United States Office of Air and November 1995
Environmental Protection Radiation
Agency Washington, DC 20460
&EPA Human Health Benefits
From Sulfate Reductions
Under Title IV Of The
1990 Clean Air Act
Amendments
Recycled/Recyclable • Printed with Vegetable Oil Based Inks on 100% Recycled Paper (50% Postconsumer)
-------
HUMAN HEALTH BENEFITS FROM SULFATE
REDUCTIONS UNDER TITLE IV OF THE
1990 CLEAN AIR ACT AMENDMENTS
Final Report
Preparedfor:
U.S. Environmental Protection Agency
Office of Air and Radiation
Office of Atmospheric Programs
Acid Rain Division
Prepared by:
Lauraine G. Chestnut
Hagler Bailly Consulting, Inc.
Under Subcontract to:
ICF Incorporated
EPA Contract No. 68-D3-0005
Work Assignment No. 2F-03 and 3F-12
November 10,1995
-------
ACKNOWLEDGEMENTS
We thank Baxter Jones of ICF and Ann Watkins of U.S. Environmental Protection Agency
(U.S. EPA) for project management and review. We thank Rebecca Holmes, Fran Sussman,
and Barry Galef of ICF and Brian McLean, Joe Kruger, Lester Grant, John Bachmann, Jim
DeMocker, Allyson Siwik, Allen Basala, Eric Smith, and Trish Toman of U.S. EPA for
helpful comments on drafts of the report. We also thank Charlie Richman of ICF and Sally
Keefe and Angela Patterson of Hagler Bailly for data analysis and research assistance. Thanks
to Robin Dennis for providing the RADM results on behalf of U.S. EPA. The report draws on
experience and analyses performed by Hagler Bailly Consulting, for similar topics for other
sponsors, during which Bart Ostro, of the California Environmental Protection Agency, Office
of Environmental Health Hazard Assessment, has made significant contributions. We also
thank Jackie Cody and Tamara Anderson for production assistance.
A previous draft of this assessment was completed September 30, 1994. The 1994 draft was
subject to the U.S. EPA peer review process. Peer reviews were provided by Morton
Lippmann, New York University Medical Center; Bernard Weiss, University of Rochester
Medical Center; David Bates, University of British Columbia; A. Myrick Freeman, Bowdoin
College; Gardner Brown, University of Washington; and Lester Lave, Carnegie-Mellon
University. Changes were made in the assessment and report based on the peer reviews of the
1994 draft. Responsibility for any remaining errors or omissions rests solely with the author.
November 10, 1995
-------
CONTENTS
List of Tables
List of Figures
List of Abbreviations
Executive Summary
S.I Background S-l
S.2 Methods '. • S-2
S.3 Results S-3
S.4 Conclusions S-8
Chapter 1 Introduction
1.1 Need for the Assessment 1-1
1.2 Purpose of the Report 1-1
1.3 Context of Health Benefits 1-2
Chapter 2 Overview of the Assessment
2.1 Schematic of the Assessment 2-1
2.2 Uncertainty and Sensitivity Analyses 2-6
2.2.1 Quantitative Uncertainty Analysis 2-7
2.2.2 Sensitivities to Key Default Assumptions 2-8
2.3 Results from the 1990 NAPAP Assessment 2-9
2.3.1 NAPAP Conclusions on the Effects of Gaseous SO2 2-9
2.3.2 NAPAP Conclusions Regarding Indirect Health Effects of Acid
Deposition • • • • 2-10
2.3.3 NAPAP Conclusions on the Effects of Acid Aerosols . 2-12
2.4 Focus of This Analysis on Sulfate Aerosols 2-13
2.5 General Limitations of the Assessment 2-14
2.5.1 Key Uncertainties hi Step 1: Estimating Changes in SO2
Emissions 2-14
2.5.2 Key Uncertainties hi Step 2: Estimating Changes hi Sulfate
Aerosol Concentrations 2-15
2.5.3 Key Uncertainties in Step 3: Matching Population to the Sulfate
Changes < 2-17
2.5.4 Key Uncertainties in Step 4: Estimating Health Effects 2-17
2.5.5 Key Uncertainties in Step 5: Estimating Monetary Valuation of
Health Effects 2-18
November 10, 1995
-------
CONTENTS »> page ii
Chapter 3 Changes in Ambient Outdoor Sulfate Concentrations
3.1 Changes in S02 Emissions . 3-1
3.2 Changes in Sulfate Aerosol Concentrations 3-6
3.3 Matching Population to Atmospheric Sulfate Changes 3-11
Chapter 4 Quantification of Health Effects Changes
4.1 Background on Health Effects Studies 4-1
4.1.1 Types of Health Effects Studies 4-1
4.1.2 Advantages and Limitations for Assessment Purposes 4-3
4.2 Summary of Health Effects Evidence for Sulfate Aerosols 4-7
4.2.1 Epidemiology Study Findings 4-7
4.2.2 Clinical Study Findings 4-8
4.2.3 Animal Toxicological Study Findings 4-10
4.3 Issues in Applying Epidemiology Results in this Assessment 4-11
4.3.1 The Effects of Sulfates versus Other Particulates 4-13
4.3.2 Health Effects Thresholds 4-15
4.3.3 Uncertainty in the Estimates 4-16
4.3.4 Interpretation and Aggregation of Daily Results 4-16
4.4 Selection of Concentration-Response Functions 4-18
4.4.1 Study Selection Criteria 4-18
4.4.2 Mortality 4-19
4.4.3 Chronic Respiratory Disease 4-24
4.4.4 Acute Morbidity 4-28
4.4.5 Summary of Selected Concentration-Response Functions 4-35
Chapter 5 Monetary Valuation of Health Effects Changes
5.1 Introduction 5-1
5.1.1 Monetary Valuation Concepts for Health Effects 5-1
5.1.2 WTP Estimation Techniques for Health Risks 5-2
5.2 Issues in Applying WTP Estimates for this Assessment 5-3
5.2.1 Issues in Applying Available WTP Estimates for Premature
Mortality 5-4
5.2.2 WTP to COI Ratios 5-5
5.3 Monetary Valuation Estimates for Premature Mortality Risks 5-7
5.3.1 Summary of Available WTP Estimates 5-8
5.3.2 The Potential Effect of Age on WTP for Changes in Mortality
Risks 5-11
5.3.3 Monetary Estimates Selected for this Analysis . . . 5-15
5.4 Monetary Valuation Estimates for Morbidity • • • • 5-17
5.4.1 Adult Chronic Bronchitis 5-18
5.4.2 Respiratory Hospital Admissions 5-20
November 10, 1995 ',
-------
CONTENTS > page in
5.4.3
5.4.4
5.4.5
5.4.6
5.4.7
Cardiac Hospital Admissions
Restricted Activity Days
Asthma Symptom Days
Lower Respiratory Symptom Days
Summary of Selected Morbidity Values
5-21
5-21
5-22
5-23
5-24
Chapter 6 Results and Conclusions
6.1 Annual Results Based on Default Assumptions 6-1
6.2 Aggregate Health Benefits 1997 to 2010 6-7
6.3 Sensitivity Analyses Results 6-10
6.4 Conclusions . ; 6-13
Chapter 7 References
November 10, 1995
-------
TABLES
S-l Quantification Steps for this Assessment of Health Benefits Due to Sulfate Aerosol
Reductions S-3
S-2 Estimates of Annual Human Health Benefits of Title IV for the Eastern United States
with Default Assumptions S-5
S-3 Estimates of Annual Human Health Effects Benefits of Title IV for Ontario and
Quebec, Canada with Default Assumptions S-6
S-4 Sensitivity Analyses Results S-9
2-1 Quantification Steps for this Assessment of Health Benefits Due to Sulfate Aerosol
Reductions 2-6
3-1 EPA Forecasts of Annual Utility S02 Emissions by State 3-4
3-2 Estimated Reduction in Annual Utility SO2 Emissions in 2010 Attributable to Title IV
by State 3-5
3-3 Average Reductions in Median Annual SO4 Concentrations by State/Province Due to
Title IV 3-13
4-1 Comparison of Selected Mortality Study Results 4-22
4-2 Selected Coefficients for Human Health Effects Associated with Sulfate Concentration
Changes 4-36
4-3 Key Omissions, Biases, and Uncertainties 4-37
5-1 WTP/COI Ratios 5-7
5-2 Recommended Ranges of VSL Estimates 5-9
5-3 Summary of Selected Monetary Values for Mortality Effects 5-16
5-4 Summary of Selected Monetary Values for Morbidity Effects 5-24
6-1 Estimates of Annual Human Health Benefits of Title IV for the Eastern United States
with Default Assumptions 6-2
6-2 Estimates of Annual Human Health Effects Benefits of Title IV for Ontario and
Quebec, Canada with Default Assumptions 6-3
6-3 Mean Estimated Health Effects Benefits of Title IV by State 6-6
6-4 Mean Annual Health Benefits Estimates 1997 to 2010 6-8
6-5 Total Present Value in 1995 of Mean Health Benefits 1997 to 2010 with Default
Assumptions 6-9
6-6 Sensitivity Analyses Results 6-11
November 10, 1995
-------
FIGURES
2-1 Processes Involved in Acid Deposition 2-2
2-2 Alternative Measures of Paniculate Matter in the Atmosphere 2-3
2-3 Overview of Human Health Effects Resulting from SO2 Emissions 2-4
2-4 Illustration of Potential Changes in SO2 Emissions 2-16
3-1 U.S. Utility SO2 Emission Levels: 1990 through 2010 3-3
3-2 RADM 50th Percentile Annual Sulfate Concentration 1985 Base Case 3-7
3-3 RADM 50th Percentile Annual Sulfate Concentration 1997 with Title IV 3-8
3-4 RADM 50th Percentile Annual Sulfate Concentration 2010 without Title IV .... 3-9
3-5 RADM 50th Percentile Annual Sulfate Concentration 2010 with Title IV 3-10
5-1 Value of a Statistical Life as a Function of Age 5-13
November 10, 1995
-------
ABBREVIATIONS
jag/m3
ADD
ASD
BAD
CEUM
CHA
COH
COI
GIS
IT
H2S04
HIS
LRS
MRAD
NAAQS
NAPAP
NH4HS04
(NH4)2HSO4
NOX
PM10
PM2.5
ppm
RAD
RADM
RHA
RRAD
S02
S04
TSP
U.S. EPA
VOC
VSL
WTP
micrograms/cubic meter
Airway Obstructive Disease
asthma symptom day
bad asthma day
Coal and Electric Utilities Model
cardiac hospital admission
coefficient of haze
cost of illness
geographic information system
hydrogen ion
sulfuric acid
Health Interview Survey
lower respiratory symptom
minor restricted activity day
National Ambient Air Quality Standards
National Acid Precipitation Assessment Program
ammonium bisulfate
ammonium sulfate
nitrogen oxide
particulate matter with an aerodynamic diameter of 10 microns or less
participate matter with an aerodynamic diameter of 2.5 microns or less
parts per million
restricted activity day
Regional Acid Deposition Model
respiratory hospital admission
respiratory restricted activity day
sulfur dioxide
sulfate
total suspended particulates
United States Environmental Protection Agency
volatile organic carbon
value of a statistical life
willingness to pay
November 10, 1995
-------
EXECUTIVE SUMMARY
S.1 BACKGROUND
Tide IV of the Clean Air Act Amendments of 1990 calls for a 10 million ton reduction in
annual emissions of sulfur dioxide (SO2) in the United States by the year 2010, which
represents an approximately 40 percent reduction in anthropogenic emissions from 1980
levels. Implementation of Title IV is referred to as the Acid Rain Program; the primary
motivation for this section of the Clean Air Act Amendments is to reduce acid precipitation
and dry deposition.1 This assessment has been prepared at the request of the U.S.
Environmental Protection Agency (U.S. EPA), Acid Rain Division, to quantify the expected
human health benefits associated with the SO2 emissions reductions required under the Acid
Rain Program. This assessment is intended to contribute to assessments of costs and benefits
of the Clean air Act, such as the studies called for under Sections 812 and 901 of the 1990
Amendments. The Act requests that benefits and costs be quantified to the extent possible
given available scientific and economic information. This report, therefore, focuses on
quantification of potential health benefits of Title IV in both numbers of specific health
effects expected to be reduced and their monetary valuation.
This report provides estimates of the human health benefits expected to resulrfroni changes in
ambient sulfate aerosol concentrations in the eastern United States. Title IV requirements are
expected to result in significant reductions in SO2 emissions in the eastern United States.3
This will mean lower gaseous S02 concentrations close to major emissions sources, lower
sulfate aerosol concentrations (including acid and nonacid aerosols) throughout the region, and
lower acid precipitation throughout the region. This report focuses on ambient sulfate aerosols
because the potential human health benefits of this pollutant reduction have not been fully
quantified in previous analyses, because the potential human health benefits are substantial,
and because a quantitative assessment is feasible for sulfate aerosols, given available scientific
and economic information. This report does not attempt to quantify various other possible
human health benefits of Title IV, such as those that might result from nitrogen oxide
reductions and "piggy back" toxics or particulate reductions.
1 Throughout this report the terms "acid rain" and "acid precipitation" include dry deposition.
2 SO, emissions are also controlled under Title I of the Clean Air Act
— November 10,1995 '.
-------
EXECUTIVE SUMMARY > S-2
S.2 METHODS
Sulfate aerosols are a substantial share of total ambient fine paniculate matter in the eastern
United States. A large body of epidemiology literature examines the relationship between
ambient paniculate matter and health effects. Some of these studies have specifically
examined sulfate aerosols, and many have examined more broad measures of paniculate
matter such as PM2 5 (paniculate matter with aerodynamic diameter of 2.5 microns or less) or
PM10 (paniculate matter with aerodynamic diameter of 10 microns or less). Scientific debate
and uncertainty continue concerning the extent to which sulfates may or may not be the key
causative constituent of this observed association between health effects and paniculate matter.
Sulfate aerosols, and especially that portion of sulfate aerosols that is acidic, continue to be
considered one of the likely causative agents in the observed association between paniculate
matter and health effects in the eastern United States. In this assessment, the available
epidemiology evidence is applied on the presumption that sulfate aerosols are at least a
contributing causative constituent of PM2 5. This assessment does not assume that sulfate
aerosols are the only causative constituent of PM2 5.
This assessment also relies on available economic information for estimates of willingness to
pay (WTP) for changes in risks of specific health effects. Economic values for changes in
risks of human health effects should reflect the full costs to the affected individual and to
society. The full costs of an adverse health effect include financial losses such as medical
expenses and lost income (referred to as the cost of illness), plus less tangible costs such as
pain and discomfort, restrictions on nonwork activities, and inconvenience to others. WTP, as
a monetary measure for a change in health risk, is defined as the dollar amount that would
cause the affected individual to be indifferent to experiencing an increase in the risk of the
health effect or losing income equal to that dollar amount. WTP measures of monetary value
for changes in health risks thus exceed health care and other out-of-pocket costs that are
associated with illness or premature death, because WTP reflects these as well as other less
tangible effects of illness or premature death on a person's quality of life.
Table S-l lists the five major quantification steps in this assessment and gives a brief
explanation of the quantification method selected for each step. Other related assessments are
ongoing at the U.S. EPA, such as the Section 812 studies concerning the costs and benefits of
the Clean Air Act Amendments as a whole and the review of the National Ambient Air
Quality Standards (NAAQS) for paniculate matter. Although there are many similarities in the
general approaches being taken in the health benefits components of these other assessments
and in this assessment for Title IV, many of the details of the assessment methods may differ.
Many of these differences stem from the fact that this assessment focuses on SO2 emissions
and sulfate aerosols only, while the NAAQS assessment considers all sources of ambient
paniculate matter and the Section 812 studies consider not only all sources of ambient
paniculate matter but all air pollutants regulated under the Clean Air Act
November 10, 1995
-------
EXECUTIVE SUMMARY * S-3
Table S-l
Quantification Steps for this Assessment of
Health Benefits Due to Sulfate Aerosol Reductions
Quantification Steps
1. Changes in SO2 emissions in the United
States
2. Changes in atmospheric sulfate aerosol
concentrations in the eastern United
States and eastern Canada
3. Numbers of people residing at each
location where atmospheric sulfate
concentrations change in the eastern
United States and Canada
4. Changes in sulfate-related health effects:
changes in numbers of cases of each
type of health effect
5. Monetary valuation of changes in health
Selected Quantification Method
Use ICF Resources (1994) estimates for the United
States of 1985 emissions, 1997 emissions with
Title IV, and 2010 emissions with and without
Title IV (prepared for EPA)
Use EPA's Regional Acid Deposition Model
(RADM) runs for each of the SO2 emissions
scenarios
Match the RADM 80 km x 80 km grid to
population data using a Geographic Information
System; population based on 1990 Census data for
block groups (Chapter 3)
Use concentration-response functions derived from
selected epidemiology studies on health effects of
sulfates or PM2 5 (Chapter 4)
Use selected willingness-to-pay estimates from the
available economics literature for changes in health
risks or health effects (Chapter 5)
S.3 RESULTS
Table S-2 summarizes the estimates of annual human health benefits for the sulfate aerosol
reductions attributed to Title IV in 1997 and 2010 for the 31-state eastern United States area.
Table S-3 gives the results for Ontario and Quebec. These estimates are based on the default
quantification assumptions, some of which are changed in the sensitivity analyses discussed
below. The mean total annual estimated health benefit (in 1994 U.S. dollars) for 1997 in the
United States is $10.6 billion, and rises to $40.0 billion by the year 2010, when Title IV
requirements are expected to be fully implemented.
The health benefit estimates are dominated by premature mortality and chronic bronchitis. The
numbers of cases in these health effects categories are relatively small, but the high monetary
values per case result in large monetary benefits for these categories. Premature mortality
reductions account for about 88 percent of the total health benefits. Chronic bronchitis
reductions are an additional 9 percent of the total. Together they represent about 97 percent of
the total estimated health benefits.
November 10, 1995
-------
EXECUTIVE SUMMARY > S-4
The largest numbers of cases reduced are for asthma symptom days, restricted activity days,
and days with acute lower respiratory symptoms. The restricted activity days are net of days
in the hospital and asthma symptom days because these health effects categories may
substantially overlap. The lower respiratory symptom days are net of the fraction of restricted
activity days that might also be attributed to lower respiratory symptoms. In 2010, the
estimated reduction in the number of asthma symptom days because of Title IV is about
6 million in the eastern United States; net restricted activity days prevented is about 9 million;
and the estimated number of days with acute lower respiratory symptoms prevented, net of
restricted activity days, is about 19 million. Together, these represent about 3 percent of the
total monetary health benefits.
Estimates of reductions in health effects in Canada are based on estimates of changes in
sulfate aerosol concentrations in Canada predicted to result from changes in SO2 emissions
generated in the United States. The estimated benefits for Canada occur primarily in the
Windsor-Quebec corridor, where the greatest share of the Canadian population likely to be
affected by the transport of SO2 emissions from the eastern United States is located. The
estimates for Canada represent an additional 9 percent of the Title IV benefits in 1997
estimated for the United States population. The estimates for Canada do not increase
substantially from 1997 to 2010 presumably because the upwind locations in the United States
that affect this area of Canada see their greatest reduction in SO2 emissions in the first phase
of the Title IV program. In 2010, the estimates for Canada add an additional 2 percent to the
2010 estimates for the United States population.
There are many sources of uncertainty and potential error in the mean estimates of health
benefits for Title IV reported. Table S-4 shows results of some specific sensitivity analyses
conducted to determine the potential effect on results of different assumptions than those
selected for the mean estimates. The uncertainty and sensitivity analyses reported here cover
only the uncertainties in the concentration-response functions and in the monetary valuation of
health effects. Additional uncertainties also exist in the estimates of changes in S02 emissions
and ambient sulfate concentrations that are used as inputs to the health benefits estimates.
The uncertainty and sensitivity analyses reported here are those that are reasonably amenable
to quantitative treatment. It is important to recognize that there are many sources of
uncertainty that are not possible to quantify, and that these sensitivity tests are therefore not a
comprehensive treatment of all possible sources of uncertainty. What these tests provide,
however, is an indication of how the results might change if we found that some of the key
default assumptions in the health effects quantification and valuation procedures were
inappropriate.
Most of the selected concentration-response and monetary value estimates are based on
statistically derived results. These estimates therefore have some quantified statistical
uncertainty based on the estimated statistical variance in the results. For all of the health
effects and monetary value estimates, low and high as well as central estimates were selected
. , November 10, 1995 .
-------
Table S-2
Estimates of Annual Human Health Benefits of Title IV
for the Eastern United States with Default Assumptions
(millions of 1994 dollars)
Health Effect
Premature
Mortality
Chronic
Bronchitis
(new cases)
Respiratory
Hospital
Admissions
Cardiac
Hospital
Asthma
Symptom
Days
Restricted
Activity Days
(net)
Days with
Lower
Respiratory
Symptoms
(net)
Total Annual
Health
1997
Annual Number of Casts Prevented
20th
Percentik
408
1,648
663
510
791,232
1,202,785
2,028,424
Mean
2,568
3,864
805
673
1,604341
2,467,066
5,002,393
80th
Percentik
5.714
6,590
918
867
2373,697
3,809,253
7,259,946
Annul Monetary Value
20th
Percentile
$1,428.0
$507.5
$5.7
$4.6
$20.9
$70.6
$31.8
$3,219.1
Mean
$9307.2
$974.0
$11.3
$9.4
$56.9
$147.0
$56.7
$10,562.3
80th
Percentik
$19,999.0
$1377.5
$17.1
$13.9
$93.2
$228.6
$90.0
$20,684.1
2010
Annual Number of Case* Prevented
20th
Percentik
1,539
6,179
2,501
1,924
2,983,490
4,514,939
7,614,168
Mean
9,678
14,564
3,036
2,552
5,951,693
9,283,999
18,619,000
80th
Percentik
21,544
24,715
3,462
3,270
8,950,470
14,298,930
27,251,920
Annual Monetary Value
20th
Percentik
$5386.5
$1,903.0
$21.5
$17.5
$78.7
$265.0
$1 19.3
$12,131.5
Mean .
$35,234.8
$3,705.8
$42.4
$35.7
$212.9
$554.7
$212.8
$39,999.0
80th
Percentik
$75,404.0
$5,165.3
$64.6
$52.5
$351.3
$857.9
$338.0
$77,915.5
w
I
{/J
T
00
-------
g
Ift
Health Effect
Premature Mortality
Chronic Bronchitis
(new cases)
Respiratory Hospital
Admissions
Cardiac Hospital
Admissions
Asthma Symptom
Days
Restricted Activity
Days (net)
Days with Lower
Respiratory
Symptoms (net)
Total Annual Health
Benefits
Table S-3
Estimates of Annual Human Health Benefits of Title IV
for Ontario and Quebec, Canada with Default Assumptions
(millions of 1994 dollars)
1997
Annual Number of Cases
Prevented
20th
Percentile
35
140
56
43
66,915
97,734
164,822
Mean
217
329
68
57
133,825
199,194
401,231
80lh
Pereeatile
483
562
78
73
200,746
309,526
589,916
Annual Monetary Value
20th
Percentile
$122.5
$43.3
$0.5
$0.4
$1.8
$5.7
$2.6
$273.3
Mean
$801.2
$83.3
$0.9
$0.8
$4.8
$12.0
$4.5
$907.6
80th
Perceutile
$1,690.5
$117.4
$1.4
$1.2
$7.9
$18.6
$7.3
$1,746.9
2010
Annual Number of Cases Prevented
20th
Percentile
37
150
60
46
71,594
104,568
176347
Mean
232
355
73
61
142,267
215,270
433,821
80th
Percentile
517
601
83
78
214,783
331,168
631,165
Annual Monetary Value
20th
Percentile
$129.5
$46.3
$0.5
$0.4
$1.9
S6.1
$2.8
$290.8
Mean
$839.2
$91.0
$1.0
$0.9
$5.1
$13.0
$4.9
$955.0
80th
Percenrite
$1,809.5
$125.8
$1.6
$1.3
$84
$19.9
$7.8
$1,868.1
m
m
C/3
G
C/J
o\
-------
EXECUTIVE SUMMARY * S-7
based on the estimated statistical variance and analyst judgment. In general, the selected high
and low estimates represent plus and minus approximately one statistical standard error.
It is not appropriate to combine all the "low" estimates or all the "high" estimates to calculate
upper and lower bounds on the final estimates, because it is highly unlikely that either all the
lows or all the highs would be correct. Such extreme assumptions would significantly
overstate the statistical uncertainty in the estimates. Instead, we have assigned probability
weights to the low, central, and high estimates which when incorporated in the calculation
process allow determination of a probability distribution for the total health benefit results.
The results of this procedure are shown in Tables S-2 and S-3 along with the mean estimates
for the estimated annual health benefits of Title IV in 2010 for the eastern United States and
Canada. All of these estimates are based on the default assumptions, with each estimate
representing a different selected point in the estimated probability distribution calculated for
the total health benefits. The 20th percentile of the distribution for 2010 in the eastern United
States is about $12 billion in benefits with the default assumptions. This means that
20 percent of the estimated values of benefits are below this amount and 80 percent are above
it The 80th percentile of the distribution is about $78 billion in benefits with the default
assumptions. This means that 20 percent of the estimated values of benefits are above this
amount and 80 percent are below it.
Each of the sensitivity tests illustrated in Table S-4 represents estimates of mean annual health
benefits in 1994 dollars. Each is calculated hi the same way that the default mean was
calculated, except for the specified assumption change. A comparison with the default mean
therefore illustrates the effect of the change in the assumption. There is considerable
uncertainty about whether there is a "safe" level of sulfate aerosol exposure that does not
cause any harmful health effects. There is no definitive quantitative evidence that such a
threshold exists, but neither is there proof that any amount of sulfate aerosol exposure causes
some harmful effect hi at least some people. We selected alternative threshold assumptions of
5.0 ug/m3, 3.6 jig/m3, and 1.6 jig/m3 annual median sulfate concentrations to illustrate the
potential effects of alternative threshold assumptions on the results of this analysis. The results
indicate that with a threshold of 5.0 ug/m3 annual health benefits are substantially reduced
relative to the default mean, falling very close to the 20th percentile default estimate. At
thresholds above 5.0 the health benefit estimates would diminish even more. A threshold of
3.6 |ig/m3 results in a mean health benefit estimate mat falls about midway between the
default mean and the 20 percentile default estimates. At a threshold of 1.6 (or lower), the
health benefit estimate is virtually unchanged from the default mean. This illustrates the
significance of the threshold question and shows that this continues to be an important
research issue from the standpoint of evaluating the health benefits of pollution emission
reductions.
There is a possibility that the sulfate-based concentration-response functions may be somewhat
upwardly biased because of the typical collinearity between sulfates and other fine paniculate
November 10, 1995
-------
EXECUTIVE SUMMARY »• S-8
constituents in the ambient air. For this sensitivity test we multiply the sulfate based
concentration-response functions by 0.4, which is the average ratio between measured sulfates
and measured PM2 5 in the eastern United States. This is the maximum adjustment that would
be required if the sulfate coefficients represented the total effects of all PM2 5. This adjustment
reduces the annual health benefit estimate in 2010 in the eastern United States to
about $18.5 billion, which is higher than the 20th percentile estimate with the default
assumptions. The true sulfate effect is probably between this and the mean default estimate
because the sulfate coefficients probably do reflect some, but are unlikely to reflect all, of the
effects of other harmful constituents of PM2 5 as well as the effects of sulfates alone.
S.4 CONCLUSIONS
The results of this assessment show that the potential health benefits of reductions in
exposures to sulfate aerosols in the eastern United States as a result of the SO2 emissions
reductions required by Title IV are substantial. Based on what we believe is a reasonable
interpretation of the available epidemiology and economic evidence on potential health effects
of sulfate aerosols and their monetary value, we estimate that the annual health benefits of
Title IV required reductions in SO2 in 2010 in the eastern United States are more likely than
not to fall between $12 billion and $78 billion, with an estimated mean value of $40 billion.
There is reason to expect some possible upward bias at the higher end of this range, and the
results of the sensitivity analyses suggest that there is a good chance that the benefits in 2010
fall between $12 billion and the estimated mean of $40 billion. Annual health benefits for
eastern Canada resulting from U.S. reductions in SO2 emissions would add as much as one
billion dollars to the U.S. benefit totals in both 1997 and 2010.
We have been careful throughout the report to highlight key assumptions and uncertainties
that exist in the quantification procedures used in this assessment, especially in the health
effects quantification and valuation portions of the assessment which are the focus of this
report. Most of these uncertainties cannot be resolved without substantial new research on
several topics.
November 10, 1995
-------
EXECUTIVE SUMMARY *• S-9
Table S-4
Sensitivity Analyses Results
Assumptions
Estimated Annual Health
Benefits (billions of 1994 dollars)
United States 1997
Threshold = 5.0 ng/m3 SO4
Threshold = 3.6 ng/m3 SO4
Threshold = 1.6 |ig/m3 S04
Selected SO4 Health Risks x 0.4
$3.1
$6.7
$10.8
$4.8
United States 2010
Threshold = 5.0 ^ig/m3 SO4
Threshold = 3.6 ng/m3 SO4
Threshold = 1.6 ng/m3 SO4
Selected SO4 Health Risks x 0.4
$15.0
$28.3
$39.3
$18.5
Canada 1997
Threshold = 5.0 ng/m3 S04
Threshold = 3.6 ng/m3 SO4
Threshold = 1.6 |xg/m3 SO4
Selected SO4 Health Risks x 0.4
$0.0
$0.0
$0.7
$0.4
Canada 2010
Threshold = 5.0 ng/m3 SO4
Threshold = 3.6 ng/m3 SO4
Threshold = 1.6 ng/m3 SO4
Selected SO4 Health Risks x 0.4
$0.0
$0.0
$0.9
$0.5
November 10, 1995
-------
CHAPTER 1
INTRODUCTION
1.1 NEED FOR THE ASSESSMENT
Title IV of the Clean Air Act Amendments of 1990 calls for a 10 million ton reduction in
annual emissions of sulfur dioxide (SO2) in the United States by the year 2010, which
represents an approximately 40 percent reduction in anthropogenic emissions from 1980
levels. Implementation of Title IV is referred to as the Acid Rain Program; the primary
motivation for this section of the Clean Air Act Amendments is to reduce acid precipitation
and dry deposition.1 This assessment has been prepared at the request of the
U.S. Environmental Protection Agency (U.S. EPA), Acid Rain Division, to quantify the
expected human health benefits associated with the SO2 emissions reductions required under
the Acid Rain Program. This assessment is intended to contribute to the assessments of costs
and benefits of the Clean Air Act, such as the studies called for under Section 812 of the
1990 Amendments. The Act requests that benefits and costs be quantified to the extent
possible given available scientific and economic information. This report, therefore, focuses
on quantification of potential health benefits of Title IV in both numbers of specific health
effects expected to be reduced and their monetary valuation.
1.2 PURPOSE OF THE REPORT
This report provides estimates of the human health benefits expected to result from changes in
ambient sulfate aerosol concentrations in the eastern United States. Title IV requirements are
expected to result in significant reductions in SO2 emissions in the eastern United States.2
This will mean lower gaseous SO2 concentrations close to major emissions sources, lower
sulfate aerosol concentrations (including acid and nonacid aerosols) throughout the region, and
lower acid precipitation throughout the region. This report focuses on ambient sulfate aerosols
because the potential human health benefits of this pollutant reduction have not been fully
quantified in previous analyses, because the potential human health benefits are substantial,
and because a quantitative assessment is feasible for sulfate aerosols, given available scientific
and economic information. This report does not attempt to quantify various other possible
human health benefits of Title IV, such as those that might result from nitrogen oxide
reductions and "piggy back" toxics or particulate reductions.
Throughout this report the terms "acid rain" and "acid precipitation" include dry deposition.
2 SO2 emissions are also controlled under Title I of the Clean Air Act.
November 10, 1995
-------
INTRODUCTION * 1-2
Sulfate aerosols are a substantial share of total ambient fine particulate matter in the eastern
United States. A large body of epidemiology literature examines the relationship between
ambient particulate matter and health effects. Some of these studies have specifically
examined sulfate aerosols, and many more have examined more broad measures of particulate
matter such as PM2 5 (particulate matter with aerodynamic diameter of 2.5 microns or less) or
PM10 (particulate matter with aerodynamic diameter of 10 microns or less). Scientific debate
and uncertainty continue concerning the extent to which sulfates may or may not be the key
causative constituent of this observed association between health effects and particulate matter.
Sulfate aerosols, and especially that portion of sulfate aerosols that is acidic, continue to be
considered one of the likely causative agents in the observed association between particulate
matter and health effects in the eastern United States. In this assessment, the available
epidemiology evidence is applied on the presumption that sulfate aerosols are at least a
contributing causative constituent of PM2 s. This assessment does not assume that sulfate
aerosols are the only causative constituent of PM23.
This assessment also relies on available economic information for estimates of willingness to
pay (WTP) for changes in risks of specific health effects. Economic values for changes in
risks of human health effects should reflect the full costs to the affected individual and to
society. The full costs of an adverse health effect include financial losses such as medical
expenses and lost income (referred to as the cost of illness), plus less tangible costs such as
pain and discomfort, restrictions on nonwork activities, and inconvenience to others. WTP, as
a monetary measure for a change in health risk, is defined as the dollar amount that would
cause the affected individual to be indifferent to experiencing an increase in the risk of the
health effect or losing income equal to that dollar amount. WTP measures of monetary value
for changes in health risks thus exceed health care and other out-of-pocket costs that are
associated with illness or premature death, because WTP reflects these as well as other less
tangible effects of illness or premature death on a person's quality of life.
1.3 CONTEXT OF HEALTH BENEFITS
Health effects benefits due to reductions in ambient sulfate aerosols, which are the focus of
this report, are just one category of potential benefits due to Title IV. The potential benefits of
the Title IV provisions include a wide range of environmental impacts, including
improvements or reductions in:
human health effects
effects on aquatics ecosystems, including effects on recreational fishing
visibility aesthetics
effects on materials
effects on terrestrial ecosystems, including effects on forests and crops.
November 10, 1995
-------
INTRODUCTION *• 1-3
Each of these effects involves complex chemical, atmospheric, biological, psychological, and
economic processes. Some of these processes are fairly well understood at this time and others
are not. A practical and policy-relevant assessment must recognize the complexities and
uncertainties inherent in current scientific knowledge of these processes, but it must also
synthesize, simplify, and interpret available information into conclusions that will be useful
for policy-makers.
November 10, 1995
-------
CHAPTER 2
OVERVIEW OF THE ASSESSMENT
2.1 SCHEMATIC OF THE ASSESSMENT
Figure 2-1 shows the processes involved in the formation of acid deposition. This report
focuses on SO2 emissions. SO2 is a gas that is released when fuels containing sulfur, such as
coal, are combusted. SO2 interacts with other elements in the atmosphere to form secondary
sulfate aerosols.1 The resulting sulfate aerosols are called secondary pollutants because they
are not emitted directly, but are formed later.2 The transformation into sulfate aerosols begins
within fairly short distances from the source. Sulfate aerosols can be transported long
distances through the atmosphere before deposition occurs. Some of them are acidic sulfate
aerosols, which are a primary constituent of acid deposition in the eastern United States.
Figure 2-2 shows the relationships among the most common measures of particulate matter in
the atmosphere. Different measures are used in different contexts, and many of these terms are
used throughout this report. Total suspended particulates (TSP) represent all airborne
particulate matter. Particulate matter under 10 microns in aerodynamic diameter (PM10) are
particles small enough to be inhaled into the airways of the lungs. PM10 is sometimes called
thoracic particulate matter. A smaller size category for particulate matter is fine particles,
which are particles with aerodynamic diameter of 2.5 microns or less (PM2 5).
Most sulfate aerosols are part of PM2 5 and most acid aerosols, in the particle phase, are
sulfate aerosols. The term acid aerosol is often used to refer to all airborne acids, including
those in the vapor phase such as nitric acid (NAPAP, 1991). Such vapors are outside the
definition of any of these particle measures. All acidic sulfate aerosols are particles rather than
vapors. Sulfate aerosols make up the largest single component of fine particulate matter in
most locations in the eastern United States. Measures of average sulfate aerosol concentrations
are about 40 percent of measures of average fine particulate matter levels hi the eastern
An aerosol consists of liquid or solid particles in air.
Some sulfate aerosols are emitted directly from combustion sources. These are called primary
sulfate aerosols, but they make up a very small percentage of total ambient sulfate aerosols.
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT »• 2-2
Figure 2-1
Processes Involved in Acid Deposition
SOURCES
NO,
A S02
VOC S(}9 NO,
Wet
Deposition
AliiUi
Natural
RECEPTORS
Anthropogenic
Source: National Acid Precipitation Assessment Program (NAPAP), 1991, p. 174.
i
United States (Dockery et al, 1993)3. Sulfate concentrations are lower in most of the western
United States, where fuels with lower sulfur content are more commonly used.
Figure 2-3 shows an overview of the major pathways by which SO2 emissions may cause
human health effects. A comprehensive quantitative assessment of the human health benefits
of Title IV must analyze each of these pathways. The Title IV requirements will result hi
reductions in SO2 emissions, relative to what would have been emitted in the absence of Title
3 The 1995 review draft of the PM Criteria Document (US EPA, 1995) reports an average ratio of
0.47 in the eastern U.S. and 0.37 in the central U.S. Our definition of "eastern" includes 31 states, some of
which fall in what is commonly called "central" U.S. The 0.4 estimate is therefore reasonably consistent.
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT >• 2-3
Figure 2-2
Alternative Measures of Participate Matter in' the Atmosphere
1 This figure shows the overlaps in the different measures, but is not drawn to scale in terms of
typical relative proportions in the atmosphere. Such proportions vary from place to place and
time to time.
2 The term acid aerosols has been used to refer to acids present in the atmosphere in the vapor
phase such as nitric acid (NAPAP 1991). Such vapors fall outside the definition of any of these
paniculate measures. In rare circumstances, such as in the formation of acid fogs, acid aerosols
can become larger than PM]0 (NAPAP 1991).
November 10, 1995
-------
Figure 2-3
Overview of Human Health Effects Resulting from SO2 Emissions
SO2 Emissions
Ambient Air
Concentrations Close to
Source (within -20 km)
Z
o
3
I
Ambient Levels of SO2
and Primary Sulfates
Ambient Air
Concentrations at
Greater Distances
Ambient Levels of
Secondary Particles,
Including Sulfates and
Acid Aerosols
Human Exposure
(Respiration)
Human Health Effects
Monetary Valuation of
Human Health Effects
Acid Precipitation
to Soils and
Aquatic Systems
Mobilization of
Toxics
Human Consumption
of Toxics
Human Health Effects
Monetary Valuation of
Human Health Effects
O
<
m
1
m
3
m
on
CO
m
CO
to
1
-------
OVERVIEW OF THE ASSESSMENT >• 2-5
IV. Changes in S02 emissions result in changes in human exposures to potentially harmful
substances in the ambient air, both near and far from the S02 source, and through the effects
of acid deposition on mobilization of toxic substances in soils and water.
An important point illustrated in Figure 2-3 is that when SO2 emissions are reduced, potential
benefits to human health occur along several avenues. Reductions in ambient air levels of
gaseous SO2 and sulfate aerosols mean reductions in these potentially harmful pollutants in
the air that people breathe. Once the sulfate aerosols are deposited on soils and aquatic
systems, the acidic portion of these aerosols can contribute to the mobilization of toxic
substances already present in the environment. A reduction in acid deposition thus means a
reduction in the chance that these substances will be present in the water and food that
humans consume.
For reasons discussed in subsequent sections of this chapter, this report focuses on the human
health benefits of the expected reductions in exposure to atmospheric sulfate aerosols caused
by the Title IV required SO2 emissions reductions. Table 2-1 lists the five quantification steps
in this assessment and gives a brief explanation of the quantification method selected for each
step. Some of the rationale for selecting these methods is explained in subsequent sections of
this chapter. Subsequent chapters explain the selected quantification methods in detail.
Other related assessments are ongoing at the U.S. EPA, such as the Section 812 studies
concerning the costs and benefits of the Clean Air Act Amendments as a whole and the
review of the National Ambient Air Quality Standards (NAAQS) for paniculate matter.
Although there are many similarities in the general approaches being taken in the health
benefits components of these other assessments and in this assessment for Title IV, many of
the details of the assessment methods may differ. Many of these differences stem from the
fact that this assessment focuses on SO2 emissions and sulfate aerosols only, while the
NAAQS assessment considers all sources of ambient paniculate matter and the, Section 812
studies consider not only all sources of ambient participate matter but all air pollutants
regulated under the Clean Air Act.
The results of this health benefit assessment based on the selected default assumptions are
reported in two ways. First, they are reported as annual estimates for the years 1997 and
2010. Title IV, Phase I, is expected to be implemented by 1997, and Title IV is expected to
be fully implemented by 2010. The estimated 1997 sulfate concentrations without Title IV are
based on Regional Acid Deposition Model (RADM) runs for 1985 emissions estimates,
assuming no significant change from 1985 to 1997 in the absence of Title IV. RADM results
for 1997 estimated S02 emissions with Title IV are compared to 1985 RADM results to
calculate Title IV health benefits in 1997. The 2010 estimates are based on ICF Resources
estimates of S02 emissions with and without Title IV for 2010 and on estimates of ambient
sulfate aerosol concentrations from RADM for each of the 2010 emissions scenarios. The
annual estimates are based on 1990 population and income levels and are reported in 1994
dollars.
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT + 2-6
Table 2-1
Quantification Steps for this Assessment of
Health Benefits due to Sulfate Aerosol Reductions
Quantification Steps
1. Changes in SO2 emissions in the United
States
2. Changes in atmospheric sulfate aerosol
concentrations in the eastern United
States
3. Numbers of people residing at each
location where atmospheric sulfate
concentrations change in the eastern
United States and Canada
4. Changes in sulfate-related health effects:
changes in numbers of cases of each
type of health effect
5. Monetary valuation of changes in health
Selected Quantification Method
Use ICF Resources (1994) estimates of 1985
emissions, 1997 emissions with Title IV, and 2010
emissions with and without Title IV (prepared for
EPA)
Use EPA's Regional Acid Deposition Model
(RADM) runs for each of the SO2 emissions
scenarios
Match the RADM 80 km x 80 km grid to
population data using a Geographic Information
System; population based on 1990 Census data for
block groups (Chapter 3)
Use concentration-response functions derived from
selected epidemiology studies on health effects of
sulfates or PM2S (Chapter 4)
Use selected willingness-to-pay estimates from the
available economics literature for changes in health
risks or health effects (Chapter 5)
Second, the results are reported as 1995 present value estimates of the total health benefits
expected from 1997 though 2010. Health benefits due to Title IV for the years between 1997
and 2010 are interpolated from the RADM-based estimates in proportion to the emissions
estimates available for the years between 1997 and 2010 for the scenarios with and without
Title IV. Aggregate estimates of total health benefits are reported undiscounted and discounted
with two alternative discount rates, both adjusted for average expected population and real
income growth.
2.2 UNCERTAINTY AND SENSITIVITY ANALYSES
Any quantitative assessment of this nature is subject to considerable uncertainty due to the
complexities of the physical and economic processes involved, and limits in our technical
capabilities to fully characterize current interactions and predict future changes. It is important
that analysts attempt to characterize the uncertainty in the results of such an assessment so
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT > 2-7
that policy makers can give appropriate consideration to the results in their decision making
processes. This report addresses uncertainty in the following ways:
> Limitations and assumptions in the quantification process are clearly stated and
explained.
* A quantitative uncertainty analysis is conducted based on estimated statistical
variance in some of the underlying relationships upon with the assessment is
based.
> Sensitivity analyses illustrate the effects of changing key default assumptions
on the mean results of the assessment.
There are many different valid ways to characterize and present quantitative uncertainty in an
assessment of this type. This assessment has used an approach very similar to that developed
by Rowe et al. (in press) in a quantitative model to estimate environmental effects of
electricity generation in New York. The quantitative uncertainty analysis is based on
variations hi results within and across selected studies, but specific results are selected as most
likely correct and are given probability weights that reflect some analyst judgment as well as
empirical evidence.
2.2.1. Quantitative Uncertainty Analysis
The available epidemiology and economics evidence regarding health effects associated with
air pollutants is subject to considerable uncertainty. Within a given study there is statistically
measurable uncertainty in the estimated concentration-response coefficients or monetary value
estimates, and there are differences in results obtained from different studies looking at the
same or similar health effects. This assessment uses a quantitative uncertainty analysis similar
to the approach developed by Rowe et al. (in press). For each concentration-response
relationship and each monetary value estimate presented hi this report, low, central, and high
estimates are selected. The central estimate is typically selected from the middle of the range
reported hi the study, or group of studies, that has been selected as providing the most reliable
results for that health effect based on the study selection criteria.
These ranges of estimates are not intended to reflect absolute upper and lower bounds, but
rather they are ranges of estimates that are reasonably likely to be correct, given available
epidemiology and economics study results. For example, ranges based on a single study are
selected as plus and minus one standard error, not the absolute highest or lowest results
obtained. When several different "reliable" studies are available for a given health effect, the
selected range reflects the variation in results across the studies. The reader should be aware
that there is analyst judgment in selecting these ranges and that the ranges do not reflect all
the uncertainty hi the estimates because some of the uncertainty is not quantifiable. This is,
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT »• 2-8
however, an attempt to give a more realistic presentation than is given when only point
estimates are reported.
Each low, central, and high estimate is also assigned a probability weight (the weights
summing to 100 percent for each quantified health effect and for each monetary value
estimate). These probability weights, combined with the low, central, and high estimates, are
used to estimate a probability distribution of the total health benefits estimate, which is
calculated by multiplying estimated numbers of health effects by the monetary value per case,
and summing across all the health effects categories. Calculating a probability distribution for
the total health benefit estimate provides an alternative to simply summing all the low
estimates or all the high estimates to obtain total low and high estimates. Such simple
summing results in a misleadingly large range of values, because it is highly unlikely that all
the low estimates (or all the high estimates) are correct. When the low, central, and high
estimates are based on results from different studies all judged as equally reliable, an equal
probability weight is given to the low, central, and high estimates. When only one study result
is selected, the range selected is often plus and minus one statistical standard error of the
selected central result. When a standard error is used, the probability weight given to the
central estimate is 50 percent, with 25 percent each to the high and low estimates. In a few
cases less weight has been given to a high or low estimate based on analyst judgment that
there is reason to suspect that particular estimate is less likely to be correct than the other
available estimates.
Mean, low, and high values for changes in cases of each health effect and for their monetary
values were calculated for the estimated change in sulfate concentrations, using the low,
central, and high values and the probability weights assigned to each. These calculation were
executed using the @RISK supplemental program for such applications with the Lotus 1-2-3
spreadsheet program (Palisade Corp., 1994). This program selects a sample of all the possible
combinations of low, central, and high estimates sufficient to estimate a probability
distribution for the total health benefit estimate. From this estimated distribution, we have
selected low and high values that represent the 20th percentile and the 80th percentile on the
probability distribution of the total estimated health benefit. This means, for example, that
there is an 60 percent probability that the "true" value falls between these low and high
results, given the magnitudes and the probabilities selected for each of the low, central, and
high concentration-response and monetary value estimates.
2.2.2 Sensitivities to Key Default Assumptions
Throughout the report the assumptions and uncertainties in this analysis are acknowledged. In
some cases it is possible to define alternative assumptions and to determine how the results
are affected if a default assumption were determined to be incorrect. This is an important
process for identifying the most important assumptions with regard to their effect on the
bottom line, and the results are reflected in the conclusions of the report.
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT »• 2-9
2.3 RESULTS FROM THE 1990 NAPAP ASSESSMENT
Three categories of potential human health effects associated with SO2 emissions and
subsequent secondary pollutants were considered in the 1990 NAPAP State of the Science and
Technology reports and the NAPAP 1990 Integrated Assessment:
+ direct health effects of gaseous SO2
* indirect health effects of toxic chemicals released into the environment as a
result of acid deposition
* direct health effects of acid aerosols in the ambient air.
2.3.1 NAPAP Conclusions on the Effects of Gaseous SO3
SO2 is a criteria air pollutant under the Clean Air Act, and National Ambient Air Quality
Standards (NAAQS) have been set to protect public health and welfare. The current primary
NAAQS for SO2 are:
annual average of 0.03 ppm
24-hour average of 0.14 ppm.
Ambient concentrations of SO2 have been substantially reduced in the United States since
1970, and most of the population now lives in areas that meet the primary NAAQS.
Remaining nonattainment areas are limited to geographical areas in the immediate vicinity of
a few major point sources.
Much of the recent SO2 health effects research has focused on acute exposures of asthmatics,
who are believed to be more sensitive to S02 than other people. Aggravation of asthma
symptoms in some individuals who are exercising and who already have asthma has been
demonstrated in clinical studies with short-term SO2 exposures at concentrations close to those
that occasionally occur currently in some locations in the United States. Graham et al. (1990)
cite conclusions reached by the U.S. EPA that at current S02 emission levels in the United
States, the only health effect of any concern due to short-term peaks of ambient S02
concentrations is the aggravation of asthma symptoms in exercising asthmatics.
NAPAP (1991) reported the U.S. EPA's conclusions that SO2 concentrations high enough to
cause well-documented short-term effects on individuals with asthma currently occur only
within about 12 km of a few major point sources in the United States. Graham et al. (1990)
report that approximately 100,000 exercising asthmatics may be exposed once each year to
S02 concentrations high enough and for long enough to cause a reaction in some asthmatics
(0.5 ppm for 5 minutes was the assumption used). Graham et al. cite clinical evidence that
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT *• 2-10
approximately 25 percent of asthmatic subjects may have a doubling of airways resistance
while exercising when exposed to 0.5 ppm of SO2. Not all reactive asthmatics will have
symptoms severe enough to be noticeable to them. More recent evaluations (U.S. EPA, 1994)
indicate that only about 10 to 20 percent of mild or moderate asthmatics are likely to exhibit
lung function decrements in response to SO2 exposures of 0.2 to 0.5 ppm during moderate
exercise that would be of distinctly larger magnitude than typical daily variations in lung
function or average changes in lung function experienced in response to other often
encountered stimuli (e.g., cold/dry air, moderate exercise, etc.). A more substantial percentage
(20 to 25 percent) of such asthmatics exposed to 0.6 to 1.0 ppm of SO2 experience respiratory
function decrements and severity of respiratory symptoms that exceed typical daily variations
or response to other commonly encountered stimuli that produce short-lived
bronchoconstrictor effects like S02.
A further reduction in SO2 emissions, beyond current levels, due to Title IV means that this
health effect can be expected to be reduced. Because of the limited geographic scope of this
effect however, the economic benefit of reducing this effect is relatively small. If we assume
an average monetary value of $34 (see Chapter 5) for preventing a day with aggravated
asthma symptoms, the annual aggregate value of preventing this effect would be no more than
$1,000,000 even if all 25,000 affected asthmatics have noticeable symptoms and if the Title
IV emission reduction eliminates all of this negative health effect.
The analysis and conclusions reported by NAPAP appear to be sufficient for estimating an
upper bound on the likely benefits of Title IV due to reductions in short-term effects of peak
SO2 exposures on exercising asthmatics. This category of health benefits for Title IV appears
to be relatively small and is fairly well established. It does not appear to warrant further
quantitative analysis at this time.
2.3.2 NAPAP Conclusions Regarding Indirect Health Effects of Acid Deposition
NAPAP (1991) provides a summary of the analysis and conclusions reported by Grant et al.
(1990) of potential indirect human health effects due to acid deposition. The pathway for such
potential effects is illustrated on the right-hand side 01 Figure 2-2. The mechanism is that acid
deposition can cause potentially harmful substances already present in soils or aquatic systems
to be mobilized. These substances may then ultimately be consumed by humans through food
or water. Such consumption in sufficient quantity may cause adverse health effects.
Grant et al. (1990) assessed the likelihood that current levels of acid deposition could be
associated with significant human health effects as a result of the mobilization of
methylmercury, lead, cadmium, arsenic, aluminum, copper, selenium, and asbestos. This
assessment was made difficult by complexities and uncertainties about the physical processes,
the multiple sources of these substances in the environment, and human exposures. The
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT * 2-11
conclusions are therefore tentative, but they appear to be reasonable given currently available
information. '
Grant et al. (1990) concluded that at current acid deposition levels, lead and methylmercury
are the only substances considered that may be causing measurable health effects as a result of
acid deposition. Some subpopulations of individuals are already exposed to high levels of
these compounds because of circumstances unrelated to acid deposition, and it is feasible that
further exposure due to the mobilization effects of acid deposition might result in adverse
health effects. For lead, critical health effects include slowed fetal physical and neurological
development, neurobehavioral deficits in young children, including decreased IQ, and
hypertension in adults. Critical health effects due to methylmercury include fetal psychomotor
retardation and paresthesia in adults.
Only a small segment of the population is likely to be at any appreciable risk because of
incremental lead or methyimercury exposures as a result of current levels of acid deposition.
The population segments judged to be at some potential risk are as follows:
*• Those for whom subsistence fishing is a significant source of food and who fish
primarily at acidified lakes may be at risk of harmful effects due to methylmercury in
fish. High concentrations of methylmercury have been measured hi fish at acidified
lakes hi the upper Midwest and Northeast.
v Young children and developing fetuses within pregnant women who consume acidified
drinking water (without pH or corrosivity treatment) may be exposed to potentially
harmful concentrations of lead if the soil or water distribution system contains lead
that is leached by the acidified drinking water. These are primarily individuals whose
drinking water comes not from municipal systems but from rainwater, surface water,
or shallow wells.
Grant et al. (1990) estimate that the first group may contain as many as 10,000 individuals
and that the second group may contain approximately 11,000 children and 29,000 women of
childbearing age. Estimates of how many of these individuals might be expected to suffer
adverse effects were not made, but clearly it would be some fraction of the total. Although
some potential health effects of these substances are severe, the number of people estimated to
be at any risk of elevated exposure at current acid deposition levels is small.
Uncertainty about the current extent of these health risks due to acid deposition cannot be
reasonably reduced at this time without an investment of very significant research resources.
Further quantitative analysis of this category of potential health effects does not appear to be
warranted at this time.
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT + 2-12
2.3.3 NAPAP Conclusions on the Effects of Acid Aerosols
Graham et al. (1990) reviewed the available laboratory, clinical, and epidemiological evidence
on the human health effects of acid aerosols. NAPAP (1991) summarized the conclusions of
this review, which are that (1) there is evidence of harmful respiratory effects for human
subjects exposed to some types of acid aerosols and (2) there is not sufficient information
available to conduct a quantitative assessment of the current level of health effects due to acid
aerosols in the United States.
Acid aerosols are a mixture of several pollutants. In the eastern United States, the
predominant fraction of acid aerosols appears to be acidic sulfates, which include sulfuric acid
(H2SO4), ammonium bisulfate (NH4HSO4), and ammonium sulfate [(NH4)2HSO4]. NAPAP
(1991) notes that the hydrogen ion (H*) may be the species of concern with respect to human
health, but this remains uncertain. Most of the available laboratory and clinical evidence
regarding health effects of acid aerosols focuses on acidic sulfates, especially H2SO4. NAPAP
summarizes the available clinical and laboratory evidence on acidic sulfates as follows:
, * Controlled acute exposures to acidic sulfates can cause decreased lung function
and reactivity responses in some asthmatics.
* Controlled acute exposures to acidic sulfates can alter mucociliary clearance of
the lungs in nonasthmatic and asthmatic humans. This may affect the ability of
the lungs to clear inhaled particles, including infectious organisms.
* Long-term exposures of laboratory animals to acidic sulfates reveal changes
related to the development of chronic bronchitis, including reduced mucociliary
clearance and morphological changes.
Epidemiology research concerning acid aerosols has been quite limited because of little
availability of data on ambient acid aerosol concentrations. NAPAP (1991) notes that there is
epidemiology evidence that air pollution mixtures known to contain acid aerosols are
associated with both mortality and morbidity, but that it is not possible to determine to what
extent this association is due to the presence of acid aerosols. Four aerosol pollution measures
that have been used in these types of studies are TSP, PM10, PM2 5, and sulfates. NAPAP
notes that statistically significant associations between mortality and all of these aerosol
measures have been found in macroepidemiological studies, and somewhat more consistency
has been found in the results for fine particles and sulfates. The macroepidemiological studies
compare average mortality rates across locations with different average pollution
concentrations.
The NAPAP 1990 Integrated Assessment appropriately concluded that there is not sufficient
information available at this time to conduct a credible quantitative assessment of the health
effects of acid aerosols in the United States. This is the result of limited data availability for
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT * 2-13
current concentrations of acid aerosols, as well as limited quantitative evidence on the specific
health effects that might be expected for a given concentration of acid aerosol exposure.
The NAPAP assessment, however, did not address the question of whether a quantitative
assessment is feasible for sulfate aerosols in general, rather than for just acidic sulfates. This
is really the more relevant question with regard to the health benefits of Title IV, because the
required reductions in SO2 emissions will result in reductions in all sulfate aerosols, not just
acidic sulfates.
2.4 Focus OF THIS ANALYSIS ON SULFATE AEROSOLS
The focus of this analysis is on the potential human health benefits of the reduction in
ambient concentrations of sulfate aerosols, including acidic sulfates, that is expected as a
result of the Title IV required reductions in S02 emissions. This is the middle path in
Figure 2-3. This focus was chosen for four primary reasons:
» A quantitative assessment of the health benefits of reducing ambient sulfate aerosol
concentrations is feasible given available information, but has not yet been conducted
for the type of change in ambient concentrations expected as a result of Title IV.
> A large available body of epidemiology literature concerning the association between
ambient aerosol pollutants, including sulfates, and human health effects allows a
quantitative assessment to be performed using a modest amount of research resources.
» The required reduction in SO2 emissions is substantial relative to current emission
levels, and the resulting reduction in ambient sulfate aerosol concentrations is also
expected to be substantial.
> Given the potential for reductions in risks of mortality, chronic respiratory disease, and
acute morbidity as a result of reductions in sulfate aerosol concentrations, and the long
distance and wide ranging dispersion of sulfate aerosols, there is a possibility of
substantial health benefits.
The other two branches of Figure 2-3, direct effects of gaseous SO2 and indirect effects of
acid deposition, were examined in detail in the 1990 NAPAP analyses. The results suggest
that the number of people at potential risk of health effects due to these pathways under
current conditions is fairly limited. The potential that these risks will be reduced as a result of
Title IV should not be disregarded in a comprehensive assessment of Title IV benefits, but it
does not appear that there are sufficiently different data or analysis approaches available today
to warrant further analysis of these potential health effects pathways at this time.
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT * 2-14
2.5 GENERAL LIMITATIONS OF THE ASSESSMENT
Detailed discussions of the assessment approach, assumptions, and limitations are provided in
Chapters 3, 4, and 5. In this section, we introduce and highlight what we believe are the key
difficulties, limitations, and uncertainties hi this assessment and therefore in the results.
This health benefits assessment relies on results of two other analyses conducted for or by
EPA. These are the ICF Resources (1994) estimates of emissions of SO2 with and without
Title IV, and the EPA estimates of resulting ambient sulfate aerosol concentrations using
RADM with the ICF emissions estimates as input. Each of these analyses relies on specific
applications of detailed models developed for these and other purposes, which are briefly
described in Chapter 3. Detailed discussions of these analyses, key assumptions, and
uncertainties and limitations are provided elsewhere (e.g., ICF Resources, 1994; Chang et al.,
1990; Dennis et al., 1990; Dennis et al., 1993), and are not the focus of this report. This
report focuses on the approach used to quantify and value health effects associated with these
previously estimated changes hi ambient sulfate aerosol concentrations, and we provide a
thorough discussion of the strengths and limitations of the health effects quantification and
valuation procedures. We do not provide a detailed assessment of the analyses conducted
previously that we rely upon hi this report, but it is important to acknowledge that there is
uncertainty in each of these analyses and results, which adds additional uncertainty to the final
results of this analysis.
2.5.1 Key Uncertainties in Step 1: Estimating Changes in SO2 Emissions
To estimate the benefits of Title IV it is necessary to make some assessment of what would
have happened in the absence of the Title IV requirements. The benefits of Title IV are then
calculated based on the difference between SO2 emissions levels with and without the Title IV
requirements for each future year included in the analysis. Possibly the greatest uncertainty in
the first step in the analysis is in estimating what SO2 emissions would have been over time
in the absence of the Title IV requirements.
We refer to the estimate of what emissions would have been without the Title IV
requirements as the reference case emissions estimates. For this analysis, we use SO2
emissions estimates developed by ICF Resources (1994) for the EPA's Acid Rain Division for
with Title IV and without Title IV scenarios. These estimates go through the year 2010. The
reference case emissions estimates show a slight increase in total annual emissions between
1995 and 2005, and are fairly flat after 2005.
Figure 2-4 illustrates the potential significance of this reference case estimate for the
calculation of Title IV benefits. The intent of this figure is to illustrate the potential
importance of this question. It is not drawn* to scale based on actual quantitative estimates.
With Title IV, we expect SO2 emissions will decline sharply hi 1995 and come close to the
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT > 2-15
Title IV target by the year 2000. Potential use of banked emissions allowances will probably
mean that the Title IV target will not be entirely met until 2010. Point C represents expected
emissions in 2010 with Title IV. Without Title IV, SO2 emissions might be higher, lower, or
at the same level as in 1990. Reference Case I illustrates that if without Title IV emissions
would have risen slightly by 2010, then the emissions reduction attributable to Title IV would
be the difference between point A and point C. If, on the other hand, emissions would have
decreased slightly in the absence of the Title IV requirements, as shown in Reference Case II,
the emissions reduction attributable to Title IV would be the difference between point B and
point C. Thus, the predicted reference case of what emissions would have been in the absence
of Title IV can make a big difference when it comes to estimating the benefits of Title IV.
Even without the Title IV requirements, there are many regulatory and economic factors that
are expected to affect SO2 emissions over the next several decades. NAPAP (1991) reports
that future trends in SO2 emissions without Title IV, would be expected to eventually result in
emissions as low as are required under Title IV, but it is highly uncertain how fast this
reduction would have occurred. This eventual reduction in emissions, in the absence of
Title IV, would be expected to occur because of replacement of old facilities with new
facilities that must conform to the stricter New Source Performance Standards under
previously established requirements of the Clean Air Act and that have new cleaner
technologies available that are more cost-effective to install with new facilities than to retrofit
into old facilities.
2.5.2. Key Uncertainties in Step 2: Estimating Changes in Sulfate Aerosol
Concentrations
Changes in sulfate aerosol concentrations are based on the intermediate results of RADM, a
model developed to estimate acid deposition in the eastern United States as a function of SO2
emission levels in specified locations. The transformation SO2 emissions into sulfate aerosols,
and the transport of S02 and sulfate aerosols through the atmosphere, is a function of complex
chemical and meteorological interactions. RADM estimates these relationships for a sample of
representative meteorological conditions and predicts annual sulfate concentration distributions
at each location based on the estimated frequency of the defined alternative meteorological
conditions. RADM has been thoroughly evaluated and tested as reported by Chang et al.
(1990), Dennis et al. (1990), and Dennis et al. (1993). One of the most significant
uncertainties in using the RADM estimates of airborne sulfate concentrations is that average
meteorological conditions do not occur every year. This means that for any given year, the
predicted concentrations are less reliable than over a multiple year period over which average
meteorological conditions are more likely to prevail.
November 10, 1995
-------
Figure 2-4
Illustration of Potential Changes in SO2 Emissions
1990
Level
Ctl
Title IV
Target
UJ
ft
O
tn
c
Without Title IV
Reference Case I
B
Without Title IV
Reference Case II
1990
1995
2000
2005
Year
2010
2015
2020
This figure illustrates the potential effect of uncertainty in the reference case emissions estimate
C " " We hyp°lhelical es«™t** a"* are not drawn to scale based on any actual
o
m
70
m
O
Tl
m
V)
in
m
t/J
2
1
to
-------
OVERVIEW OF THE ASSESSMENT » 2-17
2.5.3 Key Uncertainties in Step 3: Matching Population to the Sulfate Changes
The RADM grid cells are used as receptor locations to estimate the change in sulfate
concentrations for the population. Residents are matched to RADM grid cells assuming that
they are all located at the centroid of their census block group. Census block groups cover
fairly small geographic areas, so the uncertainty introduced in this step is minimal. Greater
uncertainty may exist as a result of people spending a significant share of their time at
locations other than where they live. Close by locations such as travel to work create limited
uncertainty because the sulfate gradient is fairly gradual from cell to cell. Considerable error
could exist for individuals who spend a significant share of the year in locations far from their
primary residences. However, this is not likely to be a significant source of uncertainty in this
assessment relative to the other sources of uncertainty that are present.
2.5.4 Key Uncertainties in Step 4: Estimating Health Effects
Relying on available epidemiological evidence for estimating health effects associated with
human exposure to ambient sulfate aerosols has many advantages, which are discussed in
Chapter 4. The primary advantage is that it makes a quantitative assessment feasible with
limited research resources and it uses a great deal of health effects evidence that is readily
available. There are, however, several important uncertainties and limitations that result from
the limitations of the available epidemiological evidence. The three uncertainties that we
believe are the most potentially significant as a result of the limitations of the epidemiology
evidence are summarized in this section. These and other uncertainties in the health effects
calculations are discussed in more detail in Chapter 4.
First, there is uncertainty about the specific biological mechanisms that underlie the observed
relationships in epidemiological studies, which raises uncertainty about the confidence with
which the results should be interpreted as causative. Epidemiology studies are able to
demonstrate whether a statistically significant relationship exists between health effects and
pollution concentrations, but the studies do not prove that the relationship is causal. It is
possible that a statistically significant relationship is really due to some unidentified factor that
is correlated with pollution concentrations. The causation hypothesis is strengthened when
epidemiological results are supported by repeated observation in different studies and by
biological plausibility and consistency with evidence from other types of health effects studies.
Although there is laboratory and clinical evidence of health effects associated with sulfates, as
discussed in Chapter 4, the exact biological mechanisms that underlie the observed
epidemiological association have not been established. This adds some additional uncertainty
that is difficult to fully delineate when using epidemiological relationships to predict how
health may change as a result of changes in ambient sulfate aerosol concentrations.
Second, there is uncertainty about the relative harmfulness of sulfates versus other types of
pollutant aerosols that are typically present in the ambient air. Sulfates are a significant share
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT + 2-18
of the mix of fine particulate matter in the ambient air in many locations in the eastern United
States. Some epidemiology studies have included sulfate concentrations as a measure of
pollution, as well as more comprehensive measures of particulate matter such as PM2 5 or
PM10. In some cases, epidemiology studies have found a statistically stronger association
between health effects and sulfates (e.g., Plagiannakos and Parker, 1988), and other studies
have found a stronger association with the more comprehensive measures of particulate matter
(e.g., Dockery et al., 1992). Because of the typically high correlation among sulfates and other
measures of fine particulate matter in the ambient air, it is difficult to statistically isolate the
effects of sulfates alone in epidemiology studies. For this analysis, we examine the clinical,
laboratory, and epidemiology evidence as a whole to determine reasonable assumptions about
the relative contribution of sulfates to the epidemiological evidence of an association between
health effects and fine particulate matter, but this remains an important uncertainty in the
analysis.
Third, there is uncertainty about the extent to which health effects occur at lower ambient
sulfate concentrations. For sulfate aerosols, and for particulate matter in general, it remains
uncertain whether there is a threshold concentration below which health effects no longer
occur, or whether the slope of the concentration-response function diminishes significantly at
lower concentrations. Epidemiological studies do not always consider the question of
thresholds, and epidemiological data are not always sufficient for making such a
determination. Many recent epidemiology studies show a statistically significant association
between sulfate concentrations and health endpoints over ranges of sulfate concentrations that
are typical of current conditions in the eastern United States. For the mean estimates in this
assessment we adopt the default assumption that there is no threshold for health effects
associated with sulfates. Sensitivity analysis is used to show how the results might change if
in fact some threshold exists at selected alternative concentrations.
2.5.5 Key Uncertainties in Step 5: Estimating Monetary Valuation of Health Effects
There are many uncertainties in available estimates and interpretations of monetary valuation
for changes in human health effects. Although it is quite clear that changes in human health
have both financial and nonfmancial significance to human welfare, determining appropriate
monetary measures of the total effect on human welfare is a difficult task. The uncertainty in
the monetary estimates is probably greatest for premature mortality risks. Sources of
uncertainty in all the monetary estimates are discussed in Chapter 5, but here we highlight
two key uncertainties in the monetary value estimates for premature mortality risks.
The first source of uncertainty in the monetary estimates for premature mortality is that there
is little empirical economic evidence available about how health status or life expectancy
affects an individual's willingness to pay for changes in risks of premature death. Available
willingness-to-pay estimates for changes in risks of death are drawn primarily from samples of
adults of average age distributions and average health status. It is possible that many of those
November 10, 1995
-------
OVERVIEW OF THE ASSESSMENT * 2-19
at greatest risk of premature mortality because of air pollution exposure are elderly or in
relatively poor health. The available empirical evidence on this question is discussed in
Chapter 5, but considerable uncertainty remains.
The second source of uncertainty in the monetary estimates for premature mortality is that
most of the available estimates are for changes in the risks of accidental death rather than
death due to illness, which is more the issue for pollution exposure. This is because the
economic literature concerning monetary values for changes in risks of death has been able to
exploit available data on wage differentials as a function of different levels of on-the-job risks
of fatalities. It is uncertain whether individuals might have different reactions to risks due to
illness rather than accidents, and how this might affect willingness to pay to avoid or reduce
such risks. There is some evidence that risks of death due to particularly feared illnesses, such
as cancer, are considered more abhorrent than risks due to accidents, but that evidence is
limited.
November 10, 1995
-------
CHAPTER 3
CHANGES IN AMBIENT OUTDOOR SULFATE CONCENTRATIONS
This chapter presents the approaches used in this assessment to estimate the changes in
ambient outdoor sulfate aerosol concentrations by location attributable to the Title IV required
S02 emissions reductions. This chapter relies on available results from other analyses
conducted for or by the U.S. EPA for estimates of changes in SO2 emissions and changes in
ambient sulfate aerosol concentrations. In this chapter, we briefly describe these other analyses
and explain how we use the results in this analysis.
3.1 CHANGES IN SO2 EMISSIONS
ICF Resources (1994) has prepared for the U.S. EPA estimates of current and future S02
emissions by location through 2010 for a Title IV implementation scenario and for a no Title
IV scenario. The ICF Resources analysis focuses on the SO2 emissions in the utility sector,
where 85 percent of the Title IV required emissions reduction is expected. This health benefits
assessment incorporates, without modification, the ICF Resources annual SO2 emissions
estimates for the eastern United States.
The analysis uses ICF Resources' Coal and
Electric Utilities Model (CEUM). CEUM is
a large linear programming model that
develops least-cost compliance options
across the utility industry in meeting SO2
reduction targets. The model considers in
detail the interaction between the demand
for different types of fuels and the costs of
supplying and delivering the fuels, as well
as the interaction between utilities' marginal
costs of compliance and the projected
amount of allowance "banking."
CEUM uses a series of selected economic,
energy market, and utility sector
assumptions. These assumptions play an
important role in estimating emissions with and without Title IV, because factors such as
substitute fuel prices, energy demand, and economic growth can all have significant effect on
decisions by utilities about building new capacity or retrofitting plants for alternative fuel use.
Basic Features of CEUM
set of interrelated models and databases for analyzing
the coal and electric utility industries in an integrated
way
cost-minimizing linear programming model
S02 emissions is one key output: others include NO,
emissions, environmental compliance information (e.g.,
compliance costs, coal market impacts, numbers of
scrubbers used), power plant operational choices (e.g..
new plants built, fuel choice)
incorporates technical and economic relationships of
coal and electric utility markets
high degree of resolution:
• most generating units represented individually
• detailed coal supply, transportation, transmission
and utility demand segments.
November 10, 1995
-------
CHANGES IN AMBIENT OUTDOOR SULFATE LEVELS * 3-2
Emission levels are directly related to levels of electricity production, fuels used, and
compliance options employed.
Figure 3-1 shows the ICF Resources estimates of utility SO2 emissions with and without Title
IV from 1990 through 2010. Maximum allowed SO2 emissions are fairly well defined by the
Title IV requirements. There is some uncertainty about how quickly the Title IV emission
reduction goals will be met because there are provisions that allow utilities to bank unused
emissions allowances and use them at a later time. It is uncertain how much banking the
utilities will choose to do, but ICF Resources estimates that all banked allowances will be
exhausted by 2010. Uncertainty also exists in predicting the specific location of emissions
reductions because emissions allowances can be traded among emitting facilities.
Table 3-1 shows the ICF estimates of annual SO2 emissions by state for 1997 and 2010, with
and without Title IV. Both of the with Title IV estimates include an estimated response of
utilities to the opportunities provided in the Title IV program to reduce emissions more than
required in the early years of the program and to bank these as emission allowances for future
use within a limited time period. The results of the with and without Title IV forecasts show
that even with Title IV there are a few locations where SO2 emissions are expected to
increase slightly. However, there is expected to be a significant reduction in total emissions.
In 2010, with Title IV, total SO2 emissions from utilities in the East are expected to be about
7.7 million tons versus an estimated 16.8 million tons in 2010 without Title IV. The without
Title IV emissions estimates do reflect emissions reductions expected due to other Clean Air
Act Amendment requirements.
As noted in Chapter 2, there is more uncertainty in predicting what emissions would have
been in the absence of Title IV than for the with Title IV scenario. Total emission limits are
set by Title IV and utilities (as a group) are not expected to emit less than they are allowed
under Title IV, because the Title IV limits are well below 1990 emission levels. In the
absence of Title IV, there are some factors that would cause future S02 emissions to rise and
some that would cause S02 emissions to decline. In general, economic and population growth
results in greater demand for electricity, which may result in higher S02 emissions. At the
same time, as older plants are retired and cleaner electricity generation processes are
developed, S02 emissions per unit of electricity generated can be expected to decline. How
emissions would change, therefore, depends on the relative significance of these different
factors. ICF Resources estimates that in the absence of the Title IV requirements, SO2
emissions from utilities would have risen slightly from 1990 levels. They predict a slight rise
would have occurred between 1995 and 2005, and then a fairly flat trend through 2010.
Current S02 emissions vary considerably by location in the eastern United States in part
because of significantly different amounts of high sulfur content fuels used in different
locations. The reductions in S02 emissions expected as a result of Title IV are concentrated in
areas that currently have the highest SO2 emissions. Table 3-2 shows the ICF Resources
estimates of the reduction in annual S02 emissions attributable to Title IV in 2010 by state for
31 eastern states. The last column shows the emissions reduction per capita in each state.
November 10, 1995
-------
Figure 3-1
U.S. Utility SO2 Emission Levels: 1990 through 2010
I
i
.-g-
1
24
22 h
20
18
14
12
10
8
2
0
No Title IV Case
Banked Allowances
With Title IV Case
Use of Banked
Allowances
1990
1995
2000
2005
L
2010
O
DC
O
m
c/i
O
O
o
73
05
C
r-
m
m
m
C/l
-------
CHANGES IN AMBIENT OUTDOOR SULFATE LEVELS * 3-4
Table 3-1
EPA Forecasts of Annual Utility SO2 Emissions
(thousand tons) by State1
State
Maine, Vermont, New Hamp.
Mass., Conn., R.I.
New York
Pennsylvania
New Jersey
Maryland, Delaware, D.C.
Virginia
West Virginia
North Carolina,
South Carolina
Georgia
Florida
Ohio
Michigan
Illinois
Indiana
Wisconsin
Kentucky
Tennessee
Alabama
Mississippi
Minnesota
owa
Missouri
Arkansas
Louisiana
Total 31 Eastern States
1985
87
308
413
1,174
102
285
131
951
499
998
531
2,217
409
1,045
1,496
380
783
802
534
102
111
198
961
73
79
14,672
1997 (with
Title IV)
43
175
309
991
102
336
233
629
754
577
542
1,187
428
637
738
269
531
574
478
94
140
185
455
85
104
10,596
1997 (no
Title IV)
43
175
338
1,120
131
340
225
965
719
912
748
2,455
427
901
1,360
248
817
920
. 661
160
140
245
897
85
104
15,137
2010 (with
Title IV)
46
164
259
625
115
217
159
569
547
414
517
690
370
460
536
180
386
297
379
94
104
139
308
93
71
7,740
2010 (no
Title IV)
54
189
346
1,178
164
430
264
1,085
866
919
900
2,399
397
1,199
1,559
397
967
1,074
681
163
136
266
944
93
99
16,769
1 Emissions estimates from ICF Resources (1994).
November 10, 1995
-------
CHANGES IN AMBIENT OUTDOOR SULFATE LEVELS > 3-5
Table 3-2
Estimated Reduction in Annual Utility SO2 Emissions in 2010 Attributable
to Title IV by State
State
Maine, Vermont, New Hamp.
Mass., Conn., R.I.
New York
Pennsylvania
New Jersey
Maryland, Delaware, D.C.
Virginia
West Virginia
North Carolina, South Carolina
Georgia
Florida
Ohio
Michigan
Illinois
Indiana
Wisconsin
Kentucky
Tennessee
Alabama
Mississippi
Minnesota
Iowa
Missouri
Arkansas
Louisiana
Emissions
Reduction in 2010
(1000 tons)1
7
25
87
553
48
213
105
516
319
506
384
1,709
27
738
1,022
217
581
777
301
69
31
127
637
0
28
Population 1990
(1000s)
2,900
10,306
17,990
11,882
7,730
6,054
6,187
1,793
10,116
6,487
12,938
10,847
9,295
11,431
5,544
4,892
3,685
4,877
4,041
2,573
4,375
2,777
5,117
2,351
4,220
Reduction per
Capita
(10'z tons/person)
0.24
0.24
0.48
4.65
0.62
3.52
1.70
28.78
3.15
7.80
2.97
15.76
0.29
6.46
18.43
4.44
15.77
15.93
7.45
2.68
0.71
4.57
12.45
0.00
0.66
1 Emissions estimates from ICF Resources (1994). Projected 2010 reductions are the difference
between emissions with and without Title IV.
November 10, 1995
-------
CHANGES IN AMBIENT OUTDOOR SULFATE LEVELS > 3-6
It is clear that a large variability in emissions reductions by location persists even after
accounting for differences in population. The largest reductions are in the Appalachian and
Midwest regions.
3.2 CHANGES IN SULFATE AEROSOL CONCENTRATIONS
The pollutant of interest in this health benefits assessment is sulfate aerosol, which is a
secondary pollutant formed in the atmosphere in the presence of gaseous SO2 emissions and
other atmospheric constituents. The location and amount of SO2 emissions are two factors that
determine sulfate aerosol concentrations. Other factors are weather conditions, wind speed and
direction, and the presence and quantities of other elements in the atmosphere that interact
with SO2 to form sulfate aerosols.
For this analysis, we use results from
EPA's Regional Acid Deposition Model
(RADM), which include estimates of
ambient sulfate aerosol concentrations for
alternative S02 emissions scenarios'. Chang
et al. (1990) provide a detailed description
of RADM, and Dennis et al. (1990, 1993)
provide results of evaluations of RADM.
Airborne sulfate aerosol concentrations are
an intermediate result provided by RADM
for the purposes of estimating the eventual
deposition of acidic species. RADM reports
results, including ambient sulfate aerosol
concentrations, for grid cells 80 km by 80
km in size, over the entire area of the
eastern United States. S02 emission rates by
location, as estimated by ICF Resources, are
an input into RADM. The RADM estimates
used in this health benefit assessment are
the ground-level sulfate aerosol (SO4)
concentrations for the following SO2
emissions scenarios:
* Actual 1985 emissions, used to
approximate conditions when the
1990 Amendments went into effect
The Regional Acid Deposition Model
The RADM is a comprehensive model of the atmospheric
processes that lead to the formation and deposition of acidic
species. The objective of this modeling system is to provide a
scientific basis for estimating the change in deposition caused by
large changes in precursor emissions. Specifically, the RADM is
designed to (I) mathematically represent the nonlinear dynamics
both of oxidant formation from precursor emissions of NO, and
VOCs, and of scavenging of sulphur compounds, and (2)
mathematically represent the three-dimensional dynamics of
transport, transformation, and deposition, including effects of
cloud processes. The version of the model used for this analysis
(Version 2.6) is designed to report this information on grid cells
80- x 80-km in size, over a domain that extends from east of
central Texas to the south of James Bay, Canada, including all of
Florida and southeastern Canada. This version of RADM uses six
vertical layers from the ground to approximately 16 km in altitude.
Version 2.6 has been corrected for some under predicting of
sulfate levels that occurred with earlier versions.
The model operates on a mathematical frame of reference in
which concentrations are specified as functions of time at fixed
positions within the grid cells. The RADM uses the wind flow and
precipitation simulated by a mesoscale meteorological model,
called the MM-4, over an episodic period chosen to be 3 tlays.
Modules of various chemical and physical processes involving the
transport, transformation, and removal of pollutants are included In
RADM and they utilize the meteorological simulations obtained
from the MM-4. Because each run of the RADM represents a
3-day episode, a method to produce seasonal and annual estimates
using a sample of episodic runs is required. Each episode is
weighted according to Us relative importance toward seasonal and
annual wet deposition. RADM is run in each episode, and the
results are multiplied by the weighing factors to produce seasonal
and annual deposition calculations.
Estimated 1997 emissions with Title IV and banking
November 10, 1995
-------
Figure 3-2
RADM 50th Percentile Annual Sulfate Concentration
1985 Base Case
O
m
CD
00
r
m
CO
T
-------
Figure 3-3
RADM 50th Percentile Annual Sulfate Concentration (/ig/m3)
1997 with Title IV1
o
m
O
I
>
z
a
m
t/i
eg
Fn
H
O
c
o
o
o
73
00
C
r
-r,
H
m
r
m
m
r
C/5
T
oo
-------
Figure 3-4
RADM 50th Percentile Annual Sulfate Concentration (jig/m3)
2010 without Title IV
6
re
3
"
n
x
m
5
m
O
C
a
o
o
00
C
-n
>
m
T
t>J
-------
Figure 3-5
RADM 50th Percentile Annual Sulfate Concentration (pg/m3)
2010 with Title IV
2
n
I
0
n:
>
o
m
O
•n
>
H
m
r
m
P
en
I
t—t
o
-------
CHANGES IN AMBIENT OUTDOOR SULFATE LEVELS »• 3-11
+ Estimated 2010 emissions with Title IV
» Estimated 2010 emissions without Title IV.
Figures 3-2 through 3-5 illustrate the distribution of the RADM sulfate aerosol concentration
estimates across the eastern United States for each of the SO2 emissions scenarios.
RADM results used in this assessment are summarized in Table 3-3. Table 3-3 gives the
estimated reduction in median annual SO4 concentrations for 1997 with Title IV and emission
allowance banking versus the SO4 concentrations under current (1985) conditions and for
2010 with Title IV versus predicted SO4 concentrations without Title IV. These are ground-
level SO4 reductions for the 50th percentile of the annual distribution of estimated SO4
concentrations. The results in these tables are the averages of the changes in the 50th
percentile concentrations by state based on the results for the 80 km by £0 km RADM grid!
Exposures and health effects are calculated at the grid cell level in this assessment, but
averages for the states are shown here because the grid level data are too numerous.
The partial states at the western edge of the RADM grid, as shown in Figures 3-2 through 3-5
have been dropped from the quantitative assessment because the sulfate concentration changes
expected in this area are small. The RADM grid also covers the southern parts of several
Canadian provinces. Significant changes in sulfate concentrations are predicted as a result of
the expected reductions in SO2 emissions in the United States for Ontario and Quebec, so
these.have been included in the assessment. The portions of these provinces covered in the air
quality model include the areas where the vast majority of the populations of these provinces
live. The northern edge of the RADM grid is just south of the southern edge of James Bay.
3.3 MATCHING POPULATION TO ATMOSPHERIC SULFATE CHANGES
To calculate the human health benefits associated with the expected reduction in atmospheric
sulfate aerosols concentrations, it is necessary to determine the change in ambient outdoor
sulfate concentrations where people are. This requires an overlay of the population distribution
on the RADM grid to match numbers of people to the estimated changes in sulfate aerosol
concentrations.
For this analysis, we use the Geographic Information System (GIS) to match the 1990
population data from the U.S. Census (1990) and the 1991 Canadian Census to the RADM
grid, and to estimate the populations in each relevant age group residing in each of the 1330
RADM grid cells. EPA provided us with the latitude-longitude coordinates for the center of
each RADM grid cell. These were projected into lambert projected meters using standard
parameters for lambert conformal projections of the United States. This gave us an orthogonal
grid of points. We then used the THIESSEN procedure to draw grid cell boundaries
equidistant between each pair of grid cell points.
November 10, 1995
-------
CHANGES IN AMBIENT OUTDOOR SULFATE LEVELS + 3-12
For the U.S., the latitude-longitude coordinates for each centroid of each census block group,
as provided on U.S. Census Summary Tape File 3A, were then located on the RADM grid.
For Canada, the latitude-longitude coordinates for each centroid of each enumeration area, as
provided by Maplnfo Corp. under license from Statistics Canada, were then located on the
RADM grid. Total population, divided into relevant age groupings for the health effects
calculations, for each block group or enumeration area was assigned to the grid cell within
which the block group or enumeration area centroid was located.1 The error in assuming that
all the population is located at the centroid of the block group or enumeration area is small
given that the block groups and enumeration areas are small relative to the size of the RADM
grid cells. There are about 300,000 block groups in the study area, each with a total
population of about 670. An enumeration area usually contains about 125 dwellings in a rural
area and 375-400 dwellings in an urban area.
State or province identifiers for each block group or enumeration area were used to sum to
state or province2 level results after health effects estimates were calculated for each RADM
grid cell, based on the differences in predicted sulfate concentrations for the cell under
different scenarios.
This assessment estimates health benefits for changes in sulfate concentrations in 1997 and in
2010. The 1990 populations are therefore adjusted for expected average population growth
using the mid-forecasts of the U.S. Census and the World Bank population projections for
Canada. These adjustments are made at the aggregate level using national average population
growth factors.
1 Block group specific age data were used for the U.S. population. For the Canadian population,
country average age distributions (Statistics Canada, 1994) were applied uniformly to each enumeration
area.
2 The RADM grid covers virtually all of Ontario's population, but not all of Quebec is covered. The
population of Quebec used in the assessment only includes those persons living in enumeration areas
covered by the RADM grid. Approximately 99 percent of Quebec's population is included.
November 10, 1995
-------
CHANGES IN AMBIENT OUTDOOR SULFATE LEVELS * 3-13
Table 3-3
Average Reductions in Median Annual SO4 Concentrations (Mg/m3) by State/Province
Due to Title IV
State/Province
Alabama
Arkansas
Connecticut
Delaware
District of Columbia
Florida
Georgia
Illinois
Indiana
Iowa
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
19971
0.44
0.22
0.35
0.22
0.30
-0.02
0.31
0.31
0.53
0.00
0.86
0.08
0.11
0.41
0.24
0.11
-0.03
20102
1.93
0.54
0.26
0.86
1.48
1.01
1.88
0.80
1.28
0.21
2.02
0.70
0.15
1.29
0.24
0.29
0.05
State/Province
Mississippi
Missouri
New Hampshire
New Jersey
New York
North Carolina
Ohio
Pennsylvania
Rhode Island
South Carolina
Tennessee
Vermont
Virginia
West Virginia
Wisconsin
Ontario
Quebec
1997
0.24
0.16
0.21
0.22
0.29
0.30
0.51
0.44
0.41
0.24
0.84
0.21
0.42
0.72
0.03
0.13
0.09
2010
1.01
0.45
0.16
0.68
0.34
1.73
1.43
0.92
0.31
1.82
2.09
0.20
1.75
2.08
0.20
0.13
0.05
1 The 1997 reduction is estimated versus 1985 emissions.
2 The 2010 reduction is estimated versus 2010 without Title IV emissions.
November 10, 1995
-------
CHAPTER 4
QUANTIFICATION OF HEALTH EFFECTS CHANGES
4.1 BACKGROUND ON HEALTH EFFECTS STUDIES
Several different types of health effects studies are used to measure health responses to
environmental pollutants. Different types of studies provide different types of information.
Each type of study has different strengths and weaknesses, including variations in the types of
health effects that can be considered. Some types of evidence are better suited for use in a
quantitative assessment. The brief background review of health effects studies provided in this
section is intended to help place the assessment in context for those policy makers who may
not be familiar with the health effects literature. Strengths and weaknesses of the assessment
are integrally linked to those of the scientific literature upon which the quantitative assessment
is based.
4.1.1 Types of Health Effects Studies
The types of studies that provide evidence of health effects following exposure to sulfate
aerosols include epidemiology and field studies, human clinical studies, and laboratory and
toxicology studies.
Epidemiology and Field Studies
Epidemiological and field studies for sulfate aerosols typically involve estimation of a
statistical relationship between the frequency of specific health effects observed in a study
population in its normal environment and sulfate aerosol concentrations measured at stationary
outdoor monitors in the study area. These studies are therefore able to provide "concentration-
response" functions that can be used to estimate the change in the frequency of health effects
for a population in its normal environment that would be expected to occur with specific
changes in ambient outdoor sulfate aerosol concentrations. A concentration-response function
is a quantitative relationship between ambient levels (concentration) of a pollutant and the
frequency of specific health effects in a given time period (response). For example, it may
give the percentage of study subjects who report cough symptoms on a given day as a
function of the concentration of ambient sulfate aerosol on mat day.
Epidemiology and field studies often involve time-series analyses of changes in rates of health
outcomes within a specific area, sometimes for a pre-selected sample, as air pollution
concentrations fluctuate. An example of this study design would be daily observation and
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES + 4-2
recording of asthma symptoms with a pre-selected group of subjects with diagnosed asthma,
and statistical analysis to determine if there is an association between the frequency of the
symptoms and fluctuations in sulfate aerosol concentrations from day to day. Epidemiology
studies may also use cross-sectional data, looking at differences in health outcomes across
several locations at a selected point or period of time. This may involve, for example, a
comparison of the prevalence of chronic respiratory disease in different cities with different
average sulfate aerosol concentrations. Although cross-sectional studies have the advantage of
being able to consider potential effects associated with long-term exposures, it can be very
difficult to fully control for potential confounding factors. Time-series studies reduce many of
the problems associated with confounding or omitted variables because the same population
group is studied over time, but weather and seasonal variation that may be correlated with
sulfate aerosol concentrations can pose some similar problems.
Time-series and cross-sectional epidemiology studies can be either cohort studies or
population studies. Cohort studies analyze the incidence of health effects in a sample of
identified individuals usually selected specifically for the study. For example, a cohort might
be a group of study subjects who record daily symptoms for a period of time. A cohort study
might also collect data on the health status of a selected sample of individuals and then do a
follow-up on the same individuals after a specified length of time to determine what changes
in health status have occurred for each individual. Population studies, on the other hand, rely
on data available for the population as a whole rather than tracking the effects on specific
individuals. For example, a population study may analyze daily mortality rates in a given
location as they related to daily particulate matter concentrations. Another example of a
population study is a comparison of the prevalence of chronic respiratory disease in different
locations with different average pollution concentrations. In general, cohort studies are
preferred because characteristics of the individuals in the study sample can be determined,
allowing better control for other risk factors, such as smoking or diet. Population data are,
however, readily available for many types of health effects and therefore provide an
opportunity to conduct epidemiology analyses very cost effectively.
One of the strengths of epidemiology studies is that they analyze actual health effects in
human populations at ambient pollution concentrations. Subjects are studied in their normal
environment and the health effects are directly observed. A major challenge for epidemiology
studies is the difficulty in isolating with confidence the effects of a specific air pollutant such
as sulfate aerosol when this may be just one of many complex factors that influence human
health. Finding a statistically significant correlation between a health effect and exposure to
sulfate aerosols does not prove causality. To support an inference of causality, epidemiology
results need to be supported by repeated observation in different studies and by biological
plausibility and consistency with evidence from other types of health effects studies.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-3
Human Clinical Studies
Human clinical studies, sometimes also called chamber studies, examine the response of
human subjects to pollutant exposures in a controlled laboratory setting. The response of the
individual can be monitored and the environment controlled so that the effects of one
pollutant can be isolated. Clinical studies on sulfate aerosols have typically exposed subjects
to specific sulfate aerosol concentrations for one or several hours and measured responses
such as pulmonary function or respiratory symptoms, sometimes in combination with
moderate or vigorous exercise. Clinical results can provide evidence of causation because
confounding factors are well controlled.
Clinical studies are limited to consideration of short-term reversible health effects that can be
purposely induced in human subjects. Also, the health effects of short-term exposure to sulfate
aerosols may be different in the everyday environment where other pollutants are also present
and the individual's behavior and activities are quite varied. Clinical results are also limited
for the purposes of extrapolation and generalization because sample sizes are usually quite
small, making generalization somewhat difficult. Clinical results, in combination with
supporting epidemiology results, can support an inference of causality between pollution
exposure and observed health effects.
Laboratory and Toxicology Studies
Laboratory and toxicology studies use animal subjects, and sometimes human tissue or cells,
to study biological responses to pollutants in a controlled laboratory setting. Animal organs
and tissue can be directly examined for effects of acute and chronic exposures, revealing a
wealth of information about biological responses to sulfate aerosol exposures. These studies
provide a great deal of useful and important information about the specific biological
pathways and mechanisms by which pollutants cause harm to living organisms. For example,
laboratory studies may provide direct biological evidence of how a pollutant decreases the
ability of a living organism to defend against disease and infection.
4.1.2 Advantages and Limitations for Assessment Purposes
For a quantitative assessment of the human health benefits of Title IV, we want answers to
the following questions:
* How many cases of each specific type of health effect will be avoided because
of Title IV?
* How much does the exposed population value the reduction in these health
effects?
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-4
Epidemiology Advantages and Limitations
For addressing the first question, many epidemiology studies provide sufficient information to
infer a concentration-response function, which typically gives a quantitative relationship
between the incidence of a given health effect and ambient outdoor air pollutant
concentrations. A concentration-response function can be used to predict a change in the
number of cases of a given health effect for an estimated change in ambient outdoor pollutant
concentration.
Epidemiology-based concentration-response functions available in the literature pertaining to
airborne sulfates correlate observed changes in health status or symptoms with ambient
outdoor sulfate concentrations. Everyday human activity patterns and specific pollutant doses
associated with specific outdoor pollutant concentrations are implicit in the concentration-
response functions.-Changes in incidence of specific health effects can therefore be estimated
as a function of changes in ambient outdoor pollutant concentrations without conducting
detailed pollution exposure modeling, as long as we accept the assumption that the human
activity patterns will not change significantly when ambient outdoor pollutant concentrations
change. There are implicit assumptions in this approach that the relationship between outdoor
concentrations and individual exposure that exist in the original study populations are the
same in the assessment population.
Epidemiology studies are also useful in addressing the second question because they are able
to define health effects in terms of factors that can be directly related to perceived welfare,
such as risks of premature death or days with noticeable respiratory symptoms. By drawing on
available health data such as vital statistics and national health surveys, or observing changes
in health over time for a panel of study subjects, epidemiology studies are able to consider a
wide range of health effects. This includes very serious health effects such as premature
mortality or chronic disease that are not possible to study with human subjects in controlled
exposure environments. Epidemiology studies can be designed to consider potential effects of
long-term exposures to air pollutants as well as short-term effects. This is an additional
advantage over clinical studies.
The primary limitation in using epidemiology results for predicting changes in health effects
as a function of changes in ambient air pollutant concentrations is the uncertainty about
whether the causal factors for the observed association with health effects have been fully and
accurately specified. Inaccurate predictions could occur, for example, if the actual causal
factor is some unspecified pollutant that is highly correlated with the specified pollutant.
Thus, a change in the specified pollutant would not necessarily result in a change in health
effects, unless the unspecified pollutant were to also change in the same proportion. This
source of uncertainty is always present in epidemiology results to some extent, and the
potential for error is difficult to quantify because the extent of unspecified, and correlated,
causal factors is unknown.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES > 4-5
There may also be other important inaccuracies in the specification of the relationship
between health effects and ambient air pollutant concentrations in an epidemiological study.
Functional forms for the concentration-response relationship and averaging times for pollutant
concentrations are examples of things that might be misspecified, resulting in inaccuracies in
predictions of health effects changes. For example, if a linear relationship is specified but the
actual relationship is significantly nonlinear, the predicted estimates could be either too high
or too low, depending on the shape of the actual relationship. These sources of uncertainty are
also difficult to quantify because the extent of the error in the original epidemiological
specifications is unknown. Uncertainties are greater if epidemiology results are extrapolated
beyond the range of concentrations over which the original results were estimated.
Clinical Advantages and Limitations
Clinical study results provide information about the relationship between exposure and health
response obtained in a controlled environment. Thus, concerns about whether the observed
relationship is actually causal are reduced. Relationships between exposure and response are
more accurately measured in the controlled environment of the clinical study. This is in
contrast to most epidemiology studies which use ambient outdoor pollutant concentrations as
the measure of exposure. Clinical studies therefore provide potentially more accurate, or at
least more convincing, dose-response information than is obtained with epidemiology studies.
This is an important advantage of clinical study results for quantitative assessment purposes.
There are two significant limitations of clinical study results when it comes to a quantitative
assessment of changes in health effects as a function of changes in outdoor pollutant
concentrations. First, clinical study results for quantitative assessment purposes are limited
because only a small range of exposures and potential health effects can be considered in
clinical studies. Clinical studies are generally confined to short-term exposures and to health
effects that are reversible and not life-threatening. It is simply not possible to confine human
subjects to controlled environments for extended periods of time or to attempt to induce
permanent or life-threatening health effects. Clinical results are therefore unable to provide
information on the full range of potential health effects of pollutant exposures, if long-term
exposures or permanent or life-threatening health effects are suspected.
Second, using clinical study results requires some type of quantitative exposure analysis to
link changes in outdoor ambient air quality to pollutant exposures as measured in the clinical
(indoor) study setting. This typically requires some analysis or assumptions about how much
time people spend in various environments (e.g., indoor, outdoor, automobile) and about the
relationship between outdoor pollutant concentrations and pollutant concentrations in each of
the other types of human environments. Thus, several extra steps are added to the analysis
relative to what is needed when epidemiology study results based on outdoor pollutant
concentrations are used.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES + 4-6
Two quantitative assessments conducted recently for changes in ambient ozone concentrations
provide examples of assessments that have used some available clinical-based dose-response
information for acute respiratory symptoms as a function of controlled ozone exposures
(Krupnick and Kopp, 1988; Hall et al., 1989). These studies illustrate the difficulty in
applying clinical dose-response functions because of the need for detailed exposure analysis;
they also provide an interesting comparison between results obtained using epidemiology-
based concentration-response functions and clinical-based dose-response functions for the
same type of health effect: acute respiratory symptoms.
The clinical studies used in these two assessments provided data on whether respiratory
symptoms occurred with subjects exposed to controlled concentrations of ozone while
exercising for one, two, or seven hours. To utilize these results to estimate how a change in
ozone concentrations in the ambient outdoor air would affect the frequency of respiratory
symptoms for a population in its everyday environment, either extensive modeling or
assumptions must be used regarding population activity patterns and resultant ozone
exposures. Hall et al. (1989) developed a detailed ozone exposure model for the South Coast
Air Basin. They found that estimates of the frequency of respiratory symptoms based on the
clinical results and the exposure modeling were higher per unit of ambient outdoor ozone
relative to the results obtained using available epidemiology study results. Krupnick and Kopp
(1988) did not conduct detailed exposure modeling, but rather used a range of alternative
assumptions regarding activity patterns and exposures. They obtained estimates that were
either higher or lower than the epidemiology-based estimates, depending on the assumptions
used in the exposure portion of the analysis.
Laboratory Advantages and Disadvantages
Laboratory study results have the same advantage as that discussed above for clinical study
results: pollutant exposures are well controlled in a laboratory setting and variations in
confounding factors are reduced. The analyst therefore has more confidence that the observed
relationships are causal and that the measured dose-response functions are accurate.
Laboratory studies also have the potential to consider the effects of long-term as well as short-
term exposures, which extends the range of health effects that might be considered relative to
clinical studies.
Laboratory study results have three important limitations when it comes to a quantitative
assessment of changes in health effects as a function of changes in pollutant emissions.
Similar to clinical study results, using laboratory study results requires some type of
quantitative exposure analysis to link changes in outdoor ambient air quality to pollutant
exposures as measured in the laboratory study setting. Thus, an extra step is added to the
analysis relative to what is needed when epidemiology study results are used. Second,
laboratory studies often use animal subjects, which introduces considerable uncertainty when
attempting to extrapolate quantitative results to human populations, as is needed in a
quantitative assessment. Third, laboratory studies sometimes focus on health effects that are
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES »• 4-7
difficult to interpret in terms of specific illnesses or symptoms. Linkages between cellular and
biochemical concentration changes and clinical manifestation of illness are often difficult to
quantify.
4.2 SUMMARY OF HEALTH EFFECTS EVIDENCE FOR SULFATE AEROSOLS
This section provides a brief summary of the available health effects evidence concerning
sulfate aerosols and other fine particulates (PM2.5). This summary is not intended to be a
comprehensive review. Its purpose is to highlight the range of available evidence, list the
kinds of health effects that have been observed, and to focus specifically on health effects that
have been found in association with sulfate aerosols because they are the focus of this
assessment. Many of the health effects listed here have also been found to be associated with
PM2J concentrations in locations where sulfate concentrations are low, so none of the findings
reported here and elsewhere in this report should be interpreted as suggesting that sulfates are
the only harmful constituent of PM2 5.
Detailed reviews of available health effects evidence for inhalable particulate matter, including
sulfate aerosols, covering results from laboratory, clinical, and epidemiology studies, are
provided in the EPA criteria documents and other documents (U.S. EPA, 1982, 1986a, 1986b,
1989, 1995). Additional reviews of part or all of this literature include Ferris (1973), Graham
et al. (1990), American Thoracic Society (1991), Gong (1992), Folinsbee (1992), and Lipfert
(1994).
4.2.1 Epidemiology Study Findings
A detailed discussion of epidemiology study findings is presented in Section 4.4, including
identification of specific concentration-response functions selected for use in this assessment.
This section gives an overview of the types of health effects that have been observed in
epidemiology studies concerning sulfate aerosols and PM2 5.
Epidemiology studies conducted to date provide evidence of statistically significant
associations between ambient outdoor concentrations of sulfate aerosols or PM2 s, or both, and
the following human health effects:
» Premature mortality. Evidence has been found in prospective cohort and cross-
sectional studies of an association between mortality rates in different locations and
average sulfate concentrations in those locations (e.g., Pope et al., 1995). Evidence has
also been found in time-series studies of an association between daily mortality rates
and sulfate concentrations in several urban areas in the United States and elsewhere
(e.g., Dockery et al., 1992).
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES »• 4-8
Chronic respiratory disease. Prospective studies have found higher rates of chronic
respiratory disease in locations with higher PM2 5 concentrations (e.g., Abbey et al.,
1995).
Hospital admissions. Time-series studies show a correlation between daily hospital
admission rates and daily sulfate concentrations (e.g., Burnett et al., 1995).
Aggravation of asthma symptoms. Time-series studies with panels of diagnosed
asthmatics who record their symptoms and medication usage each day have found an
association between the aggravation of asthma symptoms and daily sulfate
concentrations (e.a.. Ostro et al.. 1Q91V
ooauwiauim uviwtsu UK> aggiavauuu ui <
concentrations (e.g., Ostro et al., 1991).
*• Restricted activity days. Self-reported number of days on which activities are
restricted because of illness during a 14-day recall are recorded in a national sample
through the Health Interview Survey. The frequency of such days has been found to be
significantly associated with the average PM2 5 concentrations in the city of residence
during the same 14-day period (e.g., Ostro and Rothschild, 1989).
> Acute respiratory symptoms. In a study during which a panel of healthy subjects
recorded daily respiratory symptoms, the frequency of such symptoms was found to be
correlated with daily sulfate concentrations in the study location (e.g., Ostro et al.,
1993).
Taken as a whole, the available epidemiology evidence shows a strong relationship between
sulfate aerosols, and other fine particulates, and respiratory-related illness in the United States.
The types of illness range from severe acute and chronic illnesses that are associated with
increases in risks of death to mild acute symptoms such as coughing and wheezing. There is
epidemiology evidence of health effects for both short-term fluctuations in sulfate
concentrations within a given location and long-term variations in sulfate concentrations
across locations.
4.2.2 Clinical Study Findings
Several studies have examined the health effects of humans exposed briefly through inhalation
to moderate concentrations of sulfate aerosols hi the form of sulfuric acid (e.g., Amdur et al.,
1991; Koenig et al., 1993). The effects observed in some of these acute exposure studies
include decreased pulmonary function and decreased bronchial clearance rates.1 Graham et al.
(1990) review this literature and conclude that acute exposures to some acidic sulfates can
increase airway resistance, decrease pulmonary function, and increase responsiveness to
1 This section provides just a brief overview of clinical study findings and draws upon the summary
of clinical research on this topic provided by Admur et al. (1991).
—, '. November 10, 1995 —
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES »• 4-9
bronchoconstrictors, especially in asthmatics, but that considerable variability in the results of
different studies suggests uncertainty about which exposures will reproducibly cause these
effects.
Decreased pulmonary function in the form of increased airway resistance has been noted in
some studies for both adult and adolescent asthmatics following inhalation exposure to
sulfuric acid during exercise (Amdur et al., 1991; Koenig et al., 1993). In adult asthmatics,
inhalation exposure to 450 ng/m3 sulruric acid for 16 minutes resulted in an increase in
airway resistance, whereas exposure to 100 ng/m3 caused no response (Amdur et al., 1991). In
adolescent asthmatics, inhalation exposure to 68 ng/m3 sulruric acid for 40 minutes resulted in
increased airway resistance. The increase in airway resistance of adolescent asthmatics was
greater following exposure to a combination of sulfuric acid and 0.1 ppm sulfur dioxide
(Amdur et al., 1991). No increase in airway resistance was observed following acute
inhalation exposure of nonasthmatics to 1,000 iig/m3 sulruric acid.
Bronchial clearance, a major defense mechanism employed by the body following inhalation
of irritant particles, decreased following inhalation exposure to moderate concentrations of
sulfuric acid. Studies in humans show that inhalation exposure to less than 200 ng/m3 sulfuric
acid for one hour stimulates clearance in larger airways; however, clearance is depressed in
small airways where more acid deposits. Clearance is restricted hi both small and large
airways following exposure to 1,000 ug/m3 sulfuric acid (Amdur et al., 1991). At exposure
concentrations of 100 jag/m,3 increasing the exposure time from one to two hours results in an
even greater decrease in bronchial clearance, and a persistent reduction in clearance of
particles for up to three hours following exposure (Amdur et al., 1991).
The effects noted (i.e., decreased bronchial clearance and decreased respiratory ability) are
similar to some of the effects observed in epidemiological studies, including increased
incidence of acute respiratory symptoms and depressed lung function (American Lung
Association, 1978). However, as described in Section 4.1, these studies are limited for the
following reasons:
» Exposure was limited to either sulfuric acid or a combination of sulfuric acid and
sulfur dioxide. Synergistic or antagonistic interactions between air pollutants would not
be represented in these studies.
» Because of the inherent and understandable limitations of clinical studies, health effects
resulting from chronic exposure to sulfate aerosols cannot be observed in clinical
studies. It is possible that effects over a longer duration would be more pronounced.
Continuous exposure to ambient aerosols results in the simultaneous deposition and
redistribution of particles, causing changes such as marked and persistent depression in
bronchial clearance, whereas acute exposure results in an initial rapid clearance of
inhaled particles (U.S. EPA, 1986a). Indeed, animal exposures show a pattern of
decreased clearance that continues well after exposure has ceased.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-10
The sample size of the exposed populations was quite small, making extrapolation to
the actual exposed populations difficult. Additionally, sample demographics were not
representative of actual exposed populations.
The body of work on the effects of sulfuric acid on pulmonary function is not fully
consistent, and concentration-response relationships have not yet been demonstrated
(Graham et al., 1990).
4.2.3 Animal Toxicological Study Findings
Acute Exposure Animal Studies
The effects of acute inhalation exposure to sulfate aerosols (in the form of sulfuric acid) in
animals have been described in several studies. The studies used a variety of species,
including mice, rats, guinea pigs, dogs, donkeys, rabbits, and monkeys. Although some studies
failed to cause effects in exposed animals, a number of studies show respiratory effects
increasing with concentration and decreased particle size. Observed effects, including
respiratory system damage, increased airflow resistance, and decreased function of the body's
defense mechanisms, are summarized below.
Respiratory system injury, including lesions in the bronchi, bronchioles, larynx, and trachea,
was noted in mice and guinea pigs following short-term exposure to high concentrations (60
to 125 mg/m3) of sulfuric acids (Lee and Mudd, 1979). These are, however, much higher
concentrations than are typical of ambient conditions in the United States.
Increased airflow resistance occurred in exposed animals, the magnitude of which is related to
both concentration and particle size. At concentrations below 1 mg/m3, a greater response was
observed for 0.3 urn than for 1 urn particles (Amdur et al., 1991; Chen et al., 1991).
Significant airflow resistance continued up to at least one hour following exposure, and
persisted longer than flow resistance resulting from exposure to sulfur dioxide (Amdur et al.,
1991; Chen et al., 1991).
A decreased ability in mechanisms enabling the body to respond to disease and infection was
noted (Lee and Mudd, 1979; Amdur et al., 1991). One study determined that a single
inhalation exposure to concentrations of sulfuric acid at concentrations that occur in the
ambient air decreased the body's resistance to infectious disease (Zelikoff and Schlesinger,
1992). The body's defense mechanisms are impaired as described below:
> There is a decrease in production of interferon, which provides resistance to viral
infections (Lee and Mudd, 1979).
-. November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES »> 4-11
> A decrease in bronchial clearance occurs (Amdur et al., 1991; Fujimaki et al., 1992).
As described above, bronchial clearance is one of the body's defense mechanisms.
*• The release of histamine, a compound believed to cause allergic reactions, is increased.
This suggests that sulfuric acid might be one cause of allergic diseases (Fujimaki et al.,
1992).
* Both phagocytic activity of macrophages and superoxide production are decreased
following inhalation exposure to relatively low concentrations of sulfuric acid
(Schlesinger et al., 1992). This decreased function compromises the cellular ability to
defend against infection and disease.
Although these findings are consistent to a certain degree the effects observed in
epidemiological studies, such as increased respiratory infection and decreased respiratory
function in children, it is difficult to determine their applicability for the following reasons:
*• Exposures were brief, rather than the daily long-term exposure typical of air
pollution exposure.
* Interaction among various air pollutants was not studied.
•
+ ' Extrapolation of animal findings to humans contains a degree of uncertainty.
Chronic Exposure Animal Studies
There are few laboratory studies describing the effects of chronic inhalation exposure to
sulfuric acid, however, observed effects included pulmonary damage, decreased airflow, and
decreased bronchial clearance. In one study, monkeys were exposed for two years to
160 jag/m3 of sulfur as 0.54 uM of sulfuric acid. This exposure resulted in moderate to severe
pulmonary damage, and decreased airflow (Arndur et al., 1991). Bronchial clearance was
decreased in both rabbits and donkeys following chronic exposure to concentrations of sulfuric
acid ranging from 100 to 250 (ig/m3; a continued decrease was noted in both species for up to
three months after the final exposure (Schlesinger et al., 1979; Gearhart and Schlesinger,
1988, 1989). Although there is uncertainty regarding the extrapolation of exposure
concentrations from animals to humans, these studies suggest the possibility of respiratory
injury due to chronic exposures to sulfuric acid. It is uncertain whether these exposures are
relevant to human populations in the United States at current ambient sulfate concentrations.
4.3 ISSUES IN APPLYING EPIDEMIOLOGY RESULTS IN THIS ASSESSMENT
This quantitative assessment relies on concentration-response functions from the available
epidemiology literature concerning human health effects associated with sulfate aerosols, and
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES » 4-12
in some cases PM2 5. Available epidemiology results were selected for quantitative use in this
assessment for the following reasons:
> Epidemiology results are based on studies of actual human health data and associated
pollution exposures. Extrapolations from animal responses or from artificial clinical
exposures are not necessary.
* A large available body of relevant epidemiology literature allows a quantitative
assessment to be performed using a modest amount of research resources.
* Available epidemiology results cover a wide range of suspected health effects,
including responses to long-term as well as short-term exposures.
» Reasonable assumptions can be employed when applying epidemiology results to
calculate changes in health effects as a function of estimated changes in outdoor sulfate
concentrations that allow the assessment to be conducted without doing detailed human
exposure modeling.
> Epidemiology results are available for health effects that readily lend themselves to
monetary valuation, such as premature mortality risks and self-reported symptoms.
The basic approach used in this assessment of applying epidemiology results to estimate
health effects changes associated with estimated changes in outdoor air pollutant
concentrations has been used in previous assessments for various air pollutants. These include
EPA's Regulatory Impact Analyses for particulate matter and sulfur oxides (U.S. EPA, 1984,
1986c, 1988). Other assessments of the benefits of alternative pollution control strategies that
used epidemiology results to estimate health benefits include Rowe et al. (in press), Krupnick
and Kopp (1988), Hall et al. (1989), Harrison and Nichols (1990), and Thayer (1991).
Applying available epidemiology results to construct specific concentration-response functions
for changes in ambient sulfate aerosol concentrations requires specific interpretations and
assumptions. This section presents some of the key issues that must be considered, and
explains the approaches chosen for this assessment. Whatever choices that are made on each
of these specific issues, considerable uncertainty remains in the final results. Using these types
of epidemiology results for quantitative assessments of health risks is not universally
supported by all health researchers. Concerns exist about the accuracy of the estimated
quantitative relationships in epidemiology studies, the fact that epidemiology studies can show
an association but do not prove causation, and about transferring results from a specific study
context to an assessment context that invariably has some different characteristics.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES »• 4-13
4.3.1 The Effects of Sulfates versus Other Participates
Title IV requirements will result in a large reduction in SO2 emissions, primarily from sources
located in the eastern United States. Sources in the western United States are subject to the
same emissions limits, but few sources in the western United States currently exceed the Title
IV emissions limits. The reduction in S02 emissions in the eastern half of the United States
will result in a significant reduction in ambient airborne concentrations of sulfate aerosols
over a large geographic area. For this quantitative assessment, we want to know specifically
how sulfate aerosols affect human health. However, a significant difficulty for this assessment
is that epidemiology studies are limited in their ability to isolate the effects of sulfates from
the effects of PM2 $ as a whole. Because sulfates are a significant component of PM2 5, the
two pollutant measures are typically highly correlated (Ozkaynak and Thurston, 1987).
Furthermore, only a few epidemiology studies have used data on sulfate concentrations as well
as PM2 5 concentrations so only a few direct comparisons of results are available. Most use
one or the other measure.
Sulfate aerosols are a significant share of PM2 5 in the United States. In the eastern United
States, the ratio of average measured sulfate concentrations to average measured PM2 5
concentrations is about 0.4 (Dockery et al., 1993). An important underlying issue in
interpreting available epidemiology results for this assessment is whether sulfates are different
from other fine particulates in terms of the amount or type of adverse health effects they
cause. Although it is reasonable to expect that there may be differences, available information
is not sufficient at this time to specify the differences for sulfates or any other common
constituent of PM25. Sulfate measures have been used in many epidemiology studies, but only
a few studies have made a direct comparison of results obtained when a sulfate measure is
used versus a PM2 5 measure for the same location and study population. Several such studies
have found statistically significant associations with the health endpoint for both pollutant
measures (e.g., Pope et al., 1995; Dockery et al., 1993). Some of these studies have found a
statistically stronger association between health effects and sulfates (e.g., Plagiannakos and
Parker, 1988; Ostro, 1990) and others have found a statistically stronger association with more
comprehensive measures of particulates such as PM2S or PM10 (e.g., Dockery et al., 1992;
Abbey et al., 1993a).
Two of these studies (Dockery et al., 1993; Ostro, 1990) have reported sulfate and PM25
coefficients for the same population groups as well as mean concentrations of each pollutant
measure in the study area. We can expect that if the sulfate coefficient fully reflects the
effects of all PM2 5, or is the sole causal constituent of PM2 5, the ratio of the sulfate
coefficient to the PM2 s coefficient should equal the inverse ratio of the sulfate and PM2 5
concentrations. This is true for the Dockery et al. (1993) results for premature mortality, but
not true for the Ostro (1990) results for respiratory restricted activity days. The latter suggest
that there is an effect of PM2 5 that is not fully reflected in the sulfate coefficient, but that the
additional effect per unit PM2 s is about half that for sulfate. These results are suggestive at
best, because of the high collinearity between the sulfate and PM2 5 measures, and are not
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-14
sufficient for determining differences in potency between sulfate participates and other
constituents of PM2 5.
The epidemiology evidence is abundant, however, that some or all of the constituents of
PM2 j, including sulfates, are harmful to human health. Clinical and laboratory studies provide
evidence that at least some types of sulfate aerosols are harmful to the respiratory system
when subjects are exposed to controlled amounts of sulfates alone. Thus, mere is reason to
believe that sulfates are contributing, at least in part, to the health effects observed in
association with PM2 5 and other particulate matter measures. The approach we take in this
analysis to address this issue is three tiered:
> First, for health effects that have been statistically associated with sulfate
concentrations, we select low, central, and high magnitudes of the estimated
relationships between health effect incidence and sulfate concentrations.
* Second, for additional health effects that have been statistically associated with PM2 5,
we select low, central, and high magnitudes of the estimated relationships and apply
them to the predicted changes in sulfate concentrations on the assumption that the
estimated association between health effect incidence and PM2 5 applies equally on a
per ug/m3 basis to sulfates, which are a substantial constituent of PM2.5.
* Third, we use sensitivity analysis to determine how the results of the analysis would
change if we were to assume that the estimated sulfate coefficients that form the basis
of the health effects estimates in step one reflect the effects of PM2.5 as a whole, not
just sulfates, because of the typical collinearity between sulfates and PM2 5.
For the sensitivity test on this question, we multiply all the sulfate coefficients by 0.4. This
reflects an alternative assumption that the sulfate coefficient reflects the effects of other
constituents of PM2 5 as well as sulfates. This assumption and the first assumption (that the
sulfate coefficients reflect the effects of sulfates only) most likely bound the "true" sulfate
effect. The 0.4 adjustment is derived as follows. If we presume that sulfates and PM2 5 are
100 percent correlated, then a coefficient estimated for a sulfate measure will reflect all the
effects of the nonsulfate portion as well as the sulfate portion of PM2 5. We might, for
example, have the following estimated concentration-response relationship between a health
effect (H) and sulfate levels (S), where B, is the estimated sulfate coefficient:
H = B, x S. (4-1)
If, for example, Bs equals 4, this means that for every unit change in S there are 4 health
effects observed. However, because of the collinearity between S and PM2 5, B, may actually
reflect the effects of the 1 unit of S and the 1.5 units of collinear nonsulfate PM2 5 (the ratio
of measured sulfate to PM2 5 being 0.4). Thus, if we change S by 1 unit and do not change
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES > 4-15
the nonsulfate particulates, we would obtain only a 1.6 unit change in H. Therefore, Bs must
be multiplied by 0.4, to calculate the health effects associated with a 1 unit change in S alone.
4.3.2 Health Effects Thresholds
Another important uncertainty in this assessment is whether there is a threshold sulfate
concentration below which health effects no longer occur, or whether the slope of the
concentration-response function diminishes significantly at lower concentrations. Available
epidemiological evidence is inconclusive on the question. No clear threshold has been
determined, but such a determination is very difficult with typical epidemiological data. Most
of the epidemiology studies reported here have estimated linear or log-linear functions that
suggest a continuum of effects down to the lowest sulfate concentrations observed in the study
sample, and have not attempted to identify a threshold concentration.
For this report, the default assumption adopted is that there is no threshold for health effects
associated with ambient sulfate aerosols. In a practical sense, this does not mean that health
effects are presumed to occur all the way down to zero sulfate concentrations because the
changes in consideration (i.e., those due to Title IV) do not mean the elimination of all
anthropogenic sulfate aerosols. If a threshold exists, however, it could have a significant effect
on the accuracy of the results of this analysis. Depending on the level of the threshold relative
to the estimated exposure concentrations, the existence of a threshold could reduce (but not
increase) estimated health effects and benefits.
Because the evidence on whether, and at what concentration, there is a health effects threshold
for sulfates remains inconclusive at this time, we report the results of some sensitivity
analyses conducted using different assumptions regarding possible threshold concentrations for
sulfate aerosols. We select two alternative threshold assumptions based on the low ends of the
range of sulfate concentrations over which health effects have been estimated. The highest
selected threshold for the sensitivity analysis is 5 ug/m3. This is the mean sulfate
concentration reported by Abbey et al. (1993a) for the Southern California study area for
which a statistically significant association between the sulfate measure and chronic bronchitis
incidence was not found. Another selected threshold is an annual average sulfate concentration
of 3.6 fig/m3, which is the lowest average sulfate concentration in the 151 cities included in
the Pope et al. (1995) prospective cohort study on mortality rates in the United States. The
third selected threshold for the sensitivity tests is 1.6 (ag/m3, which is the average sulfate
concentration for 50 percent of the observations hi the Southern Ontario study on hospital
admissions (Burnett et al., 1995). This study reports a statistically significant difference for
hospitalization rates between days with average sulfate concentrations of 1.6 ug/m3 versus
days with average concentrations of 4.13 ug/m3 (the next 25 percent of the observations). This
is not a direct test for a threshold, but it suggests that effects may occur at sulfate
concentrations as low as 1.6
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-16
4.3.3 Uncertainty in the Estimates
The available epidemiology evidence regarding health effects associated with air pollutants,
including sulfate aerosols, is subject to considerable uncertainty. Within a given study there is
statistically measurable uncertainty in the estimated concentration-response coefficients, and
there are differences in results obtained from different studies looking at the same or similar
health effects. For each concentration-response relationship presented in this report, we have
selected low, central, and high estimates. The central estimate is typically selected from the
middle of the range reported in the study, or group of studies, that has been selected as
providing the most reliable results for that health effect based on the study selection criteria
discussed in Section 4.4.
These ranges of concentration-response values are not intended to reflect absolute upper and
lower bounds, but rather ranges of estimates that are reasonably likely to be correct, given
available health effects data. For example, ranges based on a single study are selected as plus
and minus one confidence interval, not the absolute highest and lowest result obtained. When
several different "reliable" studies are available for a given health effect, the selected range
reflects the variation in results across the studies. The reader should be aware that there is
analyst judgment in selecting these ranges and that the ranges do not reflect all the uncertainty
in the concentration-response estimates because some of the uncertainty is not quantifiable.
This is, however, an attempt to give a more realistic presentation than is given when only
point estimates are reported.
Each low, central, and high estimate is also assigned a probability weight (the weights
summing to 100 percent for each quantified health effect). These probability weights are used
to propagate the uncertainty through the multiplication and aggregation process to total health
benefit estimates. This provides an alternative to simply summing all the low estimates or all
the high estimates to obtain total low and high estimates. Such simple summing can be
misleading because it is highly unlikely that all the low estimates (or all the high estimates)
are correct. When the low, central, and high estimates are based on results from different
studies all judged as equally reliable, an equal probability weight is given to the low, central,
and high estimates. When only one study is selected, the range used is plus and minus one
standard error from the mean results of the study. When a statistical standard error is used, the
probability weight given to the central estimate is 50 percent, with 25 percent each to the high
and low estimates. In a few cases less weight has been given to a high or low estimate based
on analyst judgment that there is reason to suspect that particular estimate is less likely to be
correct than the other available estimates.
4.3.4 Interpretation and Aggregation of Daily Results
Many of the epidemiology studies that provide information about the health effects associated
with particulate matter exposures have examined the daily incidence of a health effect such as
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES > 4-17
mortality or hospital admissions, and daily sulfate concentrations. The air quality modeling
used in this analysis predicts changes in annual average sulfate concentrations, not changes in
the daily concentration. Therefore, it is necessary to determine how changes in annual average
sulfate concentrations contribute to daily health effects.
Two types of functional forms have been used in the daily epidemiology studies. One is a
linear, function, in which the estimated coefficient gives the number of additional cases each
day as a function of changes in the daily pollution concentration. A linear function gives the
following relationship:
AC, = R x AS; x POP, (4-2)
where:
= additional cases on day i associated with a change hi sulfate
concentration
R = concentration-response coefficient between daily C and S
ASj = change in sulfate concentration on day i
POP = affected population.
To obtain the number of cases each year, we sum Equation 4-2 over 365 days:
365 365
£ AC; = RxPOP £ (AS;). (4-3)
i-l i-1
If we multiply the right-hand side of Equation 4-3 by 365/365, we obtain:
365 365 (AS )
T; AC; = RxPOP x365 V l-Jl. (4-4)
fz\ i«i 365
Thus, Equation 4-3 is equivalent to
Annual AC = R x POP x 365 x Annual average of daily changes in S.
The annual average of the daily changes in sulfate concentration is the same as the change in
the annual average sulfate concentration. A linear coefficient for the daily number of cases
due to sulfates, therefore, can be multiplied by 365 to obtain a coefficient for predicting the
number of annual cases as function of the change in the annual average sulfate concentration.
The other common functional form is one in which the estimated coefficient gives the
percentage change in the number of cases each day as a function of the daily pollution
concentration. This gives the following relationship:
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES *• 4-18
AC/C = R x ASj x POP, (4-5)
where:
C' = the average daily number of cases of C due to all causes.
Equation (4-5) is simplified by substituting the average daily number of cases per individual.
Once C' is moved to the right-hand side of Equation 4-5, ACj can be estimated.
4.4 SELECTION OF CONCENTRATION-RESPONSE FUNCTIONS
This section provides a discussion of the specific epidemiological studies selected (based on
the selection criteria discussed below) for quantitative use in this analysis. Concentration-
response coefficients are selected from these studies. Ranges of concentration-response
coefficients are given for each health effect category. The ranges are based on results from
different studies when more than one equally applicable study is identified. All of the selected
concentration-response functions are reported as functions of sulfate, based on studies that
report health effects associated with sulfates or with PM2 5.
4.4.1 Study Selection Criteria
Concentration-response functions were identified and adapted from the available epidemiology
literature. These functions allow the estimation of the change in the number of cases of each
health effect that would be expected as a result of changes in ambient sulfate concentrations.
To be included as a basis for the concentration-response functions used in this assessment, an
epidemiology study had to meet several specific criteria.
First, a proper study design and methodology were required. Studies were expected to have
data based on continuous monitoring of the relevant pollutants, careful characterization and
selection of exposure measures, and minimal bias in study sample selection and reporting. In
addition, the studies had to provide concentration-response relationships-over a continuum of
relevant exposures. Second, studies that recognized and attempted to minimize confounding
and omitted variables were included. For example, studies that compared two cities or regions
and characterized them as "high" and "low" pollution areas were not used for quantitative
purposes because of potential confounding by other factors in the respective areas and vague
definition of exposure. Third, controls for the effects of seasonality and weather had to be
included. This could be accomplished by stratifying and analyzing the data by season, by
examining the independent effects of temperature and humidity, or by other statistical
corrections.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES > 4-19
A fourth criterion for inclusion was that the study had to include a reasonably complete
analysis of the data. Such analysis would include a careful exploration of the primary
hypothesis and preferably an examination of the robustness and sensitivity of the results to
alternative functional forms, specifications, and influential data points. When studies reported
the results of these alternative analyses, the quantitative estimates that we judged as most
representative of the overall findings are those that we selected for use in this assessment.
Finally, studies that addressed clinical outcomes or changes in behavior that would lend
themselves to economic valuation were included. Estimates for endpoints such as changes in
lung function, therefore, were not included.
4.4.2 Mortality
Over the last few decides, many epidemiologic studies have found statistically significant
associations between sulfate concentrations (and other measures of participate matter) and
premature mortality among the general population. The earliest studies focused on relatively
rare episodes of extremely high pollution concentrations in the 1940s and 1950s in the United
States and hi the United Kingdom (U.S. EPA, 1982). More recent studies have found an
association at concentration levels typical of most metropolitan areas in North America.
The earliest studies of this type were cross-sectional studies examining annual mortality rates
across U.S. cities with different average sulfate concentrations, often including 100 or more
cities (e.g., Evans et al., 1984; Ozkaynak and Thurston, 1987). Very recently, two prospective
cohort studies using individual-specific data and tracking mortality for a study sample in
multiple cities over multiple years, also found an association between premature mortality and
sulfate concentrations (Dockery et al., 1993; Pope et al., 1995). Time-series studies have also
found statistically significant associations between daily mortality and daily fluctuations in
sulfate concentrations (e.g., Dockery et al., 1992).
Some skepticism remains about whether these studies reflect a true causal relationship
primarily because a biological mechanism to fully explain and verify this relationship has not
been demonstrated in clinical or laboratory research (Utell and Samet, 1993). However, the
epidemiologic studies are consistently finding a statistically significant association between
sulfates and mortality, using different study designs and locations, and over a wide range of
sulfate concentrations, including levels currently typical of many locations in the United
States. It is therefore a reasonable exercise to estimate the reductions hi premature mortality
that might occur if sulfate concentrations were reduced, on the basis of the available
epidemiologic results.
Summary of Selected Quantitative Evidence
This section does not provide a detailed review of all available literature, but focuses on
results available in the literature that are best suited for the purposes of this analysis. The
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-20
study selection process .relied on study selection criteria discussed in Section 4.4.1, and
incorporated results from prospective cohort, single-period cross-sectional, and time-series
studies. From all three perspectives the results show an association between mortality and
sulfate concentrations, and results from all three types of studies are relied upon in selecting a
range of risk estimates for use in this analysis.
Two types of long-term exposure studies have found statistically significant associations
between mortality rates and particulate matter levels in the United States. The first type is an
ecologic cross-sectional study design in which mortality rates for various locations are
analyzed to determine if there is a statistical correlation with average air pollutant
concentrations hi each location. Such studies have consistently found measurably higher
mortality rates in cities with higher average sulfate concentrations. However, concern persists
about whether these studies have adequately controlled for potential confounding factors.
Ozkaynak and Thurston (1987), Evans et al. (1984), and Chappie and Lave (1982) provide
examples of ecologic cross-sectional studies. These studies each conducted a thorough
examination of data for 100 or more U.S. cities, including average sulfate concentrations for
each city, with special emphasis on the effects of including or excluding potential
confounding factors such as occupations or migration. Plagiannakos and Parker (1988)
combined annual cross-sectional data for 7 years for 9 counties in Ontario, Canada and also
found an association between mortality rates and sulfate concentrations.
A second type of long-term exposure study is a prospective cohort study in which a sample is
selected and followed over time in each location. In 1993, Dockery et al. published results for
a 15-year prospective study based on samples of individuals in 6 cities. In 1995, Pope et al.
published results of a 7-year prospective study based on samples of individuals in 151 cities
in the United States. These studies are similar in some respects to the ecologic cross-sectional
studies because the variation in pollution exposure is measured across locations rather than
over time. These studies rely on the same type of pollutant exposure data as that used in the
ecologic studies, which is average pollutant concentrations measured at stationary outdoor
monitors in a given location. However, the mortality data are for identified individuals, which
enables much better characterization of the study population and other health risks than when
area-wide mortality data are used. Because they used individual-specific data, the authors of
the prospective studies were able to control for premature mortality risks associated with
differences in body mass, occupational exposures, smoking (present and past), alcohol use,
age, and gender.
Dockery et al. (1993) found a mortality-rate ratio of 1.26 over the 15-year study period from
the most polluted to least polluted city. Pope et al. (1995) found a mortality-rate ratio of
1.15 over the 7-year study period from the most polluted to least polluted city. Both of these
findings were statistically significant.
The two prospective cohort studies represent a very important contribution to the study of
premature mortality and sulfates (and other particulate matter measures) because the
'. November 10, 1995 .
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-21
prospective design using individually identified subject allows for better accounting of other
risk factors for an individual that might be confounding factors when attempting to isolate the
risk associated with air pollution exposure. The findings of a significant association between
mortality and sulfate concentrations in this study are very supportive of the findings in
previous single-year cross-sectional studies. The prospective studies provide evidence that
long-term exposures to higher average sulfate (and other paniculate matter) concentrations are
associated with statistically significantly higher risks of premature mortality. However, due to
limitations in the measure of exposure used in these studies, it is not possible to yet determine
the specific length of exposure required to obtain this result, or whether there may be some
latency between elevated exposure and elevated risk. This is because the studies have used
measures of sulfate and other participate matter concentrations at the beginning of or during
part of the study period as the measure of exposure. Lifetime cumulative exposures are not
known. Current period concentrations are probably correlated with lifetime exposures for
individuals residing in a given location, but quantitative extrapolation from the results based
on this exposure measure are uncertain.
The results of the two prospective studies and four selected cross-sectional studies are
summarized in Table 4-1. Results are reported in terms of the estimated percentage change in
mortality in the study sample for every ng/m3 change in average sulfate concentrations. For
example, the Pope et al. results for 151 U.S. cities indicate that for every one |j.g/m3 increase
in average sulfate concentrations where subjects live is associated with a 0.75 percent increase
in observed mortality in the 7-year study period. The cross-sectional studies typically report
results from many different specifications of the mortality regressions, because the intent of
some of these studies was to test for the effect of changes in the specification. The results
reported here are selected from the middle to low end of ranges of results reported, and are
drawn from specifications that include the significant explanatory variables identified in
addition to the air pollutant measures.
The results with respect to sulfates fall between 0.3 percent and 1.4 percent, with the
exception of the sulfate result for the 6-cities prospective study, which is substantially higher.
The results of the prospective studies are generally equal to or higher than the results of the
cross-sectional studies, which supports that the cross-sectional results are meaningful, not just
spurious statistical associations, and suggests that more accurate accounting of individual
mortality risks results in greater risk attributed to air pollution exposure. This conclusion is
tentative until more prospective cohort studies have been completed and continue to verify
this finding.
In some studies the premature mortality result is also analyzed per unit of PM2 5, and this is
also shown in Table 4-1. When estimates are reported for both pollutant measures, these are
based on estimates that do not account the other pollutant measure. They should therefore be
interpreted as different measures of the same health effect based on different but highly
collinear measures of fine paniculate concentrations. For the Pope et al. results, the ratio of
the effects of sulfate to PM2 5 exceeds the inverse of the ratio of the mean concentrations of
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES »> 4-22
Table 4-1
Comparison of Selected Mortality Study Results
Study
Pope et al.
(1995)
Dockery et
al. (1993)
Ozkaynak
and Thurston
(1987)
Evans et al.
(1984)
Chappie and
Lave (1982)
Plagiannakos
and Parker
(1988)
Study
Design
Prospective
Cohort
Prospective
Cohort
Cross-
Sectional
Cross-
Sectional
Cross-
Sectional
Cross-
Sectional
Time
Period
1982-
1989
1974-
1989
1980
1960
1960
1969
1974
1976-
1982
Study
Location
50 U.S. cities
151 U.S. cities
6 U.S. cities
98 U.S. cities
98 U.S. cities
117 U.S. cities
1 12 U.S. cities
102 U.S. cities
9 Ontario
counties
% Change in Mortality
per ng/m3
SO4
0.75%
3.25%
0.77%
0.29%
0.50%
0.54%
1.37%
0.50%
PM2.5
0.69%
1.40%
each measure in the study areas. This suggests that the sulfate effect exceeds the PM2 5 effect
on a per ng/m3 basis, but suggests that there are additional effects picked up by the PM2 5
coefficient that are not fully reflected in the sulfate coefficient. The Dockery et al. results,
however, suggest that there may be no additional PM2 5 effects other than those reflected in
the sulfate coefficient.
There have also been a substantial number of daily tune-series studies examining the
relationship between daily mortality and daily paniculate matter concentrations hi many cities
in North America. Dockery and Pope (1994) review and summarize these studies. These
studies have for the most part used TSP or PM10 as the measure of paniculate concentration.
One time-series study (Dockery et al., 1992) reports a sulfate coefficient of 0.6 percent change
in daily mortality per ug/m3 sulfate, which is within the range of results reported in
Table 4-1. Dockery and Pope report that overall, the results of the time-series studies range
from 0.05 percent to 0.15 percent higher mortality for every ng/m3 increase in 24-hour PM10.
This range falls just below the range of results reported for sulfates in Table 4-1 from the
cross-sectional and prospective studies.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-23
Evidence on Who is at Risk
The results of a time-series study in Philadelphia (Schwartz and Dockery, 1992a) provide
estimates of elevated mortality risks separately for those over and under 65 years old. These
results suggest that about 90 percent of the premature deaths associated with paniculate matter
occur in the over-65 group. This finding is consistent with the results of an early cross-
sectional mortality study (Lave and Seskin, 1977). Ostro et al. (in press) found that about
80 percent of the premature deaths associated with paniculate matter were in the over-
65 group in their Santiago, Chile, study. In the United States, about 70 percent of all deaths
are individuals 65 years old or older, so it appears that risks associated with air pollution
exposure fall in somewhat greater proportion to the elderly.
As discussed in Chapter 5, the age of the individual at risk of premature mortality may have
some bearing on the monetary value of changing that risk. For the purposes of this analysis, it
is presumed based on evidence in Ostro et al. (in press) and Schwartz and Dockery (1992a)
that 85 percent of the individuals at risk of premature mortality associated with sulfate
exposures are 65 years old or older.
The results from Pope et al. (1995) show that the greatest association is with deaths associated
with cardiopulmonary illness, and that elevated mortality risks are similar for both smokers
and nonsmokers in higher pollution locations. Some of the time-series studies (e.g., Schwartz
and Dockery, 1992a) have also found significant cause-specific mortality associations
indicating that most pollution-associated deaths are cardiopulmonary related. Some of those at
risk therefore probably suffer from chronic diseases that might be expected to shorten life
expectancy even in the absence of air pollution. This does not, however, rule out the
possibility that some of these chronic illnesses could themselves be related to air pollution
exposure.
Estimation Approach for this Analysis
For this analysis, the epidemiologic results are being used to predict how mortality rates may
change given a change in ambient sulfate concentrations. For this purpose, we select a range
of results from the three types of mortality studies. Premature mortality is a very serious
health endpoint and there is a large body of epidemiologic literature that has studied mortality
as it relates to air pollutant exposure. However, there remain many uncertainties in specific
quantitative interpretations of the results of the epidemiologic studies that have studied the
association between premature mortality and sulfate concentrations. We therefore select a
wider range of findings than those selected for most of the other health endpoints quantified
in this assessment.
We select a range of four estimates to reflect the range of results obtained hi the mortality
studies. For a lowest estimate, we select the 0.1 percent mortality effect found for PM10 in the
many time-series studies. This is at the low end of the range of mortality effects estimated
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES > 4-24
and because it is based on PM10, applying it to an estimated change in sulfate concentration
presumes that a sulfate aerosol is no more harmful than a typical PM,0 aerosol. We select a
low-central estimate of 0.3 percent based on the low end of the cross-sectional results for
sulfates. We select a high-central estimates of 0.7 percent based on the Pope et al. prospective
study. This is still within the range of the cross-sectional results. As a high estimate we select
1.4 percent based on the PM2 5 results of the 6-cities study and the highest cross-sectional
result reported in Table 4-1. Although there are results from some studies that are both lower
and higher than this range (e.g., some time-series studies find 0.05 percent or less and the
6-cities result for sulfates is greater than 3 percent), a very large share of the findings for
sulfates fall into this range. We give equal probability weights (25%) to all four of the
selected risk estimates.
The selected percentage changes in mortality must be multiplied by average annual mortality
to calculate the change in annual premature deaths per change in annual average sulfate
concentrations. For this we use the average U.S. nonaccidental mortality rate of about 8,000
per million population per year (U.S. Bureau of the Census, 1994). For example, the low-
central estimate is 0.3 percent of 8,000 divided by 1,000,000. The selected mortality risk
coefficients and calculation procedures are thus:
\
Low annual SO4 premature mortality = 8 x IQ"6 x POPj x (ASj) (4-6a)
Low-central annual SO4 premature mortality = 24 x 10"6 x POPj x (ASj) (4-6b)
High-central annual SO4 premature mortality = 56 x 10"6 x POPj x (ASj) (4r6c)
High annual SO4 premature mortality = 112 x 10"6 x POPj x (ASj) (4-6d)
where:
POPj = total population in area j
ASj = change in annual average sulfate concentration in area j.
4.4.3 Chronic Respiratory Disease
For more than two decades, there has been some evidence suggesting that higher ambient
particulate matter exposures are associated with higher rates of chronic respiratory disease.
Much of this evidence, however, has been based on cross-sectional analyses, comparing
disease or symptom prevalence rates in different communities with different average pollution
levels (e.g., Ferris et al., 1973, 1976; Hodgkin et al., 1984; Portney and Mullahy, 1990).
These studies are able to suggest a possible association, but are difficult to use for quantitative
estimates of specific concentration-response functions. This difficulty stems primarily from
uncertainty about how to characterize the relevant exposure units, in particular the time
aspects of exposure. Chronic symptoms presumably occur as a result of long-term exposures,
but cross-sectional analyses are not very enlightening about whether, for example, it is the ,
five-year average, the twenty-year average, or the number of times a given level is exceeded
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES > 4-25
that is the relevant exposure measure. Without this information, it is difficult to predict
quantitatively how risks change when exposures change.
Recently published articles (Abbey et al., 1993a, 1993b, 1995) have reported results of a 10-
year prospective cohort study conducted at Loma Linda University in California with a large
sample of nonsmoking adults. This follow-up allowed for measures of exposure preceding and
during the 10-year study period and for obtaining information on changes in chronic
respiratory disease incidence over time. Thus, development of new cases of disease were
analyzed in relation to individual-specific air pollution exposure history. This study provides
for the first time a more definitive concentration-response function for chronic respiratory
disease. Uncertainty about the potential effect of exposures that preceded the study period, and
lag times between exposure and illness onset still exists with these findings.
The Loma Linda University Study
In the first stage of the Loma Linda University study, a large sample (approximately 7,000) of
Seventh Day Adventists (selected because they do not smoke), was interviewed in 1977.
Health histories, current respiratory symptoms, past smoking and passive smoking exposure,
and residence location histories were obtained. Hodgkin et al. (1984) compared the chronic
respiratory disease status of respondents who had lived for at least 11 years in either a high or
a low pollution area in Southern California. After adjusting for sex, race, age, education,
occupational exposure, and past smoking history, residents of the higher pollution area were
found to have a prevalence of airway obstructive disease (AOD) (including chronic bronchitis,
asthma and emphysema) that was 15 percent higher than for residents in the low pollution
area. Using the same 1977 Loma Linda sample, Euler et al. (1987) report results showing a
statistically significant association between past TSP exposure, based on residence zip-code
history, and the prevalence of chronic respiratory disease.
Abbey et al. (1993a, 1993b, 1995) report the icsults of a cohort study with the Seventh Day
Adventist sample in 1987, which provides better quantitative concentration-response
information. Nearly 4,000 subjects were interviewed in 1987 who had been interviewed
previously in 1977. All were 25 years old or more in 1977. Estimates of air pollutant
exposures histories were developed based on subjects' reported residence locations from 1967
to 1987 and pollutant measures from stationary outdoor monitors closest to each residence
location over the study period. Abbey et al. (1993b) report results of the cohort study based
on TSP data from 1973 to 1987. Abbey et al. (1995) added data on PM23, based on airport
visibility data from 1967 to 1987, sulfate data from 1977 to 1987, and data on gaseous air
pollutants including ozone, nitrogen dioxide, and sulfur dioxide.
Several different health outcomes were examined including new cases of emphysema, chronic
bronchitis, or asthma, in 1987 for those not reporting any definite symptoms of these diseases
in 1977. Disease definition was based on self-reported symptoms using the standardized
respiratory symptoms questionnaire developed by the National Heart and Lung Institute for
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-26
the United States. Respondents were classified as having definite symptoms of emphysema,
chronic bronchitis or asthma if they met specific criteria for the disease diagnosis. Having
definite symptoms of any one of these three was defined as definite airway obstructive disease
(AOD). Having definite chronic bronchitis was defined as having symptoms of cough and/or
sputum production on most days for at least 3 months/year, for 2 years or more. Emphysema
and asthma required a physician's diagnosis as well as associated symptoms.
Logistic models were estimated for mean concentrations of air pollutants and for hours above
selected levels for each pollutant. The regressions included independent variables for past and
passive smoking exposure, possible symptoms in 1977, childhood respiratory illness, gender,
age and education. Abbey et al. (1993b) report a statistically significant association between
average long-term TSP exposure levels and AOD, as well as with chronic bronchitis alone.
Abbey et al. (1995) report no statistically significant associations between the gaseous
pollutants and the development of new cases of chronic respiratory disease, although
aggravation of existing disease was apparent, specially for asthma in relation to ozone
exposure. More important, the authors conclude that exposures to gaseous pollutants did not
appear to be a significant confounding factor in the measured association between paniculate
matter exposure and incidence of chronic respiratory disease.
Abbey et al. (1995) report statistically significant associations between TSP exposure and new
cases of AOD, as well as with new cases of chronic bronchitis and new cases of asthma
(which are two types of AOD); and the magnitude of the TSP results was consistent with the
previous reported results (Abbey et al., 1993b). The authors also report a statistically
significant association between new cases of chronic bronchitis and the PM2 5 measure, and
between new cases of asthma and the sulfate measure. The magnitudes of the reported odds
ratios for new cases of AOD were similar for selected changes in TSP, PM2 5, and sulfates,
but the result was statistically significant only for the TSP measure. The authors note that
there is probably more measurement error in the PM2 5 exposure estimates because of the
approximation from airport visibility, and in the sulfate exposure estimates because they were
based on data from 1977 to 1987 only.
Abbey et al. (1995) also report evidence that increased severity of AOD is statistically
significantly associated with TSP, PM2 5, and sulfate exposure for those who reported definite
symptoms in 1977. Thus, it appears that particulate matter exposure both aggravates existing
cases and causes new cases.
Selected Chronic Bronchitis Risk Estimates from Abbey et al. (1995)
We have selected the chronic bronchitis results from Abbey et al. (1995) for PM25 for
quantification of changes in risks of developing chronic bronchitis in this analysis. The
estimates used in this analysis reflect only the development of new cases, not the aggravation
of existing cases. The key assumption in this application of the PM2 5 results is that sulfates
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES »> 4-27
contribute to this risk at an equal level per ng/m3 as other constituents of PM2 5. This
assumption is partially, but not fully, supported by the Abbey et al. (1995) results. Limitations
in both the PM2 5 and the sulfate data available for this analysis contribute to the ambiguity in
the findings. The quantitative interpretation of this assumption is to apply the risk associated
with each ug/m3 of PM2 5 to the estimated ug/m3 change in sulfate concentration without any
adjustment to the risk value due to the difference in the particulate measure. Implicit in this is
the assumption that sulfates contribute to the PM2 5 effect only in proportion to their share of
total PM2.5 and that other constituents of PM2 5 are equally as harmful.
The failure to find a statistically significant relationship between sulfate concentrations and
new cases of chronic bronchitis is somewhat troubling with respect to this quantification
approach, but it is offset to some extent by the finding of a significant relationship between
sulfates and new cases of asthma (another type of AOD) and by the fact that the magnitude of
the estimated relationship between AOD as a whole and sulfates is similar to the magnitude
estimated for TSP and for PM2 5, even though the statistical significance of this relationship
was low for the sulfate and PM2.s measures.
Using the PM2 5 results for chronic bronchitis in this assessment gives a lower risk per ug/m3
than would have been obtained using other feasible quantification approaches based on the
Abbey et al. (1995) results. For example, if we applied the estimated relative risk estimate for
new AOD cases reported by Abbey et al. (1995) for 7 (ig/m3 of sulfate and attributed this risk
to sulfate alone, the risk coefficient per ug/m3 would be about 3 times higher than the
selected central estimate based on the PM2 j results. Alternatively, if we used the statistically
significant relative risk for new cases of asthma associated with a 7 |ag/m3 increment of
sulfate, the risk coefficient per ^ig/m3 would be about 5 times higher than the estimate based
on the PM2 5 chronic bronchitis results. Thus, although there is uncertainty in applying the /
PM2 j results, it is unlikely that they overstate the effect of sulfates on new cases of AOD as a
whole.
Abbey et al. report a relative risk of 1.81 for developing a new case of chronic bronchitis
during the 10-year follow-up period for an increase in average PM2 5 exposure of 45 |ig/m3.
This means mat the incidence of new cases of chronic bronchitis is 81 percent higher in
locations with average PM2 5 concentrations 45 ng/m3 higher, or 1.8 percent higher for every
1 ug/m3 increase in average PM2 5 concentrations. The 10-year incidence of new cases of
chronic bronchitis was about 6 percent (117 * 1,868 in the subsample for which PM2 5
exposures were estimated). Thus, an individual's probability of developing chronic bronchitis
in the 10-year period is 0.018 x 0.06 = 0.0011 per 1 ug/m3 increase in average PM25
concentration. We divide this individual risk by 10 to obtain an annual risk of developing
chronic bronchitis. The high and low estimates are based on plus and minus one standard
error of the estimated risk relationship. The selected low, central, and high estimates for
changes in chronic bronchitis are thus:
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-28
Low annual cases of CB = 0.5 x KX4 x POP>25j x (ASj) (4-7a)
Central annual cases of CB = 1.1 x 104 x POP>25j x (ASj) (4-7b)
High annual cases of GB = 2.0 x 10"4 x POP>25j x (ASj) (4-7c)
where:
CB = adult chronic bronchitis
POP>2Jj =? population over age 25 years in area j
AS: = change in annual average sulfate concentration in area j.
We apply the risk estimates to the adult population age 25 and over because this is the
minimum age in the Abbey et al. study group. Chronic bronchitis takes awhile to develop and
these risk estimates may not apply to younger individuals.
4.4.4 Acute Morbidity
Epidemiology studies have found health effects associated with ambient sulfates ranging from
elevated rates of hospital admissions to small differences in lung function measurements. The
studies selected as the basis for quantitative estimates for this report provide evidence with
clear clinical significance; i.e., the effects are noticeable to subjects. This means symptoms
that are noticeable to the subject and can be expected to have some impact on the individual's
well-being. For this reason, we have not included studies that look only at effects on lung
function. Although this may be a medically relevant health endpoint, it cannot at this time be
translated into changes in symptoms or illness that can be readily valued.
Respiratory Hospital Admissions
Recent evidence indicates an association between ambient sulfates and both respiratory
hospital admissions (RHAs) and cardiac hospital admissions (CHAs). Evidence of a
relationship between RHAs and CHAs and sulfates, controlling for collinear ozone
concentrations, is provided by Burnett et al. (1995) for Ontario, Canada. Additional evidence
of a relationship between RHAs and sulfates is provided by Thurston et al. (1994) for
Toronto, and by Thurston et al. (1992) for selected cities in New York. For this analysis,
specific quantitative estimates are derived from the Burnett et al. (1995) Ontario study
because they are for both RHAs and CHAs. The Thurston et al. studies are examined for
supporting evidence, but are not used quantitatively because their results are less amenable for
providing separate associations for sulfates and ozone. Supporting evidence for an effect of
particles on cardiac hospital admissions is provided by Schwartz and Morris (1995).
Burnett et al. (1995) studied the relationship between hospital admissions for respiratory and
cardiac disease and both sulfate and ozone from 1983 through 1988 in Ontario, Canada. Air
pollution data were obtained from a large network of monitors existing throughout Ontario.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES > 4-29
Admissions data from 168 acute care hospitals in Ontario below the 47th parallel were used.
After elective admissions were excluded, counts of daily admissions for all ages and for age-
specific and disease-specific categories were created. A time-series regression model was used
that removed the influences of day-of-week effects, slow moving serial correlations due to
seasonal patterns, and differences between hospitals. Ultimately, the effects of air pollution on
deviations in the expected number of admissions to each hospital on any given day were
estimated. Regression models included temperature effects and were specified with ozone and
sulfate considered alone and together as explanatory variables. The results indicated that one-
day lags of both ozone and sulfates were associated with respiratory admissions, and that
sulfates, but not ozone, were associated with cardiac admissions. The sulfate effects were
observed in both the summer and winter quarters, both males and females, and across all age
groups (Burnett et al., 1995).
Thurston et al. (1992, 1994) provide supporting evidence of an association between RHA
during summer months and either sulfate or ozone concentrations, or both. They do not report
results for models that include both ozone and sulfate, so their results for both pollutants are
likely confounded by the presence of the other correlated pollutant. However, the results are
useful for rough comparison to the Burnett et al. results. Burnett et al. (1994) found that the
mean sulfate concentration was associated with a 2.2 percent increase in daily summer RHAs
when only sulfate was included in the model, and that the mean ozone concentration was
associated with a 6.0 percent increase in daily summer RHAs when only ozone was included
in the model. The single pollutant results are similar to results obtained by Thurston et al.
(1992) for New York City, which were 3.5 percent for mean sulfate and 5.3 percent for mean
ozone. These estimates are also reasonably consistent with the findings obtained in the
Toronto study (Thurston et al., 1994).
Bates and Sizto (1989) provide some additional evidence on the issue. They estimated a
stepwise regression for respiratory hospital admissions during the summer months in Ontario.
First they included temperature, which explained 0.89 percent of the variance in RHA. Then
they added sulfate, which increase the explained variance to 3.3 percent. When ozone was
then added, the explained variance increased to 5.6 percent. This suggests that adding ozone
to the regression explains about as much of the variance as that explained by the sulfate
variable.
' •
Low, central, and high estimates of RHAs associated with sulfates are selected based on the
results of Burnett et al. (1995). Results were selected from a model that included both sulfates
and ozone in the regression, to reduce the chance of overstating the sulfate effect because of
the collinearity between sulfates and ozone in the study area. We apply a 50 percent
probability to the central estimate, and 25 percent each to the low and high estimates, which
are the central minus and plus one standard error. Specifically, Burnett et al. (1995) report a
3.5 percent increase in RHAs for a 13 jig/m3 increase in sulfate when ozone was included in
the model. The average daily RHA for the study period was 16.0 per million population.
Thus, 3.5 percent of the 16.0 daily RHA are attributed to 13 ng/m3 sulfate. Therefore, the
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES > 4-30
daily RHA per 1 ug/m3 sulfate is: 0.035 x (16.0 x 10"6) + 13 = 4.31 x 10'8. We multiply by
365 to obtain the estimated annual number of RHAs for a change in annual average sulfate
concentration. The central estimate of changes in RHA incidence is thus as follows, with the
low and high selected as the central minus and plus one standard error:
Low annual RHA = 1.3 x 10'5 x ASj x POPj (4-8a)
Central annual RHA = 1.6 x 10'5 x ASj x POPj (4-8b)
High annual RHA = 1.8 x 10'5 x ASj x POPj (4-8c)
where:
POPj = total population in area j
ASj - change in annual average sulfate concentration.
Burnett et al. (1995) also reported a statistically significant association between sulfates and
cardiac hospital admissions (CHA) throughout the year, while no association was found for
ozone. Burnett et al. (1995) report a 3.3 percent increase in CHAs for a 13 u£/m3 increase in
sulfate when ozone was included in the model. Thus, 3.3 percent of the average daily CHAs
per million population (14.4) in the study area gives the number of additional daily CHAs per
13 ug/m3 sulfate. Dividing by 13 gives the daily CHAs per ng/m3 sulfate [0.033 x (14.4 x
10'*) -5- 13 = 3.66 x 10'8]. We multiply by 365 to obtain the estimate annual number of RHAs
for a change in annual average sulfate concentration. We apply a 50 percent probability to the
central estimate, and 25 percent each to the low and high. The central estimate of CHAs is
thus as follows, with the low and high selected as minus and plus one standard error of the
central estimate:
Low annual CHA = 1.0 x 10'5 x ASj x POPj (4-9a)
Central annual CHA = .1.3 x 10'3 x ASj x POPj (4-9b)
High annual CHA = 1.7 x 10'5 x ASj x POPj. (4-9c)
Aggravation of Asthma Symptoms
Several studies have related particulate matter concentrations to exacerbation of asthma
symptoms in individuals with diagnosed asthma. Ostro et al. (1991) report results specifically
for day-to-day fluctuations in sulfate concentrations. Ostro et al. had subjects record daily
asthma symptoms during the duration of the study. An aggravation of asthma symptoms was
defined for each subject based on each individual's manifestation of asthma symptoms. This
typically meant a notable increase hi symptoms, such as shortness of breath or wheezing,
and/or in use of medication relative to what was "normal" for that individual. Daily air
pollution concentrations were then examined for correlations with day-to-day fluctuations in
asthma symptom frequency, controlling for other factors such as weather and previous-day
symptoms.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES » 4-31
Ostro et al. (1991) examined the association between several different air pollutants, including
sulfates, PM2 5, and acidic aerosols, and aggravation of asthma symptoms among adults during
winter months in Denver. A significant association was found between the probability of
moderate or severe asthma symptoms (measured as shortness of breath) and sulfate particulate
concentrations, after controlling for temperature, day of study, previous-day illness, and use of
a gas stove. Ozone concentrations were very low, near background concentrations, and do not
create a confounding influence. The results suggest the following relationship in the winter
months between sulfates and aggravation of asthma symptoms (A).
Change in daily probability of A = [0.0077 (± 0.0038)/S] x AS (4-10)
Using the reported sulfate mean for the study of 2.11 ^ig/m3 to linearize the function yields
the following calculation procedure to estimate daily probability of asthma symptoms per
asthmatic based on the Ostro et al. results,
Change in daily probability of A - [0.0036 (± 0.0018)] x AS (4-1 1)
There may be an upward bias in the Ostro et al. (1991) results because the data were collected
during winter months only. Winter months in Denver are also a period of more frequent
respiratory colds that also aggravate asthma symptoms and may in turn cause asthmatics to be
more sensitive to air pollutants. We therefore assume for the purposes of this analysis that the
measured relationship between aggravation of asthma symptoms and sulfate concentrations
applies during only half of the year. To annualize the relationship we therefore multiply by
182.5 rather than by 365.
Using an estimate of 4.7 percent for the portion of the U.S. population with diagnosed asthma
(National Center for Health Statistics, 1992) yields the following calculation procedure to
estimate annual number of asthma attacks based on the selected Ostro et al. (1991) results.
Low annual ASD = 3.3 x 10'1 x (ASj) x POPj x 0.047 (4-12a)
Central annual ASD = 6.7 x 10'1 x (ASj) x POPj x 0.047 (4-12b)
High annual ASD = 9.9 x 10'1 x (ASj) x POPj x 0.047 (4-12c)
Restricted Activity Days
Restricted activity days (RADs) include days spent in bed, days missed from work, and days
when activities are partially restricted due to illness. Ostro (1987) examined the relationship
between adult all-cause RADs in a two-week period and PM2 5 in the same two-week period
for 49 metropolitan areas in the United States. The RAD data were from the Health Interview
Survey (HIS) conducted annually by the National Center for Health Statistics. The PM2.5 data
were estimated from visual range data available for airports in each area. Since fine particles
have a more significant impact on visual range than do large suspended particles, a direct
relationship can be estimated between visual range and PM2 5.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES + 4-32
Separate regression estimates were obtained for 6 years, 1976 to 1981. A statistically
significant relationship was found in each year and was consistent with earlier findings
relating RADs to TSP by Ostro (1983). The mean of the estimated coefficient for PM25
across the 6 years indicated approximately 91,000 RAD each year per 1 million population for
each ng/m3 increase in annual average PM25, and ranged from a low of 53,000 for the 1981
coefficient to a high of 171,000 for the 1976 coefficient.
Additional work conducted by Ostro and Rothschild (1989) added ozone measures to the
regressions and found the estimated relationship between RADs and PM2 5 to be essentially
unchanged. This suggests that the RAD/PM2 5 relationship was not confounded by the
exclusion of ozone concentrations and is independent of ozone exposures. The newer work
also estimated the relationship between respiratory RAD (RRAD) and PM2 5 for employed
individuals only. It was expected that this relationship might be more stable than that between
all-cause RAD and PM25 for all adults for two reasons: (1) it is expected that pollution
induced RADs might be predominantly related to respiratory illness, and (2) workers might
define a RAD more consistently than the entire adult population. It was expected, though, that
confining the data to RRADs for workers might result in a smaller total number of predicted
restricted activity days for a given concentration of pollution, because all effects might not be
classified as respiratory and workers may be a healthier and therefore less sensitive group, on
average, than all adults. The findings are consistent with this expectation. The average of the
PM25 coefficients for the 6 years suggested an annual increase of approximately 47,000
RRAD per 1 million population for each ug/m3 increase in annual average PM2 5, and ranged
from a low of 31,000 for the 1978 coefficient to a high of 55,000 for the 1980 coefficient.
Ostro (1990) reports results also using data on RRADs for working adults. In this analysis he
matched data from EPA's Inhalable Particles Monitoring Network on sulfates and PM25,
based on particulate monitors, with the HIS data for 1979 to 1981. Data on 25 cities resulted
and the analysis shows statistically significant relationships between RRAD incidence and
both sulfate and PM2 5, in separate regressions as necessitated by the collinearity between the
two measures of fine particulate. The quantitative results were quite comparable to the Ostro
and Rothschild (1989) results for RRADs for working adults, and were also reasonably similar
for sulfates and PM2 s. Estimated annual RRADs per million population (of working adults)
was approximately 56,000 per ng/m3 sulfate or 42,000 per ug/m3 PM2 5.
For this analysis, we calculate changes in RAD incidence as a function of changes in ambient
sulfate concentrations based on the estimated relationship between RADs and PM2 5. The
Ostro (1990) results suggest that this is a reasonable assumption, the effect of which may be
to slightly understate the sulfate effect. We choose to use the PM2 s results for quantitative
purposes because the sulfate results are available for only a subset of RADs (i.e., RRADs for
working adults).
The mean results over the 6 years from Ostro (1987) for all-cause RADs for all adults (mean
coefficient = 0.0048) have been selected as the central estimate for this analysis. The mean
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES *• 4-33
results from Ostro and Rothschild (1989) for respiratory RADs for workers (mean coefficient
= 0.0158) were selected for the low estimate. This is a low estimate because it excludes some
nonrespiratory RADs that might be related to pollution exposures and is based on a healthier
than average sample (i.e., workers). The selected high estimate is the mean of the two highest
coefficients in the six year analysis (mean coefficient = 0.0076) by Ostro (1987). The reported
coefficients give percentage changes in RADs or RRADs for a 1 ug/m3 change in PM25.
Daily average estimates from the studies based on HIS data of 0.052 RAD and 0.0083 RRAD
per person are used to determine the relationship between number of RADs and PM25. For
example, the central daily individual risk estimate is thus:
0.0048 x 0.052 = 2.5 x 10"*. (4-13)
Multiplying by 365 to estimate annual changes in RAD incidence we obtain the following
low, central and high estimates for changes in annual average sulfate concentrations. The
calculations are applied to the adult population 18 years and over.
Low annual RAD = 4.7 x 10'2 x ASj x POP^gj (4-14a)
Central annual RAD = -9.3 x 10'2 x ASj x POP^,8j (4-14b)
High annual RAD = 14.6 x 10'2 x ASj x POP^ (4-14c)
where:
POP^gj = population in location j 18 years of age and older.
Acute Lower Respiratory Symptoms
Krupnick et al. (1990) and Ostro et al. (1993) report analyses of relationships between the
daily incidence of acute upper and lower respiratory symptoms among a general population
panel of adults in Southern California and daily concentrations of air pollution. These health
endpoints include some days with symptoms bothersome enough to result in a restricted
activity day, but also include days when noticeable symptoms are present but no change in
activities occurs. The statistical analyses incorporated the presence of illness on the prior day,
presence of chronic respiratory disease, daily weather conditions, indoor air pollution sources,
and controlled for autocorrelation.
The air pollution measures used hi the Krupnick et al. analysis were coefficient of haze
(COH), a measure of the visibility impairing particulates in the air, and ozone. Krupnick et al.
report a statistically significant relationship between daily COH and the daily incidence of
respiratory symptoms (upper and lower combined), after controlling for a statistically
significant ozone effect. Ostro et al. (1993) conducted separate analyses for upper and lower
respiratory tract symptoms, and added sulfates as a measure of daily paniculate matter in the
study area hi place of the COH measure. They continued to find a statistically significant
association between daily ozone and both kinds of symptoms. They found a statistically
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES *• 4-34
significant relationship between daily sulfate concentrations and lower respiratory symptoms
only, after controlling for ozone. We select these results for quantitative use in this assessment
of changes in sulfate concentrations.
Ostro et al. (1993) report an odds ratio for incidence of lower respiratory symptoms in adults
of 1.30 for a 10 ug/m3 increment of sulfates. The average daily incidence of lower respiratory
symptoms is 1.5 percent in the study sample. Thus, the average daily individual probability of
having lower respiratory symptoms is 0.03 x 0.015 = 4.5 x 10"4 per ng/m3 sulfate. To
annualize we multiply by 365. The low and high estimates are based on minus or plus one
standard error of the regression coefficient.
Low annual LRS = 6.6 x 10'2 x (ASj) x POPilgj (4-15a)
Central annual LRS = 16.4 x IV2 x (ASj) x POP218j (4-15b)
Low annual LRS = 23.0 x 10'2 x (ASj) x POP^18j. (4-15c)
Aggregatidn Procedures for Acute Morbidity Health Effects
Several of the more broad categories of acute morbidity health effects, such as restricted
activity days or days with lower respiratory symptoms, may include days on which effects
measured in another function occur, such as days spent in the hospital. To avoid double
counting, therefore, it is necessary to subtract some of these potentially overlapping
categories. Some additional adjustment will be necessary when one function is for all ages and
another is only for adults. In this case, we will assume the incidence of the effect is
proportional to the age distribution which is that 83 percent of the U.S. population is 18 and
older. The following subtractions are done before monetary valuations are applied and
summed. As discussed in Chapter 5 on monetary valuation of human health effects, each
RHA is assumed to average 6.8 days and each CHA averages 6.9 days. We assume that all
days in the hospital and all asthma symptom days are also restricted activity days and
therefore subtract these from total RADs. We also assume that all RADs are also acute
respiratory symptom days and therefore subtract a fraction of RADs from LRSs. The Ostro et
al. (1993) study reports that 28 percent of the acute respiratory symptoms are lower
respiratory tract. We therefore assume that RADs are split between upper and lower
respiratory tract in the same proportions. Net RADs and net LRSs are therefore defined as
follows:
net RADs - total RADs -. (0.83 x 6.8 x RHAs) - (0.83 x 6.9 x CHAs) - (0.83 x ASDs)
net LRSs = LRSs - (0.28 x total RADs).
These adjustments are approximate, but they do eliminate and even possibly over-compensate
for overlap in the daily health endpoints. There may remain, however, some subtle overlap
between the daily health endpoints and the chronic bronchitis and premature mortality health
endpoints. For example, some of the hospital admissions may reflect health effects that are
November 10, 1995 .
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES * 4-35
accompanied by premature death. Because as is shown in Chapter 6, the total health benefits
are dominated by the premature mortality and chronic bronchitis effects, the possible impact
on the total health benefits of such overlaps is necessarily small.
4.4.5 Summary of Selected Concentration-Response Functions
Table 4-2 lists the selected concentration-response estimates for each of the health effects
categories for sulfates. Omissions, biases, and uncertainties are summarized in Table 4-3.
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES + 4-36
Table 4-2
Selected Coefficients for Human Health Effects Associated
with Sulfate Concentration Changes
Health Effect Category
Annual mortality risk per 1 ug/m3 change in annual
average SO4 concentration.
Sources: See Table 4-1
Chronic bronchitis (CB) annual risk per 1 ug/m3 change in
annual average SO4 concentration.
Source: Abbey et al. (1995)
Respiratory hospital admissions (RHA) annual risk factors
per 1 fig/ni3 change in annual average SO4 concentration.
Source: Burnett et al. (1995)
Cardiac hospital admissions (CHA) annual risk per 1
ug/m3 change in annual average SO4 concentration.
Source: Burnett et al. (1995)
Asthma symptom day (ASD) annual risk factors given a 1
ug/m3 change in annual average SO4 concentration.
Source: Ostro et al. (1991)
Restricted activity day (RAD) annual risk factors given a
1 ug/m3 change in annual average SO4 concentration.
Sources: Ostro (1987), Ostro and Rothschild (1989)
Day with lower respiratory symptom (LRS) annual risk
factors given a 1 ug/m3 change in annual average SO4
concentration.
Source: Ostro et al. (1993)
Selected Concentration-Response
(probability weights)
L 8 x KT6 (25%)
L-C 24 x ID"6 (25%)
H-C 56 x 10-* (25%)
H 112x 10-* (25%)
For population 25 years and over:
L 0.5 x lO"4 (25%)
C 1.1 x 10-4 (50%)
H 2.0 x ID"4 (25%)
L 1.3 x 10'5 (25%)
C 1.6 x lO'5 (50%)
H 1.8 x 10'5(25%)
L 1.0 x 10'5 (25%)
C 1.3 x 10's (50%)
H 1.7 x 10'5 (25%)
For population with asthma (4.7% of
population):
L 3.3 x 10'1 (33%)
C 6.7 x 10'1 (34%)
H 9.9 x 10'1 (33%)
For population aged 18 years and over:
L 4.7 x 10° (33%)
C 9.3 x 10'2 (34%)
H 14.6 x 10'2 (33%)
For population aged 18 and over:
L 6.6 x 10'2 (25%)
C 16.4 x lO'2 (50%)
H 23.0 x 10'2 (25%)
November 10, 1995
-------
QUANTIFICATION OF HEALTH EFFECTS CHANGES *• 4-37
Table 4-3
Key Omissions, Biases, and Uncertainties
Qmissions/Biases/Uncertainties
Concentration-response
relationships
Transfer of concentration-
response relationships
Relationship between sulfates
and other measures of paniculate
matter
Zero threshold assumption
Age group assumptions
Presumed linearity of
concentration-response
Assumed independence of
baseline health incidence and
sulfate concentrations
Overall Impact
Direction
of
Potential
Error
?
?
+
+
?
9
+
Comments
Statistical association in epidemiology studies does
not prove causation. Measurement error and averting
behavior could cause downward bias. Omitted
confounding variables could cause upward bias.
Estimates are based on transfers across time and
location. Possible unaccounted for differences add
uncertainty.
Collinearity among particulate matter measures add
uncertainty to the quantitative interpretation of
sulfate based results. This uncertainty is addressed in
the sensitivity analysis.
Evidence on possible thresholds is inconclusive. This
uncertainty is addressed in the sensitivity analysis.
The effect of sulfates on mortality for different age
groups was based on the results of nonsulfate
studies. Effects on children probably understated due
to limited studies that include children.
The effect of assuming a constant risk per unit of
sulfate is difficult to assess with available
information. Error could occur in either direction.
Used average incidence to transform %
change/sulfate to the number of cases per change in
sulfate concentration. There is no bias if they are
independent.
No clear directional bias is entirely dominant, but
tendency may be toward upward bias. This is
addressed in the sensitivity analyses.
November 10, 1995
-------
CHAPTERS
MONETARY VALUATION OF HEALTH EFFECTS CHANGES
5.1 INTRODUCTION
This chapter presents monetary value estimates for the adverse human health effects expected
to be reduced because of the reduction in airibient sulfate aerosol concentrations attributable to
Title IV. Monetary value estimates are presented for an average case of each type of health
effect quantified in this assessment.1 These monetary value estimates per case are multiplied
by the estimated reduction in number of cases to obtain total monetary value estimates for
each type of health effect. These are then summed to total monetary value estimates for all
health effects benefits attributable to the sulfate aerosol reduction.
5.1.1 Monetary Valuation Concepts for Health Effects
The purpose of this assessment is to quantify the benefit to society of the reduction in health
effects expected from the Title IV required SO2 emissions reductions. Monetary values for
changes in risks of human health effects should therefore reflect the full consequences to the
affected individuals and to society.
Adverse health effects result in a number of economic and social consequences, including:
1. Medical costs. These include personal out-of-pocket expenses of the affected
individual (or family), plus costs paid by insurance or medicare, for example.
2. Work loss. This includes lost personal income, plus lost productivity whether the
individual is compensated for the time or not. For example, some individuals may
perceive no income loss because they got sick pay, but sick pay is a cost of business
and reflects lost productivity.
3. Increased costs for chores and caregiving. These include special caregiving and
services that are not reflected in medical costs. These costs may occur because some
health effects reduce the affected individual's ability to undertake some or all normal
chores, and he or she may require caregiving.
1 This chapter relies on previous literature reviews prepared for EPA. including Violette and
Chestnut (1983), Chestnut and Violette (1984), and Fisher et al. (1989).
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES * 5-2
4. Other social and economic costs. These include restrictions on or reduced enjoyment
of desired leisure activities, discomfort or inconvenience (pain and suffering), anxiety
about the future, and concern and inconvenience to family members and others.
Cost-of-illness (COI) measures include only medical costs plus work loss (Consequences 1
and 2 above), and thus do not reflect the total welfare impact of an adverse health effect.
Therefore, using COI measures in a quantitative assessment results in a clear downward bias
in the valuation of adverse health effects. COI measures, however, have the practical
advantages of being easily understood and often readily available because they are based on
available market and expenditure data.
A comprehensive monetary measure of value for changes in health risk is the dollar amount
that would cause the affected individual to be indifferent to experiencing an increase in the
risk of the health effect or losing income equal to that dollar amount. This monetary measure
is the maximum willingness to pay (WTP) to reduce the risk of the health effect and all
associated costs. WTP will thus reflect all the reasons an individual might want to avoid an
adverse health effect, including financial and nonfinancial concerns.2 WTP is a more
comprehensive measure of value than COI, but it can be more difficult to estimate.
Sometimes in this discussion of monetary valuation for health effects we distinguish between
health effects and health risks. A health effect refers to an illness or symptom, including
death, that is experienced by someone. A health risk is the quantitative probability that any
one individual might experience a given health effect. Changes in air quality cause changes in
the number of health effects in the exposed population, but from the point of view of the
individual what changes is the risk of experiencing a given health effect. This is because it is
unknown exactly which individuals might be affected. WTP estimation techniques for more
serious health effects such as premature mortality or chronic illness tend to focus on changes
in the risks of such health effects that an individual might experience. For example, WTP
studies for premature mortality do not estimate what individuals would be willing to pay to
prevent a certain death, but rather estimate what they are willing to pay for small changes in
risks of death.
5.1.2 WTP Estimation Techniques for Health Risks
WTP is typically measured by analyzing prices that are paid for goods and services. The
maximum price that an individual is willing to pay for a good or service is a measure of how
much they value that good or service. Prices cannot be directly observed for preventing health
2 Financial costs of health effects are not always borne fully by the individual but are shared
through health insurance and public health care subsidies. In some instances therefore empirical estimates
of WTP to avoid or reduce health effects may not fully reflect these shared costs. For a comprehensive
measure of WTP such shared costs should be added to individual WTP.
——. November 10, 1995 —
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES >> 5-3
risks because prevention of health risks is not directly purchased in the market. However,
there are instances when the monetary tradeoffs that people are willing to make between
income and health risks can be observed or measured. There are two general economic
approaches for measuring WTP for nonmarket goods such as health risk prevention. The first
is to analyze actual situations hi which WTP for health risks may be indirectly revealed; the
second is to have subjects respond to a hypothetical situation designed to have them reveal
their WTP.3
An example of the first approach is a wage-risk study in which wage premiums for risks of
death on the job are estimated. This is done by analyzing all the factors that determine
differences in actual wages between jobs, including on-the-job risks of death. The amount of
additional wages that people are paid per unit of additional risk of fatal injury is a measure of
the monetary value of that risk to the individual who voluntarily accepts that risk in exchange
for a given wage increment. The primary advantage of this type of study is that it is based on
actual behavior. The primary limitations are that it is difficult to find situations in which there
is a clear tradeoff between money and risk, and to statistically isolate WTP for a risk
increment from other factors involved in the specific behavior.
An example of the second approach is a contingent valuation study hi which subjects are
presented with a hypothetical situation that involves a tradeoff between income or
expenditures and a specific health risk or health effect. The subjects are then asked to estimate
what they would be willing to pay to change that risk by a specific amount. It is important
that the hypothetical situation presented to study subjects be realistic and easy to understand.
The primary concern with this type of study is whether subjects are able to give accurate
responses to hypothetical questions.4 ,
5.2 ISSUES IN APPLYING WTP ESTIMATES FOR THIS ASSESSMENT
Although WTP for changes in health risks is the conceptually correct monetary value measure
for this assessment, there are some limitations in available estimates. These limitations result
from uncertainties in the available estimates, inexact matches between the health risks for
which WTP estimates are available and the health risks of interest in this assessment, and the
lack of available WTP estimates for some of the health risks of interest.
3 This section provides a brief introduction to these estimation techniques. For more information see
Freeman (1993).
4 Contingent valuation is a somewhat more controversial technique than some other economic
valuation techniques. There are only a few instances when we are relying entirely on estimates from
contingent valuation studies for monetary valuation estimates for specific health effects in this report. Snell
et al. (1993) review the contingent valuation studies used in this report in light of recent recommended
contingent valuation guidelines.
: November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES + 5-4
WTP estimates are available for risks of death, but there are some differences between the
types of fatal risks for which WTP estimates are available and those of interest in this
assessment. WTP estimates are also available for some but not all types of morbidity of
concern in this assessment.
5.2.1 Issues in Applying Available WTP Estimates for Premature Mortality
There are several uncertainties in applying the available WTP estimates for valuing changes in
premature mortality risks. The justification for using the available WTP estimates is that they
provide estimates of what people are willing to pay to reduce their risks of premature
mortality by small amounts. The risks involved in this analysis are also small, but there are
some differences with regard to who is at risk and what the risk is. First, there is quantitative
evidence from the health effects literature that a large share of the individuals at risk are
elderly (65 years old or older). Two additional aspects of potential significance are the
potential health status of the people at greatest risk and differences regarding the expected
cause of death. There is very little available empirical evidence about how these factors might
affect the value of reducing risks of premature mortality. There is, therefore, some unresolved
uncertainty in applying available WTP estimates in this analysis.
Age
Available evidence of the effect of age on WTP for changes in mortality risks is discussed in
Section 5.3. The empirical evidence is quite limited, but it provides a basis for some
adjustment in average WTP estimates for elderly individuals for changes in mortality risks.
Most available WTP estimates for changes in mortality risks, however, are from studies in
which the elderly are not well-represented. The adjustment selected for the elderly in this
analysis, therefore, must be acknowledged as relatively uncertain. The adjustment for age
explained in Section 5.3 is based on an analysis of the available empirical evidence first
presented by Rowe et al. (in press). Other approaches for addressing the age question could be
justified, including making no adjustment; however, we selected the approach proposed by
Rowe et al. as a reasonable interpretation of limited empirical evidence.
Health Status
The available WTP estimates for changes in mortality risks are based on results from study
samples of individuals with average levels of health. Although it cannot be determined from
available epidemiologic studies, it is possible that those individuals at greatest risk of
premature mortality due to exposure to air pollutants are those who are already in poor health
for reasons unrelated to air pollution exposure. Some instances may involve chronic illnesses,
because of which the individual may already have a reduced life expectancy even in the
absence of pollution exposure. For example, Schwartz and Dockery (1992a) found increased
mortality rates due to chronic respiratory disease, pneumonia, and cardiovascular disease
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES > 5-5
associated with higher levels of paniculate matter. Some of these individuals apparently suffer
from preexisting chronic disease. There is not sufficient evidence available to say how having
a chronic illness might affect WTP for changes in mortality risks, but it is possible that the
reduced life expectancy and irreversibly reduced quality of life associated with many chronic
illnesses may result in lower WTP to reduce mortality risks.
Cause of Death
It is possible that people are more concerned about avoiding some kinds of death than others.
For example, Jones-Lee et al. (1985) results suggest that some people are more afraid of death
from cancer than of death from automobile accidents. This may be related to the perceived
pain, suffering, and expense associated with the illness that precedes death in the case of
cancer. Some studies also suggest that people find involuntary risks, such as pollution
exposure, less acceptable than voluntary risks, such as traffic accidents (Violette and Chestnut,
1983). Studies have not been able to separate these different aspects of the different risks of
death in terms of the potential effect on WTP. The most reliable WTP studies to date have
focused on accidental deaths, primarily on the job and in vehicle accidents. The types of death
of interest for this analysis are related to various illnesses, both chronic and acute. Based on
the limited evidence available about how people respond to different types of risks, it is likely
that if there is any error in applying available WTP estimates in this analysis it will be to
understate the WTP to avoid the types of risks of interest in this analysis.
For this analysis, available WTP estimates for changes in risks of death are applied to all
estimated mortality risks regardless of the cause of death. Although arguments could be made
for small adjustments in some cases, any such adjustment is overshadowed by the lever of
uncertainty in using these estimates, which cannot be reduced at this time. For example, WTP
estimates based on accidental death probably do not reflect the medical costs typically
associated with treatment of the chronic or acute illness that may precede premature death due
to air pollutant exposure. However, COI estimates suggest that average lifetime medical costs
per chronic respiratory disease patient are under $100,000 (Krupnick and Cropper, 1989). This
omission is not very significant relative to a selected range of WTP estimates of $2 million to
$7 million per fatality.
5.2.2 WTP to COI Ratios
WTP estimates are not available for some of the nonfatal health effects considered in this
analysis. In these cases, COI estimates are used and are adjusted upward by a factor of 2 to
compensate for the expected ratio of WTP to COI estimates for any given health effect. This
adjustment is based on limited available evidence on WTP/COI ratios, but we believe the
resulting adjusted health valuation estimates are less biased than would occur if only
unadjusted COI estimates were used. This section develops a general WTP/COI ratio to
escalate COI values to approximate WTP values. Because mis ratio is likely to be specific to
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES »> 5-6
each health effect, any such ratio based on existing studies must be seen as an approximation
to improve valuation and reduce known bias that would occur if unadjusted COI estimates
were used to value health effects.
This summary of the empirical evidence and the selected ratio for adjusting the COI estimates
is taken from Rowe et al. (in press). The empirical literature on this question is limited and
other interpretations could be justified, including making no adjustment at this time. Our
judgment was that an uncertain adjustment was preferable to no adjustment, because no
adjustment results in a clear downward bias in the estimates.
Three studies provide evidence on WTP/COI ratios for the same study population addressing
the same change in the same health effect. In each study, the participants were individuals
diagnosed with the health effect. These studies addressed changes in incidence of asthma
symptoms (Rowe et al., 1984; Rowe and Chestnut, 1986), increased angina symptoms
(Chestnut et al., 1988), and risks of cataracts (Rowe and Neithercut, 1987). In each study,
participants rated the importance of each of the components of WTP (listed in Section 5.1.1),
and provided WTP estimates for reducing or preventing these health effects. The participants
rated some non-COI consequences as more important to avoid than the COI consequences.
This again suggests that WTP significantly exceeds COI.
The dollar ratio results listed in Table 5-1 are based on estimated individual and social COI in
dollars, and on individual WTP in dollars. Individual COI is less than social COI because
society incurs some costs the individual does not (because of insurance coverage, sick pay,
and other types of compensation). Because social COI exceeds individual COI, the WTP/COI
ratio for individuals exceeds the ratio for society. Also available from the asthma and cataract
studies are respondent ratings of their COI as a share of their perceived total damages. From
these ratings, the individual and society WTP/COI ratios are computed and reported in
Table 5-1.
Across the three studies, the total social WTP/COI ratios range from 1.3 to 2.4. The COI in
these studies range from a few dollars to $7,000 per episode of cataracts. Based on these
results, we select a WTP/COI ratio of 2.0 for this analysis. Thus, we multiply available COI
estimates by 2.0 to approximate WTP, when actual WTP estimates are not available for a
given health effect.
Basing a WTP/COI adjustment on these study results is admittedly uncertain. The study
samples are small and the range of health effects is limited. However, we still judge that it is
preferable to make some adjustment than to make no adjustment. Making no adjustment in
COI estimates for valuation purposes results in a clear downward bias. We have selected a
fairly conservative adjustment factor, based on available evidence, to minimize the chance of
overadjusting. Additional evidence that these adjustment factors are conservative exists in the
WTP estimates for risks of death. Average COI estimates for fatalities are typically in the
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES * 5-7
Table 5-1
WTP/COI Ratios
Health Effect
Asthma Symptoms
Cataracts
Angina Symptoms
Dollar ratio
Dollar ratio
Respondent rated share of
total damages ratio
Respondent rated share of
total damages ratio
WTP/COI
Affected Individual
1.6 to 2.3
4.25
5.3
2.5-4
WTP/COI
Society
1.3 to 1.7
2.4
2.1
NA
Sources: Asthma: Rowe et al. (1984), Rowe and Chestnut (1986).
Cataracts: Rowe and Neithercut (1987).
Angina: Chestnut et al. (1988).
middle hundreds of thousands. WTP estimates per fatality are in the millions, a difference of
an order of magnitude.
5.3 MONETARY VALUATION ESTIMATES FOR PREMATURE
MORTALITY RISKS
Several economic studies have estimated average WTP in the United States for small changes
in risks of accidental death. These estimates have been widely used in benefit analysis of
public policy options that would result in changes in risks of death for the public (Viscusi,
1992). They are sometimes referred to as "value of life" estimates because they are expressed
on a per life basis. But it is important to note that they are based on WTP of the individual
for reducing his or her risk of premature death by a small amount, not on the total value of a
human life under all circumstances.
The estimates provided by these studies are average dollar amounts that individuals are
willing to pay for small reductions in risks of death. For example, one study might find an
average WTP of $300 for an annual reduction in risk of death of 1 in 10,000. These estimates
are'extrapolated to a per life basis by summing individuals' WTP over enough people that a
value per life saved is obtained. In this example, this value would be $3 million per life, the
result of $300 multiplied by 10,000 people. The term used for this estimate in much of the
economics literature is "value of a statistical life" (VSL) to denote that it is a summation of
WTP for small changes in risks of premature death.
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES * 5-8
Available estimates of WTP to prevent small changes in risks of death are based on situations
where individuals are observed making tradeoffs between probabilities of death and some
benefit, such as income. Most of these studies have estimated wage premiums associated with
different levels of on-the-job risks. Additionally, some contingent valuation studies have been
conducted in which subjects have been asked what they would be willing to pay to reduce, for
example, their risks of fatal accidents at work or in traffic accidents. A few averting behavior
studies have also been conducted that estimate costs associated with observed behaviors that
reduce risks, such as smoke detector usage in the home or seat belt usage in automobiles.
For the most part, available WTP estimates are for risks of accidental death in circumstances
where individuals are voluntarily exposed to risks (e.g., choosing a job or driving in a car).
The estimates are also drawn largely from studies of working-age adults. Some potentially
important differences exist between the contexts of these available estimates and the
environmental health risks being evaluated in the externality model. Environmental health
risks are related to illness rather than accidents and may in some cases fall disproportionately
on the elderly and those with already compromised health. The potential implications of these
differences were discussed in Section 5.2.1. The potential effect of age on WTP is discussed
in more detail in Section 5.3.2.
5.3.1 Summary of Available WTP Estimates
Four recent reviews of this literature evaluated and summarized available WTP estimates for
small changes in risks of death for potential use in analyses of public policy decisions (Fisher
et al., 1989; Miller, 1989; Cropper and Freeman, 1991; Viscusi, 1992). Each review concludes
with a list or range of "best" estimates that the authors judged as most appropriate for use in
evaluating public policy decisions that result in small changes in risks of death for the public.
All of these reviews covered basically the same body of literature, but the most recent review
(Viscusi, 1992) included a few additional studies that were not completed when the earlier
reviews were done. These reviews are consistent in many of their conclusions regarding which
of the available estimates are most appropriate for use in policy analysis, but there are also
differences. We take into consideration the conclusions, and their basis, of each of these four
reviews in selecting a central, low, and high estimate of WTP for changes in risks of death
for use in this analysis. The selected estimates for this analysis are discussed in Section 5.3.3.
Ranges of VSL recommended by the authors of each of the four reviews as best for policy
analysis are listed in Table 5-2.
Fisher et al. (1989) list 21 studies that each give a VSL estimate. The authors reject three
studies listed as "early low-range wage-risk estimates," primarily because of problems in the
risk data used. The authors also reject the "consumer market studies," which fall into the
category of averting behavior studies, because they argue that each of the estimates is clearly
downward biased because of study design problems or data limitations. They also reject one
of the "new wage-risk studies" that examined wages for police officers in the United States,
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES *> 5-9
Table 5-2
Recommended Ranges of VSL Estimates
Review
Fisher et al. (1989)
Cropper and Freeman (1991)
Viscusi (1992)
Miller (1989)
VSL Rounded to Millions (1994 dollars)
Low
$2
$2
$3
$1
High
$11
$7
$8
$4
because of the limited scope of the study sample and potential problems with the on-the-job
death rate data used. This leaves 13 VSL estimates judged by these authors as most
appropriate for use in policy analysis. These estimates range from $2 million to $11 million
(1994 dollars), and have an arithmetic mean of about $6 million. All but two of the 13 studies
are wage-risk studies. The remaining two studies are contingent valuation studies, which
obtained results of $4.1 million and $3.8 million. These results fall in the lower half of the
overall range. Fisher et al. caution that all the estimates above $8 million are based on wage-
risk studies using Bureau of Labor Statistics data for on-the-job risks. These data are limited
in that they give risk information by industry, but not by occupation. There is no specific
reason why these data would cause any upward bias in VSL results, but results that are not
verified by similar conclusions using different data sources are somewhat less robust. The
authors therefore conclude that the $2 million to $8 million range is the strongest because it
has been verified by different studies using varying data sources, but they do not rule out the
possibility that the higher estimates might be correct.
Cropper and Freeman (1991) present an adapted version of Table 1 from Fisher et al. They
deleted four of the 21 studies. The authors do not explain these exclusions, but presumably
they found them to be less appropriate for policy analysis than the remaining 17. Two of the
deleted studies were in categories that were rejected by both sets of reviewers, so then-
exclusion causes no change in the conclusions. The primary difference in the conclusions of
these two reviews is that Cropper and Freeman make a stronger statement that using the
Bureau of Labor Statistics on-the-job risk data apparently causes upward bias in the VSL
estimates, based on comparisons of results using different types of data. Excluding the
estimates based on Bureau of Labor Statistics data leaves six VSL estimates judged as "best"
for use hi policy analysis. These are from four wage-risk studies and two contingent valuation
studies. The wage-risk estimates selected by Cropper and Freeman range from $2.1 million to
$7.3 million (1994 dollars), and the contingent valuation estimates selected range from
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES *• 5-10
$3.5 million to $4.1 million. The arithmetic mean of all six selected VSL estimates is
$4.1 million.
Viscusi (1992) provides separate discussions and summaries of averting behavior, wage-risk,
and contingent valuation studies. His overall conclusion is that the most appropriate range of
VSL estimates for use in policy analysis is $3 million to $8 million in 1994 dollars. He also
rejects the available averting behavior study results for use in policy analysis because of clear .
downward biases in the study designs and data. Viscusi lists 27 VSL estimates from 22 wage-
risk studies and eight estimates from six contingent valuation studies. Similar to the
conclusions of the previous reviewers, Viscusi raises questions about some of the earlier
wage-risk studies that used inappropriate risk data and obtained relatively low VSL results. He
also raises some questions about some of the wage-risk studies that obtained results above $8
million. Viscusi concludes that the best VSL results from wage-risk studies are between
$3 million and $8 million. Viscusi suggests that the two earliest contingent valuation Studies
were exploratory and that less weight be given to these two estimates (one is very low, the
other is very high). The arithmetic mean of the remaining four contingent valuation estimates
is either $3.1 million or $5.1 million, depending on whether the median or the mean estimate
is selected from one of the studies. The range of the contingent valuation estimates is
$1.4 million to $4.3 million or $11.0 million, depending on whether the median or the mean
value is selected from one of the studies.
Miller (1989) uses a different approach than that used in the other three reviews and reaches
some different conclusions. He selects a larger number of available VSL estimates as
potentially appropriate for use in policy analysis, but makes several adjustments in the
estimates to reconcile differences in study design or limitations in data. Miller includes
29 VSL estimates as of "reasonably good quality." Included in these 29 estimates are most of
the estimates selected in the other reviews as most appropriate for policy analysis. An
important difference is that Miller includes results from eight averting behavior studies, which
are rejected by the other reviewers as likely to be biased downward. An additional four are
from contingent valuation studies, and the remaining 17 are wage-risk estimates. Miller made
several adjustments to the estimates, most of which resulted in lowering the estimates,
especially for some of the wage-risk studies with the highest results. The adjustments Miller
made included (1) converting the wage-risk results to after-tax dollars, (2) adjusting for
differences in labor risk data sources, (3) adjusting for failure to include nonfatal injury risks
in the analysis, (4) adjusting to a uniform value of time or discount rate if used, and
(5) adjusting for differences in perceived versus actual risks. The conceptual arguments for
some of these adjustments may be valid, but the reliability of the data used to determine the
exact adjustment to make is in many cases questionable. Miller concludes by choosing a mean
VSL estimate of $2.7 million (1994 dollars), and a range of $1.4 million to $4.3 million.
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES * 5-11
5.3.2 The Potential Effect of Age on WTP for Changes hi Mortality Risks
Although it has been suspected that age may be a factor in risk of death due to air pollution
exposure, until recently there has been little quantitative evidence in the available
epidemiologic literature. Schwartz and Dockery (1992a) report evidence that the measured
association between daily mortality rates and daily levels of ambient particulate matter is
greater for people over the age of 65. They provide sufficient information to estimate the
change in the number of deaths expected for people over 65 and under 65 for a given change
in ambient particulate matter. It is therefore important to consider whether average WTP for
changes in mortality risks might be different for people over 65.
This raises the question of whether WTP for changes in risks of death in the current time
period is different for people over 65 than for the average adult. There is limited empirical
evidence regarding this question, but some information is available. The expectation is that
WTP will be lower for a 65-year-old than for the average adult, because expected remaining
years of life are fewer. This expectation is based on the presumption that WTP for one's own
safety is derived from the utility one receives from one's own life and that this utility is to
some extent a function of the amount of time one expects to remain alive.
Some analysts have suggested that effects of age might be introduced by dividing average
WTP per statistical life by average expected years of life remaining (either discounted or not)
to obtain WTP per year of life (Miller, 1989; Harrison and Nichols, 1990). Such a calculation
implies very strong assumptions about the relationship between life expectancy and the utility
a person derives from life, namely, that utility is a linear function of life expectancy.
Although this might be correct, it is also plausible that this calculation will result in
significant understatement of WTP for the elderly. An understatement could result for a
number of reasons. One is that there may be a value to being alive that is independent of the
amount of time one expects to live. Another is that as one ages, the remaining time may be
more highly valued than it was in midlife.
We have identified one study that provides unconstrained empirical evidence concerning how
WTP for small changes in risks of death varies with age. Jones-Lee et al. (1985) conducted a
contingent valuation study concerning motor vehicle accidents and report an estimated WTP
function for characteristics of the respondents, including age.5 (There are some other studies
that provide some suggestive evidence regarding how WTP for reducing risks may change
with age, but each of these studies imposes some constraints on the conclusions in the form of
unverified model assumptions.)
This summary and proposed adjustment in the monetary values based on available empirical
evidence of the effect of age on WTP for changes in mortality risks is drawn from Rowe et al. (in press).
November 10, 1995
-------
"MONETARY VALUATION OF HEALTH EFFECTS CHANGES * 5-12
Jones-Lee et al. conducted a general population survey in the United Kingdom in which about
1,000 respondents were asked how much additional money they would be willing to pay for
transportation with a bus company with a better safety record. All relevant risk information
was quantitatively specified and the survey appears to have been well designed and executed.
Implied WTP per life (VSL) was calculated for each response. For example, the VSL is
$6 million when the WTP response is $240 for a reduction in risk of death of 4 in 100,000.
Variations in the implied VSL estimates across respondents were then examined as a function
of age and other characteristics of the respondents. An appropriate functional form was used
that allowed WTP to be a nonlinear function of age (age + age2).
The results show a statistically significant relationship estimated between age and VSL, which
was statistically strongest for the responses to the first bus safety questions. The results
indicate gradually increasing VSL until around age 45, then gradually declining VSL. The
results for both the bus safety questions imply that VSL for a person aged 65, all other things
being equal, is about 90 percent of VSL for a person aged 40.
The Jones-Lee et al. results with respect to age, based on the responses to the first bus safety
question, are:
VSL = Constant + 12,489 x (Age - Mean Age) - 660 x
(Age - Mean Age)2 + zBjXj, (5-1)
where:
VSL = the implicit VSL given by the respondent
= the other independent variables in the WTP regression.
The authors do not report mean age for the sample, but describe the sample as nationally
representative. For purposes of interpreting the regression results, we use 40 years as an
average age, which is close to the average age of adults in the United States. The average
VSL is reported as 1.6 million British pounds. We then calculated illustrative VSL estimates
at selected ages using the following formula:
VSL = 1,600,000 + 12,489 x (Age - 40) - 660 x (Age - 40)2. (5-2)
This calculation assumes that other factors that influence VSL do not change with age. The
risk of error due to this assumption seems small because only the age variables were
statistically significant in this regression.
To allow for simple comparison to the results of other studies, we calculated VSL at each age
using Equation 5-2. We men calculated VSL at each age as a percentage of VSL at age 40.
These percentages are plotted in Figure 5-1.
November 10, 1995
-------
Figure 5-1
Value of a Statistical Life as a Function of Age
160%
o
i
o-
n>
vo
vo
0%
Age in Years
Linear Function of Age +Moorc and Viscusi XShcpard and Zeckhauser
(1988) Model 1 (1982)
70
Shcpard and Zcckhauscr
Model 2 (1982)
75 80
3.loncs-I,cc ct al.
(1985)
•<
|
I
1
oo
O
O
w
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES > 5-14
Moore and Viscusi (1988) estimated a wage-risk premium for a sample of workers in the
United States. They defined risk on the job as the probability of a fatal accident multiplied by
the discounted remaining life years of the individual. They used a nonlinear estimation
technique to estimate both the risk coefficient and the implicit discount rate for time. They
also included an expected annual annuity variable to account for the possibility that a wage-
risk premium might not be as high if available insurance covers some of the risk to
dependents. The results showed a significant relationship between wages and risks of fatal
accidents and implied a value per statistical life of about $6.5 million (1986 dollars). The
finding of a significant (negative) relationship between wages and expected annual annuity
suggests that estimates that ignore potential death benefits may understate WTP to reduce
risks of death. The estimated discount rate was 10 percent to 12 percent.
The Moore and Viscusi model assumes a constant value per year of life, and future years are
discounted at rate r. The model, therefore, does not provide an unconstrained test of how VSL
varies with age. VSL at different ages is simply a function of the discount rate, according to
this model, and is therefore proportional to discounted remaining life years. The model
implies that WTP for small changes in current risks decreases with age throughout a person's
lifetime. How fast it declines depends on the discount rate. Moore and Viscusi define
discounted remaining life years as:
DRLY - 1/r x [1 - exp(-r x R)] (5-3)
where:
DRLY = discounted remaining life years
R = expected life years remaining.
The implications of different discount rates on WTP for changes in risks of death can be
illustrated as follows. VSL will be proportional to the discounted remaining life years
(DRLY). This means that the ratio of VSL at age 40 to VSL at age 65 will be the same as the
ratio of DRLY at age 40 and DRLY at age 65. The implications of Moore and Viscusi's
results from their linear wage function (r = 9.6 percent) with respect to the age of the worker
are shown in Figure 5-1. It should be noted that the estimates are based on a sample of
317 working adults, which included few individuals over age 60 (62 is two standard
deviations above the mean age). Also, life expectancies do not actually decline linearly with
age, as is assumed in the calculations that underlie Figure 5-1. Average life expectancy at
birth in the United States was 75 years in 1983, but was 17 years for 65-year-olds.
Cropper and Freeman (1991) provide a summary of the life-cycle consumption-saving model
that can be used to derive a theoretical definition of WTP for changes in me probability of
death. This model is based on the premise that utility is a function of consumption. The
authors note that if there is additional utility derived from survival per se, then the life-cycle
model provides a lower bound estimate of WTP. Of interest is what the model predicts in
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES + 5-15
terms of how WTP for changes in risks of death in the current time period changes as a
function of age. For a quantitative example, this depends on assumptions regarding a lifetime
pattern of earnings, endowed wealth, the rate of individual time preference, and other
parameters of the model. These will all vary for different individuals, and uncertainty exists
empirically about population averages for many of these factors. However, using reasonable
values to calibrate the model is illustrative.
Cropper and Freeman (1991) note that if consumption is constrained by income early in life,
the model predicts that VSL increases with age until age 40 to 45, and declines thereafter.
Shepard and Zeckhauser (1982) illustrate this point with numerical examples for the life-cycle
model. When they estimate the model with reasonably realistic parameters and assume no
ability to borrow against future earnings or to purchase insurance, they find a distinct hump in
the VSL function that has a peak at about 40 years and drops to about 50 percent of the peak
by 60 years. When they allow more ability to borrow against future earnings and to purchase
insurance, the function flattens and at 60 years drops only to 72 percent of the VSL at age 40.
For comparison purposes, all of the estimates discussed above are plotted in Figure 5-1 along
with the relationship between VSL and age implied by a simple linear decline with age. This
linear decline implies that VSL at age 65 is about 30 percent of VSL at age 40. This is a
much larger decline in VSL as a function of age than implied by the available empirical .
results reported above. The strongest weight should be given to the Jones-Lee et al. results
because they are based on a representative general population survey and were not unduly
constrained by an imposed functional form. However, survey results can be highly variable
and need to be interpreted cautiously until verifying results from multiple studies are obtained. •
The life-cycle model results are quite variable depending on assumptions used to quantify the
model. These assumptions have not been verified empirically. Because the model defines
utility as a function of consumption and consumption is a function of time, it is expected that
if the life-cycle estimates err it is on the side of overstating the effect of age on VSL (in other
words, reducing VSL too much at age 65 relative to age 40). The error would result if there
is some value to just being alive independent of consumption. At consumption levels above
subsistence, this is quite plausible. Therefore, these estimates should be interpreted as
representing the maximum plausible reductions in VSL as a function of age.
5.3.3 Monetary Estimates Selected for this Analysis
Obviously, there is some judgment involved in selecting central, high, and low values for the
WTP for changes in risks of death. The selected mortality valuation estimates for each age
group are summarized in Table 5-3. We selected $4.5 million as the central estimate,
$2.5 million as the low, and $9.0 million as the high for those under 65. The central estimate
of $4.5 million is consistent with the mean ($4.1 million) of the six estimates indicated by
Cropper and Freeman as most appropriate for policy analysis uses. It is within the range of
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES »• 5-16
Table 5-3
Summary of Selected Monetary Values for Mortality Effects
Population Group
>65 years
<65 years
Age Weighted Average
Selected Probability Weights
VSL Estimate (1994 dollars)
Low
$1.9 million
$2.5 million
$2.0 million
33%
Central
$3.4 million
$4.5 million
$3.5 million
50%
High
$6.8 million
$9.0 million
$7.1 million
17%
results from both wage-risk and contingent valuation estimates, and is consistent with giving
less weight to the wage-risk studies that have relied on Bureau of Labor Statistics risk data.
When these are included, the mean estimate from the Fisher et al. review is $6.3 million. In
selecting the central estimate we have given less weight to the Miller review because of the
uncertainties involved in many of the adjustments he made in the estimates. Both the study
selection and the adjustments made by Miller suggest that his conclusions are on the low side
in terms of an appropriate VSL estimate for policy analysis. The central estimate of
$4.5 million is close to the upper end of the range selected by Miller as appropriate for policy
analysis. The low estimate selected for those under 65 is just below Miller's mean VSL
estimate of $2.7 million. It is the lower end of the range selected by Fisher et al. and Cropper
and Freeman. The selected high estimate falls within the upper estimates of $11 million and
$7 million from the first three reviews summarized above. The VSL estimates discussed in
Section 5.3.1 are based primarily on samples of working age adults. A few of the contingent
valuation studies included individuals of retirement age, but this age is not well represented in
the mean VSL values. We therefore apply the selected VSL estimates from these studies to
the under 65 years old population.
Available evidence suggests that WTP for small changes in risks of death for people over age
65 can be expected to be lower than WTP for the same change in risk at age 40. However,
there is considerable uncertainty about how much lower. The most relevant direct evidence
suggests that the decline in VSL with age may be relatively small (e.g., 90 percent of the age
40 WTP at age 65). The evidence strongly suggests that a linear decline in VSL with age
significantly understates actual VSL over age 65. Based on our evaluation of the above
described evidence regarding VSL and age, we utilize the Jones-Lee et al. results to calculate
a weighted average VSL based on the approximate age distribution for the U.S. population
age 65 and older. This produces an adjustment to VSL for those 65 years old and older of
about 75 percent of the average VSL for adults under age 65. Taking 75 percent of the
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES * 5-17
estimates per statistical life selected above for adults under 65, we get a central estimate of
$3.4 million for those over 65, a low of $1.9 million, and a high of $6.8 million.
A age-weighted average VSL for this analysis is then calculated on the assumption that
85 percent of the sulfate-related deaths are people aged 65 and over (See Chapter 4). The
results are shown in Table 5-3. These are the VSL estimates applied to the predicted changes
in premature deaths associated with Title IV in this assessment.
The selection of probability weights for the low, central, and high estimates is somewhat
arbitrary because there are several uncertainties in using these estimates in this analysis for
which no quantitative information is available. The selected weights therefore reflect the
uncertainty in the underlying WTP estimates for small changes in risks of accidental death for
working-age adults, but do not fully reflect the uncertainty in applying these estimates in this
analysis. The weight selected for the central estimate is 50 percent, because the underlying
WTP estimates are predominately in the $3 to $6 million range. A weight of 33 percent is
given to the low estimate and a weight of 17 percent to the high. This reflects that the high
estimate is represented by fewer studies and a somewhat skewed distribution in the available
WTP estimates. These weights result in a weighted mean value that approximates the selected
central estimate.
5.4 MONETARY VALUATION ESTIMATES FOR MORBIDITY
WTP estimates of value are available for about half of the nonfatal health effects identified in
Chapter 4, primarily the least serious health effects. However, most of the WTP studies
completed to date have limitations because of small sample sizes and limited variation in the
health effect studied, and few of these studies have been replicated. Some interpretations and
adjustments in the results of the WTP studies have been necessary in applying them for this
analysis. These studies have been reviewed and synthesized in previous air quality benefits
studies (Rowe et al., in press; Krupnick and Kopp, 1988; Hall et al., 1989; Thayer, 1991;
Unsworth and Neumann, 1993). We rely to a large extent on these previous reviews for
specific interpretations.
When WTP estimates are not available at all, the monetary estimates are based on COI
information, and the COI values are inflated to WTP estimates, as discussed hi Section 5.2.
The COI information used in this analysis reflects medical costs and lost productivity due to
illness. The average daily wage is used as a measure of lost productivity for days when all
normal activities are prevented because of illness. Such days include days spent in the
hospital, one day for each emergency room visit, and days spent hi bed because of illness.
The average wage rate is used as a measure of the average opportunity cost of time for
employed and not-employed individuals, on the presumption that those who are not employed
value their leisure or household services at a level equal to the wage they forego in choosing
not to pursue paid employment. This approach may somewhat overstate foregone wages for
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES * 5-18
the elderly and women, who make up a large share of the not-employed group and may have
less than average earning power in the labor market. On the other hand, this approach does
not reflect any productivity losses beyond the average work-day hours, thereby understating
productivity losses for employed and not-employed individuals who perform household,
childcare, and community service work beyond the usual work-day hours. This omission,
however, is offset by the adjustment used to proxy WTP when using the COI estimates. For
these calculations, we use the 1994 median daily wage for full-time salaried workers in the
United States, which is about $93 (U.S. Dept. of Labor, 1995).
The available WTP studies provide some information on the range as well as the mean WTP
values. In general, these ranges are minus 50 percent to plus 50 to 100 percent. A range of
plus or minus 50 percent is therefore applied to the central estimates of WTP based on COI
data in this analysis to derive the low and high estimates. High and low values are selected
from the range of WTP results available when WTP studies have been conducted for those
health endpoints. The low, central, and high WTP estimates for all morbidity effects are given
equal probability weights. This reflects the limited number of empirical studies providing the
WTP estimates and the fairly extensive assumptions and approximations used in deriving all
of the estimates.
5.4.1 Adult Chronic Bronchitis
Viscusi et al. (1991) and Krupnick and Cropper (1992) conducted a set of survey exercises to
estimate WTP for reducing risks of developing chronic respiratory disease. In both studies,
respondents were presented with trade-off options for risks of developing chronic bronchitis
(or chronic respiratory disease in general) versus cost of living. Respondents were presented
with hypothetical residence location options where in some locations risks of developing
chronic respiratory disease are lower but cost of living is higher. An additional trade-off
question was for risks of developing chronic bronchitis versus risks of death in an auto
accident. An interactive computer program was used to adjust the trade-off until the
respondent reached a point of indifference between the two options. At this point, a maximum
WTP to prevent developing chronic bronchitis is revealed.
The health endpoint defined in these studies does not exactly match that defined in the Abbey
et al. (1995) study, upon which the estimates of new cases of chronic bronchitis are based
(see Chapter 4). The primary difference is the level of severity. The WTP studies defined a
severe case of chronic bronchitis. The Abbey et al. results reflect a more average case. In this
section we present the results of these WTP studies and a procedure for adjusting the results
to better reflect the level of severity of interest for this analysis.6
This adjustment procedure is based on information reported by Krupnick and Cropper (1992) and
was suggested by Alan Krupnick in personal communication.
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES > 5-19
The samples for the two studies differ. Viscusi et al. selected a representative sample of about
390 respondents. Krupnick and Cropper selected a sample of individuals who had a relative
with a chronic respiratory disease. The Krupnick and Cropper sample was smaller (about
190 respondents) and less representative of the general population (lower average age and
higher average income), reflecting a large percentage of respondents taken from the University
of Maryland staff and students. The intent of the Krupnick and Cropper study was to test for
the effect of familiarity with the disease .on WTP responses.
Both studies used a definition of chronic bronchitis that reflects a severe case. The description
of the disease included persistent symptoms of cough and phlegm, limits in physical activities,
and ongoing medical care. Krupnick and Cropper used this definition in one version, and
asked respondents to consider the risk of developing "a case of chronic respiratory disease like
your relative's" in a second version. The relatives had chronic bronchitis, asthma, or
emphysema. Respondents provided information on the severity of the relative's disease based
on the number of symptoms present. This ranged from 0 to 13, where 13 reflects the severe
chronic bronchitis case defined in the earlier questions. The analysis of WTP responses
included the effect of the severity of the relative's case on the WTP response. At the mean of
the variables, the estimated elasticity of WTP with respect to severity was 1.16. This means
that WTP increased by 1.16 percent for every 1 percent increase in the 0 to 13 symptoms
scale.
The WTP results from Viscusi et al. are more appropriate for this assessment because they are
from a study sample that is more representative of the general population. The responses
reflect the maximum amount the respondents revealed they would be willing to pay to reduce
then- annual risk of developing chronic bronchitis by a specified amount. The authors then
calculated the implicit WTP per statistical case avoided. The median response for the cost of
living trade-off was approximately $570,000, and the arithmetic mean was about $1,100,000
in 1994 dollars. The authors caution that the mean is affected by a small number of fairly
high estimates and recommend that the median is more representative of the sample. We
cautiously accept this recommendation until the accuracy of the high estimates can be further
verified in repeated studies and analyses. For a low estimate for a severe case of chronic
bronchitis we select the 20th percentile value of $340,000 and for a high estimate we select .
the 80th percentile value of $900,000.
We use an elasticity estimate for numbers of symptoms to scale the estimates for a severe
chronic bronchitis case to better reflect WTP to avoid a more typical case. The elasticity
estimate is calculated from results reported by Krupnick and Cropper for a combined analysis
of chronic bronchitis, asthma, and emphysema. Using this estimate for chronic bronchitis
assumes that the elasticity of WTP with respect to severity is similar for chronic bronchitis to
that for all three diseases combined. The mean severity rating reported for the Krupnick and
Cropper sample is 6.5, based on the 0 to 13 scale. Using the elasticity at the mean of 1.16,
this suggests that WTP for an average case is 58 percent lower than for a case at 13 on the
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES > 5-20
scale. Using this to adjust the Viscusi et al. estimates, we get a central WTP estimate of
$240,000, a low of $140,000, and a high of $380,000 for an average chronic bronchitis case.
It is important to note that these WTP estimates for preventing a new case of chronic
bronchitis reflect the perceived welfare effects of living with chronic bronchitis over the entire
course of the illness, which can span many years. It is a measure of the present value of the
welfare effect that occurs over a multiple-year period. This is somewhat different than the
other morbidity effects considered in this analysis which are short-term effects. In using the
WTP values for chronic bronchitis we are assigning the full welfare effect for the new chronic
bronchitis case in the year in which the clinical onset of the disease occurs. We do the same
with the acute morbidity effects, but in those cases the illness typically begins and ends in the
same year.
5.4.2 Respiratory Hospital Admissions
WTP estimates for respiratory hospital admissions (RHA) are not available. We therefore use
the COI approach. The American Hospital Association reports an average cost per day of a
hospital stay of $820 in 1992 dollars (as cited in U.S. Bureau of the Census, 1994). This is
inflated to $920 (1994 dollars) using the medical consumer price index. We calculated the
average length of stay in the hospital for the 13 ICD-9-CM codes7 in the Burnett et al.
(1995) study (see Chapter 4) using data from the 1992 National Hospital Discharge Survey
(Graves, 1994). We found an average length of stay for a respiratory hospital admission of
about 6.8 days, which is slightly longer than the overall average length of stay in the hospital
for all conditions of approximately 6.2 days (Graves, 1994). The length of stay is multipKed
by the average cost per day as an estimate of the medical cost of a RHA. The length of stay
is multiplied by the average daily wage (W) as an estimate of the value of lost productivity
for employed and not- employed individuals on the presumption that it is a measure of
average opportunity costs for all individuals. The medical cost and lost productivity estimates
are summed and multiplied by the WTP/COI ratio of 2 to account for additional potential pain
and suffering and activity losses not reflected in the COI numbers. The central estimate is
thus calculated as follows:
Central $/RHA = (6.8 x 910) + (6.8 x W) x WTP/COI. (5-4)
Therefore, the central estimate is $14,000 (1994 dollars), rounded to the nearest thousand.
Applying a plus or minus 50 percent adjustment results in a low estimate of $7,000 and a
high estimate of $21,000.
7 The ICD-9-CM codes included were: 466, 480, 481, 482, 483, 485, 486, 490, 491, 492, 493, 494,
and 496. The diseases they correspond to include acute bronchitis, chronic bronchitis, pneumonia,
emphysema, and asthma.
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES * 5-21
5.4.3 Cardiac Hospital Admissions
WTP estimates for cardiac hospital admissions (CHA) are not available. We therefore use the
COI approach. The American Hospital Association reports an average cost per day of a
hospital stay of $820 in 1992 dollars (as cited in U.S. Bureau of the Census, 1994). This is
inflated to $920 (1994 dollars) using the medical consumer price index. We calculated the
average length of stay in the hospital for the 4 ICD-9-CM codes8 in the Burnett et al. (1995)
study (see Chapter 4) using data from the 1992 National Hospital Discharge Survey (Graves,
1994). We found an average length of stay for a cardiac hospital admission of about 6.9 days,
which is slightly longer than the overall average length of stay in the hospital for all
conditions of approximately 6.2 days (Graves, 1994). The length of stay is multiplied by the
average cost per day as an estimate of the medical cost of a CHA. The length of stay is
multiplied by the average daily wage (W) as an estimate of the value of lost productivity for
employed and not- employed individuals on the presumption that it is a measure of average
opportunity costs for all individuals. The medical cost and lost productivity estimates are
summed and multiplied by the WTP/COI ratio of 2 to account for additional potential pain
and suffering and activity losses not reflected in the COI numbers. The central estimate is
thus calculated as follows:
Central $/CHA = (6.9 x 910) + (6.9 x W) x WTP/COI. (5-5)
Therefore, the central estimate is $14,000 (1994 dollars), rounded to the nearest thousand.
Applying a plus or minus 50 percent adjustment results in a low estimate of $7,000 and a
high estimate of $21,000.
5.4.4 Restricted Activity Days
A restricted activity day (RAD) is a measure of illness defined by the Health Interview
Survey (HIS) as a day on which illness prevents an individual from engaging in some or all
of his or her usual activities. This includes days spent in bed, days missed from work, and
days with minor activity restrictions because of illness. WTP estimates for preventing a RAD
are not available. We therefore approximate WTP for an average RAD using available COI
data and WTP estimates for days with symptoms.
RADs reflect a combination of complete activity restrictions and minor activity restrictions. It
is unknown what proportion of RADs attributable to air pollution exposure is minor rather
than severe. Recent data from the HIS indicate that about 40 percent of all RADs are bed-
8 The ICD-9-CM codes Included were: 410, 413, 427, and 428. The diseases they correspond to are
acute myocardial infection, angina pectoris, cardiac dysrhythmias, and heart failure.
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES > 5-22
disability days.9 The results of Ostro (1987) suggest that RADs associated with air pollution
exposure may be less severe on average than all RADs. We therefore presume a lower
proportion of bed-disability days for this analysis than the national average for all RADs. We
select an assumption that 20 percent of RADs due to air pollution exposure are bed-disability
days.
WTP studies do provide some information about values for preventing illness symptoms that
are probably associated with minor restricted activity days (MRADs). There are no studies
specifically addressing the WTP to avoid an MRAD; however, Loehman et al. (1979), Tolley
et al. (1986), and Berger et al. (1987) report results from survey respondents who were asked
how much they would be willing to pay to avoid a day with various specified symptoms such
as serious or minor coughing. The focus of these studies was on respiratory symptoms that
might be related to air pollution levels, but the results from each of these studies are difficult
to interpret for this analysis because there is fairly wide variability in the responses and
because the definitions of symptoms vary. However, Krupnick and Kopp (1988) note that an
MRAD is probably more severe than a single minor symptom day (congestion, cough, etc.);
hence, they concentrate on the WTP estimates for severe symptoms in Loehman et al. and
symptom combinations in Tolley et al. For a central estimate, they select $26 (1994 dollars),
which is Loehman's high median value for a severe symptom day.
Productivity losses associated with more serious RADs (bed-disability days) are estimated as
equivalent to the daily wage rate for employed individuals. We apply the same measure of
lost productivity for not-employed individuals on the presumption that it is a measure of
average opportunity costs for all individuals. This lost productivity estimate is multiplied by
the WTP/COI ratio of 2 to account for additional potential pain and suffering, additional
leisure activity losses, and potential medical costs that are not reflected in the lost productivity
estimates. Taking a weighted average of the value for more serious and more minor RADs
gives the average value for an air pollution induced RAD as follows:
Central $/RAD = [0.20 x W x WTP/COI] + [0.80 x 26]. (5-7)
Therefore, the central estimate is $60 (rounded to the nearest ten). Applying a plus or minus
50 percent adjustment results in a low estimate of $30 and a high estimate of $90.
5.4.5 Asthma Symptom Days
Krupnick and Kopp (1988) review two studies that provide monetary value estimates for
asthma symptom days. The first is a study by Krupnick (1986), which presents the medical
expenditures associated with ozone-induced asthma attacks. The expenses vary by the baseline
9 National Center for Health Statistics (1992) reports average number of restricted activity days for
all adults in the United States in 1991 was 16.1, and the average number of bed-disability days was 6.5.
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES > 5-23
number of attacks and by the assumed prices for medical services. Krupnick and Kopp use
these figures as a benchmark for calibrating estimates of WTP.
The second study (Rowe and Chestnut, 1986) is a WTP survey study that obtained asthmatics'
estimates of WTP to prevent an increase in "bad asthma days" (BADs). Each respondent
defined for himself a BAD on a 1 to 7 severity scale for asthma symptoms. After analyzing
the WTP responses, Rowe and Chestnut found WTP estimates that are about 1.8 times greater
than the medical costs found by Krupnick. Krupnick and Kopp point out that this finding is
consistent with economic logic and lends credibility to both studies. Thus, for WTP values to
prevent an asthma attack, Krupnick and Kopp rely on the Rowe and Chestnut estimates.
Rowe and Chestnut found that the WTP responses were positively associated with the baseline
number of annual attacks. The values also varied by how an asthmatic defined a BAD. For
example, when a BAD was defined as a day with any symptoms, the WTP estimate was $13
in 1994 dollars. At the higher end of the scale, when a BAD was defined as a day with more
than moderate symptoms, the WTP was $58. A central estimate is $36. We follow Krupnick
and Kopp and adopt these WTP estimates.
5.4.6 Lower Respiratory Symptom Days
Krupnick et al. (1990) estimated the number of study subjects who reported any respiratory
symptoms on a given day as a function of air pollutant levels on that day. These included
19 specific symptoms such as coughing, congestion, and throat irritation. The symptoms were
noticeable to the subjects, but did not necessarily result hi any changes in the person's
activities on that day. This health effect therefore includes but is not limited to restricted
activity days. In the procedures used to add the health effects cases, restricted activity days
are subtracted from acute respiratory symptom days because of the overlap in the definitions
of these health effects. The monetary valuation required for acute respiratory days is therefore
a value for the days on which symptoms are noticeable but do not restrict normal activities for
that day.
Loehman et al. (1979) and Tolley et al. (1986) obtained estimates of WTP to avoid a day
with a single minor respiratory symptom such as head congestion or coughing. Their median
results per day in 1994 dollars range from $6 to $17. We prefer the median results from these
studies because neither study did any adjusting for potentially inaccurate high WTP responses,
resulting in reported mean WTP estimates that far exceed the median values. The medians
may be too low relative to the average WTP that we would prefer to use in this analysis, but
there is less risk of significant upward bias in the median estimates from these studies. We
prefer to err in this direction. We select $11 as typical of the range of estimates obtained in
these two studies for minor respiratory symptoms. We select a low of $6 and a high of $17.
November 10, 1995
-------
MONETARY VALUATION OF HEALTH EFFECTS CHANGES + 5-24
5.4.7 Summary of Selected Morbidity Values
Table 5-4 provides a summary of the selected monetary values for human morbidity effects.
Table 5-4
Summary of Selected Monetary Values for Morbidity Effects
Morbidity Effect
Adult chronic
bronchitis
Respiratory hospital
admission
Cardiac hospital
admission
Restricted activity day
Asthma symptom day
Lower respiratory
symptom day
Selected probability
weights for all effects
Estimate per Incident (1994$)
Low
$140,000
$7,000
$7,000
$30
$13
$6
33%
Central
$240,000
$14,000
$14,000
$60
$36
$11
34%
High
$380,000
$21,000
$21,000
$90
$58
$17
33%
Primary Source
Viscusi et al. (1991)
Krupnick and Cropper
(1992)
Equation (5-4)
Graves (1994)
Equation (5-5)
Graves (1994)
Equation (5-7)
Loehman et al. (1979)
Rowe and Chestnut
(1986)
Loehman et al. (1979)
Tolley et al. (1986)
Type of
Estimate'
WTP
Adjusted
COI
Adjusted
COI
WTP&
Adjusted
COI
WTP
WTP
1 WTP = Contingent valuation WTP estimate.
Adjusted COI = COI x 2 to approximate WTP.
November 10, 1995
-------
CHAPTER 6
RESULTS AND CONCLUSIONS
This chapter presents the quantitative results of the health benefits assessment for Title IV
when all the pieces are put together as described in Table 2-1. First, the annual results are
presented for 1997 and 2010 using the default assumptions in the calculation of changes in
health effects and their monetary valuation for the eastern United States and southern portions
of Ontario and Quebec, Canada. These results for 1997 and 2010 are presented in 1994 U.S.
dollars and have been adjusted for expected average population growth in the United States
and Canada. These results are presented as annual totals for the U.S. and Canadian study
areas, with mean estimates from the distribution of the final results for each year presented as
well as the 20th and 80th percentiles of the distribution. State-by-state estimates of the mean
annual estimates and present value calculations for the 1995 to 2010 period are also presented.
Results of some sensitivity analyses are then presented to give a sense of the directions and
magnitudes of effects of key assumptions in the assessment calculations.
6.1 ANNUAL RESULTS BASED ON DEFAULT ASSUMPTIONS
Table 6-1 shows the mean, 20th percentile, and 80th percentile estimates based on the default
assumptions for the 31-state eastern United States area for 1997 and 2010. Table 6-2 gives
comparable results for the southern portions of Ontario and Quebec. The mean estimates are
calculated using the probability distributions assigned to each health effect category, as
discussed in Chapters 4 and 5. The 1997 estimates compare the annual median sulfate aerosol
concentrations predicted as a result of estimated SO2 emissions in 1997 under the Title IV
requirements, with predicted banking of emissions allowances incorporated into the
estimates,1 with sulfate concentrations estimated based on 1985 SO2 emissions. The 2010
estimates are based on predicted S02 emissions in 2010, after Title IV is expected to be fully
implemented, versus what S02 emissions are predicted to have been hi 2010 without Title IV
but with all other provisions of the Clean Air Act in place.
The annual estimated mean health benefits hi the eastern United States (Table 6-1) for 1997
are $10.6 billion, and they rise to $40.0 billion by the year 2010. The mean estimates for
Canada (Table 6-2) add an additional $908 million in 1997 and $955 million in 2010. The
mean results represent the estimated annual number of cases of each type of health effect
expected to be prevented as a result of Title IV versus what would have occurred without
1 The banking assumptions suggest that SO2 emissions will be lower in years between 1995 and
2000 than they would have been without banking, but that the rate of decline in emissions will be
somewhat slower after the year 2000 (see Figure 3-1).
November 10, 1995
-------
z:
I
Table 6-1
Estimates of Annual Human Health Benefits of Title IV
for the Eastern United States with Default Assumptions
(millions of 1994 dollars)
Health Effect
Premature
Mortality
Chronic
Bronchitis
(new cases)
Respiratory
Hospital
Admissions
Cardiac
Hospital
Admissions
Asthma
Symptom
Days
Restricted
Activity Days
(net)
Days with
Lower
Respiratory
Symptoms
(net)
Total Annual
Health
Benefits
1997
Annual Number of Cases Prevented
20th
Percentile
408
1,648
663
510
791,232
1,202,785
2,028,424
Mean
2,568
3,864
805
673
1,604^41
2,467,066
5,002,393
80th
Percentile
5,714
6,590
918
867
2^73,697
3,809,253
7,259,946
Aanual Monetary Value
20th
Percentile
$1,428.0
$507.5
$5.7
$4.6
$20.9
$70.6
$31.8
$3,219.1
Mean
$9307.2
$974.0
$113
$9.4
$56.9
$147.0
$56.7
$10,562.3
80th
Percentile
$19,999.0
$1377.5
$17.1
$13.9
$93.2
$228.6
$90.0
$20,684.1
2010
Annual Number of Cases Prevented
20th
Percentile
1,539
6,179
2,501
1,924
2,983,490
4,514,939
7,614,168
Mean
9,678
14,564
3,036
2,552
5,951,693
9,283,999
18,619,000
80th
Percentile
21,544
24,715
3,462
3,270
8,950,470
14,298,930
27,251,920
Annual Monetary Value
20th
Percentile
$5386.5
$1,903.0
$21.5
$17.5
$78.7
$265.0
$119.3
$12,131.5
Mean
$35,234.8
$3,705.8
$42.4
$35.7
$212.9
$554.7
$212.8
$39,999.0
80th
Percentile
$75,404.0
$5,165.3
$64.6
$52.5
$351.3
-
$857.9
$338.0
$77,915.5
CO
i
C/5
1
C/l
ON
to
-------
r
*
VO
Health Effect
Premature Mortality
Chronic Bronchitis
(new cases)
Respiratory Hospital
Admissions
Cardiac Hospital
Admissions
Asthma Symptom
Days
Restricted Activity
Days (net)
Days with Lower
Respiratory
1 Symptoms (net)
Total Annual Health
1 Benefits
Table 6-2
Estimates of Annual Human Health Benefits of Title IV
for Ontario and Quebec, Canada States with Default Assumptions
(millions of 1994 dollars)
1997
Annual Number of Cases
Prevented
20th
Pertentile
35
140
56
43
66,915
97,734
164,822
Mean
217
329
68
57
133,825
199,194
401,231
80th
Percentile
483
562
78
73
200,746
309,526
589,916
Annual Monetary Value
20th
Percentile
$122.5
$43.3
$0.5
$0.4
$1.8
$5.7
$2.6
$273.3
Mean
$801.2
$833
$0.9
$0.8
$4.8
$12.0
$4.5
$907.6
80th
Percentile
$1,690.5
$117.4
$1.4
$1.2
$7.9
$18.6
$7.3
$1,746.9
2010
Annual Number of Cases Prevented
20th
Percentile
37
150
60
46
71,594
104,568
176347
Mean
232
355
73
61
142,267
215,270
433,821
80th
Percentile
517
601
83
78
214,783
331,168
631,165
Annual Monetary Value
20th
Percentile
$129.5
$46.3
$0.5
$0.4
$1.9
$6.1
$2.8
$290.8
Mean
$839.2
$91.0
$1.0
$0.9
$5.1
$13.0
$4.9
$955.0
80th
$1,809.5
$125.8
$1.6
$1.3
$8.4
$19.9
$7.8
$1,868.1
CO
CO
I
CO
1
c/j
u>
-------
RESULTS AND CONCLUSIONS *• 6-4
Title IV. Estimates for both years and both countries are in 1994 U.S. dollars, and have been
adjusted for expected population growth based on the mid-forecasts of the U.S. Census (U.S.
Bureau of the Census, 1994). Canadian population growth estimates are from World Bank
population projections (Bos et al., 1992).
The health benefit estimates are dominated by the premature mortality and the chronic
bronchitis effects. The numbers of cases in these health effects categories are relatively small,
but the high monetary values per case result in large monetary benefits for these categories.
Premature mortality reductions alone account for about 88 percent of the total health benefits.
Chronic bronchitis reductions are about 9 percent of the total. The combination of the
premature mortality reductions and the chronic bronchitis reductions represent about
97 percent of the total health benefits.
The largest numbers of cases reduced are for asthma symptom days, restricted activity days,
and days with acute lower respiratory symptoms. The restricted activity days are net of days
in the hospital and asthma symptom days because these health effects categories may
substantially overlap. The lower respiratory symptom days are net of the fraction of restricted
activity days that might also be attributed to lower respiratory symptoms. In 2010, the
estimated mean reduction in the number of asthma symptom days because of Title IV is about
6 million in the eastern United States; the mean net restricted activity days prevented is about
9 million; and the mean estimated number of days with acute lower respiratory symptoms
prevented, net of restricted activity days, is about 19 million. Together, these represent about
3 percent of the total mean monetary health benefits.
The other categories of health effects (respiratory and cardiac hospital admissions) together
represent only about 0.2 percent of the total monetary health benefits. This is because
relatively small risks and small monetary values combine to give relatively small total benefit
amounts.
The estimates of reductions in health effects in Canada are based on estimates of changes in
sulfate aerosol concentrations in Canada predicted to result from changes in SO2 emissions
generated in the United States. The estimates for Canada are primarily in the Windsor-Quebec
corridor, where the greatest share of the Canadian population likely to be affected by the
transport of SO2 emissions in the eastern United States is located. The estimates for Canada
represent an additional 9 percent of the title IV benefits in 1997 estimated for the United
States population. The estimates for Canada do not increase substantially from 1997 to 2010
because the estimated reductions in sulfate concentrations in Canada do not change
substantially from 1997 to 2010. This is presumably because the upwind locations in the
United States that affect this area of Canada see their greatest reduction in SO2 emissions in
the first phase of the Title IV program. In 2010, the estimates for Canada add an additional 2
percent to the 2010 estimates for the United States population.
Most of the selected concentration-response and monetary value estimates are based on
statistically derived results. These estimates therefore have some quantified statistical
November 10, 1995
-------
RESULTS AND CONCLUSIONS > 6-5
uncertainty based on the estimated statistical variance in the results. For all of the health
effects and monetary value estimates, low and high as well as central estimates were selected
based on the estimated statistical variance and analyst judgment. In general, the selected high
and low estimates represent plus and minus approximately one statistical standard error.
It is not appropriate to combine all the "low" estimates or all the "high" estimates to calculate
upper and lower bounds on the final estimates, because it is highly unlikely that either all the
lows or all the highs would be correct. Such extreme assumptions would significantly
overstate the statistical uncertainty in the estimates. Instead, we have assigned probability
weights to the low, central, and high estimates which when incorporated in the calculation
process allow determination of the probability distribution of the'total health benefit results.
The mean, 20th percentile, and 80th percentile estimates shown for each year in Tables 6-1
and 6-2 are the result of this procedure. All three of these estimates for each health effect
category are based on the default assumptions, with each estimate representing a different
selected point in the estimated probability distribution calculated for the health effect category
and for total health benefits. The 20th percentile of the distribution of total health benefits for
2010 in the eastern U.S. is about $12 billion with the default assumptions. This means that 20
percent of the estimated values are below this amount and 80 percent are above it. The 80th
percentile of the distribution is about $78 billion with the default assumptions. This means
that 20 percent of the estimated values are above this amount and 80 percent are below it.
Thus, sixty percent of the distribution of the annual total health benefits in 2010 in the eastern
U.S. falls between $12 billion and $78 billion, with a mean value of $40 billion, when the
default assumptions are used and the selected probability weights for each selected low,
central, and high estimate are incorporated into the calculations.
Table 6-3 lists the estimated 1997 and 2010 mean annual health benefits by state and
province. The per capita health benefits are calculated by dividing the total annual benefits in
each state or province by the estimated 1997 and 2010 populations in each state and province
(based on national average population forecasts). These give a picture of the distribution of •
the health benefits across the region. Five states have average annual per capita health benefits
in 2010 that exceed $400. These are West Virginia, Georgia, Kentucky, Tennessee, and
Alabama. Eight more states have per capita benefits between $200 and $400. These are
Maryland, Delaware, Virginia, North Carolina, South Carolina, Ohio, Indiana, and Mississippi.
The largest per capita benefits are thus in the Ohio River Valley, the central Atlantic states,
the central and southern Appalachian states, and the eastern Gulf coast states. The lowest
benefits are in the northern states of Minnesota, Maine, Vermont, and New Hampshire.
The average annual per capita estimate for all of the eastern United States for 1997 is about
$57, and rises to about $194 in 2010. The average per capita benefit estimate for 1997 and
2010 in Canada is about $50, which is very similar to the 1997 average per capita estimate
for the eastern United States.
November 10, 1995
-------
RESULTS AND CONCLUSIONS * 6-6
Table 6-3
Mean Estimated Health Effects Benefits of Title IV by State
State
Maine, Vermont, New Hampshire
Massachusetts, Connecticut, Rhode Island
New York
Pennsylvania
New Jersey
Maryland, Delaware, D.C.
Virginia
West Virginia
North Carolina, South Carolina
Georgia
Florida
Ohio
Michigan
Illinois
Indiana
Wisconsin
Kentucky
Tennessee
Alabama
Mississippi
Minnesota
Iowa
Missouri
Arkansas
Louisiana
31-State U.S. Regional Total
Ontario
Quebec
Canadian Total
Annual Monetary Value of Health Benefits (1994
dollars)
1997
Total
(millions)
$129
$956
$1,160
$943
$341
$418
$394
$245
$412
$765
$35
$1,058
$325
$340
$512
$71
$777
$881
$312
$107
($72)
$1
$242
$123
$87
$10,562
$673
$235
$908
Average
per Capita
$41
$86
$60
$74
$41
$64
$59
$127
$38
$109
$2
$90
$32
$28
$85
$13
$195
$167
$72
$38
($15)
$0
$44
$49
$19
$57
$62
$32
$50
2010
Total
(millions)
$128
$580
$1,658
$2,633
$1,112
$1,614
$2,535
$950
$4,818
$3,508
$2,849
$3,344
$1,168
$1,713
$1,515
$334
$2,049
$2,741
$1,974
$654
$88
$176
$721
$285
$852
$39,999
$789
$166
$955
Average
per Capita
$37
$47
$76
$183
$119
$221
$339
$439
$394
$448
$182
$255
$104
$124
$226
$56
$460
$465
$404
$210
$17
$52
$117
$100
$167
$194
$68
$21
$49
November 10, 1995
-------
RESULTS AND CONCLUSIONS * 6-7
6.2 AGGREGATE HEALTH BENEFITS 1997 TO 2010
The reduction in SO2 emissions due to Title IV is expected to increase each year after 1997
until full implementation is reached in 20 iO. The first year for which ICF Resources (1994)
reports a specific estimate for an SO2 emissions reduction due to Title IV is 1997, when the
Phase I requirements are expected to be fully implemented. The health benefits will therefore
be expected to occur each year during this period, and increase each year until full
implementation of Title IV is reached. The estimates of emissions reductions expected are
based on a comparison of emissions expected with and without Title IV. ICF Resources
reported Title IV emissions reductions estimates for 1997, 2000, 2005, and 2010.
After 2010, Title IV may continue to result in lower SO2 emissions than would have occurred
without Title IV, but projections of what emissions would have been without Title IV have
not been made by EPA beyond 2010. The predicted trend in emissions for this "no Title IV"
scenario up to 2010 is fairly flat, with a very slight increase in emissions from 2000 to 2010.
Some analysts have predicted that after 2010, SO2 emissions might have begun to decline
even without Title IV requirements, because as old facilities are replaced, the new ones are
subject to more stringent new source performance standards and other permitting
.requirements. However, it remains highly uncertain as to how quickly, if at all, the Title IV
emissions limits would have been reached if Title IV had not been enacted.
The RADM was run for this assessment to obtain estimates of ambient outdoor sulfate aerosol
concentrations in the eastern United States for 1997 and 2010 with and without Title IV
scenarios. To estimate the total health benefits over the 1997 to 2010 period, annual estimates
for each year in the period are needed. We estimated the annual health benefits for 1998 to
2009 using the 1997 and 2010 health benefits estimates described in the previous section and
the emissions reductions estimates for 1997, 2000, 2005, and 2010 from ICF Resources
(1994). We assume that health benefits occur in proportion to the emissions reductions to
obtain the health benefits estimates for 2000 and 2005. For example, in the year 2000, the
predicted increase in emissions reductions over the 1997 level is about 46.6 percent of the
additional reduction expected by 2010 over the 1997 level (see Table 6-4: (6.38 - 4.07)/(9.03
- 4.07) = 0.466). We therefore estimate that 46.6 percent of the difference between health
benefits in 1997 and 2010 will be achieved in 2000 (i.e., $10.5628 + 0.466 (S39.999B -
$10.5626) = $24.2803). The same procedure was used to estimate the health benefits in the
year 2005. We then linearly interpolate between 1997 and 2000, 2000 and 2005, and 2005
and 2010 to obtain estimates of annual health benefits for each intervening year.
The resulting annual estimates of health benefits are reported in Table 6-4. The second
column shows the estimated emissions reductions for 1997, 2000, 2005, and 2010 provided by
ICF Resources. The annual health benefits estimates in the third and fourth columns for 1997
and 2010 are based on the RADM estimates of changes in sulfate concentrations under each
scenario and the health effects and monetary valuation procedures as described in Chapters 4
and 5 of this report. These are mean estimates for the eastern United States and Canada based
on the default assumptions. The last row shows aggregated health benefits from 1997 to 2010,
November 10, 1995
-------
RESULTS AND CONCLUSIONS *• 6-8
Table 6-4
Mean Annual Health Benefits Estimates
1997 to 2010
Year
1997
1998
1999
2000
2001 ,
2002
2003
2004
2005
2006
2007
2008
2009
2010
Total
Undiscounted
Estimated Reduction
in Annual SO2
Emissions Due to
Title IV1
(million tons)
4.07
6.38
7.89
9.03
Eastern United
States Estimated
Annual Monetary
Health Benefit
(millions of 1994
dollars)
$10,562
$15,135
$19,707
$24,280
$26,070
$27,859
$29,649
$31,439
$33,229
$34,583
$35,937
$37,291
$38,645
$39,999
$404^84
Ontario and Quebec
Estimated Annual
Monetary Health
Benefit
(millions of 1994
dollars)
$906
$915
$922
$930
$933
$935
$938
$941
$944
$946
$948
$951
$953
$955
$13,120
1 Based on emissions estimates from ICF Resources (1994).
November 10, 1995
-------
RESULTS AND CONCLUSIONS * 6-9
in 1994 U.S. dollars, undiscounted. Undiscounted, the aggregate health benefit for this period
is $404 billion for the United States and $13 billion for Canada.
Table 6-5 presents 1995 present values of total health benefit estimates from 1997 to 2010 for
the eastern United States and Canada using two alternative discount rates. Given uncertainty
about what the correct discount rate is for aggregating these kinds of benefits over time, we
select a 7 percent rate based on OMB recommendations for analyzing government programs
(OMB, 1992), and a possible lower rate of 3 percent based on evidence of a social rate of
discount.2 The discount rate has a significant effect on aggregate values over a time period as
long as this so it is useful to illustrate the results using alternative rates. We have made an
adjustment to each of the selected discount rates, because benefits are expected to grow over
time due to increases in real income that have not been accounted for in the annual estimates
presented previously. Applying discounting without making these adjustments would
inappropriately downward bias the present value estimates. Expected real income growth was
accounted for by deducting 0.94 percent from each discount rate based on expected annual
average growth in real income from 1997 to 2010 (U.S. Department of Commerce, 1990).
Real income growth is expected to increase health benefits by increasing willingness to pay
for prevention of health effects. We make a rough assumption here that WTP increases in
proportion to real income, although there is not sufficient empirical data to verify the
accuracy of this assumption at this time. WTP could in actuality increase in either greater or
lesser proportion to real income growth.
Table 6-5
Total Present Value in 1995 of Mean Health Benefits
1997 to 2010 with Default Assumptions
Net Discount Rate1
6.06%
2.06%
Eastern United States
(billions of 1994 dollars)
$234.7
$333.0
Ontario and Quebec
(billions of 1994 dollars)
$8.2
$11.1
1 The discount rates were derived as: 7.00% - 0.94% (average per capita income growth 1997-
2010) giving a net discount rate of 6.06% and 3.00% - 0.94% giving a net discount rate of
2.06% (U.S. Dept of Commerce, 1990).
2 Freeman (1993) presents a thorough discussion of discounting and evidence regarding appropriate
discount rates for environmental programs. Alternative arguments can be made to support alterative
discount rates. We select two rates from the range that is typically discussed.
: November 10, 1995 —
-------
RESULTS AND CONCLUSIONS * 6-10
6.3 SENSITIVITY ANALYSES RESULTS
There are many sources of uncertainty and potential error in the mean estimates of health
benefits for Title IV reported in the previous two sections. This section presents results of
some specific sensitivity analyses conducted to determine the potential effect on the results of
different assumptions than those selected for the default estimates. These alternative
assumptions reflect some of the key uncertainties identified hi the health effects quantification
and valuation chapters. The analyses reported in this section cover only the uncertainties in
the concentration-response functions and hi the monetary valuation of health effects.
Additional uncertainties also exist in the estimates of change in SO2 emissions and ambient
sulfate concentrations that are used as inputs to the health benefits estimates. These
uncertainties were discussed qualitatively in Chapter 2, but are not treated quantitatively here
because these inputs are based on analyses that have been reported elsewhere.
The uncertainty and sensitivity analyses reported here are those that are reasonably amenable
to quantitative treatment. It is important to recognize that there are many sources of
uncertainty that are not possible to quantify, and that these sensitivity tests are therefore not a
comprehensive treatment of all possible sources of uncertainty. What these tests provide,
however, is an indication of how the results might change if we found that some of the key
default assumptions in the health effects quantification and valuation procedures were
inappropriate.
The selected sensitivity tests are based on different assumptions that we think have some
nonzero probability of being accurate. A completely comprehensive range of possible results
given all the uncertainties in this assessment would include zero health benefits at the low end
and a very large number at the high end. That kind of comprehensive range is probably not
very helpful for policy makers without some guidance in understanding the likelihood that
different results within the range could be correct. We try to give this interpretation, at least
qualitatively, for each of the sensitivity test results.
Each of the sensitivity tests illustrated in Table 6-6 is discussed below. They all represent
estimates of mean annual health benefits for 1997 and 2010, in 1994 U.S. dollars. Each is
calculated in the same way that the default mean was calculated, except for the specified
assumption change. A comparison with the default means in Tables 6-1 and 6-2 therefore
illustrates the effect of the change in the assumption.
Health Effects Thresholds
As discussed hi Chapter 4, there is considerable uncertainty about whether there is a "safe"
level of sulfate aerosol exposure that does not cause any harmful health effects. There is no
definitive quantitative evidence that such a threshold exists, but neither is there proof that any
amount of sulfate aerosol exposure causes some harmful effect hi at least some people. We
selected three possible threshold levels to illustrate how this could affect the results. The
November 10, 1995
-------
RESULTS AND CONCLUSIONS »• 6-11
Table 6-6
Sensitivity Analyses Results
Assumptions
Estimated Annual Health
Benefits (billions of 1994 dollars)
United States 1997
Threshold « 5.0 ug/m3 S04
Threshold = 3.6 ug/m3 SO4
Threshold = 1.6 ug/m3 SO4
Selected S04 Health Risks x 0.4
$3.1
$6.7
$10.8
$4.8
United States 2010
Threshold - 5.0 ug/m3 SO4
Threshold » 3.6 ug/m3 S04
Threshold = 1.6 ug/m3 SO4
Selected SO4 Health Risks x 0.4
$15.0
$28.3
$39.3
$18.5
Canada 1997
Threshold = 5.0 ug/m3 S04
Threshold = 3.6 ug/m3 SO4
Threshold = 1.6 ug/m3 SO4
Selected S04 Health Risks x 0.4
$0.0
$0.0
$0.7
$0.4
Canada 2010
Threshold = 5.0 ug/m3 SO4
Threshold = 3.6 ug/m3 S04
Threshold = 1.6 ug/m3 SO4
Selected SO4 Health Risks x 0.4
$0.0
$0.0
$0.9
$0.5
existence of a threshold could only decrease, not increase, the results because it means that
further reductions in sulfate levels in areas that are already at or close to the threshold would
not yield any health benefits.
We selected alternative threshold assumptions of 5.0 ug/m3, 3.6 ug/m3 and 1.6 ug/m3 annual
median SO4 concentrations to illustrate the potential effects of alternative threshold
assumptions on the results of this analysis. As discussed in Chapter 4, none of these
concentrations has been identified as a true threshold, but each represents a mean or low end
November 10, 1995
-------
RESULTS AND CONCLUSIONS * 6-12
value for the range of concentrations considered in one of the epidemiology studies that
concentration-response functions were taken from. The threshold calculation was implemented
as follows. Any RADM grid cell with a base case 2010 (without Title IV) level of annual
50th percentile SO4 at the threshold concentration or less was assigned zero health benefits for
the Title IV emissions reductions.3 Further, health benefits were calculated only for
reductions in annual 50th percentile down to the threshold. For the 5.0 (Ag/m3 threshold, for
example, if the level without Title IV was 5.5 and the level with Title IV was 4.5, the health
benefits calculations for that grid cell were made only for the 5.5 minus 5.0 reduction of 0.5.
Any additional reduction below 5.0 was presumed to provide no health benefit.
The results indicate that with a threshold of 5.0 ng/m3 SO4, annual health benefits are
substantially reduced relative to the default mean, falling very close to the 20 percentile
estimates. At thresholds above 5.0 the health benefit estimates would diminish even more. A
threshold of 3.6 ug/m3 SO4 results in a health benefit estimate that falls about midway
between the default mean and the 20 percentile default estimates. At a threshold of 1.6 (or
lower), the health benefit estimate is virtually unchanged from the default mean. This
illustrates the significance of the threshold question and shows that this continues to be an
important research issue from the standpoint of evaluating the health benefits of pollution
emission reductions.
Lower Health Risks for Sulfates
As discussed in Chapter 4, there is a possibility that the sulfate based concentration-response
functions may be somewhat upwardly biased because of the typical collinearity between
sulfates and other fine paniculate constituents hi the ambient air. For this sensitivity test we
multiply the sulfate based concentration-response functions by 0.4, which is the average ratio
between measured sulfates and measured PM2 5 in the eastern United States. This is the
maximum adjustment that would be required if the sulfate coefficients represented the total
effects of all PM2 5. This adjustment reduces the annual health benefit estimate to about $8.5
billion, in 2010, which is close to the 20th percentile estimate with the default assumptions.
The true sulfate effect is probably between this and the mean default estimate because the
sulfate coefficients probably do reflect some, but are unlikely to reflect all, of the effects of
other harmful constituents of PM2 5 as well as the effects of sulfates alone.
3 RADM estimates that in a few locations sulfate concentrations will be higher with Title IV than
without Title IV. In general, these places have very low sulfate concentrations and may fall below the
threshold concentration under consideration. Because a cell below the threshold concentration is assigned
zero health benefits, no negative health benefits are calculated in the threshold analyses. This is why the
1997 results/using the 1.6 threshold assumption slightly exceed the mean .results when no threshold is
assumed.
• November 10, 1995
-------
RESULTS AND CONCLUSIONS > 6-13
6.4 CONCLUSIONS
The results of this assessment show that the potential health benefits of reductions in
exposures to sulfate aerosols in the eastern United States as a result of the SO2 emissions
reductions required by Title IV are substantial. Based on what we believe is a reasonable
interpretation of the available epidemiology and economic evidence on potential health effects
of sulfate aerosols and their monetary value, we estimate that the annual health benefits in the
United States of the Title IV required reductions in SO2 in 2010 are more likely than not to
fall between $12 billion and $78 billion. There is reason to expect some possible upward bias
at the higher end of this range. The results of the sensitivity analyses suggest that there is a
good chance that the annual benefits in the United States fall between $12 billion and $40
billion.
We have been careful throughout the report to highlight key assumptions and uncertainties
that exist in the quantification procedures used in this assessment, especially in the health
effects quantification and valuation portions of the assessment which are the focus of this
report. Most of these uncertainties cannot be resolved without substantial new research on
several topics. The most important empirical uncertainties hi the health effects quantification
are:
*• What is the relative harmfulness of sulfate aerosols versus other fine particulate
matter?
* Is there a threshold for health effects from sulfate aerosols, and if so, what is it?
»• Is there sufficient evidence to presume that the observed association between sulfate
concentrations and human health effects is causative?
The most important uncertainties in the monetary valuation of health effects are:
* Are WTP estimates for risks of accidental deaths in populations of average health
status applicable to premature mortality risks associated with air pollutant exposures?
* How do WTP values for premature mortality and other health risks vary for the elderly
and for those whose health is already poor?
> Will the available WTP estimates for chronic respiratory disease be verified by new
WTP research?
November 10, 1995
-------
CHAPTER 7
REFERENCES
Abbey, D.E., M.D. Lebowitz, P.K. Mills, F.F. Petersen, W.L. Beeson and R.J. Burchette.
1995. "Long-term Ambient Concentrations of Particulates and Oxidants and Development of
Chronic Disease in a Cohort of Nonsmoking California Residents." Inhalation Toxicology 7:
19-34.
Abbey, D.E., F.F. Petersen, P.K. Mills and L. Kittle. 1993a. "Chronic Respiratory Disease
Associated with Long Term Ambient Concentrations of Sulfates and Other Air Pollutants."
Journal of Exposure Analysis and Environmental Epidemiology 3(Suppl. 1): 99-115.
Abbey, D.E., F. Petersen, P.K. Mills, and W.L. Beeson. 1993b. "Long-Term Ambient
Concentrations of Total Suspended Particulates, Ozone and Sulfur Dioxide and Respiratory
Symptoms in a Non-Smoking Population." Archives of Environmental Health 48(l):33-46.
Amdur, M.O., J.-Doull, and C.D. Klaassen, eds. 1991. Casarett and Doull's Toxicology: The
Basic Science of Poisons. 4th ed. New York: Pergamon Press, Inc. pp. 857-863.
American Lung Association. 1978. Health Effects of Air Pollution. New York, NY.
American Thoracic Society. 1991. "Report of the ATS Workshop on the Health Effects of
Atmospheric Acids and Their Precursors." American Review of Respiratory Disease 144:464-
467.
Bates, D.V., and R. Sizto. 1989. "The Ontario Air Pollution Study: Identification of the
Causative Agent." Environmental Health Perspectives 79:69-72.
Berger, M.C., G.C. Blomquist, D. Kenkel, and G.S. Tolley. 1987. "Valuing Changes in Health
Risks: A Comparison of Alternative Measures." Southern Economic Journal 53(4):967-984.
Bos, E. et al. 1992. World Population Projections 1992-1993 Edition: Estimates and
Projection -with Related Demographic Statistics. Johns Hopkins University Press: Baltimore,
Maryland. November.
Burnett, R.T., R.E. Dales, D. Krewski, R. Vincent, T. Dann, and J. Brooke. 1995.
"Associations between Ambient Paniculate Sulfate and Admissions to Ontario Hospitals for
Cardiac and Respiratory Diseases." American Journal of Epidemiology 142(l):15-22. May.
November 10, 1995
-------
REFERENCES * 7-2
Burnett, R.T., R.E. Dales, M.E. Raizenne, D. Krewski, P.W. Summers, G.R. Roberts, M.
Raad-Young, T. Dann, and J. Brooke. 1994. "Effects of Low Ambient Levels of Ozone and
Sulfates on the Frequency of Respiratory Admissions to Ontario Hospitals." Environmental
Research 65:172-194. May.
Chappie, M., and L. Lave. 1982. "The Health Effects of Air Pollution: A Reanalysis." Journal
of Urban Economics 12:346-76.
Chang, J.S., P.B. Middleton, W.R. Stockwell, C.J. Walcek, I.E. Pleim, H.H. Lansford, and
F.S. Binkowski. 1990. The Regional Acid Deposition Model and Engineering Model. NAPAP
SOS/T Report 4, National Acid Precipitation Assessment Program, Washington, D.C.
Chen, L.C., S.M. Peoples, and M.O. Amdur. 1991. "Pulmonary Effects of Sulfur Oxides on
the Surface of Copper Oxide Aerosol." American Industrial Hygiene Association Journal
52(5):187-191.
Chestnut, L.G., and D.M. Violette. 1984. Estimates of Willingness to Pay for Pollution-
Induced Changes in Morbidity: A Critique for Benefit-Cost Analysis of Pollution Regulation.
EPA-230-07-85-009. Report prepared for the U.S. Environmental Protection Agency,
Washington D.C.
Chestnut, L.G., S.D. Colome, L.R. Keller, W.E. Lambert, B. Ostro, R.D. Rowe, and S.L.
2Wojciechowski. 1988. Heart Disease Patients' Averting Behavior, Costs of Illness, and
Willingness to Pay to Avoid Angina Episodes. Final Report to the Office of Policy Analysis,
U.S. EPA, Washington D.C. October.
Cropper, M.L. and A.M. Freeman III. 1991. "Environmental Health Effects." Measuring the
Demand for Environmental Quality. J.B. Braden and C.D. Kolstad (ed.) North-Holland. New
York.
Dennis, R.L., W.R. Barchet, T.L. Clark, and S.K. Seilkop. 1990. Evaluation of Regional Acid
Deposition Models (Part I). NAPAP SOS/T Report 5. In: Acidic Deposition: State of Science
and Technology. National Acid Precipitation Assessment Program. September.
Dennis, R.L., J.N. McHenry, W.R. Barchet, F.S. Binkowski, and D.W. Byun. 1993.
"Correcting RADM's Sulfate Underprediction: Discovery and Correction of Model Errors and
Testing the Corrections through Comparisons Against Field Data." Atmospheric Environment
37A(6):975-997.
Dockery, D.W. and C.A. Pope III. 1994. "Acute Respiratory Effects of Paniculate Air
Pollution." Annual Review of Public Health 15:107-132.
November 10, 1995
-------
REFERENCES *• 7-3
Dockery, D.W., C.A. Pope III, X. Xu, J.D. Spengler, J.H. Ware^ M.E. Fay, B.G. Ferris, Jr.,
and F.E. Speizer. 1993. "An Association Between Air Pollution and Mortality in Six U.S.
Cities." The New England Journal of Medicine 329(24): 1753-9.
Dockery, D.W., J. Schwartz, and J.D. Spengler. 1992. "Air Pollution and Daily Mortality:
Associations with Particulates and Acid Aerosols." Environmental Research 59:362-373.
Euler, G.L., D.E. Abbey, J.E. Hodgkin, and A.R. Magie. 1987. "Chronic Obstructive
Pulmonary Disease Symptom Effects of Long-Term Cumulative Exposure to Ambient Levels
of Particulates and Sulfur Dioxide in California Seventh-Day Adventist Residents." Archives
of Environmental Health 42: 213-22.
Evans, J.S., T. Tosteson, and P.L. Kinney! 1984. "Cross-Sectional Mortality Studies and Air
Pollution Risk Assessment." Environment International 10:55-83.
Ferris, B.G., Jr. 1978. "Health Effects of Exposure to Low Levels of Regulated Air Pollutants:
A Critical Review." Journal of the Air Pollution Control Association 28(May):482-497.
Ferris, B.G., Jr., H. Chen, S. Puleo, and R.L.H. Murphy, Jr. 1976. "Chronic Non-specific
Respiratory Disease in Berlin, New Hampshire, 1967-1973. A Further Follow-up Study."
American Review of Respiratory Disease 113:475-485.
Ferris, B.G., Jr., I.T.T. Higgins, M.W. Higgins, and J.M. Peters. 1973. "Chronic Non-specific
Respiratory Disease in Berlin, New Hampshire, 1961-67. A Follow-up Study." American
Review of Respiratory Disease 107:110-122.
/
Fisher, A., L.G. Chestnut, and D.M. Violette. 1989. "The Value of Reducing Risks of Death:
A Note On New Evidence." Journal of Policy Analysis and Management 8(1):88-100.
Folinsbee, L.J. 1992. "Human Health Effects of Air Pollution. Environmental Health
Perspectives 100 (April): 45-56.
Freeman, A.M. III. 1993. The Measurement of Environmental and Resource Values: Theory
and Practice. Washington D.C.: Resources for the Future.
Fujimaki, H., N. Katayama, and K. Wakamori. 1992. "Enhanced Histamine Release from
Lung Mast Cells of Guinea Pigs Exposed to Sulfuric Acid Aerosols." Environmental Research
58:117-123.
Gearhart, J.M. and R.B. Schlesinger. 1988. "Response of the Tracheobronchial Mucociliary
Clearance System to Repeated Irritant Exposure: Effect of Sulfuric Acid Mist on Function and
Structure." Exp. Lung Res. 14:587-605.
November 10, 1995
-------
REFERENCES * 7-4
Gearhart, J.M, and R.B. Schlesinger. 1989. "Sulfuric Acid-Induced Changes in the Physiology
and Structure of the Tracheobronchial Airways ." Environ. Health Perspect. 79:127-137.
Gong, H., Jr. 1992. "Health Effects of Air Pollution: A Reviews of Clinical Studies." Clinical
Chest Medicine 13 (June): 201-14.
Graham, J.A. 1990. Direct Health Effects of Air Pollutants Associated with Acidic Precursor
Emissions. State of Science and Technology Report 22. Prepared by the National Acid
Precipitation Assessment Program, Washington, D.C.
Grant, L.D. 1990. Indirect Health Effects Associated with Acidic Deposition. State of Science
and Technology Report 23. Prepared by the National Acid Precipitation Assessment Program,
Washington, D.C.
Graves, E.J. 1994. Detailed Diagnoses and Procedures National Hospital Discharge Survey,
1992. National Center for Health Statistics. Vital and Health Statistics 13(118). Hyattsville,
Maryland.
Hall, J.V., A.M. Winer, M. Kleinman, F.W. Lurmann, V. Brajer, S.D. Colome, R.D. Rowe,
L.G. Chestnut, D. Foliart, L. Coyner, and A.F. Horwatt. 1989. Economic Assessment of the
Health Benefits from Improvement in Air Quality in the South Coast Air Basin. Prepared for
South Coast Air Management District by California State University Fullerton Foundation,
Fullerton, California. Contract No. 5685. June.
Harrison, D. and A.L. Nichols. 1990. Benefits of the 1989 Air Quality Management Plan for
the South Coast Air Basin: A Reassessment. Prepared for California Council for
Environmental and Economic Balance by National Economic Research Associates, Inc.,
Cambridge, Massachusetts.
Hodgkin, J.E., D.E. Abbey, G.L. Euler, and A.R. Magie. 1984. "COPD Prevalence in
Nonsmokers in High and Low Photochemical Air Pollution Areas." Chest 86:830-838.
ICF Resources, Inc. 1994. Economic Analysis of the Title IV Requirements of the 1990 Clean
Air Act Amendments. Prepared for the U.S. Environmental Protection Agency, Office of Air
and Radiation, Acid Rain Division, Washington, D.C. February.
Jones-Lee, M.W., M. Hammerton, and P.R. Philips. 1985. "The Value of Safety: Results of a
National Sample Survey." The Economic Journal 95(March):49-72.
Koenig, J.Q., K. Dumler, V. Rebolledo, P.V. Williams, and W.E. Pierson. 1993. "Respiratory
Effects of Inhaled Sulfuric Acid on Senior Asthmatics and Nonasthmatics." Archives of
Environmental Health 48(3):171-175.
November 10, 1995
-------
REFERENCES »> 7-5
Krupnick, A.J. 1986. A Preliminary Benefits Analysis of The Control of Photo
chemical Oxidants. Report prepared for the U.S. Environmental Protection Agency,
Washington D.C. September.
Krupnick, A.J. and M.L. Cropper. 1989. Valuing Chronic Morbidity Damages: Medical Costs,
Labor Market Effects, and Individual Valuations. Final Report to U.S. EPA, Office of Policy
Analysis.
Krupnick, A.J. and M.L. Cropper. 1992. "The Effect of Information on Health Risk
Valuations." Journal of Risk and Uncertainty 5:29-48.
Krupnick, A.J. and R. Kopp. 1988. The Health and Agricultural Benefits from Reductions in
Ambient Ozone in the United States. Resources for the Future, Washington, D.C. Discussion
Paper QE88-10.
Krupnick, A.J., W. Harrington, and B. Ostro. 1990. "Ambient Ozone and Acute Health
Effects: Evidence from Daily Data." Journal of Environmental Economics and Management
Lave, L.B. and E.P. Seskin. 1977. Air Pollution and Human Health. Johns Hopkins
University Press for Resources for the Future. Baltimore and London.
Lee, S.D. and J.B. Mudd, eds. 1979. Assessing Toxic Effects of Environmental Pollutants. 2nd
Printing. Ann Arbor, Michigan: Ann Arbor Science Publishers, Inc. pp. 173-186.
Lipfert, F.W. 1994. Air Pollution and Community Health. New York: Van Nostrand Reinhold.
Loehman, E.T., S.V. Berg, A.A. Arroyo, R.A. Hedinger, J.M. Schwartz, M.E. Shaw, R.W.
Fahien, V.H. De, R.P. Fishe, D.E. Rio, W.F. Rossley, and A.E.S. Green. 1979. "Distributional
Analysis of Regional Benefits and Cost of Air Quality Control." Journal of Environmental
Economics and Management 6:222-243.
Miller, T.R. 1989. "Willingness to Pay Comes of Age: Will the System Survive?"
Northwestern University Law Review 83:876-907.
Moore, MJ. and W.K. Viscusi. 1988. "The Quantity-Adjusted Value of Life." Economic
Inquiry 26:369-388.
National Acid Precipitation Assessment Program (NAPAP). 1991. 1990 Integrated Assessment
Report. Office of the Director, Washington, D.C.
National Center for Health Statistics. 1992. Current Estimates from the Health Interview
Survey, 1991. Series 10, No. 184, DHHS Publication No. (PHS) 93-1512, Hyattsville, MD.
November 10, 1995
-------
REFERENCES + 7-6
Office of Management and Budget. 1992. Revised Circular No. A-94, Revised Technical
Memorandum No. 64: Guidelines and Discount Rates for Benefit-Cost Analysis of federal
Programs. Transmittal Memorandum No. 64 from Richard Barman, Director, Executive Office
of the President, Office of Management & Budget, Washington D.C. October 29.
Ostro, B.D. 1983. "The effects of air pollution on work loss and morbidity." Journal of
Environmental Economics and Management 10:371-382.
Ostro, B.D. 1987. "Air Pollution and Morbidity Revisited: A Specification Test." Journal of
Environmental Economics and Management 14:87-98.
Ostro, B.D. 1990. "Associations Between Morbidity and Alternative Measures of Particulate
Matter." Risk Analysis 10:421-427.
Ostro, B.D. and S. Rothschild. 1989. "Air Pollution and Acute Respiratory Morbidity: An
Observational Study of Multiple Pollutants." Environmental Research 50:238-247.
Ostro, B.D., M.J. Lipsett, J.K. Mann, A. Krupnick and W. Harrington. 1993. "Air Pollution
and Respiratory Morbidity among Adults in Southern California." American Journal of
Epidemiology 137(7): 691-700.
Ostro, B.D., M.J. Lipsett, M.B. Wiener, and J.C. Seiner. 1991. "Asthmatic Responses to
Airborne Acid Aerosols." American Journal of Public Health 81:694-702.
Ostro, B.D., J.M. Sanchez, C. Aranda, and G.S. Eskeland. In press. "Air Pollution and
Mortality: Results from a Study of Santiago, Chile." Journal of Exposure Analysis and
Environmental Epidemiology.
Ozkaynak, H. and G. Thurston. 1987. "Associations Between 1980 U.S. Mortality Rates and
Alternative Measures of Airborne Particle Concentration." Risk Analysis 7:449-62.
Palisade Corporation. 1994. Risk Analysis and Simulation Add-In for Microsoft Excel or
Lotus 1-2-3. Release 3.0. Newfield, New York.
Plagiannakos, T. and J. Parker. 1988. An Assessment of Air Pollution Effects on Human
Health in Ontario. Ontario Hydro, March.
Pope, C.A., III, M.J. Thun, M.M. Namboodiri, D.W. Dockery, J.S. Evans, F.E. Speizer, and
C.W. Heath, Jr. 1995. "Particulate Air Pollution as a Predictor of Mortality in a Prospective
Study of U.S. Adults." American Journal of Respiratory and Critical Care Medicine
151:669-674.
Portney, P.R., and J. Mullahy. 1990. "Urban Air Quality and Chronic Respiratory Disease."
Regional Science and Urban Economics 20:407-418.
November 10, 1995
-------
REFERENCES *• 7-7
Rowe, R.D. and L.G. Chestnut. 1986. Oxidants and Asthmatics in Los Angeles: A Benefits
Analysis-Executive Summary. Prepared by Energy and Resource Consultants, Inc. Report to
the U.S. EPA office of Policy Analysis. EPA-230-09-86-018. Washington D.C. March.
Rowe, R.D. and T.N. Neithercut. 1987. Economic Assessment of the Impacts of Cataracts.
Prepared for U.S. Environmental Protection Agency, Office of Policy, Planning, and
Evaluation. Washington, D.C.
Rowe, R.D., L.G. Chestnut, and W.D. Shaw. 1984. "Oxidants and Asthmatics in Los Angeles:
A Benefits Analysis." Evaluation of the Ozone/Oxidants Standards. Si Duk Lee (ed.) Air
Pollution Control Association. Pittsburgh.
Rowe R.D., C.M. Lang, L.G. Chestnut, D.A. Latimer, D.A. Rae, S.M. Bernow, and D.E.
White. In press. The New York Electricity Externality Study. Oceana Publications, Inc. Dobbs
Ferry, NY.
Schlesinger, R.B., M. Halpern, R.E. Albert, and M. Lippmann. 1979. "Effect of Chronic
Inhalation of Sulfuric Acid Mist Upon Mucociliary Clearance from the Lungs of Donkeys." J.
Environ. Pathol. Toxicol. 2:1351-1367.
Schlesinger, R.B., J.T. Zelikoff, L.C.Chen, and P.L. Kinney. 1992. "Assessment of
Toxicologic Interactions Resulting from Acute Inhalation Exposure to Sulfuric Acid and
Ozone Mixtures." Toxicology and Applied Pharmacology 115:183-190.
Schwartz, J., and D.W. Dockery. 1992a. "Increased Mortality in Philadelphia Associated with
Daily Air Pollution Concentrations." American Review of Respiratory Disease 145:600-604.
Schwartz, J., and D.W. Dockery. 1992b. "Particulate Air Pollution and Daily Mortality in
Steubenville, Ohio." American Journal of Epidemiology 135(1): 12-19.
Schwartz, J., and R. Morris. 1995. "Air Pollution and Cardiovascular Hospital Admissions."
American Journal of Epidemiology 142: in press.
Shepard, D.S. and R.J. Zeckhauser. 1982. "Life-Cycle Consumption and Willingness to Pay
for Increased Survival." The Value of Life and Safety. M.W. Jones-Lee (ed.) North-Holland.
New York.
Snell, B., L. Robinson, and B. Unsworth. 1993. "Comparison of Morbidity, Visibility, and
Forest Valuation Studies to Contingent Valuation Guidelines." Memorandum prepared by
Industrial Economics, Inc. for the U.S. EPA, Office of Air and Radiation, Washington, D.C.
September 30.
Statistics Canada. 1994. 1994 Canada Year Book. Ottaway, Canada.
November 10, 1995
-------
REFERENCES > 7-8
Thayer, M. 1991. Valuing the Environmental Impacts of Alternative Energy Resources: Phase
III Task I Report. Prepared by Regional Economic Research for the California Energy
Commission, Sacramento, California. July 15.
Thurston, G.D., K. Ito, C.G. Hayes, D.V. Bates, and M. Lippmann. 1994. "Respiratory
Hospital Admissions and Summertime Haze Air Pollution in Toronto, Ontario. Envrionmental
Research 65:271-290.
Thurston, G.D., P. Kinney, K. Ito, and M. Lippmann. 1992. "Daily Respiratory Hospital
Admissions and Summer Haze Air Pollution in Several New York Metropolitan Areas."
American Review of Respiratory Disease 145(4:2):A429.
Tolley, G.S., L. Babcock, M. Berger, A. Bilotti, G. Blomquist, R. Fabian, G. Fishelson, C.
Kahn, A. Kelly, D. Kenkel, R. Kumm, T. Miller, R. Ohsfeldt, S. Rosen, W. Webb, W.
Wilson, and M. Zelder. 1986. Valuation of Reductions in Human Health Symptoms and Risks.
Prepared at the University of Chicago. Final Report for the U.S. EPA, Grant
#CR-811053-01-0. January.
Unsworth, R.E. and J.E. Neumann. 1993. "Review of Existing Value of Morbidity Avoidance
Estimates: Draft Valuation Document." Memorandum prepared for the U.S. EPA, Office of
Air and Radiation, Washington, D.C. September 30.
U.S. Bureau of the Census. 1990. Census of Population and Housing, 1990: Summary Tape
File 3 on CD-ROM Technical Documentation. Washington, DC.
U.S. Bureau of the Census. 1994. Statistical Abstract of the United States: 1994 (114th
edition). Washington, D.C.
U.S. Department of Labor. 1995. Bureau of Labor Statistics. Washington, D.C.
U.S. Environmental Protection Agency. 1982. Air Quality Criteria for Paniculate Matter and
Sulfur Oxides. Volumes I-IV. Prepared by the Environmental Criteria and Assessment Office,
Research Triangle Park, NC. EPA-600/8-82-029.
U.S. Environmental Protection Agency. 1984. Regulatory Impact Analysis on the National
Ambient Air Quality Standards for Paniculate Matter. Prepared by the Strategies and Air
Standards Division, Office of Air, Noise and Radiation, Research Triangle Park, NC. February
21.
U.S. Environmental Protection Agency. 1986a. Second Addendum to Air Quality Criteria for
Paniculate Matter and Sulfur Oxides (1982): Assessment of Newly Available Health Effects
Information. EPA/600/8-86-020F. Prepared by the Environmental Criteria and Assessment
Office, Research Triangle Park, NC.
November 10, 1995
-------
REFERENCES + 7-9
U.S. Environmental Protection Agency. I986b. Review of the National Ambient Air Quality
Standards for Particulate Matter: Updated Assessment of Scientific and Technical
Information: Addendum to the 1982 OAQPS Staff Paper. EPA/450/05-86-012. Prepared by the
Office of Air Quality Planning and Standards, Research Triangle Park, NC.
U.S. Environmental Protection Agency. 1986c. Regulatory Impact Analysis on the National
Ambient Air Quality Standards for Particulate Matter: Second Addendum. Prepared by the
Strategies and Air Standards Division, Research Triangle Park, NC. December.
U.S. Environmental Protection Agency. 1988. Regulatory Impact Analysis on the National
Ambient Air Quality Standards for Sulfur Oxides (Sulfur Dioxide). Prepared by the Office of
Air and Radiation, Research Triangle Park, NC. March.
U.S. Environmental Protection Agency. 1989. An Acid Aerosols Issue Paper: Health Effects
and Aerometrics. EPA/600/8-88/005F. Prepared by the Office of Health and Environmental
Assessment, Washington, D.C.
U.S. Environmental Protection Agency. 1994. Supplement to the Second Addendum (1986) to
Air Quality Criteria for Particulate Matter and Sulfur Oxides (1982): Assessment of New
Findings on Sulfur Dioxide Acute Exposure Health Effects in Asthmatic Individuals. Research
Triangle Park, NC. Environmental Criteria and Assessment Office. EPA Report No. EPA-
600/AP-93/002. August.
U.S. Environmental Protection Agency. 1995. Air Quality Criteria for Particulate Matter.
Volumes I-III. External Review Draft. Prepared by the Office of Research and Development,
Washington, D.C. April. EPA/600/AP-95/001.
Utell, M. J., and J. M. Samet. 1993, "Particulate Air Pollution and Health: New Evidence on
an Old Problem." American Review of Respiratory Disease 147:1334-35.
Violette, D.M., and L.G. Chestnut. 1983. Valuing Reductions in Risks: A Review of the
Empirical Estimates. EPA-230-05-83-002. Report prepared for the U.S. Environmental
Protection Agency, Washington, D.C.
Viscusi, W.K. 1992. Fatal Tradeoffs: Public and Private Responsibilities for Risk. Oxford
University Press. New York.
Viscusi, W.K., W.A. Magat, and J. Huber. 1991. "Pricing Environmental Health Risks:
Survey Assessments of Risk-Risk and Risk-Dollar Trade-offs for Chronic Bronchitis." Journal
of Environmental Economics and Management 21(1 ):32-51.
Zelikoff, J.T. and R.B. Schlesinger. 1992. "Modulation of Pulmonary Immune Defense
Mechanisms by Sulfuric Acid: Effects on Macrophage-Derived Tumor Necrosis Factor and
Superoxide." Toxicology 76:271-281.
November 10, 1995
------- |