Against 'Individual Risk': A Sympathetic Critique of Risk Assessment

Abstract

"Individual risk" currently plays a major role in risk assessment and in the regulatory practices of the health and safety agencies that employ risk assessment, such as EPA, FDA, OSHA, NRC, CPSC, and others. Risk assessors use the term "population risk" to mean the number of deaths caused by some hazard. By contrast, "individual risk" is the incremental probability of death that the hazard imposes on some particular person. Regulatory decision procedures keyed to individual risk are widespread. This is true both for the regulation of toxic chemicals (the heartland of risk assessment), and for other health hazards, such as radiation and pathogens; and regulatory agencies are now beginning to employ individual risk criteria for evaluating safety threats, such as occupational injuries. Sometimes, agencies look to the risk imposed on the maximally exposed individual; in other contexts, the regulatory focus is on the average individual's risk, or perhaps the risk of a person incurring an above-average but nonmaximal exposure. Sometimes, agencies seek to regulate hazards so as to reduce the individual risk level (to the maximally exposed, high-end, or average individual) below 1 in 1 million. Sometimes, instead, a risk level of 1 in 100,000 or 1 in 10,000 or even 1 in 1000 is seen as de minimis. In short, the construct of individual risk plays a variety of decisional roles, but the construct itself is quite pervasive.

This Article launches a systematic critique of agency decisionmaking keyed to individual risk. Part I unpacks the construct, and shows how it invokes a frequentist rather than Bayesian conception of probability. Part II surveys agency practice, describing the wide range of regulatory contexts where individual risk levels are wholly or partly determinative of agency choice: these include most of the EPA's major programs for regulating toxins (air pollutants under the Clean Air Act, water pollutants under the Clean Water Act and Safe Drinking Water Act, toxic waste dumps under the Superfund statute, hazardous wastes under RCRA, and pesticides under FIFRA) as well as the FDA's regulation of food safety, OSHA regulation of workplace health and safety risks, NRC licensing of nuclear reactors, and the CPSC's regulation of risky consumer products.

In the remainder of the Article, I demonstrate that frequentist individual risk is a problematic basis for regulatory choice, across a range of moral views. Part III focuses on welfare consequentialism: the moral view underlying welfare economics and cost-benefit analysis. I argue that the sort of risk relevant to welfare consequentialism is Bayesian, not frequentist. Part IV explores the subtle, but crucial difference between frequentist and Bayesian risk. Part V moves beyond cost-benefit analysis and examines nonwelfarist moral views: specifically, safety-focused, deontological, contractualist, and democratic views. Here too, I suggest, regulatory reliance on frequentist individual risk should be seen as problematic. Part VI argues that current practices (as described at length in Part II) are doubly misguided: not only do they focus on frequentist rather than Bayesian risk, but they are also insensitive to population size.

In short, the Article provides a wide ranging, critical analysis of contemporary risk assessment and risk regulation. The perspective offered here is that of the sympathetic critic. Risk assessment itself - the enterprise of quantifying health and safety threats - represents a great leap forward for public rationality, and should not be abandoned. Rather, the current conception of risk assessment needs to be reworked. Risk needs to be seen in Bayesian rather than frequentist terms. And regulatory choice procedures must be driven by population risk or some other measure of the seriousness of health and safety hazards that is sensitive to the size of the exposed population - not the risk that some particular person (whatever her place in the exposure distribution) incurs.