but this requires that advertising has a significant market expansion effect. If consumers
can search sellers from whom they have received no ad, then advertising does not affect
the market size. It is what
Marshall (1890, 1919) calls “combative advertising” that merely
reshuffles market shares. Then it is clearly excessive. Some empirical support for combat-
ive advertising can be found in
Lambin (1976), whose results suggest that advertising does
not increase demand in an industry. When allowing for product differentiation, there is
no clear takeaway. Advertising may be insufficient even if it has no impact on the market
size. This is because it may induce large gains in consumer surplus by improving the
matching between consumers and products. I provide another perspective on the social
optimality of advertising in
Section 4.6 taking into account its non-informative roles.
Note finally that the above results are relevant for a proper welfare analysis of the
media market. The advertisers’ willingness to pay for ad space does not correctly reflect
the social benefit of advertising, unless advertisers are perfectly discriminating monopo-
lists. Conversely, because of potential inefficiencies in the market for ad space, the private
advertising cost incurred by firms may differ from the social cost of advertising, making
the welfare analysis presented here inappropriate. I return to this point in the concluding
remarks to
Section 4.5.
4.3. PRODUCT ADVERTISING
The goal of this section is to present a fairly recent body of literature that investigates the
incentives of firms to provide information about their product to consumers. This
research has adopted two distinct modeling strategies that are complementary. In a first
set of models (
Section 4.3.3), ads are viewed as informative signals about how well the
product matches the consumers’ taste. The firm can choose the characteristics of this
signal—how informative it is or what it informs the consumer about—so as to maximize
its profit. However, the firm has no private information about how good the match is.
Alternatively (
Section 4.3.4), other papers explicitly model how ads inform consumers
about their match, by disclosing the product’s attributes. In such a setting, both sides have
private information about the match since the consumer knows her tastes and the firm
knows the product’s attributes. The disclosure game is then a signaling game. The section
ends with a discussion of how we might think of misleading advertising in the context of
such disclosure games.
I first introduce the general issue of the informative content of ads by briefly summa-
rizing the seminal paper by
Nelson (1974), and presenting some relevant empirical work
both in economics and marketing.
4.3.1 Nelson's Question
In his famous early 1970s articles (Nelson (1970, 1974)), Philip Nelson has drawn our
attention to the channels through which consumers obtain product information.
143Advertising in Markets
In Nelson (1974), he develops a theory of the role of advertising in this respect. To do
this, he uses the distinction between search goods and experience goods that he intro-
duced in
Nelson (1970). Characteristics of search goods can be observed by a consumer
before she decides whether to buy or not, whereas characteristics of experience goods are
learned only once the good is consumed. His premise is that advertising may transmit
information only if this information is deemed credible by potential buyers. His analysis
abstracts from any government intervention in the form of laws on misleading
advertising.
He argues that a claim pertaining to a product is credible only if it concerns “search
attributes.” Indeed, there is no point in telling major lies about such attributes because
they would at best fool the consumer into incurring some visit cost but not into buying.
The consumer may find out the truth before she buys. He concludes that ads for “search
goods” should contain a lot of information, whereas ads for experience goods should not.
And yet, in view of the anecdotal evidence provided by Nelson, ad expenditures devoted
to the promotion of experience goods are much larger than what is spent on advertising
search goods.
Nelson (1974) is best known for solving this puzzle. The key insight is that the infor-
mation provided by experience good ads is indirect. I return to this point in the next
section that discusses the role of advertising as a signal. My focus here is on Nelson’s pre-
dictions on direct information in ads. His conclusions are that ads for experience goods
should contain hardly any information, whereas ads for search goods should be very
informative. Furthermore, the information conveyed about search goods should be cred-
ible even if there are no laws on misleading advertising. The body of research I discuss
below explores these issues through models of information disclosure. I next look at some
of the available evidence regarding the informative content of advertising.
4.3.2 Empirical Evidence on Advertising Content
The analysis of Nelson (1974) predicts that much of the ad money is spent on ad messages
for experience goods that contain little or no direct information. He provides various
indirect evidence that this is the case. Verifying it directly requires some data on the actual
content of ad messages and some coding strategy that allows for measuring the amount of
information it includes. Shortly after the publication of Nelson’s article, though with very
different motivations, some researchers in marketing have undertaken to gather such data
and to construct a coding method. This line of research, called “content analysis,” was
initiated by
Resnik and Stern (1977). They define 14 categories of “information cues”
that include such items as price, quality, performance, availability, nutrition or warran-
ties. The methodology consists of counting the number of categories present in each ad
message, in order to measure its informativeness. The content of advertisements in dif-
ferent media has been investigated in this fashion. For instance,
Abernethy and Franke
144 Handbook of Media Economics
(1996) present summary statistics for a set of previous studies on US media: they find that
the average number of categories per ad over four different studies of US television is
1.06, with only 27.7% of ads including at least two and 37.5% having no category what-
soever. For seven studies on US magazines, the average was 1.59, where 25.4% have
three or more categories and 15.6% have no category at all. For newspapers, they find
a much higher average, but still no more than 38.5% of advertisements inform about
more than four categories. Newspaper ads are much more likely to include price infor-
mation than magazine ads, 68% against 19%. The overall picture is that many ads include
little information, if any, although a majority of ads include some information.
In recent years, there have been renewed efforts to gather and exploit data on the
content of ads (e.g.,
Anderson et al., 2013, 2015; Bertrand et al., 2010; Liaukonyte,
2012
). Anderson et al. (2013) provide a systematic analysis of the content of television
ads for over-the-counter analgesics in the US. They have constructed an original dataset
by coding the content of all ads over a 5-year period, recording all specific claims that are
made about the medicine (such as fast or strong) as well as whether these claims were
comparative or not (this particular industry being characterized by an extensive use of
comparative claims in advertising). By contrast with the
Resnik and Stern (1977) meth-
odology, they do not gather information cues into general categories but rather add up
the total number of information cues in a message to measure its informativeness. Over
the period of investigation, 30 different attributes were mentioned, of which they keep
the top 23 in terms of amount of dollars spent advertising them. They also relate these
attributes to the actual characteristics of the drugs as established in medical publications.
They use a simple model of a firm’s choice of the informativeness of its ad messages,
where the firm trades off the amount of persuasive (non-informative) content with
the amount of information. The marginal benefit of an extra second of persuasion is some
random variable, the realization of which varies across ads: the smaller it is, the more
information cues the firm chooses to include in the ad. They find that ads for objectively
higher quality products include more cues. Ads for brands with large market shares pro-
vide less information and comparative ads are more informative. Finally, ads for brands
with large generic counterparts embody less information.
The results in
Anderson et al. (2013) cannot be directly compared to those of mar-
keting studies on advertising content. The focus is actually different. By using general
categories that can apply to a broad range of different products, content analysis allows
for comparing the informative content of ads across media, over time, across countries or
across industries. The results, however, are not very conclusive. Besides, the observation
that ads are not very informative because they cover only a small number of the 14 pos-
sible categories is questionable: categories are broad and only a small subset of them might
be relevant for a particular advertised product. The most convincing takeaway is that the
percentage of ads covering no information cue category is rather large. The approach in
Anderson et al. (2013) rather focuses on the incentive of firms to provide direct
145Advertising in Markets
information in ads. They derive predictions about which firms are more likely to include
more information in their advertisements. Looking at a particular industry allows for
having measures of informativeness that are more directly comparable across brands:
products are sufficiently close that the relevant information is similar. It remains hard
to make statements about the informativeness of ads in absolute terms because it is
not clear what the most informative ad possible would look like, but comparisons among
brands and products seem reasonable.
The theory of indirect information on experience goods proposed by
Nelson (1974)
provides one explanation of why many ads are completely uninformative. It tells us noth-
ing, however, about the extent to which ads should include some information when that
information is credible. Nor does it predict the nature of the information that a firm
might choose to disclose or to keep silent. The remainder of this section presents various
theoretical frameworks that help address these issues.
4.3.3 Match Advertising
Typically, we view advertising as an effort to make a firm’s product more attractive to
potential consumers. Intuition then suggests that a proper disclosure strategy consists
of revealing favorable features while hiding unfavorable ones. This, however, implicitly
assumes that all consumers agree on what is favorable and unfavorable news about a prod-
uct. Yet consumer tastes are often heterogeneous and products are horizontally differen-
tiated. As a result, information that is good news to some buyers turns away others. The
social value of information is then to facilitate a good match between products and
buyers. The empirical study by
Anand and Shachar (2011) on ads for television shows
provides an interesting illustration of this role of advertising. They find that exposure
to an ad can make it less likely that a viewer watches the advertised program and that
seeing one ad reduces the probability that a viewer watches the wrong program by
10%. I now discuss a seller’s incentive to provide detailed product information that
improves the matching.
It is useful to start with two methodological remarks. First, an obvious reason for not
disclosing information is that it might be too costly. The literature has typically ignored
these costs to focus on the demand-driven incentives for not disclosing information.
There is actually a difficulty in specifying how disclosure costs are related to the
“quantity” or the nature of the information that is being revealed (costs are introduced
in
Section 4.5.1). A second point is that, as I will explain in Section 4.3.4, information
transmission involves some strategic complexity that often leads to multiple equilibria.
Many researchers who have been concerned about the provision of march information
have circumvented this problem by assuming that the firm can completely control the
information that is inferred by buyers from looking at the ad, subject to some constraint
that it is consistent with Bayesian updating. The advertiser can then select the inference
146 Handbook of Media Economics
that is most profitable. This requires in particular that the firm holds no private informa-
tion about the value of the match. The analysis presented in this subsection follows this
modeling strategy.
In much of the following discussion, I use a setting where a monopoly seller produces
a good with constant marginal cost c 2 0,1½Þ. For simplicity of exposition, there is only
one buyer, who has unit demand. The buyer’s willingness to pay is r 0, which is the
realization of a random variable. Initially, r is unknown to both parties but it is common
knowledge that it has a uniform distribution on [0,1].
4.3.3.1 Direct Information About Experience Goods
The early contributions on the transmission of match information (
Lewis and
Sappington, 1994; Meurer and Stahl, 1994
) assume that consumers cannot acquire
pre-purchase information about products through their own endeavor. This is consistent
with the classification of products as experience goods in
Nelson (1970). Still, in contrast
to Nelson’s view, it is assumed that the disclosed product information is directly available
in the ad and is credible. Following the literature, I assume that this credibility is achieved
through certification.
20
Typically, laws on misleading advertising may achieve this as long
as the claims can be verified by a court.
21
A stripped-down version of Lewis and Sappington (1994) may be presented in the
simple monopoly setting above. Before the buyer decides whether to purchase the prod-
uct or not, the firm announces a price and, if it so chooses, provides some product infor-
mation. If it provides such information, the buyer observes a signal σ 2 0, 1½, which is
equal to her true valuation r with probability α and is uniform on [0,1] independent
of r with the complementary probability 1 α. In short, the buyer learns r perfectly with
probability α and learns nothing otherwise but does not know which situation she is in.
Then the larger α is, the more informative is the ad.
From the standard uniform distribution, the prior expected value of the match is 1/2.
After observing the signal σ, the buyer’s expected willingness to pay is ασ +1=21α
ðÞ
.
This expected willingness to pay is increasing in α if σ > 1=2 and decreasing in α if
σ < 1=2.
Figure 4.2 shows the inverse demand curve given the posterior beliefs of the
buyer, where price, p, is on the vertical axis and the probability that the product is pur-
chased, q, is on the horizontal axis. As demonstrated by
Johnson and Myatt (2006), the
firm’s problem may be analyzed graphically by letting the inverse demand curve pivot
around the point A in the figure with coordinates (1/2,1/2). Indeed, if the firm sells with
probability q < 1=2, it charges a price above 1/2, because of the standard uniform
20
As should become clear from my discussion of cheap talk in Section 4.3.5.1, in the models discussed in this
subsection, information could be credible even without certification.
21
Lewis and Sappington (1994) actually mention other possible ways of transmitting credible information
like allowing consumers to test the product themselves.
147
Advertising in Markets
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.