the ideal advertising should make demand more elastic for coffee and tea. In this
sense, the coffee-and-tea advertising is not considered successful, which might
already have been reflected in the historical expenditures on coffee. The real
advertising expenditure on coffee and tea in 2004 was only about a quarter of those
in 1984 and a third of those in 1974. Overall, advertising seems to have done the
right job for milk and soft-drink producers.
Conclusions
Showing that advertising can influence own-price elasticity through combinations of
its shift effect, shift-related and interaction-related rotation effects, this article
provides an elasticity form measure of the interaction-related rotation effect in five
demand models and tests for its existence in the non-alcoholic beverages. Results
are mixed in that compensated double-log model and the Rotterdam models fail to
reject the null hypothesis of no price-advertising interaction while linear model,
semi-log model, and the AIDS model indicate rejection. Interaction-related rotation
effects were found to be not robust to a change in model specification. This
confirms Hauser and Wernerfelt’s (1989) result that functional forms used to model
advertising and price interactions influence conclusions about its direction.
Since both model I and Kinnucan et al. (2001) use the Rotterdam model with
similar data - although the former allows a price-advertising interaction effect and
the latter does not, their results are comparable.10 For example, model I’s estimated
own-price parameters of -0.029, -0.067, -0.059, and -0.040 for the four beverages in
their respective order compare favorably to their counterparts of -0.047, -0.057,
17