Is advertising socially excessive? Economists have been reluctant to consider this question, for there appeared to be no rigorous method of analyzing cases in which advertising influences tastes. Some recent studies have made fundamental contributions to this debate (Avinash Dixit and Victor Norman, 1978; Yehuda Kotowitz and Frank Mathewson, 1979), but no consensus has emerged. Dixit and Norman were the first to establish a rigorous methodology. They evaluated social welfare (SW= consumer's surplus + profits) under preand post-advertising market equilibria, using pre-advertising tastes for one ranking and post-advertising tastes for another ranking. When the SW rankings were identical, Dixit and Normal claimed that unambiguous inferences could be drawn. In general, they concluded that the profitmaximizing level of advertising will be excessive under a variety of product market structures. Franklin Fisher and John McGowan (1979) criticized Dixit and Norman for including advertising in the utility function but ignoring the direct effect that advertising has on utility in their welfare analysis. Dixit and Norman (1979) courvlered that advertising merely shifts the preference ordering over goods, and that it is this preference yardstick that matters. Welfare conclusions about advertising can be generated less controversially within the framework suggested by George Stigler and Gary Becker (1977). They treat utility as if it were generated by which individuals produce with purchased goods, time, human capital, and firms' advertising. Their commodities can be interpreted similarly to what Kevin Lancaster (1966) calls a characteristic, the terminology used herein.' As an example, consider tennis. We buy racquets and balls, not to own them per se, but rather to enjoy playing the game. Playing the game actually generates utility, and this characteristic is produced by an individual's racquet, balls, human capital (skill level), and time. Thus preferences are ordered over characteristics, not over goods. Advertising can lower the shadow price of a characteristic by increasing the ability of the purchased good to produce the desired characteristic. This is accomplished by providing the consumer with different information or beliefs about the product than previously held. Suppose a perennial tennis champion states a preference for a particular brand of racquet. This enhances the self-image of amateur owners of that brand and increases their enjoyment of the game without changing their expenditures. The per unit cost of the characteristic will decline, making current owners of the racquet likely repeat customers and current shoppers more likely to choose the star's preferred brand. Observe that advertising has increased the demand for a market good that produces the now cheaper char-

The persistently large number of uninsured, roughly 40 million per year since 1993, continues to elicit bipartisan policy interest. Coverage-expansion proposals without mandates, by far the most common since the defeat of the Clinton plan, must address risk-pooling realities in private markets. Insurers have strong financial incentives to segment risks and minimize pooling of heterogeneous risks, and narrow risk-pooling will diminish the adequacy of premium subsidies based on income alone, at least for higher-risk individuals. The current debate over flat tax credits and the non-group market is a case in point (Blumberg, 2001; Center for Studying Health System Change, 2002; Jack Hadley and James D. Reschovsky, 2002). We, along with nine other teams, were asked to develop a proposal that would expand coverage in a large and creative way (see Holahan et al., 2001). The proposal we developed would subsidize low-income individuals and families but also addresses the issue of inefficient and inequitable risk-pooling.