taken to apply to anything else which shifts the demand curve outward. Inter? preted in these terms, for example, D-N have apparently shown that, for a variety of market structures, the equilibrium amount of research and develop? ment in product improvement is excessive. Surely something is wrong here. The basis of D-N's approach is to evaluate the welfare effect ofa change in output accompanying a change in advertising according to both preadvertising and postadvertising tastes. Thus, supposing U(x0) to be the preadvertising level of utility derived from the output x0 and i//(*i) to be the postadvertising level of utility from the output xi9 D-N proceed to compare U(Xi) with U(x0) and i//(*i) with ifj(x0). Using a geometric analysis for the case ofa monopolist, they demon? strate that U(xt) < U(x0) and ijj(Xi) < tjj(x0) and use this result to conclude that advertising is excessive whether judged by the tastes embodied in U(x) or by those embodied in i//(x). This result is correct, as far as it goes, but it does not remove the impediment to making welfare judgments when tastes change. In terms ofthe notation above, we believe the fundamental question to be: How does ifj(Xi) compare with U(x0)7 Knowing that U(xt) < U(x0) and i//(*i) < ifj(x0) does not allow one to infer that i//(*i) < U(x0) unless it is also known, or one is willing to assume, that tjj(x0) < U(x0) or ijj(Xi) < U(xt). Since this latter requirement is equivalent to knowing or assuming that advertising does not increase welfare, the D-N analysis can provide no information as to whether additional advertising is beneficial or not. The same point may be put in another way. In their analytic (as opposed to geometric) setup, D-N (p. 6) assume that demand is generated by a utility function whose arguments consist of the output of the numeraire, the output
Stored only when both papers are in the corpus — 53 citations exist upstream.