Annual Reviews www.annualreviews.org/aronline Annu.Rev. EnergyEnviron. 1993. 18:319-42 Copyright©1993by AnnualReviewsInc. All rights reserved

ECONOMIC VALUATION OF AIR POLLUTION ABATEMENT: Benefits from Health Effects Luis A. Cifuentes Department of Engineering and Public, Policy, Carnegie Mellon University School of Engineering, Catholic University of Chile

&

Lester B. Lave Graduate School of Industrial Administration Public Policy, Carnegie Mellon University

& Department of Engineering and

KEYWORDS:energy economics, energy policy, mortality,

morbidity, externality

adders

CONTENTS INTRODUCTION .......................................... METHODS TO ESTIMATE BENEFITS ........................... ESTIMATING THE BENEFITS OF ABATEMENTUSING THE DIRECT METHOD ........................................ Local or National Estimates? ................................. Estimating the Health Effects of Air Pollution ...................... Estimation of the Marginal Benefits of Pollution Abatementfor the United States SUMMARY .............................................

319 320 322 322 323 335 338

INTRODUCTION The environmental effects of energy production and use have been studied for more than two decades. Most studies have focussed on identifying and quantifying the effects, without trying to ascribe a dollar value to them (1-8). A few studies have assigned monetary values to the effects (9, 10). Interest in the area has increased rapidly in the 1990s, due in part to concern 319 1056-3466/93/1022-0319502.00

Annual Reviews www.annualreviews.org/aronline 320

CIFUENTES& LAVE

about global climate change and using market mechanisms to improve environmentalquality (11). Economistsfinally convincedthe USCongressand regulators that effluent fees are efficient waysto manageexternalities from environmentaldischarges (12-17). In order to implementeffluent fees, economists have been called uponto estimate the dollar value of these externalities (18). Wereview the state of the art of estimation of benefit from air pollution abatement, and propose national average estimates for the benefits of abatementfor several air pollutants. Economists were unprepared for the rapidity with which public service commissionsfrom about 30 states turned to the use of "externality adders"---estimates of the social loss from having an additional ton of each pollutant released into the environment. Since no consensus existed, Public Utility Commissions(PUCs) have adopted bizarre estimates that bear resemblance to the marginal benefits of abatement that they are intended to approximate. For example, Massachusetts, NewYork, and California set disparate externality adders, none of which is closely tied to estimates of the economicbenefits of abatement (see for exampleRef. 19). METHODS TO ESTIMATE

BENEFITS

Figure 1 showsthe flow from emissions to ambient concentrations to physical effects and finally to social valuation of the effects. This frameworkassumes that valuation is done by humans. If humansvalue wilderness and species diversity, they have social value; animals, plants, and buildings are valued only through humanperceptions and values. The damagesfrom air pollution are taken to be the same as the benefits from abatement. Environmental discharges can offend consumers because of odors, obscured visibility, and other directly perceived consequences.The discharges can also damagehealth, materials, and agricultural and ornamental plants. The most important damages considered in past studies have been direct effects on the health of the populationboth morbidity (illness and disability) and mortality (death). Amongmorbidity effects are increased incidence asthmaattacks, cough, eye irritation, etc. Economists have used both direct and indirect methods to estimate the value of changes in environmentalquality. The direct or "damagefunction" methodof estimating marginal benefits requires working through each of these steps: (a) estimating the change in environmental quality resulting from changes in emissions, (b) estimating the effects (damages) from degradation in environmental quality, (c) estimating whatever change behavior occurs because of the degraded environment, and (d) valuing the

Annual Reviews www.annualreviews.org/aronline AIR POLLUTION ABATEMENTBENEFITS 321

Annual Reviews www.annualreviews.org/aronline 322

CIFUENTES & LAVE

damages. Each of these steps is difficult, requiring extensive data and analysis. Uncertainties complicate the interpretation and use of each step and the resulting estimates (20, 21). It is importantto give an explicit characterization of uncertainty, both for the resulting social costs and each individual step. Manyanalysts tend to overlook behavior change that mitigates damage. For example, home-ownerswill not continue to plant ornamental plants that are highly sensitive to local air pollution, and farmers will not try to grow crops that are continually damagedby pollution. Also, by separating the effects and pathwaysof each pollutant, interaction effects betweenpollutants can be lost. Several "indirect methods"have been developedto finesse the difficulties and uncertainties of the direct method. These indirect methods focus on market valuations or observed behavior, assumingthat consumersare aware of environmentalquality and that their preferences are reflected in observed market prices and behavior. Althoughthe simplicity of the indirect methods is appealing, we question their basic assumptions, as well as the interpretations of their results. Thus, we focus on the direct methodin the rest of this analysis. For reviews of the indirect methods, see (10, 22-24). ESTIMATING THE DIRECT

THE BENEFITS METHOD

OF ABATEMENT

USING

Weconsider only the health effects on humansof sulfur oxides (SOx), nitrogen dioxide (NO2), particles, and ozone. These health effects are the most important. Local or National

Estimates?

In evaluating the marginal benefits of pollution abatement, we have two options. Wecould analyze the local effects of air pollution, and then attempt to generalize to other sites. Unfortunately,this generalization is not straightforward: it would require analyzing several representative sites and then averaging the results. It is not evident that the resulting estimates wouldbe meaningful. Alternatively, we could estimate "average" values for a state or region. Deriving "average" values for a region results in more uncertain estimates, because the details of topography, meteorology, and population distribution are ignored. However,these estimates are intended to represent a broader range of situations. Our objective is to obtain average data values that can be used nationally as a reference. Weestimate the relationship between emissions and ambient concentrations and the size of the population exposed to the resulting change in ambient concentrations on a national basis.

Annual Reviews www.annualreviews.org/aronline AIR POLLUTIONABATEMENT BENEFITS Estimating

the Health Effects

323

of Air Pollution

The first step is to establish that air pollution causes morbidityand mortality. The second step is to estimate the quantitative magnitude. Studies on the health effects of air pollution can be classified as clinical or epidemiological. EPIDEMIOLOGIC STUDIES The standard methods of epidemiology consist of cohort- and case-controlled studies. A population-based approach can be used wherestatistical controls are substituted for controls in data collection in order to utilize standard demographicand public health data, e.g. Lave & Seskin (25). The cost of cohort studies has limited sample size, and so the studies have focussed on investigating causality (26). One study design uses each subject as his or her owncontrol by contrasting the incidence of symptoms on polluted days with those on days of good air quality (27). Most of the estimates of the concentration-response relationship come from population-based studies [Lave & Seskin (25), Schwartz & Marcus (28), Portney & Mullahy (29), etc]. Because these studies utilize collected for a different purpose, they are subject to criticism. For example, there is no measure of cigarette smokingin the studies of Lave &Seskin, Schwartz & Marcus, and Ozkaynak & Thurston (30). Thus, these studies are criticized for not having measuresof important variables that are known to affect the prevalence of the health conditions being studied. In some studies, e.g. Lave & Seskin, population data on smoking rates are used. In time-series studies, the assumption is that the smokingrate does not change over time. However,if people tend to smoke more when the air is polluted, the estimated effect of air pollution will be overstated. The reverse is true if people smokeless whenthe air is polluted. A general problemfor almost all the studies is that the measurementfor air pollution is that of outdoorair at a few points in the city or metropolitan area. Since people spend less than 10%of their time outdoors (not in vehicles), this air pollution measureneed not be closely associated with the air they actually breathe. In cross-sectional studies (whichcontrast air quality and health measuresacross cities), the implicit assumption is that the air quality measurementsin each area are about equally goodin characterizing the quality of the air people breathe. That seems a heroic assumption. For time-series studies, the assumption is that the air pollution measures are equally good at characterizing the quality of air breathed during each day of the week. But this seems suspicious, especially whencontrasting weekdays and weekends. Onweekdays~more people breathe the central city air; whereason weekends,people spend their time in different activities, such

Annual Reviews www.annualreviews.org/aronline 324

CIFUENTES & LAVE

as sports, gardening, hiking, etc, and they may be outdoors a greater proportion of the time. The cohort, population, clinical, and time-series studies all have their unique advantages and problems. With care, each can be used to measure different effects. Cross-sectional analyses can modelboth chronic and acute effects (31); their maindrawbackis the possibility of missing variables such as smokingor diet. If a causal variable that is correlated with the air-pollution variables is left out of the analysis, the estimatedeffect of air pollution will be biased. Time-series analysis has fewer problems with missing variables. Its problems stem more from confounding effects of differences in weekday and weekend moods and activities. People may feel less tense on the weekends,as work, commuting,and manyother activities are curtailed and air quality is better. Asthmaattacks, heart attacks, and other morbidity are associated with tension and activities that decrease during the weekend,as do air pollution levels. This can obscurethe relationship betweenair pollution and health. CLINICAL STUDIES Clinical studies are conducted on a sample of volunteers under well measured and controlled conditions. These can assess the magnitude of the effects with precision. However,their key advantage is also their maindisadvantage whenapplied to real life, because of three points: 1. Artificial conditions: In order to produceresults that can be replicated, the dose in a clinical study is carefully controlled. Ambientair quality in real settings is a complexmixture that maybear little relationship to the air quality used in the laboratory setting. There are two additional reasons whythe clinical setting mayhave little to do with real life: (a) Clinical studies are extremely expensive to conduct and so sample size is necessarily small, usually about half a dozen subjects. These subjects cannot represent the diverse mix of age, sex, ethnicity, and prior exposuresthat characterize the general population. (b) The subjects are volunteers, almost alwayshealthy ones. Humansubject committees are reluctant to approve an experiment wherethere is a nontrivial chancethat the subject could be harmedor killed by the experiment. Almost inevitably, this means that sensitive persons cannot be subjects in the clinical study. 2. Averting behavior: People in clinical studies are subjected to very special conditions, and they follow a protocol (for example,exercise for sometime, rest, exercise again), despite the appearanceof symptoms.In real life, those people likely.would avoid, or at least mitigate, the effect of air pollution: they might rest or stop activity or even movetheir residence to a less polluted place. This averting behavior is not captured in clinical studies, although it maybe captured in epidemi~logical studies.

Annual Reviews www.annualreviews.org/aronline AIR POLLUTIONABATEMENT BENEFITS 325 3. Valuation of effects: Epidemiologicalstudies measureeffects that people are able to perceive; thus people could conceivably put a value on such effects. Clinical studics measurephysiological changesthat people sometimes do not perceive; thus they could not put a value on them. Basedon the previous discussion of epidemiological and clinical studies, we rely on epidemiological studies to obtain the quantitative estimates. Clinical studies are primarily importantin probing the existence and causality of an effect. MORTALITY EFFECTS OFAIRPOLLUTION The severe episodes of air pollution that occurred in Donora, Pennsylvania, in 1948, and in London, England, in 1952, left no doubt about the lethal potential of air pollution. Several epidemiological studies have reported statistically significant relationships betweenexcess mortality and several measuresof particulate matter. To our knowledge,only one study (32) has reported mortality effects for a pollutant other than particulate matter or sulfur oxides (ozone). However,there not yet enough evidence to forge an unequivocal link between ozone and excess mortality. Therefore, we consider only particulate matter and sulfur oxides in our analysis of mortality effects of air pollution. The usual terminologyfor air pollution’s mortality effects is a shorthand that is misleading to almost everyone. Rather than "deaths due to air pollution," we should be careful to discuss "premature deaths due to air pollution." A first metric to measurethe mortality effects of air pollution is the reduction in "premature" deaths as air quality improves. A second metric supplementsthe first in noting the numberof "additional life years" due to air-quality improvements.A third metric supplementsthe first two by measuring the numberof additional "quality-adjusted life years" gained by improving air quality (33, 34). Since everyone dies eventually, interest centers on premature deaths. The problem with this measure is that the death maybe premature by only days or even minutes. Thus, it is helpful to knowthe number of life years lost due to premature death. However, not all life years are the same. Granting someonewith endstage emphysema an additional year of life maynot be worth as muchas giving this person an additional year of active, healthy life. These metrics becomeimportant because of the nature of the epidemiological studies done to quantify air pollution’s effects. Time-series studies relate deaths to the level of air pollution on a daily basis. Such analyses seek to explore: Howmanypremature deaths would occur if there were an increase of one microgram per cubic meter of fine particles? Howmuch longer wouldthese individuals have lived if the level of fine particles had not increased that day? If their life expectancy were a few days, the air pollution is "harvesting" deaths that likely wouldhave occurred within days,

Annual Reviews www.annualreviews.org/aronline 326

CIFUENTES & LAVE

absent the air pollution. The additional premature deaths uncoveredby this day-to-day analysis maynot be important to public policy. Twomechanismsmight be at work. In the first, people on the edge of dying have their death advanced by a few days. In the second, the entire population is injured, manifestated primarily by an increase in mortality that day. If an air-pollution episode decreases everyone’s life expectancy by a few days, it is muchmoreimportant than if it decreased the life expectancy only of those dying during the episode. One way to test which mechanismis at work is to examine the pattern of deaths following the episode. If, for example, the next week shows a dip in the mortality rate equal to the numberof excess deaths during the episode, one might conclude that air-pollution episodes lead to harvesting. If, instead, the mortality rate remains slightly abovenormalor declines only to the expected level, the air-pollution episode wouldappear to be causing large reductions in life expectancy. The point is that time-series studies cannot be content with estimating the immediateresponse to air pollution, since these immediate deaths might be due predominately to harvesting. Cross-sectional studies contrast the mortality rate across urban areas, attempting to find how many premature deaths are associated with a difference in air quality. These studies hypothesize that differences in air quality have persisted for manyyears. Thus, these analyses appear to be answering the question: If someone was born, grew up, and worked in an area with slightly worse air quality, what would be the decrease in life expectancy?For these studies, a prematuredeath is hypothesizedto represent manymore life years lost than does a premature death in a time-series analysis. Studies using time-series data are numerous. Several have studied the "Londonfogs" (28, 35-38), while others have studied NewYork (39-42), Steubenville (43), Philadelphia (44), and Los Angeles(32, 45). Cross-sectional studies are also numerous, with manyof them following the work by Lave &Seskin (25, 30, 46-48). Manyof these studies found a significant relationship betweenmeasuresof particulate matter (usually Total Suspended Particles--TSP) and sulfate with excess mortality. Based on our previous discussion, we rely on cross-sectional studies to estimate the magnitudeof the effects. Weuse two cross-sectional studies to quantify the impact of air pollution on mortality: Lave &Seskin (25) and Ozkaynak& Thurston (30). Both studies found a significant association betweenmortality and both particulate matter and sulfate concentrations. The data used in these studies span a period of 20 years, from 1960 to 1980. Someanalysts have concluded that the Lave & Seskin study mayhave overestimatedthe effects of these pollutants on mortality by a factor of two (47). Since the Ozkaynak&Thurston study generally found

Annual Reviews www.annualreviews.org/aronline AIR POLLUTIONABATEMENT BENEFITS 327 larger effects, our averageof the results of these twostudies can be considered to be a reasonableupperboundof the mortality effects of these pollutants. The USCongress directed the Environmental Protection Agency(EPA) to set primaryambientair-quality standards that protect the health of the population with an amplemarginof safety. For morethan 20 years, EPAhas been carrying out this directive, relying on the most sensitive clinical and epidemiological studies. For air quality that surpasses the primary air-quality standards, one should expect that there will be no health effects. However,no one can show that the standard constitutes a threshold. Thestudies on whichwe relied were doneat levels of air quality substantially worsethan those currently prevailing. A re-examination of the Lave & Seskin work showsthat the estimated health effects of air pollution werenot significant whenair quality wasin compliance with the primaryair-quality standards. Thus, in our judgment,the health effects of a small changein air quality, whenair quality exceedsthe primary,are likely to be very small or even zero. In particular, we hypothesize that the concentration-responsecoefficient for air quality belowthe primaryair-quality standard is one-tenth of that for muchworsepollution. Therefore, we use the original mortality coefficients dividedby 10. Particulate matter has been measuredhistorically in several ways: British Smoke(BS), Total Suspended Particles (TSP), by the Coefficient of (COH),Fine Particles (FP, particles less than 2.5 microns in diameter), morerecently as PM10(particles less than 10 microns in diameter). In order to arrive at a meaningfulestimate, we need to convert the different measures into a commonone, PM10. To convert from TSP to PM10 , we have used the EPAestimate (49) that PM10is between 0.5 and 0.6 of TSP, with meanof 0.55. For fine particles (FP), we have used the estimate obtained by Trijonis (50) using nationwide data, that gives FP = 0.30 TSP, which results in PM~o= FP/0.55 using the previous relationship. VALUATION OFMORTALITY EFFECTS If the analysis is to estimate the dollar benefits of pollution abatement, the decrease in premature deaths, in life years or quality-adjusted life years lost, mustbe translated into dollars. How much is society willing to pay to reduce the risk of a person dying prematurely from air pollution? This question has been misstated by many investigators to the confusion of themselves and the general public. The usual misstatement to is ask the "dollar value of a life." The value of a life is a moral and social question that is not related to the chancethat a small proportion of people will have their life expectancy diminished by exposureto air pollution. The question is not howmuchwouldwe pay to save BabyJane. There is no wayof knowingwhowill die prematurelydue to air pollution. Indeed, there is no wayto identify after the fact whodied due to air pollution. This topic is

Annual Reviews www.annualreviews.org/aronline 328

CIFUENTES & LAVE

inherently controversial, but the misstatement has vastly increased the confusion (52-57). Wemention the main aspects of the problem here; more in-depth treatment can be found elsewhere(34, 58, 59). Economists use two alternative approaches to find a dollar value for a premature death. The first approach is based on measurementsof economic productivity of the individual at risk, and is usually referred to as the "humancapital" approach. It is based on two premises: that the value of individuals to society is whatthey produce,and that productivity is accurately measuredby earnings. Besides being ethically questionable, this approach is not consistent with the fundamental premise of welfare economicsthat each individual’s ownpreferences should be used to establish the economic values used in cost-benefit analysis (60). The second approach is based on an individual’s willingness to pay to reduce his risk of death; it is called the "willingness to pay" approach. It assumesthat individuals treat longevity as any other good, and that it is possible to estimate the value they place on life expectancy by looking at the trade-offs they madebetweenreductions in the risk of death and other goods whose value can be measured in monetary terms. For example, some people choose to drive small, light cars that are far more dangerous in a crash than large, heavy cars. Somepeople accept dangerous jobs (such as painting the steeples of churches or commercial fishing in Alaska), with knownhigher accidental death rates, because they receive a higher wage. Whenit is possible to estimate the increased risk and compensation from each choice, then the individual’s willingness to pay, or the amountof compensationhe or she requires, is revealed by these choices. This method is theoretically correct, in contrast to measuresof earnings. Weuse it in our estimates. In a recent review of several willingness to pay studies, Fisher et al (61) present a range of values from $1.6 million to $8.5 million per life. They place more confidence in the lower bound than in the upper. Moore & Viscusi (57) present a best estimate of $5 million using recent data from the National Institute for OccupationalSafety. Since the analysis indicates that the prematuredeaths tend to be concentrated amongthe elderly, society maynot be willing to pay as muchto prevent a premature death as it would for young workers. Thus, we regard the $4 million dollar figure in this context as a plausible upper bound, and $1 million as a plausible lower bound. Tables 1 and 2 present the meanvalue for the mortality coefficient for each study, for particulate matter and sulfate, respectively. The low and high columns correspond to the meanestimate of the low and high scenarios for valuation THE SOCIAL COSTS OF PREMATUREMORTALITYDUE TO AIR POLLUTION

Annual Reviews www.annualreviews.org/aronline AIR POLLUTIONABATEMENTBENEFITS

+++

329

Annual Reviews www.annualreviews.org/aronline 330

CIFUENTES & LAVE

Table 2 Marginalbenefits of reducing SO4concentrations

Reference Lave&Seskin, 1977(25) Lave&Seskin, 1977(25) Ozkaynak& Thurston, 1987 (30) Average

Yearof data 1960 1969 1980

~ Marginaleffects Mortality Std err 0.299 0.245 0.660

0.110 0.100 0.150

0.401

0.257

Valuationof life Low High (S/life saved) IE + 6 1E + 6 1E + 6

4E + 6 4E + 6 4E + 6

bMarginalbenefit Low High 299 245 660

1,196 980 2,640

401

1,605

3 of changeof SOnambient aMarginal effectsper 100,000people,per/xg/m concentration bThousands of dollarper ,ttg/m3 of SO4,per 100,000 people. of the effects. The average mortality coefficient was calculated as the average of the three coefficients. The standard error of this average coefficient includes the standard error of each coefficient and the standard error that results from averaging the three coefficients. This method of averaging gives the same weight to each study, which we believe is appropriate in this case. A similar procedure was followed in the calculation of the average coefficients in other cases. Although EPAsets ambient air-quality standards for sulfur dioxide, both clinical studies and epidemiological studies indicate that small particles, including sulfates, have a greater effect on health at current concentrations (25, 30, 38, 62).

MORBIDITY EFFECTSOF AIR POLLUTIONMany researchers have investigated the association of air pollution and respiratory morbidity. Clinical studies have provided a causal relationship for most of the symptoms. We study the effects of each pollutant separately, and try to assess the extent of the effects of each pollutant on humans as completely as possible. We base our assessment on a review of epidemiological studies, for the reasons discussed before. From each study; it is possible to obtain a concentrationresponse relationship. Whenthese relationships are not linear, we linearize them about the average ambient concentration of the pollutant in the latest years. This assumes that there are no thresholds at current levels of air pollution. After we have calculated the marginal increments of each effect, we check for double counting of the effects. As we discuss below, some measures of morbidity may include others, so we need to subtract them when appropriate. Once we have obtained the net marginal effects, we value them using several contingent valuation measures. We finally aggregate them to obtain the total marginal benefit of abatement of each ambient pollutant.

Annual Reviews www.annualreviews.org/aronline AIR POLLUTIONABATEMENT BENEFITS 331 MORBIDITY ENDPOINTS Several endpoints have been used when studying morbidity effects. Clinical studies tend to use very precise measures of pulmonaryfunctions, such as Forced Vital Capacity (FVC)or Peak Expiratory Flow Rate (PEFR), and some clearly identifiable symptoms, such eye or throat irritation, or shortness of breath. Epidemiologicalstudies have looked at individual symptomstoo, but they have also looked at the overall effects that such symptoms(or a combination of them) have on people. Somemeasures have been developed to represent those effects. Restricted Activity Days (RAD)is the most comprehensive measure, and includes days spent in bed, days missed from work, and other days whennormal activities are restricted due to illness. WorkLoss Days (WLD)includes days missed from work or school. Bed Disability Days (BDD)is the most serious the three morbidity measures, and includes days spent in bed only. Someof these measuresare included in others, so in determining the net marginal effects of a given pollutant, we have to subtract the included measures when appropriate, to obtain the net effect. The presence of a symptom,if severe enough, maylead to a RAD.Weassume that the cases of sinusitis and asthma attacks produce a RAD,so wheneverthese measures are considered together we subtract the asthmaand sinusitis cases from the RADscount. Weassume that less severe symptoms, such as eye or throat irritation, do not lead to a RAD. PARTICULATE MATTER EFFECTS Several studies of air-pollution episodes have confirmed an association between particulate matter and morbidity effects (63-65). They have measured mainly pulmonary function, so they are not appropriate to use for valuation purposes. Recent epidemiological studies have also tried to find a relationship between morbidity measures and particulate matter (27, 66-72). Evaluating these studies requires some care. The associations between the morbidity measures and air-quality measuresare weak. Evenif current air-quality levels caused this morbidity, we would expect the observed associations to be weak: The air monitors are at sites somemiles from most of the subjects; they measureoutdoor air quality, not the air in buildings or transport vehicles where people spend the vast majority of their time. In addition, the morbidity measures, such as "restricted activity days," are subjective, and no two individuals are likely to classify a set of symptomsin precisely the same way. Lackof statistical significance for an association, despite the large sample sizes, meansthat association is extremely weakor there is no effect. The weaknessof the association could be due to the irrelevance of the air-pollution measureor differences in interpretation of what constitutes a "minor restricted activity day." In three papers, Ostro and coworkers (68-70) found associations between

Annual Reviews www.annualreviews.org/aronline 332

CIFUENTES & LAVE

fine particles (FP) and Work Loss Days (WLD), Restricted Activity (RADs), Respiratory-related RADs(RRADs), and Minor Restricted Activity Days (MRADs).However, the results were generally obtained in only one or two years, out of six years of data. Combining the estimates for the six years (1976-1981) reveals that only the results for RADs(69) and RRADs (70) are significant. Whittemore & Korn (27) found a significant association of particulate matter and asthma attacks, but the magnitude of this effect was negligible, so we do not consider it. Portney & Mullahy (71) investigated the relationship between TSP and chronic respiratory disease, asthma, and emphysema, but the results were not significant. Krupnick et al (72) found a relationship between any symptom and coefficient of haze (a measure of particulate matter). However, the morbidity measure is quite general; we can conclude only that morbidity is associated with haze in this study. After considering these studies, we conclude that the only morbidity measure we can use is RADs. The coefficient for RADsis included in Table 1. VALUATION OF THE PHYSICALDAMAGES In contrast to studies to estimate the value of averted deaths, not many studies have estimated how much people are willing to pay for a reduction in morbidity symptoms. The most widely cited study is Loehman et al (73). Krupnick & Kopp (74) offer review of several other studies. Table 3 shows the values we have used. All the values from the original studies were updated to 1991 dollars using the general consumer price index. From Ostro & Rothschild (70), we infer that MRADsare approximately two-thirds of the RADs count. Therefore, the valuation used for RADsis one-third of a restricted activity day plus two-thirds of a minor restricted activity day. Since there is no measure for a minor restricted activity day, we used the shortness of breath symptom from Loehman. For the restricted activity day, we use the average lost wage for one day (75). Although it has been shown that the willingness to pay function is not linear in the number of days for which the symptoms are

Table 3 Value of symptoms,in 1991 dollars Symptom

Valuation(19915) Low High

Cough Eyeirritation Sinusitiscase Asthmaattack RAD

4.33 3.27 27.30 11.79 19.45

13.02 13.10 93.30 53.69 47.97

Reference (73) (74) any symptom (74) (74) (73) plus our own calculations

Annual Reviews www.annualreviews.org/aronline AIR POLLUTIONABATEMENT BENEFITS 333 avoided (73), we have used the values for a reduction of one day in each symptom,since we are estimating marginal benefits. OZONE EFFECTSOzone is one of the main components of photochemical air pollution of metropolitan areas. It is formedby the oxidation of nitrogen oxides in the presence of sunlight and reactive volatile organic compounds (VOCs). There is no doubt that, at high concentrations, ozone produces acute effects on the respiratory tract that result in a variety of symptoms. Clinical studies have found a significant relationship betweenozoneexposure and several acute respiratory conditions, including cough, shortness of breath, nose and throat irritation, and chest discomfort (76-79). Cohort epidemiological studies have found significant impacts of ozone exposure on pulmonaryfunctions of children, even at levels encountered in normal ambient concentrations (80, 81). Time-series epidemiological studies have also found a relationship between acute effects and ozone exposure (27, 29, 72, 82, 83). Only one study that we knowof, however, has found relationship betweenozoneexposure and chronic respiratory conditions (71). For a complete review of the health effects of ozone, see (84). Webase our analysis on the following endpoints: 1. RADestimates are based on Portney & Mullahy (29). This study based on symptomsidentified by adults nationwide over a two-week period as part of the National Health Interview Survey (NHIS)during 1979. 2. Estimates of asthma attacks are based on Whittemore& Kom(27), based on daily asthma-attack diaries collected by the EPAin Los Angeles, from 1972 to 1975. The frequency of the attacks increased with oxidant levels and particulate matter levels. 3. Eye irritation and coughdays estimates are based on Schwartzet al (83). This is a re-analysis of the Los AngelesStudent Nurse data from Hammer et al (82), using a new modelthat considers the time-series nature the data. It replicated the results concerningozone. 4. Sinusitis incidence is taken from Portney &Mullahy (71). Using data for the 1979 NHIS,they found a significant association between the incidence of sinusitis and the average concentration of ozone during the past five years, both for the wholepopulationand for a cohort of residents who had lived for more than five years in the same location. A combination of emphysema,chronic bronchitis, and asthma was found not to be significantly associated with the concentration of ozone. Table 4 showsthe marginal benefits of abatementof ozoneconcentrations. To avoid double counting in the benefits, we have subtracted the sinusitis

Annual Reviews www.annualreviews.org/aronline

Annual Reviews www.annualreviews.org/aronline AIR POLLUTIONABATEMENT BENEFITS 335 and asthma cases from the RRADs measure. For asthma attacks, we assume that the asthmatic population is 3%of the total population (74). NITROGEN DIOXIDE EFFECTS The literature on direct health effects of oxides of nitrogen is less definitive. There is somesuggestion of health effects from epidemiological studies. Harrington (85) found a relationship between NO2exposure and acute respiratory disease in children, but the relationship was not monotonic. Schwartz et al (83) found a relationship between eye irritation and NO2exposure in the same data with which Hammeret al (82) failed previously. However,there is little indication of effects from laboratory studies (86-88). Therefore, we disregard the direct effects NO2on morbidity.

Estimation of the Marginal Benefits of Pollution Abatement for the United States After having estimated the marginal impacts of changes in average ambient concentrations of air pollutants, we need to link those concentrations with pollutant emissions, the point at whichregulation will be enforced. Since we are constructing national estimates, we have analyzed the relation between emissions and ambient concentrations by focussing on national emissions and average national air quality. In particular, we estimated the relationship between the ambient concentration of each pollutant and anthropogenic emissions of its precursors via linear regression analysis. The dependent variable was EPA-estimatednational ambient concentration. The independent variable was EPA-estimated total national emissions of the appropriate pollutants. The data are annual for several years, dependingon the availability of EPAestimates (89, 90). Our regressions results are shown in Table 5. The slope of the regression line is shownin column3. Since particulate matter has been measured historically as TSP, we had to convert TSPconcentrations to PM~0concentrations using the relation presented before. EPAdoes not provide sulfate concentrations, so we derived SO4levels from the TSP concentrations as a constant fraction of those concentrations; this estimate was regressed on SO2emissions. Weregressed the remaining fraction of PM10concentrations (PM10 less the SO4fraction) on TSP emissions. NO2concentrations were regressed directly on NO2 emissions. Weregressed ozone concentrations on VOCsemissions and NO2 ambient concentrations simultaneously, relating them later to NO2emissions through the coefficient of NO2concentrations to emissions. Since the EPA data for ozone concentrations is the average of the annual second highest daily maximum one-hour concentration at each monitor, and the epidemiological studies use the average of the daily maximum one-hour concentra-

Annual Reviews www.annualreviews.org/aronline 336

CIFUENTES

& LAVE

Table5 Regressioncoefficients for ambientconcentrationsas functionof national emissions Pollutant concentration 3) PM~ominusSO4(p,g/m 3) SO,,(p.g/m NO2(ppb) Ozone(avg. daily max.) (ppb) Ozone(avg. daily max.) (ppb)

Pollutant emissions TSP (M-ton) SO2 (M-ton) NO2 (M-ton) VOCs (M-ton) NO2 (M-ton)

Slopea

Elasticity t statistic

R2

Years of data

2.54

0.84

11.3

0.94

1976-1989

0.37

1.05

13.4

0.95

1975-1989

0.51

0.42

2.5

0.92

1979-1989

0.58

0.29

0.97

0.79

1976-1989

0.72

0.38

2.67

0.79

1976-1989

a Slopes are expressed in units of pollutant ambient concentration per million-ton of emissions of the appropriate pollutant (for example, the units of the slope in the second line are /zg/m3 of SO4over M-ton of SO2emissions).

tions during the period of study, we scaled the coefficients according to an estimated ratio between the two measures of 3.13. Unfortunately, ambient concentrations of the various criteria pollutants are highly correlated. Similarly, emissions of the relevant pollutants are highly correlated. Thus, there is little ability to separate the effects of emissions of each pollutant on the ambient concentrations of related pollutants, e.g. volatile organic compounds and nitrogen dioxide versus ozone. This difficulty, however, is small compared to problems in the quality of emissions and ambient air-quality data. Emissions of each pollutant are estimated by EPAon the basis of few data, especially for the earlier years. Ambient air quality are averages from reporting monitoring stations. However, there were few stations in the early years, and stations were not sited with a view to getting a national average. Thus, the EPAestimates for both dependent and independent variables must be considered to have a large random component. Regression results are uncertain, since the EPA data on which they are based have major uncertainties. POPULATION EXPOSED The final stage to compute the marginal benefits of emissions abatement is to determine the total population that would be exposed to a change in ambient concentrations. For particu!ate matter and sulfate effects, we have assumed that the population affected is the total population of the United States living in metropolitan areas, around 193 million (91). For ozone morbidity effects, we assume that the population affected is the adult population living in metropolitan areas, around 144 million.

Annual Reviews www.annualreviews.org/aronline AIR POLLUTION ABATEMENTBENEFITS

337

TableIf Marginalbenefits of reducingpollutant emissions Pollutant emissions TSP TSP TSP SO2 VOC NO2

Effect

Marginalabenefit Target Emission cMarginalbenefit Low High population coeff,b Low High (K-$/Conc/100,000p) (Millionpeople) (Conc/M-ton) (199IS/Ton)

PM~omortality PM~omorbidity Total SO,~mortality Ozonemorbidity Ozonemorbidity

118 94

472 231

193 144

2.541 2.541

401 92 92

1605 280 280

193 144 144

0.37 0.58 0.72

579 343 922 287 77 96

2315 845 3161 1146 234 290

~ Marginal benefit of abatement of ambient concentrations of the corresponding pollutant b See Table 5 ¢ Marginal benefit of emissions abatement, in dollars per ton. Lowand High correspond to the meanof the low and high valuation scenarios.

MARGINAL BENEFITS OF EMISSIONS ABATEMENT Table 6 shows the calculations of the marginal benefits of emissions abatement. It should be noted that these values include only the mortality and morbidity effects discussed previously. Howcertain are these estimates? Figure 2 shows the cumulative distribution function (cdf) of the particulate matter and sulfate marginal benefits. The cdf was derived using the Demos modeling software (92), considering the uncertainty of each of the steps. As it is apparent from the figures, the estimates have substantial uncertainty associated. What can be done to CumulativeProbability 1.00

0.75

0.50

0.25

-1000

J~

0

1000

TSP

2000

3000

4000

Marginal Benefits (S/Ton) Figure2 Marginalbenefit of emissionsabatementfor particulate matterandsulfur dioxide, for the highvaluationscenario.Mortalityeffects only.

Annual Reviews www.annualreviews.org/aronline 338

CIFUENTES& LAVE

reduce this uncertainty? Muchof it comesfrom the uncertainty concerning the health effects of ambient concentrations. Continuing research in this area should provide somewhatbetter estimates of the impact of air pollution on humanhealth. The link between emissions and concentrations, and the population exposed, do not add muchuncertainty in our analysis, although this step is one that has muchvariability across different geographicareas. For example, Rowe(93) presents exposure coefficients for the United States that vary from 10 to 80 person-lxg/m3 of sulfate per ton of SO2emitted. The 95%confidence interval of our exposure coefficient goes from 61 to 82 person-p~g/m3 per ton of SO2. SUMMARY The USEnvironmental Protection Agency, state and local environmental agencies, and public utility commissionsseek estimates of the physical damageand dollar valuation of damagecaused by air pollution. Both to estimate externality adders and improveregulatory decision-making, reliable estimates are needed. Wehave estimated the marginal benefits of air-pollution abatement, due to health effects, of the most importantpollutants: suspendedparticles, SO2, NO2,and 03. For our estimation of abatement benefits we use the direct or damagefunction approach. Twomajor issues concern (a) whether there is a practical threshold for air-pollution effects and (b) interpreting the effects of time-series compared to cross-sectional studies. Proving the existence of a threshold, a level of air pollution belowwhichthere are no significant health effects, is essentially impossible. Even the studies with large sample sizes and relatively good air-pollution exposure data cannot either prove or disprove the existence of a threshold. For example, even small mischaracterizations of exposure or omissions of confounding variables will obscure a quantitatively small association betweenair pollutants and health. Although time-series studies appear to have fewer problems with confounding than do cross-sectional studies, they still have problems. In addition, existing studies of the association of daily air-pollution levels with the daily death rate do not distinguish between"harvesting" (the advancing by a few days or weeks a death that would have occurred anyway) and longer-term effects. Thus, the premature deaths that time-series studies estimate would be averted by improved air quality probably represent a muchsmaller numberof quality life years lost than the premature deaths estimated from cross-sectional studies. For morbidity effects, time-series studies are appropriate, given the acute nature of the effects measured. Given these considerations, we have reviewed the relevant literature,

Annual Reviews www.annualreviews.org/aronline AIR POLLUTION ABATEMENT BENEFITS 339 estimating the number of premature deaths and morbidity effects associated with decreased air quality. Since we regard these estimates as uncertain, we have presented an explicit analysis of the uncertainty of the resulting estimates. Since this uncertainty is great, we look to future health studies to provide more confident estimates. More than two decades of research has shown that air pollution levels in many US cities in the 1960s decreased life expectancy and had other health effects. As the methods have improved, air quality has improved even faster, except for ozone levels. For air-quality levels in US cities in the 1990s, the magnitude and importance of health-related air-pollution effects are uncertain. Reducing uncertainty requires (a) better measures of air-pollution dose (such as the use of personal monitors or having a biological measure of cumulative exposure), (b) standardized measures of morbidity or other health effects of concern, (c) control for other factors affecting the health measure or randomization, and (d) large sample size to isolate possible subtle effects. Carrying out a study with these characteristics would be exceedingly expensive. Thus, it is unlikely that the health effects of air pollution will be known with much more certainty in the future, unless different methods are used. Literature Cited 1. Sagan, L. A. 1972. Humancosts of nuclear power.Science 177:487-93 2. Lave, L. B., Freeburg, L. C. 1973. Healtheffects of electricity generation fromcoal, oil andnuclear fuel. Nucl. Safety 14:409-28 3. Sagan, L. A. 1974. Health costs associated with the mining,transport, and combustionof coal in the steamelectric industry. Nature250:107-11 4. Hamilton,L. D. 1974. The Health and EnvironmentalEffects of Electricity Generation--A Preliminary Report. Upton, NY:BrookhavenNatl. Lab. 5. Budnitz,R. J., Holdren,J. P. 1976. Social andenvironmentalcosts of energy systems. Annu. Rev. Energy 1:553-80 6. Comar,C. L., Sagan, L. A. 1976. Health effects of energy production and conversion. Annu. Rev. Energy 1:581-600 7. Holdren,J. P., Morris, G., Mintzer, I. 1980. Environmental aspects of renewableenergy sources. Annu. Rev. Energy5:241-91 8. Hamilton, L. D. 1984. Health and environmental risks of energysystems.

9.

10.

11. 12. 13. 14. 15.

16.

In Risks and Benefits of Different EnergySystems, pp. 21-57. Vienna: Int. At. EnergyAgency Lave, L. B., Silverman,L. P. 1976. Economiccosts of energy-related environmentalpollution. Annu.Rev. Energy 1:601-28 Fisher, A. C., Smith, V. K. 1982. Economic evaluationof energy’senvironmentalcosts with special reference to air pollution. Annu.Rev. Energy7: 1-35 Natl. Acad. Sci. 1991. Policy Implications of GreenhouseWarming. Washington,DC:Natl. Acad. Press Pigou, A. C. 1920. The Economicsof Welfare. London:Macmillan.1st ed. Kneese,A. V., Schultze, C. L. 1975. Pollution, Prices, and Public Policy. Washington,DC:BrookingsInst. Schultze, C. L. 1977. The Public Use of Private Interest. Washington,DC: TheBrookingsInst. Baumol,W. J., Oates, W. E. 1988. The Theoryof EnvironmentalPolicy. NewYork: CambridgeUniv. Press. 2nd ed, Stavins, R. N. 1988. Project 88:

Annual Reviews www.annualreviews.org/aronline 340

17. 18.

19. 20. 21.

22.

23.

24.

25. 26.

27.

28.

29. 30.

CIFUENTES

& LAVE

Harnesing Market Forces to Protect Our Environment--Initiatives for the NewPresident. A public policy study sponsored by Senator Timothy E. Wirth, Colorado, and Senator John Heinz, Pennsylvania. Washington, DC Hahn, R. W., Hird, J. A. 1991. The costs and benefits of regulation: Review and synthesis. Yale J. Regul. 8:233 ff. Dept. Publ. Util. 1992. Investigation by the Departmentof Public Utilities on its ownmotion as to the environmental externality values to be used in resource cost-effectiveness test by electric companies subject to the Department’s jurisdiction. Report D.P.U. 91-131. Commonw.Mass. Chernick, P., Caverhill, E. 1991. Methods of valuing environmental externalities. Electr. J. 4:46-53 Krupnick, A. J., Portney, P. R. 1991. Controlling urban air-pollution: Abenefit cost assesment. Science 252:52~27 Hall, J. V., Wirier, A. M., Kleinman, M. T., Lurmann, F. W., Brajer, V., Colome, S. D. 1992. Valuing the health benefits of clean air. Science 255:812-17 Pearce, D. W., Markandya, A. 1989. Environmental Policy Benefits: Monetary Valuation. Washington, DC: OECDPubl./Inf. Cent. Mitchell, R. C., Carson, R. T. 1989. Using Surveys to Value Public Goods: The Contingent Valuation Method. Washington, DC: Resourc. Future Braden, J. B., Kolstad, C. D. 1991. Measuring the Demandfor Environmental Quality. Amsterdam: North Holland Lave, L. B., Seskin, E. P. 1977. Air Pollution and HumanHealth. Baltimore, Md: Johns Hopkins Univ. Press Fen-is, B. G., Anderson, D. O. 1962. The prevalence of chronic respiratory disease in a NewHampshire town. Am. Rev. Respir. Dis. 86:165 ft. Whittemore, A. S., Korn, E. L. 1980. Asthmaand air pollution in the Los Angeles area. Am. J. Pub. Health 70:687-96 Schwartz, J., Marcus, A. 1990. Mortality and air pollution in London:A time series analysis. Am. J. Epidemiol. 131:185-94 Portney, P., Mullahy, J. 1986. Urban air quality and acute respiratory illness. J. Urb. Econ. 20:21-38 Ozkaynak, H., Thurston, G. D. 1987. Associations between 1980 U.S. mortality rates and alternative measuresof airborne particle concentration. Risk Anal. 7:449-61

31. Evans, J. S., Kinney, P. L., Koehler, J. L., Cooper, D. W. 1984. The relationship between cross-sectional and time series studies. J. Air Pollut. Control Assoc. 34:551-53 32. Kinney, P. L., Ozkaynak, H. 1991. Associations of daily mortality and air pollution in Los Angeles County. Environ. Res. 54:99-120 33. Zeckhauser, R., Shepard, D. S. 1976. Where now for saving lives? Law. Contemp. Probl. 39:545 34. Zeckhauser, R., Shepard, D. S. 1984. Principles for saving and valuing lives. In Technological Risk Assesment, ed. P. F. Ricci, L. A. Sagan, C. G. Whipple, pp. 133-68. The Hague: Matinus Nijhoff 35. Cummings,B. T., Walker, R. E. 1967. Observations from a ten-year study of pollution at a site in the City of London. Atmos. Environ. 1:49-68 36. Mazumdar, S., Schimmel, H., Higgins, I. T. T. 1982. Relation of daily mortality to air pollution: Ananalysis of 14 London winters 1958/59-1971/ 72. Arch. Environ. Health 37:213-20 37. Ostro, B. D. 1984. A search for a threshold in the relationship of air pollution to mortality: A reanalysis of data on London winters. Environ. Health Perspect. 58:397-99 38. Thurston, G. D., Ito, K., Lippmann, M., Hayes, C. 1989. Re-examination of London, England mortality in relation to exposure to acidic aerosols during 1963-1972 winters. Environ. Health Perspect. 79:73-82 39. Glasser, M., Greenburg, L, 1971. Air pollution and mortality and weather, NewYork City 1960-1964. Arch. Environ. Health 22:334 ff. 40. Schimmel, H., Greenburg, L. 1972. A study on the relationship of pollution to mortality, NewYork City 19631968. J. Air Pollut. Control Assoc. 22:607 ff. 41. Buechley, R. W., Riggan, W. B., Hasselblad, V., Van Bruggen, J. B. 1973. SO2levels and perturbations in mortality. A study in NewYork-New Jersey metropolis. Arch. Environ. Health 27:134 ff. 42. Schimmel, H., Murawski, T. J. 1976. The relation of air pollution to mortality. J. Occup. Med. 18:316 ff. 43. Schwartz, J., Dockery, D. W. 1992. Particulate air pollution and daily mortality in Steubenville, Ohio. Am. J. Epidemiol. 135:12-25 44. Schwartz, J., Dockery, D. W. 1992. Increased mortality in Philadelphia associated with daily air pollution con-

Annual Reviews www.annualreviews.org/aronline AIR POLLUTION centrations. Am. Rev. Respir. Dis. 145:600-4 45. Shumway, R. H., Azari, A. S., Pawitan, Y. 1988. Modeling mortality fluctuations in Los Angelesas I~tnctions of pollution and weather effects. Environ. Res. 45:224-41 46. Chappie, M., Lave, L. B. 1982. The health effects of air pollution: A reanalysis. J. Urb. Econ. 12:34676 47. Evans, J. S., Tosteson, T., Kinney, P. L. 1984. Cross-sectional mortality studies and air pollution risk assesment. Environ. Int. 10:55-83 48. Lipfert, F. W. 1984. Air pollution and mortality: Specifications searches using SMSA-baseddata. J. Environ. Econ. Manage. 1 l:208-43 49. US Environ. Protect. Agency. 1982. Review of the National Ambient Air Quality Standard for Particulate Matter: Assesmentof Scientific and Technical Information’. Research Triangle Park, NC 50. Trijonis, J. 1983. Development and application of methodsfor estimating inhalable and fine particles concentrations for HI-Vol data. Atmos. Environ. 17:999-1008 51. Deleted in proof 52. Mishan,E. J. 1971. Evaluation of life and limb: A theoretical approach. J. Polit. Econ. 79:687-705 53. Broome, J. 1978. Trying to value a life. J. Public Econ. 9:91-100 54. Broome, J. 1985. The economic value of life. Economica52:281-94 55. Jones-Lee, M. W., Hammerton, M., Philips, P. R. 1985. The value of safety: Results from a national sample survey. Econ. J. March:48-72 56. Garen, J. E. 1988. Compensating wagedifferentials and the endogeneity of job riskiness. Rev. Econ. Stat. 70:9-16 57. Moore, M. J., Viscusi, W. K. 1988. Doubling the estimated value of life: Results using newoccupational fatality data. J. Policy Anal. Manage.7:47690 58. Kahn, S. 1986. Economicestimates of the value of life. IEEE Tech. Soc. Mag. 5:24-31 59. Viscusi, W. K. 1991. Strategic and ethical issues in the valuation of life. In Strategy and Choice, ed. R. J. Zeckhauser, pp. 359-87. Cambridge, Mass: M1TPress 60. Cropper, M. L., Freeman, A. M. III. 1991. Environmentalhealth effects. In Measuring the Demandfor Environmental Quality, ed. J. B. Braden, C.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71. 72.

ABATEMENT BENEFITS

341

D. Kolstad, pp. 165-211. Amsterdam: North Holland Fisher, A., Chestnut, L. G., Violette, D. M. 1989. The value of reducing risks of death: A note on newevidence. J. Policy Anal. Manage. 8:88-100 Dockery, D. W., Schwartz, J., Spengler, J. D. 1992. Air pollution and daily mortality: Associations with particulates and acid aerosols. Environ. Res. 59:362-73 Brunekreef, B., Lumens, M., Hock, G., Hofschreuder, P., Fischer, P., Biersteker, K. 1989. Pulmonaryfunction changes associated with an air pollution episode in January 1987. J. Air Pollut. Control Assoc. 39:1444-47 Dassen, W., Bmnekreef, B., Hoek, G., Hofschreuder, P., Staatsen, B., et al. 1986. Decline in children’s pulmonary function during an air pollution episode. J. Air Pollut. Control Assoc. 36:1223-27 Dockery, D. W., Ware, J. H., Ferris, B. G. Jr., Speizer, F. E., Cook, N. R. 1982. Changein pulmonaryfunction in children associated with air pollution episodes. J, Air Pollut. Control Assoc. 32:937-42 Ware, J. H., Ferris, B. G. Jr., Dockcry, D. W., Spengler, J. D., Stram, D. O., Speizer, F. E. 1986. Effects of ambient sulfur oxides and suspended particles on respiratory health of preadolescent children. Am. Rev. Respir. Dis. 133:834-42 Dockery, D. W., Speizer, F. E., Stram, D. O., Ware, J. H., Spengler, J. D., Fen’is, B, G. Jr. 1989. Effects of inhalable particles on respiratory health of children. Am. Rev. Respir. Dis. 139:587-94 Ostro, B. D. 1983. The effects of air pollution on work loss and morbidity. J. Environ. Econ. Manage. 10:37182 Ostro, B. D. 1987. Air pollution and morbidity revisited: A specification test. J. Environ. Econ. Manage.14:8798 Ostro, B. D., Rothschild, S. 1989. Air pollution and acute respiratory morbidity: An observational study of multiple pollutants. Environ. Res. 50:238-47 Portney, P., Mullahy, J. 1990. Urban air quality and chronic respiratory disease. Reg. Sci. Urb. Econ. 20:407-18 Krupnick, A., Harrington, W., Ostro, B. 1990. Ambient ozone and acute health effects: Evidence from daily data. J. Environ. Econ. Manage.18:118

Annual Reviews www.annualreviews.org/aronline 342

CIFUENTES

& LAVE

73. Loehman,E. T., Berg, S. V., Arroyo, A. A., Hedinger, R. A., Schwartz, J. M., et al. 1979. Distributional analysis of regional benefits and cost of air quality. J. Environ. Econ. Manage. 6:222--43 74. Krupnick, A. J., Kopp, R. J. 1988. The Health and Agricultural Benefits of Reductions in Ambient Ozone in the United States. Report QE88-10.Washington, DC: Resourc. Future 75. Table C4: Establishment data, Earnings, Not seasonally adjusted. 1992. Employ. Earn. 39:156 76. McDonnell, W. F., Horstman, D. H., Hazucha, M., Seal, E. Jr., Haak, E. D., et al. 1983. Pulmonaryeffects of ozone exposure during exercise: Dose responsecharacteristics. J. Appl. Physiol. 54:1345-52 77. Kulle, T. J., Sauder, L. R., Hebel, J. R., Chatam, M. D. 1985. Ozone response relationships in healthy nonsmokers. Am. Rev. Respir. Dis. 132: 36-41 78. Horstman, D. D., Folinsbee, L. J., lves, P. J., Abdul-Salaam, S., McDonnell, W. F. 1990. Ozone concentration and pulmonary response relationships for 6.6 hour exposures with five hours of moderate exercise to 0.08, 0.10 and 0.12 ppm. Am. Rev. Respir. Dis. 142:1158-63 79. McDonnell, W. F., Kehrl, H. R., AbduI-Salaam, S. 1991. Respiratory response of humans exposed to low levels of ozone for 6.6 hours. Arch. Environ. Health 46:145-50 80. Spektor, D. M., Lippmann, M., Lioy, P. J., Thurston, G. D., Citak, K., ct al. 1988. Effects of ambient ozone on respiratory function in active, normal children. Am. Rev. Respir. Dis. 137: 313-20 81. Spektor, D. M., Thurston, G. D., Mao, J., He, D., Hayes, C., Lippmann, M. 1991. Effects of single- and multiday ozone exposures on respiratory function in active normalchildren. Environ. Res. 55:107-22

82. Hammer,D. I., Hasselblad, V., Portnoy, B., Wehrle, P. F. 1974. Los Angeles student nurse study. Arch. Environ. Health 28:255-60 83. Schwartz, J., Hasselblad, V., Pitcher, P. 1988. Air pollution and morbidity: A further analysis on the Los Angeles student nurses data. J. Air Pollut. Control Assoc. 38:158-62 84. Lippmann, M. 1989. Health effects of ozone: Acritical review. J. Air Pollut. Control Assoc. 39:672-95 85. Harrington, W., Krupnick, A. J. 1985. Short term nitrogen dioxide exposure and acute respiratory disease in children. J. Air Pollut. Control Assoc. 35:1061-67 86. Morrow, P. E., Utell, M. J. 1989. Responses ~f Susceptible Subpopulations to Nitrogen Dioxide. Report 23. Health Effects Inst. 87. Off. Air Qual. Plann. Stand. 1987. Review of the National Ambient Air Quality Standards for Ozone. Preliminary Assesmentof Scientific and Technical Information. USEnviron. Protect. Agency 88. Utell, M. J., Frampton, M. W., Roberts, N. J., Finkelstein, J. N., Cox, C., Morrow, P. E. 1991. Mechanisms of Nitrogen Dio~ide Toxicity in Humans. Report 43. Health Effects Inst. 89. Counc. Environ. Qual. 1992. Environmental Quality: 22nd Annual Report. 90. US Environ. Protect. Agency. 1991. National Air Quality and Emissions Trends Report 1990. Report EPA450/4-91-023. 91. US Census Bur. 1991. Statistical Abstract of the U.S. Washington, DC 92. Morgan, M. G., Henrion, M. 1990. Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. Cambridge: Cambridge Univ. Press 93. Rowe, M. D. 1980. Human exposure to sulfates from coal-fired power plants. J. Air Pollut. Control Assoc. 30:682-83