Benefit-cost analysis judges health and safety regulations according to whether the monetized benefits of risk reductions, as measured by individual willingness-to-pay (WTP), exceed their costs. One way to complement such analyses is to take into account the distributional effects of health and safety policies. Our goal in this paper is to estimate the parameters of individuals’ social welfare functions (SWFs) defined over environmental health risks — specifically, risks of cancer and lung disease. We do this by confronting people with choices between environmental programs that result in higher average but more equally distributed health risks, and programs that would deliver lower average but less equally distributed health risks. We use the responses to parameterize an Atkinson SWF for cancer risks and a similar function for risks of lung disease. This SWF could be used to evaluate programs that would alter the distribution of environmental health risks in a population. The analysis also produces an inequality index (the Atkinson index) for health risks that reflects the preferences of our sample for equality of outcomes.
Our empirical estimates of public preferences for environmental health risk distributions come from a national internet survey with more than 900 completions, administered in August 2015. The survey asked respondents to choose between environmental programs that result in different mean health risks in a population and different distributions of these risks. Respondents made these choices (a) in a situation in which they (and their families) were not affected by the choices, and (b) in a situation in which they were affected, to see how this altered their preferences. We also used “leaky bucket” experiments to elicit respondents’ preferences for income inequality and a repeated coin toss question to gauge risk aversion. In addition to the base case survey, we used four alternative survey treatments to examine the effect of the scale of the risks, the nature of the health risks (lung disease versus cancer) and the effects of the order of questions on responses.
The results of our survey suggest that people are willing to accept a program that results in a higher total environmental health risk provided this risk is equally distributed across the population. Specifically, the median respondent is willing to accept a 50 percent increase in mean health risk (e.g., total environmental cancer cases) if these risks are distributed equally in the population. Interestingly, this result is the same whether the respondent and his family are affected by the program or not. When we compare preferences for income equality versus equality in the distribution of health risks, we find that the proportionate sacrifice people are willing to accept in the mean outcome to ensure equality in the distribution of outcomes is greater for health than for income: inequality aversion is higher for health risks than for income.