Final Report: Mortality Risk Valuation And Stated Preference Methods: An Exploratory StudyEPA Grant Number: R824711
Title: Mortality Risk Valuation And Stated Preference Methods: An Exploratory Study
Investigators: Krupnick, Alan J. , Alberini, Anna , Belli, Robert , Cropper, Maureen , Simon, Nathalie
Institution: Resources for the Future , University of Colorado at Boulder , University of Maryland and World Bank , University of Michigan
EPA Project Officer: Lee, Sonja
Project Period: October 1, 1995 through September 30, 1997
Project Amount: $114,822
RFA: Valuation and Environmental Policy (1995) RFA Text | Recipients Lists
Research Category: Economics and Decision Sciences
Summary/Accomplishments (Outputs/Outcomes):I. Introduction
Recent analyses of the benefits and costs of environmental regulations, such as the U.S. Environmental Protection Agency's (USEPA) Retrospective Cost-Benefit Analysis of the 1970 Clean Air Act and the USEPA's Regulatory Impact Analyses for Ozone and Particulates pivot on the estimates of the benefits from reducing mortality risks. To compute such benefits, analysts have relied on estimates of the value of a statistical life (VSL), which in turn are derived from an individual's willingness to pay (WTP) to secure a beneficial change in risk, or willingness to accept (WTA) to forgo an adverse change in risk.
In practice, estimating WTP or WTA for changes in mortality risk is often problematic. First, reductions in the mortality risks of interest are usually not traded in regular markets. This suggests that WTP for a reduction in those risks be elicited using stated preference approaches, such as contingent valuation (CV). However, in recent CV surveys , WTP for reductions in mortality risks was often found to be inconsistent with predictions from standard economic theory. In particular, for small risks, WTP appears to be insensitive to the order of magnitude of the risk. This problem has been attributed to respondent difficulty in understanding the nature of probability. With few exceptions , CV surveys about mortality risk reductions have focused on accidental death.
Even when it is possible to observe actual choices involving tradeoffs between risks and income, as is often done in wage-risk and other revealed preference studies [3, 4], it is difficult to disentangle the risk attribute from other attributes of the commodity or action under investigation. This may be one reason for the wide range ($100,000 to several millions of dollars) of VSL estimated from the wage premia paid to workers bearing the risk of workplace accidents.
Moreover, it is often argued that mortality risks associated with pollution fall disproportionately on the elderly, the poor, and persons with compromised health. This suggests that it may be inappropriate to apply to these persons VSL amounts estimated from wage-risk studies, where the population under observation is, for the most part, males in their prime age who voluntarily accept workplace risks.
Perhaps the most important shortcoming of the estimates of VSL from compensating wage studies and the available CV literature [e.g., 5] is that they are based on accidental death risks. By contrast, exposure to pollution can result in delayed effects, with cancer or other disease developing years after exposure to the offending substance, or in mortality risk changes that apply only to a person late in life. Deaths from disease can be slow and painful, perhaps preceded by a period of time in which the individual's quality of life is significantly decreased. In the presence of latent and late-in-life risks, the current age of an individual and the age at which the individual is to experience the risk reduction brought about by an environmental policy should be important determinants of this individual's WTP to secure such risk reductions. If these factors latency, late-in-life risks and death type influence perception of risks, and perception of risks in turn affects the amount an individual is prepared to pay for a reduction in those risks, the estimates of VSL currently available are probably not applicable to contexts involving pollution.
Unfortunately, such relationships and concepts have not been adequately explored in the previous valuation literature. In this paper, we report on an on-going research project in which we attempt to explore these issues using the method of contingent valuation. We do so by creating a survey instrument that elicits WTP for risk reductions to be incurred over 10 years -- effective immediately -- and for reductions in the probability of dying between age 70 and 80.
The paper is organized as follows: section 2 briefly discusses the literature on which we build; section 3 describes the survey instrument we use to elicit WTP for mortality risk changes; section 4 discusses the results obtained from two recent pretests conducted in the U.S. and in Japan and section 5 concludes.
II. Using Surveys to Value Risk Reductions.
Absent transactions in regular markets in which people's preferences about risk can be observed, researchers have resorted to stated-preference, survey-based approaches to value mortality risk reductions. Contingent valuation is an increasingly popular if controversial method. In a typical CV survey, respondents are confronted with a hypothetical policy that would bring about a change usually an improvement -- in the present level of environmental quality, public good or risk. Respondents are asked to report their maximum willingness to pay to secure this improvement. Average WTP, multiplied by the relevant population, yields an estimate of the benefits of the proposed policy.
In most recent CV surveys, the payment question has been asked using a "dichotomous choice" format, calling for a simple "yes" or "no" answer. Respondents are asked whether they would pay $X for the change, or whether they would vote in favor of or against the policy, should approval result in a cost of $X to the respondent's household. The amount $X is varied across respondents.
CV surveys have occasionally been used to place a value on the mortality risk reductions to be attained from environmental, transportation safety and health programs[1,2,5,6]. These studies have found that while many respondents report positive WTP amounts to secure such risk reductions, a considerable fraction of the respondents is likely to have WTP equal to zero. Some respondents fail to grasp the basic notions of probability, and others ascribe similar WTP amounts to grossly different risk reductions. With few exceptions [2,6], most of these studies have dealt with accidental death risks, as opposed to risks involving latency or late-in-life risk. Recent attempts to value extensions to life expectancy conditional on survival to later ages  have been fraught with problems including the use of telephone surveys, which preclude the use of visual aids, ill-defined commodities and implausibly large risk reductions .
Our work attempts to fill some gaps in the valuation literature by developing two survey instruments to elicit WTP for an immediate reduction in risk and for a reduction in risk to be incurred at a later age. In contrast to earlier studies, our survey instruments are administered in-person, and make extensive use of visual aids. The first instrument is devised to elicit WTP for an immediate reduction in risk, and for a reduction in risk to be incurred at a later age. The second survey instrument is devised to elicit WTP for extensions in life expectancy, emphasizing that an extension in life expectancy can be obtained through many different "paths" in the probability of living past a given age, assuming that the individual has survived to such age in the first place. We do so in order to explore whether individuals can deal more easily with valuing time than with valuing small changes in probabilities. While the life expectancy survey is still in an early, developmental stage, the mortality questionnaire is currently being pre-tested. We describe both questionnaires in more detail below and report preliminary results from two pre-tests of the mortality risk survey instrument.
III. The Mortality Risk Questionnaire
A. Structure of the Questionnaire
Throughout our survey, we are motivated by two important concerns: (1) that respondents find the commodity to be valued meaningful, and (2) that they accept that mortality risks can be mitigated at a cost and that many people, if not themselves, perform such mitigation as part of everyday life. The first section introduces probabilities and proposes simple practice questions to familiarize the respondents with these concepts. The second section presents respondents with leading age- and gender-specific causes of death and introduces common risk-mitigating behaviors. We then educate our respondents about the costs associated with these behaviors and move on to the valuation part of the survey. The fourth section then elicits information about WTP for risk reductions of a given magnitude, taking place at a specified time. This is followed by an extensive debriefing section.
We begin the WTP section by introducing a base-line risk tailored to the respondent's age and gender. The respondent is asked how he would feel if he learned that his chances of dying over the next 10 years were a specified amount and is asked to accept this risk as his own. The respondent is then asked to consider two risk reductions occurring over the next 10 years. The first risk reduction reduces the baseline risk by 5 in 1,000 while the second change reduces risk by only 1 in 1,000. Before asking the his WTP, the respondent is asked to create his own base-line risk graph by darkening squares on a grid. To value each of these risk reductions, we use the double-bounded approach .
Our WTP questions differ from those in earlier CV surveys in four respects. First, they differ in the timing of the risk reductions, and, second, in the attention given to the timing of the payment. Third, the baseline risk is tailored to the individual according to age and gender group. Fourth, we depart from the recent CV literature and the NOAA panel recommendations  in that, instead of specifying in detail the payment vehicle and policy under which the risk reduction is to be delivered, we keep the circumstances surrounding the risk reduction extremely abstract.
We argue first, that people in a somewhat different setting (where they are asked to choose between two programs that save different numbers of lives at different time periods , are willing and able to make choices in an abstract setting. In addition, we argue that being specific about the attributes of the risk and mitigation approach may lose as many people as it gains because some respondents will not believe that the specifics apply to them. While we do provide the respondent with some specific examples of mitigating activities that could produce the risk reductions in question, we emphasize that the activity could take any number of forms, allowing respondents to focus on the risk reduction itself. To ensure that the respondent understands the implications of the risk reduction in question, he is now asked to erase the appropriate number of squares shaded earlier.
We have chosen to use an annual payment made over 10 years because the risk reductions to be valued are considered "large" for any one year (1 in 1,000 and 5 in 1,000). Obtaining risk reductions of these magnitudes would require repeated actions over time from the respondent. We ask respondents about risk reductions of two sizes to see if they are able to differentiate between small probabilities and to see if the value of a risk change increases with the size of the reduction.
Our final series of dichotomous choice questions focuses on future risk reductions. These questions are especially important for valuing environmental improvements related to conventional air pollutants and carcinogens, since the benefits accruing to the population would tend to occur later in life as reduced deaths in any given year from heart disease, chronic obstructive pulmonary disease, and cancers. The WTP questions are preceded by a question concerning the chances of surviving to future ages. Specifically, the respondent is asked to assess his chances of living to age 70. This question encourages the respondent to think about his future. A variety of surveys [7,12] have shown that individuals are reasonably good at recreating life table information of this sort and are able to value risk changes occurring in the future.
The respondent is then shown a risk graph with the chances of dying between ages 70 and 80 for the average person and is asked, through a series of dichotomous choice questions, his willingness to pay today for a future risk reduction beginning at age 70 of 5 in 1,000 (from 360 to 355 in 1,000). The reduction is indicated on the risk graph using darker colors, and is fully exogenous to the respondent. Finally, the strength of the respondent's conviction is checked by the remaining questions. The respondent is reminded that there is a chance he may not survive to age 70, making a payment today useless. He is then given the opportunity to revise his bid. The final question asks how certain the respondent is that he would be willing (or not willing) to pay the amount stated. A detailed series of debriefing and demographic questions follow the main body of the questionnaire.
B. Evidence from Two Pre-tests
Thus far we have developed and refined the mortality risk questionnaire based on a total of 27 personal, "think-aloud" interviews lasting approximately one hour each and are in the process of completing a 60-person pre-test of the survey instrument. In addition, the Fuji Research Institute, a non-profit research group receiving funding from Japan's Ministry of International Trade and Industry (MITI), is employing a very similar questionnaire in Tokyo, developed using the same protocols. The principal difference between the Japanese version and that used domestically is that the Japanese questionnaire uses a dichotomous choice format with two follow-up questions. The Japanese instrument has been pre-tested on a total of 316 individuals in Tokyo. The primary objective of this research was to develop a viable survey instrument capable of eliciting reasonable and consistent measures of WTP for mortality risk reductions. While administering a full survey using this new questionnaire falls outside of the scope of this project, the preliminary, small-sample results of both pre-tests are quite interesting.
1. Results from the U.S.
The Survey Research Center at the University of Maryland has conducted a number of interviews using the survey instrument described above with respondents from the College Park, MD area. Using maximum likelihood estimation techniques, we fit a series of distributions to the double-bounded WTP responses for each of the three risk categories, including the normal, log normal, Weibull, exponential, logistic and log logistic distributions. The mean and median WTP estimates and the distribution with the best fit for each of these categories appear in Table 1 below. Assuming a 5 percent discount rate, we find VSL estimates ranging from $132,000 to $402,000.
According to these preliminary results, the mortality risk questionnaire shows exceptional promise. Although the results violate the strict proportionality of risk and utility one would expect from standard neoclassical expected utility theory, all observations passed the internal scope test in that respondents were willing to pay more for larger risk reductions than for small ones. Furthermore, although most respondents felt the baseline risk was too high, they accepted the quoted risk for the purposes of the survey. Also, all subjects, except those answering no to both dichotomous choice questions, felt that paying today for a product that would not affect one's health until age 70 was reasonable. The sample size at this stage, however, precludes the use of statistical tests.
Table 1: Preliminary Evidence from the U.S.
|Commodity||Median WTP||Std Error||Mean WTP||Best Fit||Implied VSL1|
|5 in 1,000 risk change over next 10 years||$197||$86||$367||Log Normal||$304,000|
|1 in 1,000 risk change over next 10 years||$52||$35||$220||Log Normal||$402,000|
|5 in 1,000 risk change from age 70 to age 80||$86||$52||$214||Wei bull||$132,800|
2. Results from Japan
While we are only just beginning the analysis of the data from the Japanese pretest, the preliminary results are equally encouraging. The survey was administered to 316 residents in the Tokyo area, a large enough sample to allow statistical inference. Out of the 316 completed interviews, only 6 failed the internal scope test and almost 80 percent of the respondents accepted the baseline risk as their own. Of those who were not able to accept the baseline risk, 57 percent felt their risk was higher than that given. When questioned about whether it is reasonable to pay today for a risk reduction that would not occur until age 70, over half of the subjects responded "no" or "don't know". Using the same estimation techniques described above, we find VSL amounts ranging between $40,000 and $386,000 (see Table 2 below).
Table 2: Preliminary Evidence from Japan
|Commodity||Median WTP||Standard Error||Best Fit||Implied VSL1|
|5 in 1,000 risk change over next 10 years||$113||$13||Weibul l||$175,000|
|1 in 1,000 risk change over next 10 years||$50||$6||Weibul l||$386,000|
|5 in 1,000 risk change from age 70 to age 80||$22||$4||Weibul l||$40,000|
IV. The Life Expectancy Questionnaire
Our second survey instrument is devised to elicit WTP for extensions in life expectancy. Unlike previous studies , it emphasizes that a given extension in life expectancy can be attained through many different "paths" in the probability of living past a given age, assuming that the individual has survived to such age. With this survey instrument, we explore whether individuals can more easily value time than value small changes in probabilities of dying.
After introducing the respondent to the notion of life expectancy and providing a formal definition of the term, the respondent is asked to estimate the probability he or she will live to age 70, 80 and 90. We then introduce the respondent to survival functions and ask the respondent to familiarize himself with this concept by extracting information from a graph depicting the survival curve for the average person in his age and gender group. Specifically, we ask the respondent to find the probability that an average person his age survives to age 70, 80, and 90. The respondent is then given the opportunity to revise his own probability estimates of surviving to each of these ages. We then explain the relationship between life expectancy and the survival curve to the respondent by showing him that life expectancy is simply the area under the curve. This is followed by a series of practice problems in which the respondent is asked to compare a series of graphs and determine which one shows the highest life expectancy.
Once the respondent has mastered the notion of life expectancy, we discuss actions, such as an immunization shot or diet that can change life expectancy by changing the position of the survival curve beginning at different ages. We ask the respondent to identify the graph that best shows the effect of an action that improves one's chances of surviving to every future age beginning today, beginning in 10 years and beginning at age 70. In the next part of the survey we elicit WTP for a specified extension in life expectancy including extensions of three days, 2 weeks, one month, and two months. As in the mortality risk questionnaire, the intervention that would lengthen the respondent's life expectancy is rather abstract. Finally, we ask respondents to consider life expectancy changes that would be effective starting at age 70, and close with debriefing and demographics questions.
Mortality risk reductions associated with improvements in air quality are not easily valued. These mortality risks are generally realized later in life. Only one study to date has been able to incorporate this characteristic. In addition, CV studies of mortality risk present convincing evidence that small changes in probabilities are not being successfully communicated to respondents. Our work, however, seems to successfully bridge this gap in the literature. Not only have we developed two survey instruments that focus on mortality risks realized in the future, but the questionnaires are administered in-person with extensive use of visual aids. Furthermore, tests of cognition are imbedded in the instruments. Preliminary results from the pre-tests of the mortality risk questionnaire indicate that individuals are able to distinguish between different magnitudes of small probabilities and are able to make judgements on future risks. The life expectancy questionnaire is in a think-aloud stage and has not been subjected to extensive pre-testing. We intend to continue testing and refining both questionnaires and eventually to administer them to full samples of randomly selected individuals.
- Graham, John D. and James K. Hammitt (1996), "Willingness to Pay for Health
Protection: Insensitivity to Probabilities?" HSPH, Cambridge, MA, July.
- Smith, V. Kerry and William H. Desvousges (1987), "An Empirical Analysis of
the Economic Value of Risk Changes," JPE 95, 89-113.
- Viscusi, W. Kip (1993), "The Value of Risks to Life and Health," JEL, XXXI,
- Blomquist, G. (1979), "Value of Life Saving: Implications of Consumption
Activity," JPE, 87, 540-58.
- Jones-Lee, M.W., M. Hammerton and P.R. Phillips (1985), "The Value of
Safety: Results of a National Sample Survey," EJ, 95, 49-72.
- Mitchell, R. C. and R. T. Carson (1986), "Valuing Drinking Water Risks
Reductions Using the Contingent Valuation Method: A Methodological Study of
Risks from THM and Giardia," Report CR810466-01-6 prepared for USEPA,
- Johannesson, M. and P. Johansson (1996), "To Be, or Not to be, That Is the
Question: An Empirical Study of the WTP for an Increased Life Expectancy at an
Advanced Age," Jour Risk Unc, 13, 163-174.
- Krupnick, A., et al. (1997), "Mortality Risk Valuation and Stated Preference
Methods: An Exploratory Study," paper prepared for the Southern Economic
Association meetings, November 21-23, Atlanta, GA.
- Hanemann, W. M., J. B. Loomis and B. Kanninen (1991), "Statistical
Efficiency of Double- bounded Dichotomous Choice Contingent Valuation," AJAE,
- Arrow et al. (1993), "Report of the NOAA Panel on Contingent Valuation,"
Federal Register 58, 4601-14.
- Cropper, M. L., S. K. Aydede and P. R. Portney (1994), "Preferences for Life
Saving Programs: How the Public Discounts Time and Age," Jour Risk Unc, 8,
- Hurd, M. and K. McGarry (1996), "The Predictive Validity of the Subjective Probabilities of Survival in the Health and Retirement Survey," Econ. Dept. SUNY Stony Brook, March.