A Senior Research Scientist in the Centers for Risk Analysis and Health Decision Science at Harvard University, Lisa A. Robinson’s work focuses on the conduct of benefit-cost analysis and the valuation of non-market outcomes.
It is RegBlog’s fifth anniversary, and also five years since we first wrote about the U.S. Environmental Protection Agency’s (EPA’s) efforts to update the values it uses for mortality risk reductions. It is an especially good time to revisit these issues, because the EPA recently asked the Environmental Economics Advisory Committee (EEAC) of its Science Advisory Board to comment on the results of implementing the 2011 recommendations. These values are important because they dominate the benefit estimates for many environmental, health, and safety regulations, and—under Executive Orders 12866 and 13563—U.S. agencies are required to demonstrate that the benefits justify the costs of their major regulations unless a statute requires otherwise. Both the criteria the EPA uses and the values that result are likely to influence the estimates applied in analyses conducted world-wide.
The terminology used to describe these values has led to widespread confusion. The value in question is not the value the government places on saving an individual’s life. Rather, it represents an individual’s willingness to trade off spending on risk reductions against spending on all the other things she could use that money to buy. It is the population average of individuals’ rates of substitution between wealth and survival probability in a defined period, often the current year.
This rate of substitution has traditionally been measured in dollars per one unit change in probability (the value per statistical life, or VSL), but could just as easily be measured in dollars per one per million risk change (a “micromort”) or other units. For example, if an individual is willing to pay at most $900 to reduce her risk of dying in the current year by 1 in 10,000, her VSL is $9 million ($900 ÷ 1/10,000 = $9 million). To promote better understanding, the EPA is investigating alternative terminology, such as the “value of risk reductions” or the “value of mortality risk reductions.” The “value of a standardized mortality unit” has been used in some global health research.
The EPA’s current approach for estimating these values is based on a review conducted more than 20 years ago. The EPA has since completed substantial additional research and asked its independent advisory panels to review several proposals to update its values. These efforts led most recently to a detailed meta-analysis of studies that meet relatively rigorous selection criteria. The preliminary results yield a proposed central estimate of $10.3 million (in 2013 dollars and income levels); this value may change as a result of the ongoing EEAC review. In comparison, for the same year, the EPA’s current approach leads to a VSL of $9.7 million, the U.S. Department of Transportation (DOT) recommends a VSL of $9.2 million, and work conducted for the U.S. Department of Health and Human Services (HHS) suggests a VSL of $9.0 million.
The similarity of these values raises the question of whether the VSL should be standardized across federal agencies, especially since each agency is drawing from the same body of research. Both the current EPA proposal and the DOT and HHS values rely on the use of selection criteria that reflect the influence of the EPA’s 2011 review and previous work, suggesting that it may be possible to reach consensus on a set of criteria for evaluating the quality and applicability of available estimates.
Greater standardization would result in more comparable analytic results, allowing decision-makers and others to distinguish more clearly the differences in regulatory impacts across regulations without the potential confusion caused by the application of different VSLs. Of course, such standardization could also include decisions on when differentiation might be appropriate. For example, as more research becomes available, it may become possible to differentiate the values used for risks of different types.
We have clearly made significant strides in determining how to best synthesize and apply the available research. But how much progress have we really made? Synthesis of existing studies will only get us so far. Each agency applies its recommended population-average value across all of its regulations. But what if a regulation affects only the very young or the very old? Or addresses a particularly fearsome risk, such as cancer or terrorism? We know relatively little about the effects of various individual and risk characteristics on these values, and the information that we have is often inconsistent. More primary research is needed to tailor the estimates to differences in the populations and risks addressed.
In addition, we have a limited understanding of the value of reductions in nonfatal health risks. Few U.S. studies are available that address individuals’ willingness to pay for reductions in different types of morbidity. While we generally assume that these values are significantly smaller than the VSL, even for relatively severe effects, more research is needed to confirm this belief. In addition, small unit values can become quite large totals if multiplied by a sizable affected population.
Benefit-cost analysis is intended to encourage evidence-based policy decisions, providing information on the preferences of the affected individuals. All of the reviews noted above have indicated the need for new primary research to promote better understanding of these preferences, but agencies have not made a concerted effort to fund this work. Although there are many pressing research needs, as a society we have made a significant commitment to using benefit-cost analysis to support our major regulatory decisions. More research is needed to help fulfill this commitment.
This essay is part of RegBlog’s sixteen-part series, RegBlog@5.