Some economic ideas thrive only in the darkness: they are simply too weird and half-baked to withstand public scrutiny. Perhaps no concept better exemplifies this than the value of a statistical life, the sum of money that supposedly measures the value of a life saved or sacrificed by a government program. Few nonspecialists pay attention to the slow trickle of research articles on the topic, although the spotlight descends every few years, usually because of a scandal involving government proclamations that some people are worth more than others. (The prime case was the dispute that broke out over the economic analysis in the third IPCC report, which pegged the monetary value of an American life at ten times that of a Chinese or Nigerian life. The numbers were subsequently crossed out.)
Today’s New York Times draws our attention to the topic in a less charged context, the discrepancy between VSL’s posted by different federal regulatory agencies. It is a small issue in the larger scheme of things but still illuminating in what it reveals about the economic profession. (I should also admit that my interest is probably larger than yours, since I wrote a book on the subject 15 years ago.)
It is quite true, as the article reports, that agencies assign monetary values to human life in benefit-cost analyses. It is true that their numbers are based, more or less, on the work of Kip Viscusi and a few of his colleagues. It is also true that the numbers are drawn from regressions on wages, with the intention of identifying a wage premium for physical risk after controlling for other factors that affect labor market outcomes. Those parts of the story are correct.
What’s more important is what is missing:
1. The theoretical model underpinning all work in the field is vintage 1960s supply-and-demand analysis. There is no search and matching, no transaction costs, no bilateral bargaining, no repeated interaction between employer and employee, no unemployment. None of the VSL honchos, from Viscusi on down, have bothered to try to construct a model of wage compensation for risk using the methods all serious labor economists have now adopted. In fact, they would have a hard time doing this: equimarginal outcomes (of which wage compensation for risk is one) are few and far between in modern labor analysis.
2. The regression methods employed in the VSL studies are primitive and sloppy. They rarely account for obvious endogeneities in observed labor market phenomena. They are notoriously sensitive to omitted variables (as I showed in a paper I coauthored back when.) They don’t even take care of the basics, like accounting for the effects of grouped data (such as average injury rates assigned to individuals according to their industry or occupation) on significance estimates. It’s a real methodological backwater.
3. Practitioners seldom discuss the embarrassing fact that their wage regressions produce different values of life for different subsamples, without any obvious justification. Men’s lives are worth more or less than women. Whites are worth more than blacks. The lives of unionized workers are worth more than those not in unions, except when it’s the other way around. In fact, subsample results jump all of the map, a sure sign that the econometrics are not leading us to a stable underlying relationship.
4. The whole enterprise is conceptually vacuous. Are we identifying the willingness to pay for safety or the willingness to accept risk? (What difference do occupational safety regulations, which ostensibly set the default, play?) Why should wage rate differentials be entrusted with assigning the value of life and health? On what theoretical ground are these choices privileged over all other means of assigning values? How do we reconcile the view that workers are fully compensated for the risks they face at work with the all-too-evident politics of health and safety regulation? Why would workers forego this compensation by demanding tougher regulation? Why don’t employers factor in the ability to pay lower wages when they assess new regulatory demands? Why aren’t they funneling gobs of money to their safety consultants in order to cut wage costs without government prodding in the first place? Why is there such a complete disconnect between the world of the VSL “experts” and the one seen all over the world in the politics and law of workplace safety?
In the wake of the financial crisis, macroeconomists have been on the defensive. How could they have missed the buildup of asset bubbles and financial vulnerabilities, cheering on the very policies that brought us to the brink? But it isn’t only this particular tribe: as the VSL morass demonstrates, there are other clumps of cultishness in the profession—fields and subfields whose reputations are built on mutual citation and a willingness to set aside serious intellectual standards. In the end, the problem is that economics is too often able to insulate itself from hard criticism: there is seldom a career cost to being shown to be in error. The questions on the table are why and what to do about it.