Finally, theoretical settings involving search goods allow for analyzing the choice of
disclosing price information in ads. A firm may choose to disclose only horizontal match
information with no price, when it can fine-tune the amount of product information it is
transmitting to the consumer. In a cheap-talk context, the provision of soft quality infor-
mation may be enough to reassure the consumer that the price is not too high. Prices
could also have a signaling role, along the lines of the material in the next section,
and firms might choose to hide them because showing them might be an adverse signal
about product information.
4.4. ADVERTISING AS A SIGNAL
From the analysis of product advertising, it appears that firms that sell high-quality prod-
ucts want to make this information visible to consumers to avoid the standard inefficien-
cies associated with adverse selection. Although they may be helped in this respect by the
legal restraints on misleading advertising, it is often not feasible to make verifiable quality
claims to which the law would lend the hoped-for credibility. A substantial body of eco-
nomic research in the wake of
Nelson (1974) has explored the indirect role that ad expen-
ditures can play in making this information credible. I now review this literature, starting
with Nelson’s main arguments and moving on to the game theoretic analysis of the sig-
naling role of advertising. I end with the related topic of advertising as a coordination
device.
4.4.1 Nelson Again
As I emphasized in the previous section, Nelson (1974) argues that ads cannot provide
much direct information about experience goods. His main point is that, even if ads have
a very limited information content, they can still transmit information to consumers indi-
rectly. The general argument is that consumers expect a firm that advertises more to be a
“better buy” (
Nelson, 1974, p. 732). He offers three arguments for why this should be
the case.
The main argument is that a firm that is more efficient in “… producing the utility
that consumers seek” benefits more from advertising and reaching more consumers
(p. 732). What Nelson calls a greater efficiency is the ability of a firm to provide a certain
level of quality at a lower cost. It is unclear how this should be formally captured. One
simple way of making this formal would be to consider the difference s c, where s is
quality and c is marginal cost. A firm for which s c is larger earns a higher per-consumer
surplus and hence has a greater incentive to advertise to get more customers. Nelson’s
argument also requires that a higher s c yields a higher consumer surplus. Then con-
sumers should rationally be attracted by firms that advertise more.
Nelson provides two additional arguments to support the view that advertisements
are indirectly informative. The first is that advertising might signal high-quality products
166 Handbook of Media Economics
because they are more likely to be purchased repeatedly. Nelson bases his reasoning on
the idea that consumers tend to forget the name of the brands they have previously pur-
chased and liked, and one role of advertising is to remind them of the brand’s name.
A firm that sells a bad product would not find it profitable to remind consumers of its
brand name because consumers would not purchase again a product they have not
enjoyed in the first place. Hence, consumers who have not tried the product yet should
infer that a product that is advertised a lot has good quality. There is some form of exter-
nality exerted by informed consumers who have experienced the product’s quality before
on uninformed ones.
44
The other possibility suggested by Nelson is that advertising indirectly informs con-
sumers about how well an experience good is matched to their tastes and needs. The idea
here is that a consumer expects that the products that are advertised in the media she
watches, reads, or listens to are products that she is potentially interested in and will want
to buy again if she tries them. This is a standard targeting argument. Yet, if a firm has the
incentive to post its ad in the right media, then it should also select to make a proper
cheap-talk claim in its ad about which customers should like the product. The same out-
come could then be achieved with cheap talk along the lines of the analysis in
Section
4.3.5.1
. Indeed, as I have already pointed out, Nelson mentions that repeat purchases
may constitute a basis for credible cheap talk in ads for experience goods.
Nelson’s central message is that advertising may inform even if it contains no direct
information. An alternative view could be that such advertising is purely persuasive or
constitutes some form of complementary good as discussed in the introduction.
Ackerberg (2003) uses a structural model of consumer behavior to test the two alternative
views. He studies advertising for Yoplait yoghurts in the US. In his model, advertising
may either affect the consumer by informing her that the product is high quality (if more
ads enhance demand through this channel, it is interpreted as quality signaling). But
advertising can also affect utility through a prestige effect, where more advertising means
more prestige. It is possible to identify the two effects because consumers with no expe-
rience of the product are affected by both effects whereas consumers who have already
consumed the product are only affected by the prestige effect. Using longitudinal user
exposure data, he finds that only the information effect is significant.
Nelson stresses the self-enforcing nature of the mechanisms he describes but does not
use any formal equilibrium concept. In particular, he does not explore all the strategic
dimensions involved with attempts by firms to manipulate the consumers’ beliefs or
how these beliefs might react to unexpected changes in a firm’s behavior. Although
the reformulations described below are not entirely consistent with Nelson’s original
44
Linnemer (2002) proposes a model of advertising as a signal where some consumers are informed. See also
Lauga (2012), that I discuss in Section 4.6.2, on the role of advertising as a reminder.
167
Advertising in Markets
constructions, his work has had a considerable influence on the theoretical analysis of
advertising.
4.4.2 Quality Signaling and Money Burning
I here discuss the main insights from this substantial literature, drawing on Bagwell (2007,
subsections 6.1 and 6.2)
and Tirole (1988, chapter 2, subsection 2.6.1.2). Typically, the
frameworks used in the literature consider the signaling of quality both through adver-
tising and price. To build intuition, and following Bagwell, I start with a static setting.
4.4.2.1 Static Quality Signaling
The following framework is inspired by
Overgaard (1991). It is somewhat related to
Nelson’s conjecture that consumers should expect firms that are “efficient” to advertise
more. A monopoly firm sells a product whose quality is either high (s ¼h) or low (s ¼‘).
After learning its exogenous quality, the firm chooses a price p 0 and advertising expen-
diture A 0. Then consumers observe (p, A) but not s and form beliefs about the prob-
ability that quality is high, β ¼bp, AðÞ. Then they purchase quantity q ¼D β, p, AðÞ,
where D is differentiable on 0, 1½IR
2
+
, strictly increasing in β and A, and strictly
decreasing in p. Marginal production cost is constant and given by c
s
for quality s.
Throughout I assume that, for a given perceived quality β, a larger marginal cost
implies a higher profit-maximizing price and a lower profit-maximizing level of ad
expenditures. I also assume that both are increasing in β. The intuition for why a firm
with a higher marginal cost might advertise less is that it finds it less profitable to expand
its demand (which is also why it charges a higher price). This is what
Tirole (1988) calls
the Schmalensee effect after
Schmalensee (1978). Finally, D is assumed to be such that
the profit function p c
s
ðÞD β, p, AðÞA is single peaked in p and A and reaches some
strictly positive value.
Note that here advertising has a direct impact on demand, which is consistent with
Nelson’s argument. It could, for instance, be an increase in the number of consumers that
become aware that the firm exists as in the monopoly model discussed by
Bagwell (2007,
p. 1752)
, although it is important here that all consumers who are reached by an ad can
observe A perfectly (which is not the case if consumers are reached randomly by ads sent
by the firm
45
). Also note that actual product quality only enters the firm’s profit through
its production costs. Hence, if c
h
¼c
‘
, there cannot be a separating equilibrium because
whatever profit the high-quality firm can obtain by convincing consumers that it is high
quality can also be achieved by the low-quality firm with the same consumer beliefs.
45
There could possibly still be a signaling role for advertising, to the extent that consumers receiving mul-
tiple duplicates of the ad would infer that A is large. See
Hertzendorf (1993) and Moraga-Gonzalez (2000)
for settings where advertising intensity is observed with noise.
168
Handbook of Media Economics
Now consider what would happen if the firm could use only one signal. First assume
that it must spend some fixed amount A on advertising. Let p
s
(β) denote the profit-
maximizing price for a quality s firm with consumer beliefs β .Ifc
h
> c
‘
, then
p
h
β
ðÞ
> p
‘
β
ðÞ
for any β. Now consider a separating equilibrium
46
where firm type s
charges price p
s
*, with p
h
6¼p
‘
. After observing p
s
*, consumers believe quality is s so
β ¼1 if price is p
h
* and β ¼0 if price is p
‘
*. As is standard in the literature,
I concentrate on the least cost separating equilibrium (which is selected by standard equi-
librium refinements). It is the separating equilibrium at which the profit of the high-
quality firm is the largest.
In a separating equilibrium, the low-quality firm charges p
‘
(0) and earns its full infor-
mation monopoly profit, denoted here π
‘
.
47
If π
‘
is strictly below the profit that a type
firm earns by charging p
h
(1) while consumers believe it is high quality, then there is an
equilibrium at which the high-quality firm earns its full information monopoly profit as
well while charging p
h
¼p
h
1ðÞand this is obviously the least cost separating equilibrium.
If, on the contrary, by charging p
h
(1) while consumers believe it is high quality, a low-
quality firm can earn more than π
‘
, then there cannot be a separating equilibrium at
which firm type h charges p
h
(1) (the low type would then choose to mimic the high type).
Because profit is single peaked in price, and p
‘
1ðÞ< p
h
1ðÞ, a low-quality firm’s profit is
decreasing at p
h
(1) if consumers believe its product is high quality. Hence in order to
obtain a separating equilibrium, it suffices to increase p
h
* sufficiently above p
h
(1) so that
a type ‘ firm would earn no more than π
‘
by deviating to p
h
* while consumers believe it is
high quality.
48
The analysis of the case c
h
< c
‘
is the symmetric of the analysis above. Now,
p
h
1ðÞ< p
‘
1ðÞand type ‘’s profit for β ¼1 is strictly increasing at p
h
(1). Then, to obtain
the least cost separating equilibrium, it suffices to decrease p
h
* just enough below p
h
(1) so
that a low-quality firm earns no more than π
‘
by deviating to p
h
* to convince consumers it
is high quality.
The logic applied to price signaling can readily be adapted to the case of signaling
through advertising assuming that the firm is forced to charge a price p independent
of its quality. Let A
s
(β) be the profit-maximizing level of ad expenditures for firm type
s if consumers hold beliefs β. Let A
s
* be the type s level of ad expenditures in the least cost
separating equilibrium. Once again, the choice of the low type is not distorted from its
46
I consider sequential equilibria. As in other discussions in this chapter, I do not fully describe the equi-
librium: in particular I usually omit specifying beliefs off the equilibrium path that sustain the equilibrium
behavior.
47
This is because the high-quality firm has no incentive to pretend it is low quality, so an equilibrium does
not require that the low-quality price be distorted.
48
Such a price p
h
* indeed exists, thanks to the single-crossing condition that the price derivative of profit with
a high marginal cost is larger than the price derivative of profit with a low marginal cost. Similar single-
crossing conditions are implicitly used repeatedly in the arguments below.
169
Advertising in Markets
full information level, A
‘
¼A
‘
0ðÞ, and I again call the corresponding profit π
‘
.Ifc
h
> c
‘
,
then A
h
1ðÞ< A
‘
1ðÞ(from the Schmalensee effect) and the low type’s profit if consumers
believe it is high quality is strictly increasing at A
h
(1). Then the least cost separating equi-
librium is obtained by distorting A
h
* downward from its full information profit-
maximizing level A
h
(1) just enough that a low-quality firm could not earn more than
π
‘
by deviating to A
h
* and having consumers believe it is high quality. An analogous logic
is at work when c
h
< c
‘
so that in the least cost separating equilibrium, ad expenditures by
a high-quality firm are distorted upward from the full information profit-maximizing
level.
Bagwell (2007) characterizes the least cost separating equilibrium in his setting
adapted from
Overgaard (1991). He shows that both signals are used jointly. If c
h
> c
‘
,
then price (respectively advertising) is distorted upward (respectively downward) from
its full information profit-maximizing level. The reverse distortions arise if c
h
< c
‘
. Only
the last case is consistent with the prediction of
Nelson (1974) that “good buys” advertise
more. When the high quality is more costly to produce, it is unclear which product is a
good buy in the sense of Nelson and his prediction holds only if the low quality is a good
buy. Besides, as I already mentioned, if costs are identical for both qualities, then there
is no separating equilibrium although the high-quality firm is efficient according to
Nelson’s interpretation.
Another important difference with Nelson is that, here, product information is trans-
mitted through pricing as well as advertising. Recall that Nelson assumes there are no
laws on misleading advertising so the price cannot be used as a joint signal with adver-
tising (posting a price in the ad would be pure cheap talk). In Nelson’s argument, con-
sumers go to firms that advertise a lot because they expect a low price for the quality they
will get. I return to this idea in
Section 4.4.3.
Finally, as shown by
Bagwell (2007), advertising would not be used as a signal if it was
purely dissipative so that it would have no direct impact on demand. The separation of
types in the equilibria described above is achieved because it is more costly for the low-
quality firm than for the high-quality firm to distort its advertising level. Consider, for
instance, the case where c
h
< c
‘
. If advertising was purely dissipative, then if the high-
quality firm is willing to spend an amount A on advertising so that consumers believe
it is high quality, then so is the low-quality firm.
49
As I now show, things are quite dif-
ferent if the product can be purchased repeatedly.
50
49
In the reverse case, since both full information profit-maximizing levels are zero, it is not possible to distort
them downward.
50
Linnemer (2012) shows that dissipative advertising may be a signal of quality in a static setting where qual-
ity and cost are not perfectly correlated so private information is two-dimensional. He uses a continuum of
types with full support and his results are somewhat reminiscent of Nelson’s prediction that more efficient
firms advertise more.
170
Handbook of Media Economics
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.