1My interest in these issues was in part provoked by Posner's brief discussion of them, especially R. A. Posner, The Economic Analysis of Law, Little Brown (1977), pp. 149-152, which raises but does not resolve a number of the central issues of this paper.

2Posner (supra, note 1, pp. 119-161, especially 143). For an extensive and in part critical discussion of the validity of the economic approach to law, see (1980) 9 J. Legal Stud.

3This point is made explicitly in J. Broome, `Trying to Value a Life', (1978) 9 J. Pub. Econ. 91-100 in his discussion of the utilitarian approach to the valuing of life. My approach is similar to that discussed (but not accepted) in that part of his paper, save for the crucial difference that I do and he does not introduce Von Neumann-Morgenstern utility. See also J. Broome, `Trying to Value a Life: A Reply', (1979) 12 J. Pub. Econ. 259-262; J. M. Buchanan and R. L. Faith, `Trying Again to Value a Life', (1979) 12 J. Pub. Econ. 245-248; M. W. Jones-Lee, 'Trying to Value a Life', (1979) 12 J. Pub. Econ. 249-25& and A. Williams, 'Note on Trying to Value a Life', (1979) 12 J. Pub. Econ. 257-258.

4Readers familiar with such arguments will realize that the result is not a rigorous one; it is possible to construct (bizarre) utility functions for which it does not hold.

5See for instance Linnerooth and Bloomquist (G. Bloomquist, The Value of Human Life: An Empirical Perspective', (1981) 19 Econ. Inq. 1) and their references. For an interesting discussion of value of life see G. Tullock, The Logic of the Law, Basic Books (1971), pp.155-158.

6One way to increase the realism of the model, at considerable cost in simplicity, is to allow the individual to allocate a fixed budget of time (24 hours a day) between time used to earn income and time used, with consumption goods, to produce utility. This would be particularly appropriate in those cases in which the same injury that reduced income also increased leisure.

7Strictly speaking, I will be talking about utility functions and expenditures in one or another of two different probability states ('states of the world'): one in which the injury occurs and one in which it does not. In the state in which the injury occurs, some expenditures occur before the injury. Since the individual only knows which state he is in when the injury occurs (or after the time when it could occur has passed), it is convenient, although not strictly accurate, to refer to expenditures or utility functions before or after the accident, and I shall sometimes do so.

8Broome, supra, note 3, tries to discuss a utilitarian approach without introducing choice under uncertainty. He thus (deliberately or not) makes it impossible to discuss the utility of life save in his preferred context of a money payment sufficient to compensate for a certainty of immediate death. But this is impractical because of the declining marginal utility of money, a point he has already made.

9J. Von Neumann and 0. Morgenstern, Theory of Games and Economic Behaviour, Princeton University Press (1953), pp. 15-29.

10Expected utility is defined as the probability of each outcome in a lottery times the utility of that outcome, summed over all outcomes. It corresponds, in terms of utility, to the expected return of a lottery in terms of money.

11In this case, at least, Broome, supra, note 3 is correct in arguing that money is an insufficient yardstick for measuring life, at least as long as we restrict ourselves to costs and benefits to a single individual and do not permit interpersonal utility comparisons.

12Readers with a taste for rigour may fill in for themselves a more precise statement of the theorem whose proof I have sketched here.

13One source of market failure might be `reverse moral hazard'. The individual who has sold his right to collect damages will increase the precautions he takes to prevent or minimize injury; while such action may be desirable from an efficiency standpoint it also means that what he sells has a lower expected value to the buyer than it would have had to the seller, which may prevent the sale.

14One obstacle might be the lack of any insurable interest on the part of the buyer (I owe this point to an anonymous reviewer). While this legal doctrine may be justified in other contexts on the grounds that my ownership of an insurance policy on someone whose life is of no value to me gives me an incentive to increase its value by increasing his probability of death, it is worth noting that in this case the (desirable) incentive effects on the seller may be more important (see note 13).

15Readers who are familiar with the literature on liability will realize that I have deliberately ignored a large number of issues (such as standards of negligence) in order to concentrate on one. Those familiar with the conceptual literature on valuing a life (in particular Broome (supra, note 3) and the subsequent debate) will appreciate my reasons for trying to avoid philosophical arguments about the meaning and relevance of probabilistic considerations and expected utilities wherever possible. Any serious attempt to set forth my views on the relevance of Paretian and utilitarian comparisons to issues of choice and welfare would have made this a much longer and very different paper.