This paper defends two main claims. First: although it is easy to lose sight of this, what cost-benefit analysis really demands, in order to approve of a prospective policy, is that it be possible for those who would gain through the policy change to compensate those who would lose through it. And second: in cases where a policy change does, or can reasonably be expected to, lead to someone's death, the demand of compensability is much harder to satisfy than economists (...) typically think. More specifically, their standard move—maintaining that compensability should be judged ex ante, and that it is thus really just risk of death, not death itself, that must be (and often is) compensable—is unsuccessful. Then, the implications of these claims are briefly explored. (shrink)