the coefficients of each of the measures should have the predicted sign even when they are all included in the sample multiple regression. The absolute values of both the t values (which measure significance) and the coefficients will be considerably lower in the multiple regression than in the simple regressions. The inclusion as independent variables of several measures of the same independent variable win produce that result. Table 4 records the multiple regressions of each market characteristic on the other market characteristics and durability. All of the signs are in the right direction.
There is one more exceedingly strong test of our theory. It is not only a test of the relevance of information variables in determining market characteristics, but a test of the precise formulation of the relationship we have been using and a test of whether R is the exclusive reason for the relationship between the market characteristics. If R is the exclusive reason for the relationship between market characteristics, one can predict the value of the correlation coefficients as well as their Sip.28 For most of the correlation coefficients R passes this test fairly well (see Table 5). The few exceptions can be easily explained.29 All in all, this evidence strongly supports the theory that advertising is information.
Assuming a translog production function, quantity demanded (Q) is the following function of prices* (P) and advertising expenditures (A).
lnQ = α0 + α1lnP + α2lnA + (A.1)
α3lnPlnA + α4(lnP)2 + α5(lnA)2
The price elasticity of demand in absolute terms (εp) is given by:(A.2)
The advertising elasticity of demand (εA) equals:(A.3)
Schmalensee and others have shown that a firm that is maximizing profits will operate such that:30(A.4)____________________