theoretical advantages inherent in demand systems, the overall satisfactory
significance in its estimates, and a more reasonable size of the rotation effect. In
addition, Duffy (2001) found that the AIDS model provided the “most suitable
framework for investigating advertising effects” in U.K. alcoholic drinks markets, a
finding that helps to justify our selection of the AIDS model.
For milk, the computed interaction-related rotation effect is -0.073,
indicating a 10% increase in the milk advertising (note that most of the milk
advertising is generic advertising) would reduce the slope (in absolute value) of milk
demand by 0.73%, a number not seen in the literature. Similarly, a 10% increase in
the advertising of coffee and tea would decrease the slope of its demand by 1.16%.
Conversely, advertising is found to increase the slope of soft-drink demand. A 10%
increase in the soft-drink advertising would increase the slope of its demand by
0.49%. As a robustness check, the AIDS model was estimated with the data prior to
1976 deleted. The price-advertising interaction terms hold significant at the 5%
level, and D.W. statistics come closer to two.
To put the results into perspective, figure 2 plots the two representative cases
of demand curve rotation due to advertising, clockwise rotation for milk and
counterclockwise rotation for soft drinks. In figure (2a), a 10% increase in the milk
advertising rotates the demand curve D0 clockwise to DR by reducing the size of its
slope by 0.73% (measured at the mean advertising level). When measured at the
mean price level, the 10% increase in the milk advertisings increases milk demand
by 0.19%. What (2a) implies is that since advertising makes milk demand less
elastic, it must also be true that an increase in price increases advertising’s ability to
15