Abstract – In an analysis of whether to replace STOP signs by YIELD signs, the value of a life lost was pegged at $1,500,000 and the value of time at $6.71/hour. These numbers imply that when the sum of traffic delays accumulated by many drivers is equal in duration to the average lifetime lost in a fatal crash (37.3 years), the cost of such delay is higher than the cost of an average lost life. Most find this to be disturbing. If so, why is it that estimates of the value of time and life allegedly based on people’s preferences are at odds with what most prefer? A search for the root of this problem leads to Schelling’s distinction between the value of death to those who die and the value of the probability of dying to those who live. He thinks that while it is not possible to put a value on one’s own death, it is possible to put a value on changes in the probability of one’s own death. I think that this distinction does not solve the problem. If it is impossible to have preferences for consequences that would have to be experienced posthumously, it cannot help to make the event of death still more remote by a dimly perceived probability. People may be willing to express preferences and econometricians may be eager to interpret them. But, inasmuch as these preference are vacuous, they have no interpretation and attempts to do so may lead to the noted inconsistency. Consistent use of a wildly incorrect value of life in cost-benefit analyses involving risk leads to consistently incorrect conclusions. Instead of using a questionable value of life in dispassionate-looking computations, it may be better to give legitimacy to public decisions more directly by a mechanism akin to a ballot or a jury.