The analysis in Saak (2012) shows that a firm may keep on advertising a product indef-
initely to accumulate some goodwill that results from better consumer information. One
possible testable prediction would be that this is more likely to be the case for niche
markets than for mass markets. Still it would be necessary to identify to what extent this
persistent advertising activity may be attributed to information rather than some alterna-
tive non-informative sources of goodwill.
As discussed by
Bagwell (2007, subsection 8.1), there is a substantial empirical liter-
ature that suggests that advertising does not impact household behavior much, if the
household has some previous experience with the product (e.g.,
Deighton et al.,
1994; Erdem and Keane, 1996; Ackerberg, 2001
). In his reduced-form analysis of house-
hold purchases of Yoplait yoghurts,
Ackerberg (2001) makes the identifying assumption
that, if advertising had a persuasive impact or was a complementary good, then it should
affect the household’s choice whether or not it has had some previous experience with
the product or the brand. He finds that advertising only has a significant impact on pur-
chase decisions of new customers of Yoplait yoghurts. This supports the view that the
main role of advertising is to provide these consumers with information that induces then
to purchase the product. Although this is broadly consistent with the theoretical setting in
Saak (2012) (in particular with his assumption that yields persistent advertising in a niche
market), the exact nature of that information remains to be characterized. As discussed in
Section 4.4.1, Ackerberg (2003), in his structural model, proposes to interpret this new
information as quality signaling. Some information about the content of the ad messages
would be needed in order to know whether they might contain direct information.
Ippolito and Mathios (1990) use a natural experiment to indirectly control for ad content.
They study the impact of the 1985 termination of the ban on health claim advertising by
cereal producers in the US on cereal brand consumption. They find that the end of the
ban led to a significant increase in the consumption of fiber cereals.
Short of having detailed content information or some appropriate natural experiment
as in
Ippolito and Mathios (1990), it is not clear whether an improved perceived quality
should be interpreted as information or as resulting from some persuasion or complemen-
tary good effect of advertising (see the discussion in
Section 4.6.1 above). A recent mar-
keting literature explores whether goodwill can be attributed to an improved perceived
quality or, rather, to an improved product awareness, which is unambiguously better
information.
Clark et al. (2009) use a consumer survey dataset over the period
20002005. It provides on a yearly basis a brand awareness score and a perceived quality
score for top brands (in terms of sales) in 25 broad product categories. They estimate
a two-equation recursive model of perceived quality and brand awareness inspired by
Nerlove and Arrow (1962). Advertising expenditure contributes to the evolution of both
perceived quality and brand awareness. They use the panel structure of their data to solve
for endogeneity and account for unobserved heterogeneity across brands. They find that
advertising effects brand awareness positively and significantly, whereas its impact on
196 Handbook of Media Economics