Valuing the health effects of air pollution

]DECi notes i ki*'V- FII fill(liings Valuing the health effects of air pollution That high levels of air pollution are badfor health and can cause premature death is well known. But policymakers need more specific information, espe- cially about the costs and benefits of reduction measures, to set priorities Air pollution is a serious and growing prob- Although air pollution affects both death - - lem in many developing countries, particular- and illness rates, greater social costs are likely CC ST BE ME FIT ly in rapidly expanding cities, where pollu- to be associated with premature death. AJALYS S IS OFTEN tion levels are often several times the maxi- Chronic exposure to air pollution can lead to mum level recommended by the World premature death by increasing the rate at At- E!;SI-N-IAL Health Organization. That high levels of air which lung tissue ages, by contributing to INPU T ) Al R pollution are bad for health and can cause chronic obstructive lung disease, and bvexac-y premature death is well known. But policY- erbating cardiovascular disease. Short-term PO)LLUJTI Or makers need more specific informarion to set peaks in air pollution (acute exposures) can priorities. Should available funds go to air increase the chance that a person in a weak- pollution reduction-or would they be better ened state (a person with pneumonia, for El FO1`T!.i spent on sanitation, education, AIDS preven- example) or a susceptible person (a person tion, or any of a host of other pressing con- with asthma, for example) will die. cerns? A cost-benefit analysis of air pollution The most accurate way to measure the reduction measures is a necessary input to impact of air pollution on death rates in a such a decision. However, answers vary great- given area is to conduct epidemiologic stud- ly depending on the method used. This note ies for that area. To capture both acute and describes some of the pitfalls that commonly chronic effects, a large cross-section of indi- arise when attempting to estimate the dollar viduals should be followed prospectively for value of benefits associated with a given at least ten vears, measuring both the air pol- reduction in pollution levels in low-income lution concentrations to which they are countries. Because cost-benefit estimates are exposed and other factors that affect the risk often necessary despite these pitfalls, boxes of death (smoking behavior, body mass describe how to prepare a rough estimate of index, family history). the benefits of pollution reduction, using Because such studies are expensive, only Delhi, India as an example. two have been completed to date, both in the United States (Dockery and others 1993; Pope and others 1995). These studies show Measuring the effects of air how pollution alters the survival function pollution on premature death (the probability that a person survives to each age in a community) and allow the number Air pollutants affect health in a variety of of life-years lost as a result of air pollution to ways, from itchy eyes to premature death. be calculated. FROM THE DEVELOPMENT ECONOMICS VICE PRESIDENCY OF THE WO)19LD A A gp PRIL 1996  Because they are less expensive, retrospective How appropriate is it to transfer results from studies that use aggregate data are common. epidemiologic studies conducted in industrial Cross-sectional studies correlate aggregate varia- countries to developing countries? At a mini- tions in air pollution levels across counties with mum, such projections must be made for dis- death rates by county. Cross-sectional studies are ease-specific deaths. Extrapolating the effect of generally unreliable, however, for two reasons. air pollution on total death rates from a study in First, if there has been recent immigration, his- the United States makes no sense if the distribu- toric air pollution levels in a county may not tion of deaths by cause differs greatly between measure the concentrations to which people cur- the United States and the country of interest. rently living in the county were exposed. For instance, the effect of a change in particu- Second, it is usually hard to control for con- late concentrations on total deaths comes mainly founding factors (smoking behavior, dietary risk from its effect on respiratory and cardiovascular factors) at the county level. deaths. In the United States half of all deaths are Time series studies, which correlate daily vari- caused by cardiovascular disease or respiratory ill- ation in air pollution with variation in daily mor- ness. In Delhi, India, by contrast, fewer than 20 tality in a given city, avoid the problems men- percent of all deaths are attributable to these tioned above, but at best measure the effects of causes. Even if people in Delhi reacted to a given acute exposure to air pollution on death rates. change in particulate pollution in exactly the Because they measure primarily the effect of acute same way that people in the United States did, Prospective studies exposure to air pollution on people who are in a the effect of the change on total mortality would are generally more weakened state and are therefore susceptible to be lower in Delhi (box 1). The figures in the last death as a result of a sudden rise in pollution con- row of the table in box 1 represent the most accu- centrations, such studies measure only some of rate estimates of the benefits of reducing air pol- more expensive the effects of air pollution on premature death. To lution in Delhi based on extrapolations from than retrospective the extent that many of these people would have U.S. studies. These figures may, however, repre- Cross-sectional or had short life expectancies even in the absence of sent lower-bound estimates, for two reasons. pollution, time series studies may overestimate First, they represent only the gains from reducing time series studies the life-years lost as a result of air pollution. Such acute exposure to air pollution. Second, people studies can provide a lower-bound estimate of the in Delhi may be more susceptible to the effects of reduction in premature death that would accom- air pollution than people in the United States pany a reduction in air pollution, however. because of lower levels of baseline health and more limited access to medical care. Can results obtained for one country be applied to another? Estimating the dollar value of In the absence of epidemiologic studies of the lives lost from air pollution region in question, some researchers advocate Economists calculate the value of a reduction in the use of dose-response functions from other risk of death by determining how much an indi- countries. The only prospective cross-sectional vidual is willing to pay for it. One approach to studies of the effects of chronic exposure to pol- measuring willingness to pay is to infer the value lution on mortality come from the two U.S. from compensating wage differentials in the studies mentioned earlier. Time series studies labor market. The theory behind this approach have also been conducted, primarily in the is simple: other things being equal, workers in United States and Europe (other parts of the riskier jobs are compensated with higher wages world have been studied, such as Santiago, Chile for bearing more risk. A second approach to by Ostro and others 1995, China by X. Xu and measuring willingness to pay is to ask people others 1994 and Z. Xu and others forthcoming, what they would pay to reduce their risk of and Sao Paulo, Brazil bySaldiva and others dying. This is referred to as the contingent valu- 1994). ation approach.  Recent compensating wage studies in the developing countries. Premature death in these United States have produced estimates of the countries is typically valued by using estimates value of a statistical life in the range of of the value of a statistical life from the United $1.9-$10.7 million (1990 dollars). Contingent States and adjusting for income differences valuation studies have produced slightly lower between the two countries, or by calculating for- estimates of $1.2-$9.7 million (1990 dollars). gone earnings. Which method is preferable? Compensating wage studies measure compensa- Economic theory tells us that the willingness- tion for risk of instantaneous death for people to-pay approach yields a value of a statistical life who are about 40 years old and thus measure that exceeds the value of forgone earnings as long the approximate value of 35 years of life as people are risk averse (Rosen 1988). In the (Viscusi 1993). Because death from air pollu- United States, for example, estimates of the value tion reduces life-years by less than 35 years on of a statistical life are typically five to ten times average, labor market estimates should be higher than the value of forgone earnings. If peo- adjusted accordingly. ple in other countries were equally risk averse, Compensating wage and contingent valua- then it would be appropriate to multiply the tion studies have not yet been conducted in value of forgone earnings in the country of inter- -ox I Estimating theefects of airpl. -lutio remature death Properly controlled Del-ihi; t capia6londia, s s extr eel high leves ofparticulate air plution, with levels of total suspended articulates (TSP) som{etimes reaching lve times the maximum level suggested by the World for, results from -. Heath tazaon 0 W+AOz.; industrial countries 1;How c'ab the effcts far01yn re- . p-ons he of air polluion o:sn total mortali in Delhi be calculated? Aplying results from dose- -esponse ltcu a for U.S. cities can seriously misrepresen athe effet of pollution on health can be extrapolated ..in. Delhi.On aeage a i micO m (g change PM1articulate mater measuring tO microns or less to developing '-rin diater) resulin ha- I percentchange in tota mlity in the United States (Ostro 1994). Applying this relatio itp bt -De wkuld imply th reidng average annual TSP from its 1991 level of 374.95 micro- countries gramspe cubic meter4sg/)tlthe WHO's upper gUideline ievel (90 pgA-3) would result in a 15.05 per- c ne in notraurna d or 84 fewer deaths (ee table. This finding suggests that nearly all deaths in Dei esutinro peumia chi osru ie lmonary disease, and cardiovascular dis- ease (8244 of 80) wuid be e ate smply dig the level of TSP to 90 p/in. One explana- tonr tisipsibleresul is tht the lead-ingauses f inthe Unted States and India differ. Th minimize tis tpe of erro and to obain mor reable estimates, we can apply dose-response func- -tions es ed forspeci caus f deh t h Dlhi figuret. According to Schwartz and Dockery ( 992), --a i00 pg/ change in T resul in a 1 per change : ie number of deaths from chronic obstruc- tive pulmoiary ise an II percent cn in the number of deaths from pneumonia, and a i0 percent .ane in the ber of de0 from cariovascular dsease Applying these figures to Delhi, owering the avere anual TPtt te WI guideline would result in 2,712 fewer deaths, a 5 percent reduction in nn aumamrta-it- - Estimated number of deaths avoided in Delh by reducing averge annual total suspended particullate; lel: Study) Number of Percentage reduction Number of case of d.eath d , 11 in mortality deaths avoided Total nontraurna 54,783 15.05 8,244 Card.lo asidcA disee 28.50 1.625 Pneum nia 14644 31.34 515 Brec tanasthrna 0 54 14- 572 8,402 5.00 2,712  est by the same factor found in U.S. studies. It is, Policy Research Working Paper 1453. World b15ank, however, likely that risk aversion varies with liv- Policy Research Department, Washington, D.C. ing standards and that the value of a statistical life Pope, C.A., and others. 1995. "Particulate Air Pollution as in developing countries is a smaller multiple of a Predictor of Mortality in a Prospective Study of U.S. forgone earnings than it is in the United States. Adults." American Journal of Critical Care Medicine Until further research provides estimates of 151(3):669-74. the value of a statistical life in developing coun- Rosen, Sherwin. 1988. "The Value of Changes in Life tries, one approach to valuation is to use forgone Expectancy."JournalofRisk and Uncertainty 1:285-304. earnings as a lower bound to the value of Saldiva, P.H.N., and others. 1994. "Association between changes in life expectancy and projections of the Air Pollution and Mortality Due to Respiratory Diseases value of life from U.S. studies as an upper in Children in Sao Paulo, Brazil: A Preliminary Report." bound, as shown in box 2. . Environmental Research 65:218-25. Compensating -by Maureen L. Cropper Saldiva, P.H.N., and others. Forthcoming. "Air Pollution wage studies and and Nathalie B. Simon and Mortality in Elderly People: A Time Series Study in Sao Paulo, Brazil." Archives of Environmental Health. contingent valuation Further readin Schwartz, Joel, and Douglas Dockery. 1992. "Increased studies calculate the g Mortality in Philadelphia Associated with Daily Air value of reduced Dockery, Douglas, and others. 1993. "An Association Pollution Concentrations." American Review of Betreen Air Pollution and Mortality in Six U.S. Cities." Respiratory Disease 145:600-4. life expectancy New EnglandJournal ofMedicine 329(24): 1753-59. Viscusi, W Kip. 1993. "The Value of Risks to Life and Ostro, Bart. 1994. "Estimating the Health Effects of Air Health." Journal of Economic Literature 31:1912-46. Pollutants: A Model with an Application to Jakarta." Xu, Xiping, Jun Gao, Douglas W Dockery, and Yude Policy Research Working Paper 1301. World Bank, Chen. 1994. "Air Pollution and Daily Mortality in Policy Research Department, Washington, D.C. Residential Areas of Beijing, China." Archives of Ostro, Bart, Jose Miguel Sanchez, Carlos Aranda, and Environmental Health 49(4):216-22. Gunnar S. Eskeland. 1995. "Air Pollution and Xu, Z.Y., and others. Forthcoming. "Air Pollution and Mortality: Results from a Study of Santiago, Chile." Daily Mortality in Shenyang." Thos DECot Valsingepa fed b n Maueen L.CopendiNahaiB.SmnntePocyRsrhDprmntfteWrlBnk andd roug ncsariyrpreseainto the vilew of rh -eaWrld Bave ad it " member conrisandtheeoe seou[ d no Dehittribuedto the bound Ba Unk orirs afiiresDCntes reh produced of rhe Reeac AvsrStf.Wwecmyorq estionsandcme) t;peaee mal wh e e to the author we tEvyAfrRD yearssavedPepare for Wol B~T ank rstaff Gennvereaoilncm nranctrignQhip$0058ndaanulitrraef5 Thigas DCoestwats preparied by iaurenL Cropper arndNtaieB. Simo~ni thes Poicy Reeac Dexarmpe,t we etimae thedBak DECnotes transmit key researchi findings to Bank Group managers and staff. They are drawn from the work of individual Bank researchers and do not necessarily represent the views of the World Bank and its member countries-and therefore should not be attributed to the World Bank or its affiliates. DECnotes are produced by the Research Advisory Staff. We welcome your questions and comments; please e- mail them to the authors or to Evelyn Alfaro, RAD. Prepared for World Bank staff