The value of statistical life (VSL) can be inferred through real‐world wage–fatality risk trade‐offs made across different occupations. This paper shows that the VSL based on the wage‐risk trade‐off tends to be biased upward if it does not account for the diversity of workers’ unobservable skill to cope privately with job risk. This upward bias arises because the highest required wage differential among the workers is divided by their average risk across the population.

Lab experiments have gone to extremes to isolate and repress other-regarding behavior in extensive-form bargaining games, with limited success. Consider, for example, Elizabeth Hoffman et al.’s (1996; hereafter HMS) Anonymous Dictator game. This game controls self-interested strategic behavior by giving a person complete control over the distribution of wealth, and complete anonymity from all others including the experimenter. While theory predicts people with complete control and complete anonymity will offer up nothing to others, in fact they still share the wealth in about 40 percent of the observed bargains. Such other-regarding choice is another example in which individual behavior differs from that predicted by subgame perfection, and supports the call for a new “behavioral game theory” (Colin F. Camerer, 1997). Herein we extend the work of HMS to reveal a setting in which 95 percent of dictators follow game-theoretic predictions. In contrast to previous studies, our design has people bargain over earned wealth rather than unearned wealth granted by the experimenter. We argue that just as rewards must be salient (Kyung Hwan Baik et al., 1999), the assets in a bargain must be legitimate to produce rational behavior. Our results support this conjecture. Dictators bargaining over earned wealth were more selfinterested than observed in previous studies; and when they had complete anonymity, selfless behavior is essentially eliminated.