Estimating the Net Benefits of Environmental, Public Health and Safety Regulations

Collectively, U.S. federal regulations to improve human health, safety, and the environment avert many thousands of deaths, and reduce non-fatal harms by hundreds of thousands of cases, but also impose costs up to billions of dollars annually. To become legally binding, almost all high-cost regulations must pass a "cost-benefit test": the monetized value of the harms averted by the rule should exceed (or at least “justify”) costs to regulated industries and consumers. Cost-benefit analysis (CBA) has been controversial for its ethical underpinnings and for its need to estimate risks and costs with incomplete data, but ironically, the mechanics of how to quantify effects of both risk reduction and cost on human welfare has received less attention. This project presumes that two changes to current methods of how society values benefits and costs might change which regulations pass or fail the cost-benefit test, and alter the optimal level of stringency for any given regulation. First, typical estimates of the value of a statistical life deliberately exclude all consideration of altruism or “shared purpose,” using only the estimated private value of reducing a small risk of one's own death to value public programs benefiting the entire nation. Secondly, CBA implicitly dictates that the total benefit of a program is proportional to the number of lives saved, regardless of whether some people face much higher mortality risks, and CBA also considers only the regulation's total cost, even if costs affect some businesses or consumers disproportionately. 

Several survey experiments probe these simplified assumptions and offer principled, quantitative alternatives. The researchers estimate the "social benefit of one life prolonged” (including altruism and shared purpose) as a complement to conventional willingness-to-pay estimates for private mortality risk reduction.  They do so by querying subjects about the perceived desirability of hypothetical regulations in which the scale of tradeoffs is billions of dollars in costs shared by everyone and hundreds of randomly saved lives, not a personal tradeoff between a few dollars and a tiny fraction of one life. These tradeoffs are posed to elicit from each respondent either a user-defined acceptable cost for a fixed number of lives saved or a user-defined acceptable number of lives saved for a fixed regulatory cost; one experiment asked for both tradeoffs from each person. These respondent choices are then converted into an imputed value of the social benefit of prolonging one human life. Other experiments probed whether responses improved with contextual information about other life-prolonging benefits and costs, or with putting the regulatory impacts in more familiar terms (e.g., as per person costs), and if the fixed-lives tradeoff was preceded by eliciting the value of prolonging a single life (unit asking). Experiments also tested whether the imputed value of prolonging a single life differs when the magnitudes of the fixed values vary for the same person (for example, one person offers three tradeoffs, for either 10, 100, or 1,000 lives prolonged, or $100 million, $1 billion, or $10 billion), and isolated the effects of "paternalistic" altruism (concern for others' longevity even if those others might prefer greater risk at less regulatory cost) versus "nonpaternalistic" altruism (considering others' net benefits including their costs). Two other survey experiments tested the assumption that individual levels of risk and cost can be summed over the population without being disaggregated (i.e., that individuals regard the welfare effects of risk or cost at any level as linear), using subjects' ratings of how dire they view varying hypothetical individual probabilities of harm and varying personal costs. This experiment reveals whether there are de minimis levels of either risk or cost (those that can sensibly be rounded to zero), as well as intolerably high levels whose effects are not merely proportional to those at lower levels. If the relationship between risk (or cost) and its magnitude is non-linear, then the fundamental axiom of CBA (that the distribution of individual effects about the average risk or cost does not matter) may be invalid.