In an essay published earlier this year in the Journal of Benefit-Cost Analysis, former OIRA director Cass Sunstein grapples with a tough question: should the value of a statistical life vary for different populations?
The value of a statistical life is a key tool analysts use to understand the economic implications of regulatory actions at the federal level. Sunstein calls the value of a statistical life (VSL) “the workhorse of cost-benefit analysis,” saying “it is the principal driver of benefits in multiple domains, whether we are speaking of highway safety, road safety, food safety, cigarettes, or pandemics.”
The value of a statistical life has a long history. The statistic was first an estimate of how much it would cost to replace a worker. It later evolved into the sum of future earnings of an individual. Its current approach is drawn from risk of death reductions based on the relative safety of workplaces.
The modern method is focused on labor market data. Economists use the relative danger of different workplaces combined with wage data to see what wage premium is paid to people to take on additional risk of death. For instance, a welder who works on skyscrapers is usually paid more than one who works in a shop, all other things being equal. This extra payment represents a market for risk of death reduction, where workers will take lower pay for lower risk of death and will require higher pay for higher risk of death.
One complication of this approach is that lower-income people have less money, so they are willing to pay less to reduce their risk of death than upper-income people. This is because minute reductions in risk of death could cost the same as a meal or even a bill payment, something low-income people would not want to trade off.
This unearths an insight about regulatory policy: it can in theory be quite paternalistic, forcing poor people to spend their limited resources on minute reductions in risk of death that they would never make on their own.
Despite all this, federal agencies and most analysts in general use a uniform value of a statistical life. This means that low-income, middle-income, and upper-income people’s willingness to pay for risk of death reduction is simply assumed to be the same.
Sunstein’s article talks about the implications of this assumption. He first tackles the question of subsidies, saying that low-income people, all other things being equal, will benefit from subsidies allocated according to this principle, though less than they would from a cash transfer or some alternate uses of these funds. This is because they are likely to receive some benefit from the subsidy, even if it is not as much as middle- or upper-income people.
Note that this is only true if they are not the ones paying the bill for the subsidy. If low-income people are footing the bill, they could end up less well-off than they would if the program were not in place.
As for regulations, Sunstein argues that it all depends on incidence. If low-income people accrue much of the benefits of a regulation and little of the costs, they will of course benefit. But if they pay a large proportion of the costs and accrue little of the benefits, they will do poorly under a situation where their value of a statistical life is assumed to be higher than it is.
Overall, what I take from this essay is that decomposition of benefits and costs matters. We should know who benefits and who will pay for a regulation not only for equity reasons, but for efficiency reasons, too. This will give us a better idea of what really happens when we put a regulation in place.