Notes:
1 Real information actually plays a role similar to the match-products-to-buyers effect discussed by
Bagwell (2005, p.19).
2 The two models used by Schmit and Kaiser and Chung and Kaiser were the same in functional
form. The latter study used per capita fluid milk sales as the dependent variable while the former one
used per capita retail fluid milk/cheese demand instead.
3 Since Farr et al. (2001) and Tremblay and Okuyama (2001) argued that advertising could affect
equilibrium consumption through its influence on supply (price competition), we don’t rule out the
possibility of a price-advertising interaction relationship on the supply side. The analysis done in this
article is strictly on the demand side.
4 Exogenous price is a common finding in the empirical literature (e.g., Brester and Schroeder, 1995;
Kinnucan et al., 1997).
5 Green and Alston (1990) show that all of the previously reported formulae for AIDS elasticities are
incorrect when LA-AIDS is estimated instead of the true AIDS with a few exceptions including
constant group price, i.e., d ln P* is independent of individual goods’ prices. This condition is
satisfied since this article assumes exogenous prices.
6h
6 Note - (cii +γid ln Ai)/ wi is the Hicksian own-price elasticity (ηi ) for the Rotterdam model.
The Marshallian own-price elasticity (ηim ) is equal to - (cii +γid ln Ai) / wi - bi . The effect of
advertising on the Marshallian own-price elasticity is derived as
∂ lnηim / ∂ln Ai = (ηih /ηim)∂lnηih / ∂ ln Ai .
7 In 2005, the volume shares of the three largest non-alcoholic beverage categories by volume were
carbonated soft drinks (43.8%), bottled water (21.6%), and fluid milk (17.8%), according to Beverage
Marketing Corporation.
8 Results of t-statistics were much improved from using OLS to SUR, but remained alike from SUR
to iterative SUR.
9 For completeness purpose, all ci’s, dii,s, and γ's were used to calculate the own demand elasticities
but only significant cii ,s, dii ,s, and γi S were used to calculate the own advertising elasticities and the
interaction-related rotation effects except the dii in model J to avoid a negative own advertising
elasticity for coffee and tea; price, demand, and advertising took their mean levels when they were
needed.
10 The difference of the two datasets is reported in detail in the section IV.
19