informative advertising than on directly informative advertising. This is because it is the
amount spent on the former that conveys the information whereas ads that contain direct
information merely need to reach consumers once. He finds some empirical support for
this view by providing some empirical evidence that ad expenditures for experience
goods are much larger than for search goods. The substantial theoretical investigations
on this issue do not fully confirm Nelson’s predictions. If prices may be used to signal
a product’s high quality (in particular through introductory offers), they often achieve
this objective more effectively than massive advertising expenditures. If high quality is
more costly to produce than low quality, it may even be the case that the seller of a
high-quality product chooses to spend little on advertising to credibly signal its high qual-
ity (thus signaling that it aims at selling little at a high price).
A higher potential for repeat purchases of the product (for instance, if the product is
non-durable like food items) makes massive advertising of an experience good more
attractive to a high-quality seller. It is only in this case that a firm producing a high-quality
item that is more costly to produce than a low-quality one sometimes chooses high ad
expenditures to convince consumers its quality is high. It may then even use advertising
that is purely dissipative and has no direct impact on its demand.
In practice, ads often do not include price information so that large ad expenditures
can then induce consumers to incur the shopping cost that will uncover the price as in the
search good settings considered in
Sections 4.3.3 and 4.3.5. I am not aware of any
research that investigates quality signaling in such a framework. I have illustrated however
how, in this case, ad expenditures may signal a low price for a product of known quality.
When there are potential increasing returns in retailing, ads may help consumers coor-
dinate on buying from large retailers that offer low prices. This coordinating role of
advertising may also arise for products characterized by consumption externalities, either
because of network effects or fashion.
As a final remark, it should be stressed that in order to get definite predictions about
the indirect informativeness of advertising, it is typically necessary to refine the set of
equilibria. Indeed, the outcome hinges very much on how consumers interpret the firm’s
advertising behavior. This may even impact which media the firm chooses to use, as is the
case in the simple flyers versus TV example in
Section 4.4.3. This may also apply to the
non-informative content of the ad. Signaling quality is not merely a matter of conveying
the information that a large amount has been spent on an ad campaign. The way in which
it is transmitted may matter as well. Furthermore, as I explain in the next section, it is also
critical that consumers pay attention to these ads while they are exposed to an increasing
quantity of information of all sorts.
4.5. ADVERTISING TECHNOLOGY
Much of the discussion up to this point ignores the specifics of how the information con-
tained in an ad reaches consumers. I now provide an overview of some of the main
176 Handbook of Media Economics
insights in the literature regarding advertising technology. I start with a discussion of
advertising costs. I then introduce the targeting of ads and finish with the issues raised
by information congestion.
4.5.1 Advertising Costs
I first consider the cost for an advertiser to reach consumers, which I already introduced
in
Section 4.2, and then I discuss how the disclosure costs (that may be more than the cost
of just reaching consumers) affect the firm’s incentives to reveal product information.
4.5.1.1 Advertising Costs and Reach
In his groundbreaking analysis of informative advertising,
Butters (1977) is very careful in
specifying the technology through which advertising reaches consumers. He assumes
each ad has a constant cost b > 0 and that it lands on one out of a population of n con-
sumers, where each consumer is equally likely to be reached. It follows that each ad
reaches each consumer with probability 1/n and, for n large, if a is the number of ads
sent per consumer, the probability that a consumer is reached by at least one ad is approx-
imately ϕ ¼1 e
a
. This is a concave function of a, meaning that the impact of an addi-
tional ad on the reach decreases as the number of ads sent increases. This in turn begets
advertising cost as a strictly convex function of reach, A ϕðÞ¼bn ln 1 ϕðÞ. In other
words, as noted by
Stahl (1994, p. 164), although the marginal cost of sending an addi-
tional ad is constant, the marginal cost of increasing the reach is strictly increasing.
The technology used by Butters is best interpreted as advertising by means of flyers
that are randomly handed out to consumers, where each flyer reaches one and exactly one
consumer. Now if we think of advertising through media outlets, an ad printed or aired
once reaches a potentially large population of consumers. This means that the probability
that any one consumer is reached by one ad is larger than 1/n. As discussed by
Stegeman
(1991, p. 216)
,ifk/n denotes the probability that a consumer is reached by an ad, then,
still taking n to be large, she is reached by at least one ad with approximate probability
ϕ ¼1 e
ka
if the number of ads per consumer is a. This again is a concave function
inducing a rising marginal cost of increasing the reach.
This discussion suggests that there are decreasing returns to advertising intensity, con-
trary to what
Kaldor (1950) postulates (see Bagwell, 2007, p. 1713). Bagwell (2007,
pp. 1731–1732)
discusses a series of empirical works that looks at the impact of advertising
on sales. The evidence from that literature is that there are diminishing returns to adver-
tising. It is important here to make a few remarks on these empirical findings and how
they relate to the theory.
First, what Kaldor had in mind was rather a persuasive form of advertising, as I discuss
in the introduction. The increasing returns he envisaged had more to do with the impact
of advertising on consumer tastes (or perceived tastes) than with the reach technology.
Even if we focus exclusively on informative advertising, the role of advertising in the
Butters environment, which is to ensure potential buyers become aware of the firm’s
177Advertising in Markets
existence, is just one possible informative role. Consider, for instance, the signaling role
described by
Nelson (1974). In his formal analysis, he assumes that the indirect informa-
tion kicks in only if consumers are exposed to enough repetitions of the same ad and he
uses this to argue that experience goods are advertised more than search goods. Then, still
using the Butters technology, reach is determined by the probability that a consumer is
exposed to more than m > 1 ads. Then it is at least theoretically possible that for a low
enough ad activity, the impact of one more message on the reach probability is increasing
in the total number of messages. The empirical work by
Lambin (1976) does find such a
threshold (see
Bagwell, 2007, p. 1732).
A second point is that different media might be used by advertisers depending on how
broad a reach they are trying to achieve. As I suggested in my flyers versus TV example in
Section 4.4.3, using TV advertising might be less costly than flyers when trying to reach a
large population. For instance,
Porter (1976) finds evidence that the impact of an increase
in TV advertising on industry profit is much stronger than the impact of an increase in
overall advertising activity (see
Bagwell, 2007, p. 1733).
Finally, an important ingredient for advertising costs is the pricing behavior of the
media. The actual pricing is not publicly observed, but list prices typically incorporate
some form of quantity discounts. However, an ad that is more likely to reach many con-
sumers (a larger k in my analysis above), say because of a large viewership, is typically
priced higher, implying a higher b.
4.5.1.2 Disclosure Costs
Now I abstract from the choice of reach by advertisers and focus on how the cost of dis-
closing information affects the firm’s decision. My analysis of disclosure in
Section 4.3
assumes no cost and emphasizes that a firm might choose not to reveal information about
its product even if such revelation involves no cost.
A first question concerns the revelation of quality information. One takeaway from
Section 4.3.4 is that, if product information only concerns product quality, then it is fully
revealed by the firm. It is fairly straightforward to show that if there is some fixed cost of dis-
closing quality information, then producers of lower quality products keep quiet so the con-
sumer only learns that quality is not so good without knowing quality precisely. One natural
question is how this private choice of quality revelation compares to what would be socially
optimal.
Jovanovic (1982) analyzes a monopoly setting where disclosing quality is costly and
finds that too much quality information is revealed as compared to the social optimum.
To understand the result, consider the following simple setting. A monopolist sells a
product to a buyer with unit demand whose willingness to pay is the product’s quality s
that may be either s
‘
or s
h
, where 0 < s
‘
< c < s
h
and c is the unit production cost. The firm
knows s while the buyer only knows that s ¼s
h
with probability α 2 0, 1ðÞ. Let s be the
expected quality given the buyer’s beliefs. As in the standard persuasion game, the firm
may provide certified information about its product quality through advertising.
178 Handbook of Media Economics
Now consider the incentives of a high-quality firm to reveal that s ¼s
h
. By disclosing
quality, it can charge the consumer s
h
rather than s. Hence it will do so as long as the
associated cost is no more than s
h
s ¼ 1 αðÞs
h
s
‘
ðÞ. By contrast, the social benefit
from disclosure is to avoid the product being purchased when it is low quality. The asso-
ciated expected gain in social surplus is 1 αðÞc s
‘
ðÞ< 1 αðÞs
h
s
‘
ðÞ. Thus, for a dis-
closure cost strictly between 1 αðÞc s
‘
ðÞand 1 αðÞs
h
s
‘
ðÞ, the firm chooses to
disclose although it is not socially desirable. An important takeaway from this analysis
is that policies aimed at encouraging quality disclosure are at best wasteful (because they
involve subsidies or enforcement costs
55
) and can actually be detrimental to the market
outcome. Of course, with a different specification of demand, so consumer surplus is
non-zero, they could be desirable for consumer protection. Furthermore, as
Jovanovic (1982, p. 42) notes, if quality is endogenous, then forced disclosure may have
a beneficial incentive role by increasing the firm’s benefit from improving its quality.
When thinking about disclosure costs, one obvious question is how they are related to
the quantity of information contained in the ad. The nature of the information provided
may matter as well and it is not clear in practice that more information is necessarily asso-
ciated with a larger disclosure cost. Even if it is the case, we need a measure of the quantity
of information that can be applied to a general enough class of disclosure strategies.
Manduchi (2013) addresses this issue in the experience good match disclosure problem
discussed in
Section 4.3.3.1. A monopolist sells a product to a buyer with unit demand
where the match is either r ¼r
h
with probability q 2 0, 1
ðÞ
or r ¼r
‘
. Production cost is
c > 0 where r
‘
< c < r
h
. The firm may devise any binary information system such that
the buyer receives one of two signals σ 2 b, g
fg
, where σ ¼g is a good signal in the sense
that it is more likely to be observed by a high-valuation buyer than by a low-valuation
buyer. Absent disclosure costs, the firm would choose to have σ ¼g observed with prob-
ability 1 if r ¼r
h
and with probability 0 otherwise. Manduchi uses an entropy-based mea-
sure of information called “mutual information.” Using a disclosure cost that is linear in
the quantity of information, he finds that the firm reveals information only if there is
enough ex ante uncertainty about the match so that q is in some intermediate range.
In that range, the information provided is mostly used to screen out a low match con-
sumer if q is low, whereas for q large enough information is used to make sure a high
match consumer buys. The probability of a sale is less (respectively larger) than q if q
is small (respectively large) enough.
Next I turn to targeted advertising.
55
Note, however, that the mere possibility for the firm to disclose certified information requires some
enforcement of laws against misleading ads. The analysis here implicitly assumes the corresponding costs
are charged to the firm. Forced disclosure laws may involve extra costs.
179
Advertising in Markets
4.5.2 Targeted Advertising
When a firm faces a heterogeneous population of consumers, an obvious way to save on
advertising costs is to target advertising, which allows for reaching primarily those con-
sumers who have a more inelastic demand and not wasting advertisements on people who
would not buy the product in any case. This can be achieved by advertising in different
media that reach different consumer populations, by advertising differently to different
geographic areas, by exploiting customer information or websites’ tracking information
(see the chapters in this volume on the economics of Internet media (
Peitz and Reisinger,
2015
) and user-generated content (Luca, 2015)).
On top of the obvious cost advantage of targeting, this practice allows a firm to
enhance its market power by making its demand less elastic. This is the main takeaway
from studies that have considered the use of targeted advertising by a monopoly firm,
such as
Esteban et al. (2001). Similar insights may be obtained in a competitive environ-
ment, as illustrated by the simple setting of
Iyer et al. (2005). They consider a differen-
tiated product duopoly with zero production costs. Each firm has a captive consumer base
with measure h > 0 that is only interested in buying its product with reservation price r.
There is also a measure s ¼1 2h > 0 of shoppers who just buy the cheapest product
available. Reaching a measure m of consumers with advertising costs Am, consumers only
buy from a firm if they have received an ad from that firm.
In equilibrium, firms mix over prices. If targeted advertising is not feasible, then the
firm can guarantee itself a profit of rh A by charging r and only serving its home base. It
turns out this is the equilibrium profit if it is positive (that is, for A rh). It is immediately
apparent that if advertising can be targeted, a firm can improve over this by advertising
only to its home base and still charging r, thus earning hrAðÞ. This is viable for a wider
range of advertising costs, A r. In equilibrium, firms do send ads to shoppers with some
probability, but prices are stochastically larger than what they would be if advertising was
not targeted. Results are similar whether or not firms can price discriminate.
This reduction in a firm’s demand elasticity is, however, not the only consequence of
the targeting of ads.
Ben Elhadj-Ben Brahim et al. (2011) study targeted advertising in the
Hotelling setting with advertising reach analyzed by
Tirole (1988), discussed in
Section 4.2.5 of this chapter. There is one firm at each end of the Hotelling segment with
identical and constant marginal costs. Each firm can select its advertising intensity for each
location but must charge a uniform price. Advertising costs to achieve a reach ϕ are
A ϕðÞ¼1=2ðÞaϕ
2
. Ben Elhadj-Ben Brahim et al. (2011) first show that, in equilibrium,
each firm chooses two advertising intensities: a high one for locations at which, if per-
fectly informed, consumers would select the firm’s product at the equilibrium prices, and
a lower one for the remaining locations that are farther away from the firm. Note that the
only benefit from advertising to the latter set of consumers comes from selling to those
who do not get an ad from the competitor (which is preferred by these consumers if they
are perfectly informed). It follows that, for low advertising costs (low a), because each
180 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.