The randomly assigned risk of induction generated by the draft lottery is used to construct estimates of the effect of veteran status on civilian earnings. These estimates are not biased by the fact that certain types of men are more likely than others to service in the military. Social Security administrative records indicate that, in the early 1980s, long after their service in Vietnam had ended, the earnings of white veterans were approximately 15 percent less than the earnings of comparable nonveterans.

In On the Basing-Point System, Bruce Benson, Melvin Greenhut, and George Norman (1990)1 properly correct a misunderstanding by Jacques Thisse and Xavier Vives (1988)2 concerning my 1982 paper on basing-point prices. But BGN then take exception with several of the conclusions I had reached, or that they thought I had reached. Although I have insufficient space to deal with all the points they raise, I will address the most important ones. In 1981 most economists believed that a model of profit-maximizing basing-point pricing would have to assume collusion. My model, as BGN seem to agree, showed that it need not; atomistic competition at one production site accompanied by local monpolies elsewhere may lead to basing-point pricing and freight absorption without collusion.3 Conversely, TV showed that the practice as observed empirically will not arise if all production sites contain noncooperative local monopolies. BGN establish yet another important theoretical result-noncooperative oligopoly at one site could generate the f.o.b. prices (or f.o.b. plus transport) that, in my model, led local monopolies elsewhere to adopt freight-absorbing basing-point prices. The three models would seem to be complements. I. A Paradox

A breadwinner's demand for life insurance depends on the demographic structure of his or her household. The author captures the relationship by extending Menahem E. Yaari's life insurance framework to include the preferences of all household members explicitly. In many households, the insured is the husband and the beneficiaries are his wife and offspring. Their demand for insurance of the husband's life is derived from a life cycle model in which income is uncertain. The results are intuitively appealing in that they describe the explicit calculation made by purchasers of life insurance. Empirical estimates based on observed life insurance ownership also are encouraging.

In a recent article (Yew-Kwang Ng, 1987), Ng demonstrates that taxes on goods whose utility is derived entirely from their market value (diamond goods) can provide revenue to the government with no cost to the taxpayers aside from the administrative costs of collecting them. Such a tax decreases the quantity of the good produced while proportionally increasing the value per unit quantity. Since the utility of the good depends only on its value, the smaller quantity produced (after the tax is imposed) produces the same total utility to consumers as the previous larger quantity. The revenue collected by the government is paid, in effect, from the reduced cost of producing a smaller quantity of the good. Ng mentions a number of earlier discussions of goods whose utility comes in part from their price. He is apparently unaware of an analysis which is both much earlier and much closer to his than any he cites. In Chapter 13 of the Principles of Political Economy, David Ricardo wrote:

There are qualifications to the theoretical comparison between Health Maintenance Organizations (HMOs) and conventional insurance (CI), presented in a recent contribution to this Review by Yael Benjamini and Yaov Benjamini (1986). These qualifications make difficult the acceptance, as an unambiguous prediction of economic theory, the authors' conclusions regarding the relative attractiveness of the two methods of insurance to groups of individuals with heterogeneous demands. The authors' arguments are summarized as follows. For homogeneous insureds, an HMO can resemble ideal medical insurance depending on the extent to which the prearranged medical care provided in each health state approximates the desired level for the homogeneous group. Under such conditions, an HMO would be superior to a CI plan in that it is free of the moral that causes a welfare loss under the latter. However, if insureds are heterogeneous in terms of tastes, or any other factor that affects... a welfare loss will also result under the HMO. Because the quantity of medical care provided for a given health state is fixed for an HMO, it cannot satisfy the divergent medical care demands of all insureds in a heterogeneous group. In contrast, the greater flexibility afforded by CI enables insureds with divergent demands to more closely achieve their desired levels of medical care consumption. The welfare loss due to moral hazard is assumed to be little affected by divergent demands, and no other potential effects of demand heterogeneity on the desirability of CI are mentioned. The implication of the authors' analysis is that greater heterogeneity among insureds generally increases preferences for a CI plan over an HMO.' The analysis presented by Benjamini and Benjamini (1986) contains two omissions that pertain to the effect of heterogeneous demand on the desirability of a CI plan. First, heterogeneous demand causes cross-subsidization under a CI plan over and above that due to the incidence of health state. Low demanders subsidize high demanders under CI, whereas, no such cross-subsidization exists under HMOs. This makes ambiguous the effect of heterogeneous demand on general preferences for the two methods of insurance. Second, there appears to be no theoretical basis for assuming that the size of the welfare loss under CI is virtually unaffected by heterogeneous demand. Heterogeneous demand can be shown to increase the welfare loss under a CI plan when this loss is more accurately measured. Thus, the effect of heterogeneous demand, at least in theory, is for this reason much more ambiguous than the Benjamini and Benjamini article implies. The first point is demonstrated with the use of Figure 1. A single, precisely defined unhealthy state, X, and the divergent medical care demands of two representative individuals, A and B, are assumed. A and B are identical in terms of health risk but have differing demands for medical care because of differences in income, tastes, etc. Their respective demands for medical care under a specific CI plan are depicted in Figure 1. Zero administrative costs and a coinsurance rate of .2 are assumed. The price per unit of care is assigned a value of 10 and is equal to marginal cost, which is assumed constant. Under these constraints, individual A de-