Kinnucan and Zheng (2004) showed that the effect of advertising on the own-price
elasticity in absolute value (η) depends not only on the extent to which advertising
expenditure (A) rotates the demand curve (a rotation effect), but also on the shift in
the curve (a shift effect). Specifically, when prices are assumed exogenous, this
relation can be written as
∂ lnη ∂ ln ∆
(1) -----=--α
∂lnA ∂lnA
where q and p stand for quantity and prices, respectively, ∆ = -∂q / ∂p is the
demand curve’s slope in absolute value, ∂ln stands for logarithmic partial
differential, and α = ∂ ln q / ∂ ln A is the horizontal relative shift in the demand
curve due to a small change in advertising, i.e., the shift in the quantity direction
holding prices constant.4 A clockwise (counterclockwise) rotation, for example,
implies that ∂ ln ∆ / ∂ ln A is less than (greater than) zero.
Because this shift effect (the commonly known “advertising elasticity”) is
generally positive, it will either reinforce or offset the rotation effect depending on
the latter’s sign. For example, if ∂ln ∆/ ∂ln A > 0, the effect of this type of
advertising on the own-price elasticity is ambiguous, dependent on the relative
magnitude of α . Conversely, if ∂ln∆ / ∂ln A < 0, then ∂ lnη/ ∂ln A is
unambiguously negative in the presence of a positive shift effect. The upshot is that
the shift effect complicates the interpretation of advertising’s effect on the own-price
elasticity, especially in situations where the advertising is designed to make demand
more price elastic. Stated differently, the shift effect biases the results in favor of
making demand less price elastic, regardless of the advertising’s original intent.