Most current environmental policy analyses use Value of a Statistical Life (VSL) figures inferred from workplace safety and traffic accident contexts to compute the benefits of environmental programs that avoid premature deaths. There is considerable debate about the appropriateness of this practice, in part because the effect of cause of death may be partly confounded with latency, initial risks, and competing risks. Preference for reducing risks can be also affected by individual-assessed risk attributes that are rarely controlled in valuation studies. This paper explores reasons for differences in preferences for mortality risk reductions (if any), and establishes the magnitude of the effects of such risk attributes as compared to other sources of VSL heterogeneity. In our conjoint choice experiments, cause of death, the size of the risk reduction, and latency, the “price” of the risk reduction and the mode of delivery of the risk reduction are explicit attributes of the alternatives to be examined by the respondent. Our statistical models also control for actual and perceived exposure to risks, initial risks, risk attributes such as dread, and sensitivity to and controllability of specific risks. We find that there is significant heterogeneity in the valuation of mortality risks and thus in the VSL. The VSL increases with dread, exposure to risk, and the respondents' assessments of the baseline risks. It is higher when the risk reduction is delivered by a public program, and increases with the effectiveness rating assigned by the respondent to the mode of the risk reduction. Even when we control explicitly for all of these factors, the cause of death per se accounts for a large portion of the VSL. All else the same, the fact that the cause of the death is “cancer” results in a VSL that is about one million euro above the amount predicted by dread, exposure and other risk perception variables. The VSL in the road safety context is about one million euro less than what is predicted by dread, exposure, or beliefs compared to VSL for the respiratory risk context. The effect of cause of death is thus as large as the effect of other sources of VSL heterogeneity. Our respondents do not seem to discount future risks.
