefits in the analysis, and because the uncertainties in both the incidence and value of premature mortality are
the most important sources of uncertainty in the overall analysis, it is useful to examine the mortality benefits
estimation in greater detail. The analytical procedure used in the main analysis to estimate the monetary benefits
of avoided mortality assumes that the appropriate economic value for each incidence is a Value from the currently
accepted range of the value of a statistical life. As documented previously, the estimate value per predicted
incidence of excess premature mortality is modeled as a Weibull distribution, with a mean value of $4.8 million
and a standard deviation of $3.2 million. This estimate is based on 26 studies of the value of mortal risks.
Although there is considerable variation in the analytical designslnd data used in the 26 underlying studies, the
majority of the studies involve the value of risks to a middle-aged population. Most of the studies examine
differences in wages of risky occupations, using a wage-hedonic approach. Certain characteristics of both the
population affected and the mortality risk facing that population are believed to affect the average willingness to
pay (WTP) to reduce the risk. The appropriateness of a distribution of WTP estimates from 26 studies for
valuing the mortality-related benefits of reductions in air pollution concentrations therefore depends not only on
the quality of the studies (i.e., how well they measure what they are trying to measure), but also on (1) the extent
to which the subjects in the studies are similar to the population affected by changes in pollution concentrations,
and (2) the extent to which the risks being valued are similar.
If the individuals who die prematurely from air pollution are consistently older than the population hi the
valuation studies, the mortality valuations based on middle-aged people may provide a biased estimate of the
willingness to pay of older individuals to reduce mortal risk. There is some evidence to suggest that the people
who die prematurely from exposure to ambient particulate matter tend to be older than the populations in the
valuation studies. In the general U.S. population far more older people die than younger people; 88 percent of
the deaths are among people over 64 years old. It is difficult to Establish the proportion of the pollution-related
deaths that are among the older population because it is impossible to isolate individual cases where one can say
with even reasonable certainty that a specific individual died because of air pollution.
There is considerable uncertainty •whether older people will have a greater willingness to pay to avoid
risks than younger people. There is reason to believe that those over 65 are, in general, more risk averse than the
general population, while workers in wage-risk studies are likely to be less risk averse than the general
population. Nfbre risk averse people would have a greater willingness to pay to avoid risk than less risk averse
people. Although the list of recommended studies excludes studies that consider only much-higher-than-average
occupational risks, there is nevertheless likely to be some selection bias in the remaining studies — that is, these
studies ire likely to be based on samples of workers who are, on average, more risk-loving than the general
population. In contrast, older people as a group exhibit more risk averse behavior.
In addition, it might be argued that because the elderly have greater average wealth than those
younger; the affected population is also wealthier, on average, than wage-risk study subjects, who tend to be blue
collar workers. It is possible, however, that among the elderly it is largely the poor elderly who are most
vulnerable to pollution-related mortality risk (e.g., because of generally poorer health care). If this is the case, the
average wealth of those affected by a pollution reduction relative to that of subjects in wage-risk studies is
uncertain.
Although there may be several ways in which job-related mortality risks differ from air pollution-
related mortality risks, the most important difference may be that job-related risks are incurred voluntarily
whereas air pollution-related risks are incurred involuntarily. There is some evidence (see, for example, Violette
and Chestnut, 1983) that people will pay more to reduce involuntarily incurred risks than risks incurred
388
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
voluntarily. If this is the case, WTP estimates based oh wage-risk studies may be downward biased estimates of
WTP to reduce involuntarily incurred air pollution-related mortality risks.
Finally, another important difference related to the nature of the risk may be that 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.
. , f! " ' .
The direction of bias resulting from the age difference is unclear, particularly because age is confounded
by risk aversion (relative to the general population). It could be argued that; because an older person has fewer
expected years left to lose, his WTP to reduce mortality risk would be less than that of a younger person. This
hypothesis is supported by one empirical study, Jones-Lee et al. (1985), that found the value of a statistical life at
age 65 to be about 90 percent of what it is at age 40. Citing the evidence provided by Jones-Lee et al. (1985), a
recent sulfate-related health benefits study conducted for EPA*(Hagler Bailly Cclsuiting, 1995) assumes that the
value of a statistical life for those 65 and over is 75 percent of what ifis for those under 65.
I ' •'
There is substantial evidence that the income elasticity of WTP for health risk reductions is positive (see,
for example, Alberini et'al., 1994; MitcheUg^ Carson, 1986; Loehman and Vo Hu De, 1982; Gerking et al.,
1988; and Jones-Lee et al., 1985), althoi^fllere is uncertainty about the exact value of this elasticity).
Individuals with higher incomes (or grafter wealth) should, thenj be willing to pay more to reduce risk, all else
equal, than individuals with lower incomes or wealth. Whether the average income or level of wealth of the
population affected by PM reductions is likely to be significantly different from that of subjects in wage-risk
studies, however, is unclear, as discussed above.
Finally, there is some evidence (see, for example, Violette and Chestnut, 1983) that people will pay more
to reduce involuntarily incurred risks than risks incurred voluntarily. If this is the case, WTP estimates based on
wage-risk studies may be downward biased estimates of WTP to reduce involuntarily incurred pollution-related
mortality risks.
The need to adjust wage-risk-based WTP estimates downward because of the likely upward bias
introduced by the age discrepancy has received significant attention (see Chestnut, 1995; ffic, 1992). If the age
difference were the only difference between the population affected by pollution changes and thfc subjects in the
wage-risk studies, there might be some justification for trying to adjust the point estimate of $4.8 million
downward. Even in this case, however, the degree of the adjustment would be unclear. There is good reason to
suspect, however, tHapfiiere are biases in both directions. Because in each case the extent of the bias is unknown,
the overall direction of bias in the mortality values is similarly unknown. Adjusting the estimate upward or
downward to compensate for any one source of bias could therefore increase the degree of bias. Therefore, the
range of values from the 26 studies is used in the primary analysis without adjustment.
Examining the sensitivity of the overall results to the mortality values can help illuminate the potential
impacts of alternative mortality valuations. As mentioned above, a contractor study performed for EPA used one
approach to evaluate the economic value of sulfate-related human health improvements resulting from CAAA90
Title IV acid rain controls. That study assumed that 85 percent of the people dying from sulfates (an important
component of participate matter) were over 65, and that people over 65 have a willingness to pay to avoid a
389
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
mortal risk that is 75 percent of the values that middle-aged people have. Using this approach, the value of an
average statistical life (using a weighted averages) is reduced to 79 percent of the previous value.
An alternative mortality valuation approach emphasizes differences in the amount of remaining life
foregone, measured as life-years-lost (LYL), by individuals dying at different ages. Undermis approach
mortality in older people, with shorter life expectancies than younger people, will have a smaller number of LYL
than younger people. Implementing this approach requires three separate estimates: (1) the value of each life-
year-lost; (2) the number of people dying at each specific and age, and (3) the number of iife-years-lost by people
of specific ages dying of air pollution. At this time, the state of the epidemiological science and economic
valuation science are insufficiently advanced to allow confident use of the results of this method. However, the
following section briefly discusses each of these issues, and suggests possible approaches for each factor. In spite
of the substantial uncertainties and lack of available information, this section presents an example of a
preliminary estimate of the present value of avoided premature mortality using this approach.
Moore and Viscusi (1988) suggest one approach for determining the value of a life-year-lost. Assume
that the reported willingness to pay to avoid a mortal risk is the sum of the value of a single year times the
number of years of expected life remaining for an individual. They suggest that a typical respondent in a mortal
risk study may have a life expectancy of an additional 35 years. Using a mean estimate of $4.8 million, their
approach would yield an estimate of $137,000 per life-year-lost. A variation on this approach assumes that the
mean value of a mortal risk is the present value of the value for each additional life year. If an individual
discounts future additional years using a standard discounting procedure, the value of each life year lost must be
greater than the value assuming no discounting. Using a 35 year life expectancy, a $4.8 million value of a
statistical life, and a 5 percent discount rate, the implied value oflach life year lost is $293,000. The Moore and
Viscusi procedure is identical to this approach, but uses a zero discount rate.
Using the'LYL approach, the estimates of the value of a statistical life' vary by age. For example, if the
mean value of avoiding a statistical death is $4.8 million for a person with a life expectancy of 35 additional
years, the resulting value using a five percent discount rate for a person between 65 and 74, who has a life
expectancy295 of 14 years, is 60% ($2.9 million) as much. Using a zero percent discount rate, the value is 40%
($1.9 million) as large. Older people, who have shorter life expectancies, have smaller values. For comparison,
the approach used in the acid rain benefits study is 63% as large for all people greater than 65.
The second component required to implement a life-years-lost approach is an estimate of the number of
deaths in each age cohort. In order to do this, it is preferable to have age-specific relative risks. Many of the
epidemiological studies do not provide any estimate of such age-specific risks. In this case, the age-specific
relative risks must be assumed to be identical.
Some of the available epidemiology studies on PM do provide some estimates about age specific relative
risks. The information that is available suggests that the increase in relative risk is greater for older people. Most
of the available information comes from short-term exposure studies. There is considerable uncertainty in
applying the evidence from short-term exposure studies to results from long-term (chronic exposure) studies.
Moreover, the available studies offer different estimates of the relationship between the relative risks at various
ages, and some of the information (such as risks decreasing for people ages 65 to 74) is counter-intuitive.
However, using the available information on the relative magnitudes of the relative risks, it is possible to form a
preliminary assessment of the relative risks by different age classes.
An life expectancies and death rates used in this analysis come from the 1992 U.S. Vital Statistics.
390
-------
- Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
The analysis presented below uses two alternative assumptions about age specific risks; (1) there is a
constant relative risk (obtained directly from the health literature) that is applicable to all age cohorts, and (2) the
relative risks differ by age, as estimated from the available literature. The age-specific relative risks used in the
example below assume that the relative risks for people under 65 are only 16% of the population-wide average
relative risk, the risks for people from 65 to 74 are 83% of the population-wide risk, and people 75 and older have
a relative risk 55 percent greater than the population average. Details of this approach is'provided in Post and
Deck,(1996).
The third component of the life-years-lost approach is an estimate of the number of life years lost by a
person dying prematurely at each given age. The approach developed for this analysis2* assumes that exposure
to elevated levels of PM increases the probability of dying at a specific age. Increasing the probability of dying at
each age (using the age-specific relative risks) lowers the life expectancyfor each age cohort. The average
number of life-years-lost will depend on the distribution of ages in the population in a location. Using identical
age specific relative risks and data for Los Angeles County, each premature mortality is a loss of 15.6 years of life
(on average across all age cohorts). Using the age-specific relative risk estimates developed for this analysis,
each premature mortality is a loss of 9.8 years.
The present value benefits estimates for PM-related mortality using these alternative approaches are
shown in Table 120. Table 120 is based on a single health study: Pope et al., 1996. Alternative studies, or the
uncertainty approach used in the primary analysis, would result in a similar pattern of the relationship between
valuation approaches.
Table'i^/ Alternative Estimates of the Present Value o
(based on Pope et aL/1996, pi trillions of 199Q dollars):
Itn PM,
, .
Palliation Procedure -"''
\ssiiming 85$ of people are *64,'vwtn a value of a statistical life = 75% „
ofaveisjge value "" ^ , \^:_ ^ '",. °" '""''''""' \ - >' ,
Jfe Years Lost approaches ~ ~ ''''•" ' " „ " ' ,
,s.> N " i x , -<• • t \ f>" ', -, - / <. s
Sia^i relative jrisk» yalaation using :5%idiscounting
Approximate age-specific relative iisk, valuation using 5% discounting
Approximate aee-soecific relative risk, valuation using 0% discouhtine
Present Value ofPM
ae reflects compounding of benefits from 1971 to 1990, using a S percent discount rate..
$21.0
$16.5
$12.4
,$9:5 '
61
The life-years-lost approach develops two important aspects of uncertainty in the mortality estimates used
in the primary analysis. There is another important uncertainty about these estimates. The life-years-lost
216 The basic approach developed here is identical to the approach used in an EPA study of the Phase I NO, emission provisions of the acid
rain bill (EPA, 1994). That approach, based on unpublished work by John Evans, estimated 11.7 life years lost per death associated with pollution.
391
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
approach used here assumes that people who die from air pollution are typical of people in their age group. The
estimated value of the quantity of life lost assumes that the people who die would have had an average life
expectancy. However, it is possible that the people who die from air pollution are already in ill health, and that
their life expectancy is less than a typical person of their age. If this is true, than the number of life years lost per
PM-related death would be lower than calculated here, and the economic value would be smaller.
The extent to which adverse effects of particulate matter exjosure are differentially imposed on peopte of
advanced age and/or poor health is one of the most important current uncertainties in air pollution-related health
studies. There is limited information, primarily from the short-term exposure studies, which" suggests tiuf it least
some of the estimated premature mortality is imposed disproportionately on people who are elderly and/or of poor
health. The Criteria Document (EPA, 1996) identifies only two studies which attempt to evaluate this
disproportionality. Spix et al. (1993) suggests that a small portion of the'PM-associated mortality occurs in
individuals who would have died in a short time anyway. Cifuentes and Lave (1996) found that 37 to 87 percent
of the deaths from short-term exposure could have been premature by only a few days, although their evidence is
inconclusive. . ;
Long-term studies provide evidence that a portion of the loss of life associated with long-term exposure is
independent of the death from short-term exposures, and that the loss of life-years measured in the long-term
studies could be on the order of years. Estimates have been made that the chronic exposure studies suggest a
loss of life on the order of 1 to 2 years (Lippmann, 1995), but there are no published results derived from the data
used in the chronic exposure studies. ' :
If much of the premature mortality associated with PM represents short term prematurity of death
imposed on people who are elderly and/or of ill health, the estimates of the monetary benefits of avoided
mortality may substantially overestimate society'si totaliwiulngness to pay to avoid particulate matter-related
premature mortality. On the other hand, if the premature mortality measured in the chronic exposure studies is
detecting excess premature deaths which are largely independent of the deaths predicted from the short term
studies, and the disproportionate effect on the elderly and/or sick is modest, the benefits measured in this report
could be underestiinates of the total value, ^tibis time there is insufficient information from both the medical
and economic sciences to satisfactorily resolve these issues from a theoretical/analytical standpoint. Until there is
evidence from the physical and social sciences which is sufficiently compelling to encourage broad support of
age-specific values for reducing premature mortality, EPA will continue to use for its primary analyses a range of
values for mortaltiy risk reduction which assumes society values reductions in pollution-related premature
mortality equally regardless of who receives the benefit of such protection.
392
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
Economic Valuation References
Chestnut, L.G. 1995. "DoUars and Cents: The Economic and Health Benefits of Potential Particulate
Matter Reductions in the United States." Prepared for the American Lung Association.
Currier, R. 1995. Abt Associates Inc. Memorandum to Jim Neumann, Industrial Economics, Inc. and Jin
DeMocker, USEPA. Applicability of existing cost of illness values for use in the Section 812 :
benefits analysis and recommendations regarding valuing chronic cough using cost of illness
estimates. May 16.
Dickie, M. et al. 1991. Reconciling Averting Behavior and Contingent Valuation Benefit Estimates of
Reducing Symptoms of Ozone Exposure (draft), as cited in Neumann, J.E., Dickie, M.T., and R.E.
Unsworth. 1994. Industrial Economics, Incorporated. Memorandum to Jim DeMocker, U.S.
EPA, Office of Air and Radiation. March 31.
Hagler Bailly Consulting, Human Health Benefits from Sulfate Reductions Under Title IV of the 1990
Clean Air Act Amendments, submitted to U.S. Environmental Protection Agency / Office of Air
and Radiation / Office of Atmospheric Programs / Acid Rain Division, November 1995.
Industrial Economics, Incorporated (ffic). Memorandum to Jim DeMocker, Office of Air and Radiation,
Office of Policy Analysis and Review, U.S. Environmental Protection Agency, November 6, 1992.
Jones-Lee, MlW., et al. "The Value of Safety: Result of a National Sample Survey." Economic Journal
95:49-72. March 1985:
Krupnick, AJ. and M.L. Cropper. 1992. *The Effect of Information on Health Risk Valuations," Journal
of Risk and Uncertainty 5(2): 29-4f*
Krupnick, AJ. and RJ. Kopp.^The Health and Agricultural Benefits of Reductions in Ambient Ozone in
the United States." Resouces for the Future Discussion Paper QE88-10, Washington, D.C.
August 1988.
Moore, MJ. and W.K. Viscousi. 1988. "The Quantity-Adjusted Value of Life". Economic Inquiry
26(3):369-388.
Neumann, JdE. and R.E. Unsworth. 1993. Industrial Economics, Inc. Memorandum to Jim DeMocker,
„ U.S. EPA, Office of Air and Radiation. Revisions to the proposed value of life methodology for
the Section 812 retrospective. May 3.
Neumann, J.E., Dickie, M.T., and R.E. Unsworth. 1994. Industrial Economics, Incorporated.
Memorandum to Jim DeMocker, U.S. EPA, Office of Air and Radiation. Linkage between health
effects estimation and morbidity valuation in the Section 812 analysis — draft valuation document.
March 31.
Post, Ellen and L. Deck. 1996. Abt Associates Inc. Memorandum to Tom Gillis, U.S. EPA, Office of
Office of Policy, Planning and Evaluation, U.S. EPA. September 20, 1996.
,393
-------
Appendix I: Valuation of Human Health and Welfare Effects of Criteria Pollutants
U.S. Department of Commerce (USDOC), Economics and Statistics Administration. 1992. Statistical
Abstract of the United States, 1992: The National Data Book. 112th Edition, Washington, D.C.
' " ' .,&:>
U.S. Environmental Protection Agency (USEPA). 1994. The benefits of reducing emissions of nitrogen
oxides under Phase I of Tide IV of the 1990 Clean Air Act Amendments. Draft Final Report.
Prepared by National Economic Research Associates, Inc. for the Air Policy Branch, Office of
Policy, Planning and Evaluation, U.S. EPA. January 21.,
U.S. Environmental Protection Agency (USEPA). 1995. The impact of the Clean Air Act on lead
pollution: emissions reductions, health effects, and economic benefits from 1970 to 1990 (draft):
* Prepared by Abt Associates Inc. for Economic Analysis and Innovations Division, Office of
Policy, Planning and Evaluation, U.S. EPA. ContracTNo. 68-D2-0175, W.A. 3-05T"January 19.
m ,„ ~ _
U.S. Environmental Protection Agency (USEPA). 1995. Human Health Benefits From Sulfate
Reductions Under Title IV of the 1990 Clean Air Act Amendments. Prepared by Hagler Bailly
Consulting, Inc. for the Office of Atmospheric Programs, Office of Air and Radiation, U.S. EPA.
November 10.
Unsworth, R.E., and I.E. Neumann. 1993. Industrial Economics, Lie. Memorandum to Jim DeMocker,
U.S. EPA, Office of Air and Radiation. Review of existing value of morbidity avoidance
estimates: draft valuation document. September 30.
Unsworth, R.E., Neumann, J., and WlEl Browne. 1992. Memorandum to Jim DeMocker, U.S. EPA,
Office of Air and Radiation. Review of existing value of life estimates: valuation document.
November6.
Violette, D.M. and L.G. Chestnut 1983. "Valuing Reduction in Risks: A Review of the Empirical
Estimates." Report prepared for tiie U.S. Environmental Protection Agency, Washington, D.C.
EPA-230^05-83-602: ;
Viscusi, W.KL Fatal Tradeoffs: Public and Private Responsibilities for Risk. (New York: Oxford
University Press). 1992
Viscusi, Kip W. andW. Evans. "Utility Functions that are Dependent on One's Heath Status." American
Economic Review 1990.
Viscusi, W.K., Magat, W.A., and Huber, J. 1991. "Pricing Environmental Health Risks: Survey
Assessments of Risk-Risk and Risk-dollar Tradeoffs," Journal ofEnvironmetnal Economics and
Management 201:32-57.
394
-------
Appendix J: Future Directions
Research Implications
In virtually any benefit analysis of environmental issues, the state of scientific information limits
the degree of coverage possible and the confidence in benefit estimation. For most benefit categories,
further scientific research would allow for a better quantification of benefits. One of the major outcomes
of the retrospective analysis is a clear delineation of the major limitations in the scientific and economics
literature in carrying out an analysis of this scope. Often, a list of research needs is generated in studies
such as this, but there is no clear internal mechanism to address these needs. With this study (and the
ongoing Section 812 program), a process has been initiated where identified research needs are to be
integrated into EPA's overall extramural research grants program, administered by the Office of Research
and Development, particularly in the areas of air toxics and ecological effects. It is hoped that the research
projects that flow from this process will enable future analyses to be more comprehensive.
Certain of the limitations in the retrospective analysis are directly related to the historical nature of
the analysis, such as sparse information about air quality in the early 1970's in many areas in the country.
Other important limitations are related to the effects of elevated airborne lead concentrations, which has
beenvirtaaUyelirm^atedbymeremdVafofleadfromgasoUne. A better understanding of these
relationships would improve our understanding of the historical impact of the Clean Air Act, but would
only indirectly contribute to developing future air pollution policy. However, most of research that will
reduce the major gaps and uncertainties needed to improve the Section 812 analyses will be directly
relevant tcTEPA's primary ongoing mission of developing and implementing sound environmental'policies
to meet the national goals established in fee Clean Air Act and other legislation.
There are a number of areas of biological, physical and economic research areas that the EPA
Project Team identified as particularly important for improving future Section 812 analyses. The
following discussion is not an evaluation of the current state of scientific knowledge designed to identify
specific areas where a modest amount of targeted research efforts would likely yield important and useable
information in the near future. Instead, the research topics discussed below are likely to have a significant
impact on the overall estimates of the costs and benefits of the Clean Air Act. Significant impacts may
arise from research involving two basic types of information gaps: cost and benefit categories that are not
"currently quantified in the analysis, and reducing the uncertainty in the costs and benefit categories that are
included.
One of the challenges in preparing the retrospective analysis was modeling the integrated
relationships between emissions of many different chemicals, the subsequent mixture of pollutants in the
ambient air, and the resulting health and welfare effects of simultaneous exposure to multiple pollutants.
A major portion of the uncertainty in the analysis derives from the limited current understanding of any
interactive (synergistic or antagonistic) effects of multiple pollutants. The need to better understand these
complex issues is not a limited scientific question to improve Section 812 analyses, but is the primary
focus of EPA's current activities (organized under the Federal Advisory Council Act (FACA) process) to
develop an integrated set of attainment policies dealing with ozone, participate matter, sulfur and nitrogen
' '395 > ' ' ' '
-------
Appendix J: Future Directions
oxides, and visibility. Further research on multi-pollutant issues is likely to substantially reduce an
important source of unmeasured uncertainty in the Section 812 analyses.
Effect categories that are not included in the quantified results of the retrospective analysis are
perhaps the research subjects with the greatest potential for significantly affecting future Section 812
analyses. It is very difficult to get a balanced overall view of the impact of the Clean Air Act when there
are effects (both costs and benefits) that are not reflected in the aggregate results. Whether it is a case of
known (or highly likely) effects that cannot be adequately quantified at this time because of certain
knowledge gaps, or the potential of widespread or devastating effects we are only beginning to understand,
omitted effects greatly expand the subjective uncertainty range. Research may also lead to a better
understanding that certain potential effects are not as likely to occur, or are not as damaging, as the current
state of knowledge suggests. Physical, biological and economic science research that helps fill these gaps
and narrow the uncertainty ranges will improve the EPA's ability to better depict all the effects of the
QeanAirAct
Examples of presently omitted or underrepresented effect categories which could have a
substantial effect on the aggregate results include health effects of air toxics, health effects of chronic
exposure to criteria pollutants, ecosystem effects, any long-term impact of displaced.capital on productivity
slowdown and redirected technological innovation, and potential social costs of transitional
unemployment
Although much of the emissions reductions that occurred between 1970 and 1990 as a result of the
CAA were aimed at attaining the criteria pollutant standards, the emissions of air toxics were substantially
reduced as well. Some air toxics Were deliberately controlled because of their known or suspected
carcinogenicity, while other toxic emissions were reduced indirectly due to control procedures aimed at
other pollutants, particularly ozone and participate matter. The current Section 812 analyses was able to
present only limited information on the changes in air toxic emissions. These knowledge gaps will be even
more serious for future Section 812 analyses, which will be examining the effects of an expanded air toxic
program under the CAA Title IE, as well as toxic emissions reduced by other sections of the Act. Existing
knowledge gaps that prevented a more complete consideration of toxics in the retrospective analysis
include methods to estimate changes in acute and chronic ambient exposure conditions nationwide,
concentration-response relationships linking exposure and health or ecological outcomes, and economic
valuation methods for a broad array of potential serious health effects such as renal damage, reproductive
effects and non-fatal cancers, and potential ecological effects of air toxics.
Scientific information about the effects of long-term exposure to air pollutants is just beginning to
emerge, but continues to be the subject of intense scientific inquiry. The relationship between chronic PM
exposure and excess premature mortality (Pope et al., 1995) included in the quantified results of the
retrospective analysis is one example of such research. However, many other potential chronic effects that
are both biologically plausible and suggested by existing research are not included. Research to identify
the relationship linking certain known or hypothesized physical effects (e.g., ozone's effects on lung
function or fibrosis) with the development of serious health effects (e.g., cardio-pulmonary diseases or
premature mortality), and the appropriate economic valuation of the willingness to pay to avoid the risks of
such diseases, would reduce the uncertainty caused by a major category of omitted health effects that could
have a significant impact on the aggregate benefits estimates.
396
-------
Appendix J: Future Directions
Future research concerning certain effects that are included in the retrospective analysis will also
help to substantially reduce the range of uncertainty that is reflected in the aggregate results. As described
in Chapter 7 and Appendix I, premature mortality is both the largest source of benefits and the major
source of quantified uncertainty in the retrospective analysis. In addition to the quantified uncertainty,
there is considerable additional unquantified uncertainty about premature mortality associated with air
pollution. Much of the information base about these relationships is relatively new, more is coming out
virtually daily, and there is substantial disagreement in the scientific community about many of the key
issues. EPA's Research Strategy and Research Needs document for paniculate matter, currently under
development, will address many of these scientific issues as they relate to PM. The following selection of
highly uncertain issues could have a significant impact on both the aggregate mortality benefits estimates
and the measured uncertainty range:
• The relationship of specific pollutants in the overall premature mortality effect, including the
individual or interactive relationships between specific chemicals (e.g., ozone, sulfates, nitrates,
and acid aerosols), and particle sizes (i.e., coarse, fine and ultra-fine particles).
\
• The degree of overlap (if any) between the measured relationships between effects associated with
short term exposures and effects from long term exposure.
• The confounding effect of changes in historic air pollution, including changes over time in both
pollution levels and the composition of the pollutant mix.
• The extent to which life spans are shortened by exposure to the pollutants, and the distribution of
ages at the time of death.
• The willingness to pay to avoid the risks of shortened life spans.
After premature mortality, chronic bronchitis is the next largest health effect benefit category
included in the retrospective analysis. There is considerable measured uncertainty about both the
incidence estimation and the economic valuation. Additional research could reduce uncertainties about the
level of the pollutants associated with the observed effects, the baseline incidence used to model the
changes in the number of new cases, and the correspondence between the definition of chronic bronchitis
used in the health effects studies and the economic valuation studies.
Further research on the willingness to pay to avoid the risk of hospital admissions for specific
conditions would reduce a potentially significant source of non-measured uncertainty. The Project Team
used "avoided costs" for the value of an avoided hospital admission, based on the avoided direct medical
cost of hospitalization (including lost wages for the employed portion of the hospitalized population).
Avoided costs are likely to be a substantial underestimate of the appropriate willingness to pay, especially
for such serious health effects as hospitalization for strokes and congestive heart failure, particularly
because they omit the value of avoided pain, suffering, and inconvenience. Furthermore, in addition to
hospitalization, there is evidence that some people seek medical assistance as outpatients. It is also likely
that there are additional people adversely affected by short-term air pollution levels who seek physician
services (but stop short of hospital admissions). Revised estimates of the appropriate economic value of
avoided hospitalization and other primary care medical services could increase the total economic benefits
of this cluster of health effects sufficiently that it could be a much larger portion of the aggregate benefit
total. .
397.
-------
Appendix J: Future Directions
In addition to research to improve the understanding of the impacts of changes in air pollution on
human health and well-being, further research on the non-health impacts could result in a better depiction
of the adverse impacts on a wide array of effect categories. Improvements in air quality have likely
resulted in improvements in the health of aquatic and terrestrial ecosystems and the myriad of ecological
services they provide, but knowledge gaps prevent them from being included in (he Section 812 analysis at
this time. Additional research in both scientific understanding and appropriate modeling procedures could
lead to including such benefits as improvements in water quality stemming from a reduction in acid
deposition-related air pollutants. Such improvements would benefit human welfare through enhancements
in certain consumptive services such as commercial and recreational fishing, in addition 1o non-
consumptive services such as wildlife viewing, maintenance of biodiversity, and nutrient cycling.
Similarly, increased growth, productivity and overall health of ILLS, forests could occur from reducing
ozone, resulting in benefits from increased timber production, greater opportunities for recreational
services such as hunting, camping, wildlife observation, and nohuse benefits such as nutrient cycling,
temporary CO2 sequestration, and existence value. While (here is insufficient information to adequately
model the short-run ecological and ecosystem impacts, even less is known about the long-run effects of
prolonged exposure. Permanent species displacement or altered forest composition are examples of
potential ecosystem effects that are not reflected in the current monetized benefit analysis, and could be a
substantial source of additional benefits.
Future Section 812 Analyses
This Retrospective Study of the benefits and costs of the Clean Air Act was developed pursuant to
Section 812 of the 1990 Clean Air Act Amendments. Section 812 also requires EPA to generate an
ongoing series of prospective studies of the benefits and costs of the Act, to be delivered as Reports to .
Congress every two years. ?
Design of the first Section 812: Prospective Study commenced in 1993. The EPA Project Team
developed a list of fey analytical design issues and a "strawman" analytical design reflecting notional
decisions with respect to each of these design issues.297 The analytical issues list and strawman design was
presented to the Science Advisory Board Advisory Council on Clean Air Compliance Analysis
(ACCACA), the same SAB review group which has provided review of the Retrospective Study.
Subsequently, the EPA Project Team developed a preliminary design for the first Prospective Study. Due
to resource limitations, however, full-scale efforts to implement the first Prospective Study did not begin
until 1995 when expenditures for Retrospective Study work began to decline as major components of that
study were completed.
As for the Retrospective, the first Prospective Study is designed to contrast two alternative
scenarios; however, in the Prospective Study the comparison will be between a scenario which reflects full
implementation of the CAAA90 and a scenario which reflects continued implementation only of those air
pollution control programs and standards which were in place as of passage of the CAAA90. This means
that the first Prospective Study will provide an estimate of the incremental benefits and costs of the
CAAA90.
"* Copies of UK Prospective Study planning briefing materials are available from EPA.
398
-------
Appendix J: Future Directions
The first Prospective Study is being implemented in two phases. The first phase involves
development of a screening study, and the second phase will involve a more detailed and refined analysis
which will culminate in the first Prospective Study Report to Congress. The screening study compiles
currently available information on die costs and benefits of the implementation of CAAA90 programs, and
is intended to assist the Project Team in the design of the more detailed analysis by providing insights
regarding the quality of available data sources and analytical models, and the relative importance of
specific program areas; emitting sectors; pollutants; health, welfare, and ecological endpoints; and other
important factors and variables. , ""' _. Jp
In developing and implementing the Retrospective Study, the Project Team developed a number of
important modeling systems, analytical resources, and techniques which will be directly applicable and
useful for the ongoing series of section 812 Prospective Studies. Principal among these are the Criteria Air
Pollutant Modeling System (CAPMS) model developed to translate air quality profile data into quantitative
measures of physical outcomes; and the economic valuation models, coefficients, and approaches
developed to translate those physical outcomes to economic terms.
The Project Team also learned valuable lessons regarding analytical approaches or methods which
were not as productive or useful. In particular, the Project Team plans not to perform macroeconomic
modeling as an integral part of the first Prospective Analysis. In fact, there are currently no plans to
conduct a macroeconomic analysis at all. Essentially, die Project Team concluded, with confirmation by
the SAB ACCACA, that the substantial investment of time and resources necessary to perform
macroeconomic modeling would be better invested in developing high quality data on the likely effects of
the CAA on key emitting sectors, such as utilities, on-highway vehicles, refineries, etc. While the intended
products of a macroeconomic modeling exercise -such as overall effects on overall productivity, aggregate
employment effects, indirect economic effects-; are of theoretical interest, the practical results of such
exercises in the context of evaluating environmental programs may be disappointing.
First, the CAA has certainly had a significant affect on several industrial sectors. However, the
coarse structure of a model geared toward simulating effects across the entire economy requires crude and
potentially inaccurate matching of these polluting sectors to macroeconomic model sectors. For example,
the J/W model used for the Refrospective Study has only 35 sectors, with Electric Utilities comprising a
single sector. In reality, a well-structured analysis of the broader economic effects of the CAA would
provide for separate and distinct treatment of coal-fired utility plants, oil-fired plants, and so on.
Furthermore, the outputs of the macroeconomic model are too aggregated to provide useful and accurate
input information for the sector-specific emission models used to project the emissions consequences of
CAA programs. Again, the critical flaw is the inability to project important details about differential
effects on utilities burning alternative fuels.
The second critical problem with organizing a comprehensive analysis of the CAA around a
macroeconomic modeling approach is that the effect information produced by the macroeconomic model is
relatively unimportant with respect to answering the fundamental, target variable: "How do the overall
hecdth, welfare, ecological, arid economic benefits of Clean Air Act programs compare to the costs of
these programs?" The Project Team believes that any adverse effect, no matter how small in a global
context, should not be deemed "insignificant" if even one individual is seriously harmed. However, the
Retrospective Study results themselves have shown that, when analytical resources are limited, the
resources invested in the macroeconomic modeling would have been better spent to provide a more
complete and less uncertain assessment of the benefit side of the equation. Even on the cost side of the
399
-------
Appendix J: Future Directions
equation, it is far more important to invest in developing accurate and reliable estimates of sector-specific
compliance strategies and the direct cost implications of those strategies. This will be even more true in
the Prospective Study context when the Project Team will be faced with forecasting compliance strategies
and costs rather than simply compiling survey data on actual, retrospective compliance expenditures.
The third and most important limitation of macroeconomic modeling analysis of environmental
programs is that, unlike the economic costs of protection programs, the economic benefits are not allowed
to propagate through the economy. For example, while productivity losses associated with reduced capital
investment due to environmental regulation are counted, the productivity gains resulting from reduced
pollution-related illness and absenteeism of workers are not counted. The resulting imbalance in the
treatment of regulatory consequences raises serious concerns about the value of the macroeconomic
modeling evaluation of environmental programs. In the future, macroeconomic models which address this
and other concerns may be developed; however, until such time EPA is likely to have limited confidence in
the value of macroeconomic analysis of even broad-scale environmental protection programs.
Based on these findings and other factors, the design of the first Prospective Study differs in
important ways from the Retrospective Study design. First, rather than relying on broad-scale,
hypothetical, macroeconomic model-based scenario development and analysis, the first Prospective Study
will make greater use of existing information from EPA and other analyses which assess compliance
strategies and costs, and the emission and air quality effects of those strategies. After developing as
comprehensive a data set as possible of regulatory requirements, compliance strategies, compliance costs,
and emissions consequences, the data set will be reviewed, refined, and extended as feasible and
appropriate. In particular, a number of in-depth sector studies will be conducted to develop up-to-date,
detailed projections of the effects of new CAA requirements on key emitting sectors. Potential sectors
include, among others, utilities, refineries, and on-highway vehicles.
The first Prospective Study will also differ from the Retrospective Study in that analytical
resources will be directed toward development of a more complete assessment of benefits. The
deficiencies which prevailed in the Retrospective Study relating to assessment of the benefits of air toxics
control will be addressed. In addition, the Project Team will endeavor to provide a more complete and
effective assessment of the ecological effects of air pollution control.
400
------- |