5
Economic Valuation of Reduction in Mortality Risk Associated with Ambient Ozone

INTRODUCTION

This chapter examines the conceptual and empirical literature on the economic valuation of reductions in mortality risks generally and specifically in the context of the epidemiologic literature on the linkage between ozone and mortality risks discussed in Chapter 4.

As described in Chapter 2, the U.S. Environmental Protection Agency (EPA) follows standard practice in cost-benefit analyses by using estimates of willingness to pay (WTP) to calculate the economic value of reducing mortality risks. EPA’s current approach for estimating the benefits of reducing mortality risk has several steps. First, it estimates the reduction in individuals’ annual mortality risk stemming from the expected change in ozone concentration. Next, it calculates the annual number of deaths avoided (that is, postponed to some future year) by the postulated reduction in ozone concentrations to which the population at risk would be exposed. A central estimate of the distribution of economic values of the avoided deaths is then calculated by using a central value of a statistical life (VSL) drawn from the available WTP literature related to changes in annual mortality risk. Uncertainties in the estimates of mortality effects and valuation are then accounted for with reference to the confidence intervals around the estimates.

One important and controversial issue related to valuation is whether the effects of reducing mortality risk should be measured in terms of lives saved (deaths postponed) or years of life extension (increased remaining life expectancy). If it is the latter, then it may be more appropriate to use a monetary value per statistical life year (VSLY) or to adjust the VSL to reflect the preferences of people of different ages rather than to use the same VSL in all cost-benefit analyses. (see Chapter 2 for definitions and examples of VSL and VSLY.) EPA has included various sensitivity analyses (alternative calculations) to explore the



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 128
5 Economic Valuation of Reduction in Mortality Risk Associated with Ambient Ozone INTRODUCTION This chapter examines the conceptual and empirical literature on the eco- nomic valuation of reductions in mortality risks generally and specifically in the context of the epidemiologic literature on the linkage between ozone and mortal- ity risks discussed in Chapter 4. As described in Chapter 2, the U.S. Environmental Protection Agency (EPA) follows standard practice in cost-benefit analyses by using estimates of willingness to pay (WTP) to calculate the economic value of reducing mortality risks. EPA’s current approach for estimating the benefits of reducing mortality risk has several steps. First, it estimates the reduction in individuals’ annual mortality risk stemming from the expected change in ozone concentration. Next, it calculates the annual number of deaths avoided (that is, postponed to some future year) by the postulated reduction in ozone concentrations to which the population at risk would be exposed. A central estimate of the distribution of economic values of the avoided deaths is then calculated by using a central value of a statistical life (VSL) drawn from the available WTP literature related to changes in annual mortality risk. Uncertainties in the estimates of mortality effects and valuation are then accounted for with reference to the confidence intervals around the estimates. One important and controversial issue related to valuation is whether the effects of reducing mortality risk should be measured in terms of lives saved (deaths postponed) or years of life extension (increased remaining life expec- tancy). If it is the latter, then it may be more appropriate to use a monetary value per statistical life year (VSLY) or to adjust the VSL to reflect the preferences of people of different ages rather than to use the same VSL in all cost-benefit analyses. (see Chapter 2 for definitions and examples of VSL and VSLY.) EPA has included various sensitivity analyses (alternative calculations) to explore the 128

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 129 implications of alternative approaches in several of its cost-benefit analyses for air-pollution regulations. However, the agency has been using the same VSL estimates regardless of remaining life expectancy in its primary analyses. The Office of Management and Budget (OMB) has urged caution in adjusting VSL on the basis of age because of uncertainties in the literature (Graham 2003) but encourages sensitivity analyses with alternative measures of changes in mortal- ity and its monetary valuation. This chapter examines the theoretical and empiri- cal evidence regarding the relationship between age and WTP for mortality-risk reductions in later sections. As noted in Chapter 4, another related question is whether those most af- fected by ozone are already in such poor health that their remaining life expec- tancy is low despite a reduction in ozone exposure. If decreases in short-term ozone concentrations merely decrease “harvesting” those who are already frail and near death, the economic benefit of reducing their exposure to air pollution may be relatively small and should be reflected in a much lower VSL or VSLY than that estimated for relatively healthy people. In addition, perhaps we should be considering the preferences of those at high risk (generally, those in poor health) regarding reduction in their mortality risk. This chapter takes up the question of whether the VSL or the VSLY is more appropriate for EPA to use in valuing reductions in mortality risk associ- ated with reduced exposure to ozone. It also considers what economic theory and the empirical literature say about how WTP for reductions in mortality risk varies with the age of the person at risk, the size of the risk change, the health status of those at risk (existing chronic illness vs average health and effects on quality of remaining life), income, and other socioeconomic variables to im- prove approaches for assigning values to reductions in risk. CONCEPTUAL UNDERPINNINGS OF VALUATION OF MORTALITY RISK As introduced in Chapter 2, cost-benefit analysis for a regulatory decision is intended to help answer the question of whether some proposed policy will result in more welfare for the affected people—that is, whether the expected benefits of the regulation are worth the expected costs. The standard approach to cost-benefit analysis is to quantify both costs and benefits in terms of “opportu- nity cost.” For benefits, that means a measure of what those who benefit would be willing to forgo to obtain the specified benefit. This is the basic concept of WTP, which is taken to be a measure of how much better off the members of the benefiting population perceive themselves to be, assuming that the members of the population have full information about the benefits.1 1 There is some debate in the literature on the foundations of welfare economics as the whether WTP is an appropriate measure of the welfare changes of individuals. For a re- cent overview of this debate, see Adler and Posner (2006), especially Chapter 2. WTP as

OCR for page 128
130 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits In the early days of cost-benefit analysis for public programs, the pre- dominant approach to determining the monetary value of programs that reduced mortality risks was the human-capital approach. That approach focused on the financial loss when a person dies prematurely, as measured by lost lifetime earn- ings (see, for example, Dublin and Lotka 1930). It is conceptually similar to the cost-of-illness approach, which measures lost earnings and medical costs due to illness and premature death. Later, several economists (for example, Schelling 1968 and Mishan 1971) discussed why those indicators of the financial burden of disease and premature death do not measure the monetary value of the full effect of disease and premature death on the welfare of the population and are therefore insufficient for a full cost-benefit analysis of public policies aimed at reducing morbidity or mortality. They noted that those financial measures as- cribe no value to the lives of homemakers or retired persons, and they argued that a more appropriate way to value a program that reduces mortality risk is to determine what the reduction in risk is worth to the people who benefit. Thus, WTP is a more appropriate measure of the change in welfare in cost-benefit analysis because it reflects not just the financial effect but the value that people place on the effect on quality of life and longevity. WTP specifically measures the ex ante value of reduction in risk that pub- lic-health and public-safety programs may provide. It is not known specifically whose death will be prevented, only that the probability of death in a group of people will be reduced. Generally, WTP is expected to exceed the cost of illness, but costs paid by third parties (such as insurance-paid medical costs) will gener- ally not be included in an individual’s WTP and should in some cases be added to individuals’ WTP to determine total value. There is substantial evidence that WTP for reductions in mortality risk typically exceed by a substantial amount the expected value of lost earnings (e.g., Viscusi 1993). Some authors have argued that the appropriate measure of WTP for risk reduction is not the values expressed by the people actually affected by the pol- icy, but the values of people who do not know their actual position in society (age, income, wealth, health status, etc.) but are acting behind a “veil of igno- rance” or as “impartial spectators” (see, for example, Pratt and Zeckhauser 1996; Sunstein 2002; Hammitt 2007, p. 237). There is an extensive literature in welfare economics based on these ideas as first developed by Harsanyi (1955) and Rawls (1971). But with one significant exception, this literature has focused on the principles of distributive justice and fairness and not on the valuation a welfare measure is criticized as being dependent on the existing distribution of income and wealth and on existing preferences of individuals, which may be influenced by such things as addictive behaviors. The standard approach in cost-benefit analysis is to ac- knowledge these difficulties in principle but to base empirical measures of value on WTP whenever possible. In fact, the OMB Circular A-4 is explicit in calling for monetary measures based on WTP (OMB 2003).

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 131 of specific goods received by individuals. The exception is the paper by Pratt and Zeckhauser (1996). We see this literature as offering an interesting thought experiment for considering ethical issues regarding distributive justice and valuation. But it does not offer practical methods and techniques for the devel- opment of criteria of distributive justice or for the empirical estimation of val- ues. All of the practical methods for estimating values are necessarily based on the preferences of individuals who know their positions in society, and these values are based on expressions of WTP for changes in risk. A possible surrogate for the “veil of ignorance” or “impartial observer” perspective is the “public preference” or “social value” approach to determining benefits. This approach involves asking people to express their WTP for policies that would reduce the risk of premature mortality for specified groups of indi- viduals with certain characteristics, for example, age. Some studies have shown that people have different preferences concerning policies delivering health im- provements or risk reductions to different social groups (Cropper et al. 1994; Dolan et al. 2005). But such studies are not based on the “veil of ignorance” approach, because it is not possible to show that people could in fact ignore their own position in society relative to those whose risk changes they were being asked about. For either approach, the conceptual questions regarding appropriate aggregation of such values for a specific policy assessment have not been re- solved (see below). Basing the values used in economic analysis on the preferences of the af- fected individuals is a fundamental tenet in welfare economics. Thus, the mone- tary measure of the benefit of a reduction in mortality risk is based on the value placed on the benefit by the individuals receiving it. It is not the value held by the policy-maker or the experts. The standard approach for cost-benefit analyses is thus to measure the benefits of a mortality-risk reduction on the basis of the value to each individual of his or her own reduction in risk and then to sum these values across all affected individuals. An important conceptual question is whether, if individuals are altruistic toward others, especially family and friends, altruistic feelings should be in- cluded in the valuation of reductions in mortality risk by adding WTP values that people have for others’ risk reductions. As was discussed in Chapter 2, eco- nomic analysis (e.g., Jones-Lee 1991) shows that if the altruism involves a gen- eral regard for the well-being of others, that is, is non-paternalistic in nature, summing WTP over all affected individuals would leave the sign of net social benefits unchanged since the altruists’ benefits would be offset by the altruists’ recognition of the costs of the policy to others. Summation is appropriate only if the altruism is paternalistic in nature. Because we have no evidence to indicate the extent of paternalistic altruism regarding risk reductions, we consider that the approach of using individuals’ WTP only for their own mortality-risk reduc- tion is appropriate.

OCR for page 128
132 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits According to this approach, maximum WTP for mortality-risk reduction is defined as the amount of money that would make the person indifferent between choosing to spend the money to decrease his or her own mortality risk and for- going the decrease in risk and keeping the money to spend on other things. Ex- pressions for a person’s WTP can be derived from models of the choices made by utility-maximizing people (see Chapter 10 of Freeman [2003] for an over- view). The models are simplified representations to describe economic behavior and to predict the choices that a utility-maximizing person would make. They provide a formal definition of WTP and provide a basis for analyzing the factors that determine a person’s WTP and how it may vary. For example, the model can help to analyze the question of whether VSL varies with age, income, and other socioeconomic characteristics of individuals. The life-cycle consumption model (see Shepard and Zeckhauser 1982; Cropper and Sussman 1990) is the most comprehensive economic model that is available for analyzing how WTP for small changes in mortality risk is expected to be related to a person’s age and other factors. In the model, persons, acting as individuals, are assumed to try to maximize their expected “utility from con- sumption” over their lifetimes, given constraints involving their initial wealth, their annual income, the cost of consumption, and a subjective rate of time pref- erence. Utility from consumption is the satisfaction or enjoyment obtained from any activity that involves an expenditure of money or time. The subjective rate of time preference is the rate at which a person discounts future utility relative to current utility. Empirical evidence and common sense suggest that the value to a person today of future consumption is something less then the value of current consumption, and the difference between the two is the discount factor (which implies a subjective rate of time preference). The model postulates that a person makes economic choices on the basis of his or her expectations about future utility from future consumption and about future income, all discounted to a present value at the person’s rate of time pref- erence. People are also assumed to include in their choices their expectations about their future survival probability. The model incorporates constraints that people have because of initial wealth and income over time, and it takes oppor- tunities for borrowing, bequests, and lending into account. In a simple model, an expression of a person’s maximum WTP for a small reduction in mortality risk can be derived to illuminate the factors expected to influence the magnitude of and variation in WTP. The expression is the product of the reciprocal of the probability of surviving the current period, the present discounted value of the expected utility from consumption conditional on sur- viving the period, and the reciprocal of the marginal utility of consumption con- ditional on survival (used to convert utility into monetary terms). Marginal util- ity of consumption is the additional satisfaction or enjoyment obtained from a small increment of consumption.

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 133 To examine the effect of age on WTP for a reduction in the current risk of death, we ask how age would affect each of those terms.2 As people age, their probability of surviving the current period generally falls. Thus, their WTP to reduce their mortality risk in the current period is expected to increase because as the probability of death in the current period rises, a person is likely to be willing to forgo more current consumption to increase the chances of surviving to the next period to continue enjoying the fruits of his or her resources. That may be offset, or at least limited, by a desire to leave some bequest for one’s children and others. However, aging not only increases current risk, it decreases life expectancy conditional on surviving the current period, and this is expected to decrease WTP. This effect could easily offset the first effect on WTP. How the present value of expected utility of consumption—the second term of the equation—changes with age is ambiguous. If per-period consump- tion and its utility were constant over time, the present value of expected utility of consumption would be proportional to discounted remaining life expectancy. With rare exceptions, mortality rates rise with age. Thus, the decrease in remain- ing life expectancy with increasing age motivates the hypothesis that WTP for mortality-risk reduction should fall with age. However, if there are constraints on borrowing against future income, per-period consumption is likely to rise in the earlier years of the life cycle. If, for example, a person cannot be a net bor- rower but can lend at the riskless rate of interest, his or her consumption is likely to be constrained by limited income at the beginning of life due to lower earning potential. That will cause the present value of the utility of consumption and hence the WTP for mortality-risk reduction to increase up to some point and then to decline (Shepard and Zeckhauser 1982); a plot of WTP against age would have an inverted U-shape. The combined effect of all those factors is that the impact of age on WTP for mortality-risk reduction is ambiguous on the basis of theoretical analysis alone. It requires empirical evidence to determine whether the combined effect of these factors is that WTP increases, decreases, or remains the same over all or part of a lifetime. Similar arguments can be made about health status. Health status may af- fect life expectancy, pleasure in consumption, or both. If life expectancy is re- duced because of illness, the first term in the expression of WTP should be posi- tive; with smaller chances of surviving the current period, the person should be willing to pay more to reduce mortality risk. As with age, the effect of health on the other factors that determine WTP is ambiguous. Chronic illness, for exam- ple, could reduce the quality of life and cause a person to derive less enjoyment from consumption. In summary, the effects of age and health status on WTP are ambiguous. Under reasonable assumptions the discounted expected utility of consumption 2 See Hammitt (2007) for further discussion of the points raised in the remainder of this section.

OCR for page 128
134 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits increases with future health and life expectancy (and so decreases with age). The ambiguity arises because the expected marginal utility of consumption is also likely to increase with future health and life expectancy. Since VSL is the ratio of the discounted expected utility of consumption to the expected marginal util- ity of consumption, and since both terms move in the same direction, the effect on the ratio is indeterminate. Thus, the combined effect of lower likelihood of survival and shorter remaining life expectancy on WTP to reduce mortality risk is ambiguous. So far, we have described only the effects of a current change in risk of death on WTP (or VSL). Also potentially relevant to valuing the effect of ozone on mortality risk is the WTP (or VSL) associated with a future change in risk of death. A lag between a change in exposure and a change in risk is often referred to as a latency in the risk. To address it, the life-cycle model can be further ma- nipulated (see Alberini et al. 2006a) to show that the WTP for risk reduction in some future period is likely to be lower than the WTP for risk reduction in the current period for two reasons. First, there is a probability that the person will not survive to the age when the risk reduction is to be realized. And second, if the rate of time preference is positive, the prospect of future consumption has less current utility than present consumption. However, if WTP for some risk reduction increases with age faster than the discount rate, the current WTP for a future risk reduction would be larger than for a current risk reduction (Hammitt and Liu 2004). The models can provide insights into other key factors that affect WTP. WTP for a given risk reduction should be larger than WTP for a smaller risk reduction. However, the VSL is arithmetically the WTP divided by the risk re- duction, so that simple relationship is not enough to predict whether the VSL will be lower (or higher) for a smaller risk reduction. The theoretical literature (Hammitt and Graham 1999) suggests that for small changes in mortality risk, WTP will move roughly in proportion to the size of the risk change, leaving the VSL roughly constant over different risk changes. However, empirical studies (see Hammitt and Graham [1999] for a summary and Alberini et al. [2004] for a more recent example) often find that the VSL is larger for smaller risk changes. The expected effect of greater income or greater wealth on WTP (and therefore the VSL) is unambiguous. Higher income (or greater wealth) is ex- pected to increase WTP for a given risk reduction. No life-cycle models in the literature distinguish effects on WTP on the basis of sex, race, marital status, number of children, or other variables that one might hypothesize could affect WTP. However, the empirical literature has looked at some of those issues (see below). EMPIRICAL METHODS OF VALUING MORTALITY-RISK REDUCTIONS WTP is typically measured by analyzing the prices paid for goods and ser-

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 135 vices. The maximum price that a person is willing to pay for a good or service is a measure of how much the person values the good or service. Prices for reduc- ing or preventing mortality risks cannot be directly observed, because, for the most part, reduction or prevention of mortality risks is not directly purchased in the marketplace. However, there are instances in which the monetary tradeoffs that people are willing to make between income and mortality risks can be ob- served or measured, for example, higher wages for riskier jobs or higher prices for safer products. There are two general economic approaches to measuring WTP for non- market goods, such as changes in mortality risks. The first is to analyze actual situations in which WTP for changes in mortality risks may be indirectly re- vealed; this category of estimation approaches, which is based on behavior, is called revealed preference. The second is to have subjects respond to a hypo- thetical situation that is designed to have them reveal their WTP; this category, which is based on responses to survey questions, is called stated preference.3 Revealed-Preference Methods An example of revealed-preference methods is a wage-risk study in which wage premiums for risks of death on the job are estimated. This approach ana- lyzes the factors that determine differences in actual wages between jobs, in- cluding on-the-job risks of death. The additional wages that people are paid per unit of additional risk of fatal injury is a measure of the monetary value of the additional risk to the person who voluntarily accepts the risk in exchange for a given wage increment. In this context, this value is equal to or less than his will- ingness to accept compensation (WTA) for incurring greater risk and is equal to or greater than his WTP for a reduction in risk (Hammitt, 2002).4 The primary advantage of this type of study is that it is based on actual behavior. Some of the limitations are that it is difficult to find situations in which an observable trade- off is made between income (or expenditures) and mortality risk and difficult statistically to isolate WTP for a mortality-risk increment from other factors involved in the specific behavior or decision. Fully specifying potential con- founding factors in these kinds of decisions, such as associated decreases in rates of nonfatal injury or other benefits (such as reduced risk of property loss due to purchase of a home smoke detector), is one of the important challenges to 3 For more information on these estimation techniques, see Freeman (2003). 4 When WTP and WTA are small compared with income, the theoretical expectation is that these measures will not be much different from one another for a given change in risk. In revealed-preference studies, that is generally the case. However, in stated- preference studies, responses to WTA questions are often much higher than responses to WTP questions. Concern that respondents are not forced to consider a budget constraint when answering WTA questions has led most analysts to prefer WTP questions in stated- preference studies.

OCR for page 128
136 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits the estimation of WTP for mortality-risk reduction in revealed-preference stud- ies. Many labor-market studies have examined the tradeoff between fatality risk and income; they are referred to here as wage-risk studies. The estimation method is important because it has been a primary source of the estimates that EPA has used for monetary valuation of mortality-risk reductions in environ- mental-policy analyses. The basic premise of wage-risk studies is that workers reveal the tradeoffs that they are willing to make between risk and income in the choices of jobs that they accept. The idea is simple, but the execution of a study is complex because many factors go into a job choice and a hiring decision. The safest jobs tend to be the highest-paying jobs because they require more skills and education. Other factors must therefore be taken into account, and this is done through estimation of a wage equation in which wage rates are specified as a function of worker characteristics, job and industry characteristics, and on-the- job fatality risks. Wage-risk studies do not directly estimate a VSL but rather a market equi- librium between wages and on-the-job risk that represents the intersection of labor supply (and the worker’s marginal demand for an increment in on-the-job safety) and labor demand (and the employers’ marginal cost to supply more safety). The result is an average value of a small change in risk of on-the-job fatality. The VSL is inferred from that “risk premium” as the total of the incre- mental wages paid to each worker per year to prevent one fatality in that year. With a few exceptions (Gegax et al. 1991; Liu and Hammitt 1999; Ham- mitt and Ibarraran 2006; who asked workers their perceptions of on-the-job risks), wage-risk studies presume that workers’ perceptions about risks on the job can be accurately measured by reported risks. Further, they presume that workers have sufficient information about on-the-job risks for their choices in the labor market to reflect responses to actual differences in risks across jobs. Using the results of those studies in policy analysis also presumes that the mar- ginal value of risk reduction in the labor market is consistent with such values in the general population. For that reason, wage-risk studies that look at a broad cross-section of occupations and industries are preferred to studies that look at a narrower set of occupations and industries if the results are going to be used in policy analysis for the general population. Even so, if there is self-selection of high-risk occupations by those who are less risk averse, implying they demand less compensation than those in the general population do, then the wage-risk studies will underestimate the general-population VSL. Wage-risk studies are limited to examining the types of on-the-job fatali- ties that are recorded in available databases that cover a wide array of industries (for example, databases of the Bureau of Labor Statistics, worker compensation boards, and the National Institute for Occupational Safety and Health); these are primarily accidental deaths. Indeed, a large share of on-the-job fatalities occur while being in vehicles. Other causes of death that may be related to work, such as cancer and chronic illness related to on-the-job exposure, are more difficult to study and less likely to be included in available databases because of long lags

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 137 between exposure and death and because they are thought to be less frequent in most occupations. Stated-Preference Methods An application of stated-preference methods is a survey in which subjects are presented with a hypothetical situation that involves a tradeoff between in- come or expenditures and a change in the risk of death. In a direct stated- preference approach, subjects are asked to estimate what they would be willing to pay to change their mortality risk by a specific amount. Other stated- preference approaches ask respondents to choose among alternative scenarios or policies that have varied costs and amounts of risk reduction and thus reveal but not explicitly state their WTP values. In any stated-preference approach, it is important that the situation presented to study subjects be realistic and easy to understand. The primary concern about this type of study is whether subjects are able and have incentive to give accurate responses to questions of this nature and whether answers in the context of hypothetical situations are accurate pre- dictors of behavior and choices in situations in which the costs and conse- quences are real. Many challenges exist concerning survey design for stated-preference studies about mortality risk. Difficulties in comprehension of small risks are often discussed. Corso et al. (2001) demonstrated the importance of visual aids in communicating quantitative risk information. Krupnick et al. (2002) used some simple comprehension questions to identify respondents who might be unable to understand or work with quantitative risk information. It seems that no matter how carefully the information is presented, some fraction of the general public has difficulty in making sense of it or is unwilling to put in the effort re- quired. Another challenge is the WTP elicitation mechanism. Cues given in the survey can lead to starting point bias (Mitchell and Carson 1989) and different procedures can lead to different estimates (Champ and Bishop, 2006). Finally, some studies of illness-related health risks have used health-care changes as the risk-reduction scenario (e.g., Johnson et al. 1998; Krupnick et al. 2002; Alberini et al. 2004), and others have used relocation and cost-of-living changes (e.g., Viscusi et al. 1991). Stated-preference studies generally obtain lower mean VSLs than wage-risk studies (Kochi et al. 2006), but why this occurs is difficult to determine. Wage-risk and stated-preference studies have different strengths and weaknesses with respect to providing estimates of WTP for mortality-risk reduc- tion for use in environmental-policy analysis. Wage-risk studies have the strength of being based on actual behavior, and many economists consider this an overwhelming advantage over stated-preference methods. However, they include only working-age adults who are healthy enough to be working, and they predominantly consider accidental death—a different population and a dif- ferent type of risk from what may be associated with ozone.

OCR for page 128
138 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits Stated-preference studies have the advantage of being able to be designed to address any type of mortality risk and can capture values for any population that is capable of answering the survey questions. Thus, there is a greater possi- bility of specifically addressing the type of illness-related mortality risks that are associated with pollution exposure and of including the older population that is expected to be at greatest risk. Methods for Estimating Values of Changes in Life Expectancy The above discussion focused entirely on how revealed- and stated- preference studies estimate WTP for annual mortality-risk reductions, which have generally been summarized by reporting average VSLs. The literature on the estimation of WTP specifically for changes in life expectancy (usually re- ported as VSLYs) is smaller. In the revealed-preference literature, specifically, the wage-risk studies, variation in the average age of workers in various occupa- tions is used to identify life-expectancy differences across the working popula- tion and from that to estimate the wage premium for taking jobs that are riskier in terms of life-expectancy losses. Selection of workers into riskier jobs compli- cate the use of this information as a proxy for the VSLY in the general popula- tion and in a pollution context. Aldy and Viscusi (in press) found that on-the-job fatality rates vary by age and that taking this into account alters the estimation of wage-risk premiums by age group. Of course, since few workers are in the labor force beyond age 65, the VSLYs (and VSLs) estimated based on labor market data can only apply to the “near-elderly” (Evans and Smith 2006). Findings in this literature are reviewed below. For stated-preference studies to estimate WTP for reductions in risk de- fined as increases in life expectancy requires making respondents understand that they are valuing an increase in life expectancy over their remaining life- times (from a shift in the hazard function), rather than more time being simply tacked on to the end of life. Few have attempted this task, and evidence suggests that these attempts have not been successful (see below). In practice, most VSLY estimates used to value changes in life expectancy have been ad hoc and derived from estimates of the VSL. The most frequent are those that simply annualize the VSL, assuming a given rate of interest (time preference) and underlying life expectancy (Mauskopf and French 1991). An- other approach is to convert the annual risk reduction being valued into its equivalent change in life expectancy and, using that metric together with the estimated WTP, to calculate the VSLY (Alberini et al. 2006b). The drawback of this approach is that were respondents to realize that their annual risk reduction of (say) 5 in 10,000 translates into a few weeks of additional life expectancy, their WTP might be different. The WTP for any risk reduction can be summarized as an average VSL or an average VSLY, given estimates of the rate of time preference and remaining life expectancy for the population from which the WTP values were derived. As

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 149 cating that a quadratic specification was used is nearly significant. That means that the quadratic specification, which allows a nonlinear relationship between WTP and age, is better than a linear specification at capturing any senior- discount effect that may exist. Overall, these results suggest that a deeper search for age effects through examination of alternative specifications for age might have found such effects. These inconclusive findings are similar to the conclusions reached by Dockins et al. (2004), who, in summarizing other reviews performed by EPA, wrote that “empirical work exploring the relationship between age and the VSL has provided mixed results.” They discussed the series of studies by Alberini et al. (2004) in the United States and Canada but otherwise examined only wage- risk studies. Another review of the literature (Evans and Smith 2006) offered the same conclusion, citing Alberini et al. (2004) and DeShazo and Cameron (2004): “These empirical analyses do not offer an unambiguous message.” Revealed-Preference Results Bearing on Effect of Age on WTP to Reduce Mortality Risk Aldy and Viscusi (2007) reviewed the empirical wage-risk literature about the relationship between age and WTP to reduce mortality risk. They note that most wage-risk studies reveal little about the relationship between age and WTP for mortality-risk reduction even when an age-interaction variable is included in the specification. That is because the simpler specification implies a linear rela- tionship with age, which is expected only under the most restrictive assump- tions. Thus, an informative analysis requires at least a nonlinear specification that allows VSL to be either rising or falling with the worker’s age. In addition, the authors note that rates of on-the-job fatalities also vary with age, so a com- prehensive specification should take this into account. Although the overall on- the-job injury rate declines with age, the on-the-job mortality rate increases with age. Thus, both workers’ preferences for risk reduction and employers’ wage offers may vary with respect to workers’ age; thus, there may be different wage- risk market equilibria for different age groups. The authors report the results of their own analyses on this issue (Viscusi and Aldy 2007; Aldy and Viscusi in press) that allow the wage-risk equilibria to vary by age group. They find a consistent pattern of an inverted U-shaped rela- tionship between VSL and age, with values for the youngest (18-24 y old) and oldest (55-62 y old) being roughly similar and the peak being about twice as high, usually in the middle age group (35-44 y old). Aldy and Viscusi (in press) note that the usual cross-sectional analysis does not take into account that expec- tations about future income and life expectancies vary with year of birth, both being higher in younger workers in the sample (given trends to higher income and longer life expectancies). They adjusted for those differences, and used the age-specific estimates of on-the-job fatality risk as described by Viscusi and Aldy (2007). They report that “the age-VSL relationship still follows an in-

OCR for page 128
150 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits verted-U shape, but accounting for year of birth effectively rotates the shape so that younger workers now have a lower VSL; the curve peaks a little later in life (at age 46), and older workers now have a higher VSL. The oldest worker in our sample (age 62) now has a VSL that is 35 percent lower than a 46-y-old worker’s VSL, the peak value in this analysis.” Aldy and Viscusi (2007) note that if this result for the group of 55-62-y- olds is applied to all adults over 55 y old, they obtain a result similar to what EPA obtained when it estimated a lower VSL for those over 65 y old in a sensi- tivity analysis of the proposed Clear Skies legislation (which would create a mandatory program to reduce power plant emissions of sulfur dioxide, nitrogen oxides, and mercury by setting a national cap on each pollutant (EPA 2006b). Aldy and Viscusi (2007) present several important conclusions from their review of the evidence in the wage-risk literature: • VSL varies with age in an inverted U-shape, as has been predicted in some life-cycle consumption models. • More flexible modeling that allows differences in both preferences and on-the-job risks according to age group results in estimates of VSL that decline less in older workers than was found in some of the earlier analyses. • None of the wage-risk results support the assumption of a constant VSLY. Implications of Results Regarding Age for VSLY The weight of the evidence suggests that there is insufficient information to make a reasoned decision to stop using a single VSL and switch to either dif- ferent VSLs for different ages, a constant VSLY, which imposes a linear (dis- counted) relationship between life years remaining and the VSL, or a VSLY that varies with age groups. Nailing down these effects will take large samples and thorough studies. What seems clear from the stated-preference literature is that in a private-goods context, there is no justification for use of a constant VSLY that is applied to all life years for all age groups. None of the study results are consistent with the assumptions that underlie a constant VSLY. These results do not rule out non-constant VSLs or VSLYs. EPA’s SAB (2007) reached similar conclusions and recommended against quantitative adjustments to VSL estimates by age because of inconsistencies in the available empirical results. However, it urged EPA to support more research on the question and argued that a variation in VSL with age should not be ruled out on theoretical grounds. It also recommended against a constant VSLY, not- ing that the empirical evidence does not support the assumption that VSL is pro- portional to remaining life expectancy, which is an implicit assumption when a constant VSLY is used. We stress that the wage-risk studies discussed here do not include large numbers of people over 65 y of age, because many of them are not in the work-

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 151 force. Thus, it is difficult to draw conclusions about how the VSL may differ in those over 65 y old from the wage-risk evidence because this would necessitate extrapolating beyond the age range in the studies. Some analysts have advocated that EPA consider using quality-adjusted life years (QALYs) as a way of measuring mortality and morbidity risks in cost- benefit analyses. The QALY is a metric that blends life years saved with the quality of the life years when a person experiences acute or chronic disease. Death is generally given a value of 0, and perfect health a value of 1, although some indexes (for example, disability-adjusted life years [DALYs]) reverse this. Some QALY indices also permit scoring of health states that are “worse than death,” that is, get less than a zero score. An Institute of Medicine (IOM 2006) panel commissioned by EPA with support from OMB to examine the role of QALYs in regulatory cost-benefit analyses made several recommendations that bear on the usefulness of this metric. The IOM panel noted that QALYs provide a useful perspective on the impacts of regulatory actions but that this is only one perspective. Other perspectives are also useful, such as monetary-benefit meas- ures. The IOM panel was not asked to compare a life-years-saved metric with a lives-saved metric and did not do so. However, it did note, as has other literature (Hammitt, 2007), that cost-effectiveness with QALYs as the effectiveness measure should not be termed a “cost-utility” analysis, as in common practice, because to do so implies that the QALYs are a welfare measure (meaning, in a practical sense, that it would rank alternative projects in the same order as a benefits analysis). In fact, the conditions for this to be true are quite restrictive: for instance, life years need to be additive in the utility metric no matter who experiences them, how many are experienced by a given person, and when. The IOM report recommended strongly against a common practice used to “convert” a QALY metric (life years lost) to dollars, specifically against putting a monetary value on a QALY. The concern was about mixing the medical- practice and welfare paradigms and about how the literature on monetary values does not support treating all life years saved as equally valuable (that is, a con- stant VSLY approach). A consequence of this medical paradigm, as voiced by the committee that authored the IOM report, is that QALYs implicitly put less value on mortality reductions in older people (because their life expectancy is lower than that of younger people’s) and in those with existing chronic illness (because, owing to the quality adjustment, a life year extended for them is worth less than it would be if they were healthy). That practice is in direct conflict with the literature on WTP for mortality-risk reductions inasmuch as WTP is found to change less than proportionally with life expectancy and to be no different or even greater in the chronically ill compared with the healthy (see below). OMB requires cost-effectiveness analysis in addition to cost-benefit analy- sis in regulatory-impact analyses of major rules and suggests that cost- effectiveness analysis be considered in all cases for comparisons of alternative regulations under discussion. Cost-effectiveness analysis defines the primary goals of the regulatory effort in some unit of measure that is not monetized. Costs of the alternatives are then compared on the basis of the cost per unit of

OCR for page 128
152 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits benefit expected. For mortality risk, OMB notes that the units may be lives saved, life years saved, or QALYs, but it does not recommend one metric over the others. Unlike cost-benefit analysis, cost-effectiveness analysis does not al- low calculation of net benefits, but it can help in selecting the most efficient regulatory alternative for achieving a given goal. Thus, it is most useful in com- paring programs with goals that can be quantified with the same metric. When there are multiple important benefits of a regulatory effort and those other bene- fits can be monetized, they can be netted out from costs first, for a net cost- effectiveness analysis. Hubbell (2006) illustrates an attempt at cost-effectiveness analysis using QALYs for an air-pollution control rule (the heavy-duty engine and diesel-fuel rule). He included monetary valuation with a non-constant value per life year as an illustration but noted that there is a great deal of controversy regarding the appropriate way to infer welfare-based monetary values for VSLYs from the available VSL literature. Empirical Evidence on Effect of Health Status on WTP to Reduce Mortality Risk Few empirical studies have examined the effect of health status on WTP to reduce mortality risk. Alberini et al. (2004) and Krupnick et al. (2002) provide the most thorough examination of the question in stated-preference studies. They used the same survey instrument in the United States and in Canada and looked at the potential effect of several health-status measures on WTP re- sponses, including family history of noncancer chronic disease, family history of cancer, personal history of hospitalization for heart or lung illness, personal his- tory of high blood pressure, and widely used indices of physical and mental health (drawn from the SF-36 questionnaire). At times, a history of illness, ei- ther personally or in the family, was associated with higher WTP to reduce mor- tality risk, although not all the health measures were statistically significant and the significant measures differed between the U.S. and Canadian studies. Underlying the question of the effect of health status on WTP to reduce mortality risk is the concern that people who are chronically very ill may have such a reduced quality of life that they do not attach as much value to a policy that would extend life. The studies noted above do not address that question well, because the samples are drawn from the general population and include few very ill people. One small study that sheds a little light on the question is by Tsevat et al. (1998). They asked hospitalized patients over 80 y old and their caregivers and family members to estimate tradeoffs that they would be willing to make between improving quality of life and extending life. The authors were surprised to find that patients put a higher value on extending life (relative to improving quality of life) than did the caregivers and family members. That suggests that the WTP to reduce mortality risk is not necessarily decreased and may be increased when quality of life is diminished.

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 153 Evidence on Effect of Risk Latency on WTP There is little evidence on the effect of risk latency on WTP to reduce mortality risk, because few empirical studies have examined it. As illustrative of this literature, Alberini et al. (2006a) used results of their contingent-valuation studies in Canada and the United States to investigate the issue. The latent-risk question, which asked respondents under 60 y of age whether they were willing to pay some stated amount for a given reduction in risk starting at the age of 70 y and lasting 10 y, followed two questions on WTP for immediate risk reduc- tions. They found that delaying the time at which the risk reduction would occur by 10-30 y reduced WTP by more than 60% in respondents in both samples of people 40-60 y old. The implicit discount rates are equal to 3.0-8.6% in Canada and 1.3-5.6% in the United States, in contrast with earlier estimates of the dis- count rate in risk-reduction tradeoffs, which ranged from 0.3% (Johannesson and Johansson 1996) to 14% (Viscusi and Moore 1989). Studies to Estimate WTP for Increases in Life Expectancy Analysts have pointed out that quantifying the effect of changes in pollu- tion exposure as annual changes in numbers of deaths in a population obscures the dynamic dimension of the change in exposure and its overall effect on mor- tality over time (e.g., Brunekreef et al. 2007). A more comprehensive way to describe the effect is as a shift in the survival curve over time. For the individ- ual, that means an increase in the probability of surviving over all future periods and thus extending life expectancy. Of course, death is not prevented for the individual; rather, future survival probabilities increase, and death is more likely to occur at a later age. This can be summed up as a change in life expectancy, but it may also be described as a change in annual mortality rates over some extended period. Chapter 4 notes some unresolved issues about how best to characterize the change in mortality risk in the population as a result of a reduc- tion in pollution exposure. In particular, several authors have raised concerns that the number of deaths prevented is not a stable measure over time (e.g., Miller and Hurley 2003; Rabl 2006) and have argued that a better measure would be the increase in life expectancy or life years saved. Economic valuation of the reduction in mortality rate can be derived directly from available WTP studies, which provide WTP values to reduce annual mortality risks. However, economic valuation of changes in life expectancy are not as straightforward and, with stated-preference approaches, are difficult to communicate to respondents, although these issues could perhaps be addressed with further empirical re- search. A few stated-preference studies have tried to estimate WTP values for changes in mortality risk that are presented to respondents as changes in life expectancy (Johannesson and Johansson 1996; Johnson et al. 1998; Morris and Hammitt 2001). To define a change in life expectancy that is the same for

OCR for page 128
154 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits everyone, those studies have defined a change in risk that starts at a future age, such as 60 or 70 y, and asked respondents what they would pay now for this future risk reduction that increases their life expectancy by some specified amount. The results are difficult to interpret. Comments from respondents sug- gest that some share of them dismiss the risk reductions as too small to be worthwhile. There appears to be a tendency to view an increase in life expec- tancy as adding time at the end of life when quality of life may be diminished rather than recognizing that an increase in life expectancy means an increase in the probability of survival in all future periods. The WTP questions seem to be additionally confounded with the presentation of a risk reduction that does not begin to have benefits until some years in the future. Morris and Hammitt (2001) made an interesting comparison between presenting a continuing annual risk reduction (to half the respondents) and an equivalent increase in life expectancy (to the other half of the respondents), both starting at the age of 60 or 70 y for a hypothetical pneumonia vaccine. They presented a 0.2% annual risk reduction starting at the age of 60 y or an 11-month increase in life expectancy (after get- ting the vaccine at the age of 60 y); for another portion of the sample, they pre- sented a 0.2% annual risk reduction starting at the age of 70 y or a 5-month in- crease in life expectancy (after getting the vaccine at the age of 70 y). The substantial lag between the payment and the beginning of the risk reduction (an average of 25 y in the sample, which had a mean age of 40 y) confounds the interpretation of the results. The WTP results for the two risk presentations were not very different in those who said they would consider paying now for the future vaccine, but respondents who received the life-expectancy version were more likely to say that they would not consider getting the vaccine (33% vs 26%). The most common explanation was that the benefit was too small. Com- pared with many risk-reduction efforts that people routinely undertake (such as cancer screening and wearing seat belts), the benefit is actually large, so the re- sponse raises some questions about the presentation. For example, Corso et al. (2001) demonstrated that visual aids are useful and needed to help respondents to understand quantitative risk information. Alberini et al. (2006b) report VSLY estimates derived from WTP values obtained in a stated-preference study conducted in France, Italy, and the UK regarding annual mortality-risk changes over a 10-y period. The VSLs for the 5- in-1,000 risk change over 10 y were a mean of 2.3 million euros and a median of 1.1 million euros. Using subjects’ own perceptions of their life expectancies, the authors calculated the implied VSLY and obtained a mean of 142,000 euros and a median of 58,200 euros. However, they found that the VSLY was significantly higher in those 60 y old and older (27% above those 40-49 y old). It is notable that the 10-y mortality-risk change converts to a different life-expectancy change in each age-sex cohort. It remains unclear whether the conversion from a WTP for the risk change to a VSLY for the associated change in life expectancy is the appropriate way to obtain this value. Analysts in Europe (Bickel and Friedrich 2005) report using results from Alberini et al. (2006b) as the basis of their estimates of VSLY in the 2005 up-

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 155 date of the method for estimating externalities of energy. They used the esti- mates after presenting their argument that it is preferable to estimate the mortal- ity effect of air pollutants on the basis of changes in life expectancy. Their se- lected central estimate of 50,000 euros per life year was derived from the median WTP value for the 5-in-1,000 mortality-risk change over 10 y (median VSL, about 1.1 million euros). The ExternE authors converted the 10-y mortal- ity-risk change to its equivalent in increased life expectancy for each age-sex cohort and calculated the VSLY implicit in the WTP responses. It is notable that their selected central estimate was based on one of the lower WTP results be- cause it was based on the median rather than the mean value and on the 5-in- 1,000 10-y risk reduction rather than the 1-in-1,000 10-y risk reduction. The VSLY based on the mean value for the 5-in-1,000 reduction was about 125,250 euros (the associated VSL was about 2.3 million euros). The authors note that the WTP questions in this study include an implicit latency in that the risk re- duction occurs over a 10-y period. Using an assumed 3% discount rate, they calculated that the undiscounted VSLY based on the median VSL would be about 75,000 euros rather than 50,000 euros. They concluded that the latency was appropriate for valuation of a reduction in mortality risk posed by long-term PM exposure. As a high estimate, they used 150,000 euros, derived from the WTP responses in the UK survey to the 1-in-1,000 10-y risk reduction (the asso- ciated VSL was about 3.3 million euros). They based their low estimate on results from the study in France, which included a WTP question specifically for changes in life expectancy. Details of the estimation approach have not been published, so it is difficult to evaluate. Desaigues et al. (2004) reported VSLY estimates ranging from 20,000 to 220,000 euros, depending on the underlying change in life expectancy implied in the different questions included in the French survey. The ExternE analysts note that the WTP question for life-expectancy increases stressed that it did not mean just a few months at the end of life: “A crucial point that needs to be ex- plained very carefully in such a questionnaire is that air pollution mortality does not cut off a few months of misery at the end of life but causes ‘accelerated ag- ing’” (Bickel and Friedrich 2005, page 44). The low estimate of VSLY selected by the authors for use in the recent ExternE assessment is a value of about 20,000 euros per year of additional life expectancy. Although the ExternE ana- lysts selected a range of VSLY estimates, they used an approach that presumes a constant VSLY, which ignores the evidence that this value is likely to vary with age. FINDINGS AND RECOMMENDATIONS Finding: The charge to this committee concerns monetary valuation of mortality-risk reduction for regulatory impact analyses that are based on the basic premises of cost-benefit analysis. In this context, therefore, we focus on WTP values for mortality-risk reduction. Cost-benefit analysis, however, fo-

OCR for page 128
156 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits cuses on economic efficiency, and many other ethical and legal factors are ap- propriate to consider in policy and regulatory decisions. In general, these issues should be considered separate from the cost-benefit analysis rather than inter- jected into the valuation estimates because this endangers the neutrality of the analysis. However, there may be some instances when such interjection is ap- propriate. For example, the decision by EPA to not adjust WTP estimates for local differences in income levels is justified because to do so creates a situation that favors greater environmental protection in wealthier locations, an outcome policy makers judge to be unfair and in many cases illegal. The committee stresses that government decision-makers need informa- tion on how the WTP for mortality-risk reductions varies by risk characteristics, population characteristics, and for both risk as a private good and as a public good. How they choose to use this information, however, is not strictly a techni- cal decision, but depends on ethical precepts, legal precedent, the quality of the evidence and other factors that may be beyond the analysts’ control or purview. Recommendation: We recommend that this finding of the committee be considered within OMB and the agencies that use monetary values for mortality- risk reductions in their regulatory analyses. These agencies should develop a plan for generating the information needed to determine how WTP varies for different populations and different risk contexts. In addition, there should be an exploration and debate to determine the appropriate uses of this information. Such a debate should go beyond economic considerations and include ethical and public policy perspectives. Finding: Both economic theory and the empirical evidence are inconclu- sive about how individuals’ WTP for reducing their own risk vary with two im- portant individual characteristics: age as a proxy for remaining life expectancy and health status. Although we conclude that the empirical evidence is insuffi- cient to support a specific quantitative adjustment to the WTP for reduction in annual mortality risk based on differences in remaining life expectancy, we do not reject the concept that such adjustment may be appropriate. It is plausible that people with less remaining life expectancy are willing to devote less of their resources to reducing their mortality risk than those with more remaining life expectancy. Characteristics of the risk that may affect individual WTP values include the type of risk (such as illness or accident) and its latency. The literature is in- conclusive about the influence of risk characteristics (see below). The effects of latency on WTP values are straightforward conceptually (the WTP for a future death-risk reduction should be less than the WTP for an equivalent immediate death-risk reduction), and there is some published support for such estimates. However, the epidemiology literature is not sufficient to estimate the degree of latency in mortality response to ozone exposure. Recommendation: Despite many concerns about the accuracy of using the same WTP value (or range of values) for all mortality-risk reductions, we recommend this, with appropriate scaling to the size of the risk change, as the

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 157 most scientifically supportable approach for monetary valuation of ozone-related mortality, given the currently available information in the economics and epi- demiology literatures. Empirical evidence of how WTP varies with population or risk characteristics is not sufficiently consistent to support a change in this practice that EPA has been using for many years. Researchers should continue to explore how WTP for mortality-risk reduction may vary with personal char- acteristics (such as age and health status), with the type of risk (such as accident and illness), and with its latency. This most likely implies more stated- preference studies to include the elderly population. Finding: The use of a constant value for life years saved in the valuation of increases in life expectancy requires the assumption that WTP values for mor- tality risk reductions be consistently declining with increasing age. The empiri- cal evidence does not support that assumption and therefore does not support the use of a constant VSLY. The literature does not reject the use of a non-constant VSLY, however. And the epidemiological literature (see Chapter 4) may favor reporting life years saved. Recommendation: Unless future research produces empirical support for the assumptions that underlie a constant VSLY, EPA should not attempt to ad- just for remaining life expectancy by calculating life years saved and applying a constant VSLY. It may be appropriate to calculate and report life years saved (in addition to reporting lives saved or changes in annual mortality rates), but it is not appropriate to use a constant VSLY as a monetary valuation of life years saved, except perhaps in a bounding exercise. The committee cautions against use of such an analysis in anything but a sensitivity analysis, however (see be- low). There is likely to be good reason to use a non-constant VSLY or a non- constant VSL, once the empirical literature is sufficient to support this transi- tion. The committee stresses, however, that the status quo of using a uniform VSL should be continued until there is sufficient empirical evidence of how WTP for mortality-risk reduction varies with differences in remaining life ex- pectancy and other factors, which the committee concludes is not yet available. Finding: Most of the revealed-preference and stated-preference studies re- lied on by EPA to obtain estimates of the VSL have estimated these values for a context (such as traffic accidents or workplace accidents) and for a population that differs, for example, by age, health status, and income, from the population facing the pollution-related risks that EPA is assessing. Applying the available estimates in EPA’s assessments in different contexts of risks (such as illness vs accident) and population characteristics (such as age) introduces considerable uncertainty about how these factors affect the average WTP values. However, the current literature is inconclusive as to how and how much the WTP values may vary with those factors. Recommendation: The use of average WTP estimates selected from the literature should reflect results from both revealed-preference and stated- preference studies, take into account the strengths and weaknesses of each ap-

OCR for page 128
158 Ambient Ozone and Mortality: Estimating Risk-Reduction Benefits proach, and consider how closely the studies match the policy context in popula- tion at risk and type of risk. Given the limited studies available for different risk contexts, it is difficult to say how much the WTP values may differ, but wage- risk studies are a poor match to the population and the risk context for the ozone-mortality case. EPA should give less weight to these studies in selecting WTP estimates than it has in the past. Finding: The direction of the expected error, if any, in using the average VSL in the literature for valuing changes in ozone-related mortality risk is more likely to be toward overstating the WTP to reduce this risk. That is because greater mortality risk associated with ozone appears to be in the elderly popula- tion and because latent risk may be involved. The remaining life expectancy in this population is substantially less than that in the population as a whole, and its WTP to reduce mortality risk might also be less. However, the lower WTP as a result of lower remaining life expectancy in the elderly may be offset to some extent by higher WTP because of poorer health status or higher baseline risk. Although results in the empirical literature are not consistent, several studies suggest that WTP to reduce mortality risk does not change or declines slightly with age. The evidence suggests that, for ozone, a proportional adjustment of the VSL for remaining life expectancy (that is, using a constant VSLY) would result in WTP that is too low. Recommendation: Given the uncertainty in available VSL estimates for ozone-related mortality, it is appropriate to conduct some sensitivity analyses with alternative estimates or assumptions. The purpose of sensitivity analyses is to see whether the overall conclusions of the cost-benefit comparison are changed, for example, whether net benefits are still positive under alternative economic-valuation assumptions. In general, there is less confidence in the sen- sitivity analyses because the alternative assumptions are more speculative than the primary assumptions or deviate from the status quo, which the committee feels puts a burden of proof on those who would overturn it. By definition, sen- sitivity analyses should be given less weight in the presentation of results. How- ever, they can be included in the summary and conclusions. The selection of alternative assumptions for the sensitivity analyses can be based on either theory or evidence. For example, different published empirical estimates of the rela- tionship between WTP for mortality-risk reduction and age could be selected as illustrative of the range of published results, including estimates of VSLY or VSL that vary with age. RECOMMENDATIONS FOR FUTURE RESEARCH The above recommendations for EPA imply research needs for estimating the WTP for changes in mortality risk and for changes in life expectancy. They also imply that the research primarily should use stated-preference methods,

OCR for page 128
Economic Valuation of Reduction in Mortality Risk 159 although researchers may find ways to address the issues with further develop- ment of the revealed-preference approach. A fundamental need is to understand better how age and remaining life expectancy affect WTP for reductions in mortality risks or increases in life ex- pectancy. One step would be to ask that future (or even previous) studies report total age effects—that is, WTP by age cohort—in addition to marginal effects of age on WTP. Given the correlation of age with some of the other factors, there may be less uncertainty in the estimates of a total age effect. However, age- related income differences, sex differences, health differences, and the like would then be embedded in the estimates, and it might not be appropriate to use different VSLs that have these effects embedded. Several recent studies (e.g., Alberini et al. 2004) have looked qualitatively or quantitatively across the valuation literature to assess age and other effects. The assessments have been hampered by issues in the reporting of information and the lack of availability of the datasets produced. Future research funded by EPA should urge that its datasets be made available for meta-analysis. More fundamental research is needed to explore and develop methods for communicating and valuing changes in mortality risk that reflect the full life cycle. Studies to date have focused on WTP for annual changes in mortality risk, but the risk change of interest in most pollution-control assessments is more comprehensively described as a shift in the survival curves, which are plots of survival probability in all future periods and from which life expectancy is de- rived. That poses a challenge for stated-preference studies. A few studies have used changes in life expectancy to define mortality-risk change (e.g., Johannes- son and Johansson 1996; Morris and Hammitt 2001), but many respondents seem to dismiss these changes as time added at the end of life, when quality of life may be substantially reduced. DeShazo and Cameron (2004) included mul- tiple periods in their presentation of risk changes in an ambitious and innovative stated-preference study; analysis of their results has not yet been published. It is also important to learn more about how mortality-risk characteristics affect the valuation of reducing risk. Environmental-benefits assessments have relied primarily on estimates of WTP to reduce risk of accidental death to esti- mate values for reducing risk of illness-related death. A few studies have com- pared values of WTP (or nonmonetary preferences) to reduce risk in different contexts (e.g., Magat et al. 1996; DeShazo and Cameron 2004), and more stud- ies along these lines are needed. Studies in a public-goods context that can isolate the effect of the age of those who benefited on the WTP of respondents for programs that reduce mor- tality risk in the community might help to resolve this issue as well. But in addi- tion to such studies, we would need approaches to distinguish paternalistic from non-paternalistic altruism (as defined in Jones-Lee 1991) if double-counting of benefits is to be avoided. Conceptual analysis would be needed to understand how such results might be appropriately used in cost-benefit analysis, in which the usual paradigm is to sum private WTPs of the beneficiaries.