What should we spend to save lives in a pandemic? A critique of the value of statistical life

The value of statistical life (VSL) is a risk-to-money conversion factor that can be used to accurately approximate an individual’s willingness-to-pay for a small change in fatality risk. If an individual’s VSL is (say) $7 million, then she will be willing to pay approximately $7 for a 1-in-1-million risk reduction, $70 for a 1-in-100,000 risk reduction, and so forth. VSL has played a central role in the rapidly emerging economics literature about COVID-19. Many papers use VSL to assign a monetary value to the lifesaving benefits of social-distancing policies, so as to balance those benefits against lost income and other policy costs. This is not surprising, since VSL (known in the U.K. as “VPF”: value of a prevented fatality) has been a key tool in governmental cost-benefit analysis for decades and is well established among economists. Despite its familiarity, VSL is a flawed tool for analyzing social-distancing policy—and risk regulation more generally. The standard justification for cost-benefit analysis appeals to Kaldor- Hicks efficiency (potential Pareto superiority). But VSL is only an approximation to individual willingness to pay, which may become quite inaccurate for policies that mitigate large risks (such as the risks posed by COVID-19)—and thus can recommend policies that fail the Kaldor- Hicks test. This paper uses a simulation model of social-distancing policy to illustrate the deficiencies of VSL. I criticize VSL-based cost-benefit analysis from a number of angles. Its recommendations with respect to social distancing deviate dramatically from the recommendations of a utilitarian or prioritarian social welfare function. In the model here, it does indeed diverge from Kaldor- Hicks efficiency. And its relative valuation of risks and financial costs among groups differentiated by age and income lacks intuitive support. Economists writing about COVID-19 need to reconsider using VS