utility of blocking. Hence, the fraction of blocking consumers is ς ¼1 GvðÞ. Consider
the game in which, simultaneously, consumers decide whether to block and advertisers
decide how many ads to send. A pair (ς
, h
) constitutes an equilibrium of this game. In
this model, the cost–benefit ratios for advertisers and consumers, κ/Λ and ϕ/(1 Λ), play
a decisive role. They reflect the cost of an ad relative to the benefit of a successful match.
Johnson (2013) shows that the second-best advertising level (when consumers are free
to block advertising) is smaller than the equilibrium level h
if and only if the cost–benefit
ratio on the consumer side is larger than the cost–benefit ratio on the advertiser side. In
equilibrium, advertisers are indifferent about whether to place the marginal ad. Since the
marginal probability of trade is the same for advertisers and consumers, intuitively there is
socially insufficient advertising if consumers place a higher value on ads than advertisers.
If the reverse holds, advertisers post too many ads.
70
Improved tracking—i.e., an increase in ψ—keeping blocking decisions unchanged,
leads to more advertising if the surplus derived from the marginal ad exceeds the surplus
derived from the unconditional average ad. If advertisers increase advertising due to
improved tracking, this must mean that they gain more from the marginal ad than from
the unconditional average ad. This must continue to hold if consumers adjust their block-
ing decision.
Johnson (2013) shows that improved tracking increases advertiser profits in
equilibrium, even though this may imply more blocking. Whether consumers gain or
lose from improved tracking is ambiguous.
71
Depending on the information available to advertisers, when there are multiple plat-
forms, advertisers may waste impressions by hitting the same consumer more than once even
if tracking is perfect on each platform. If, however, platforms share cookie information—a
practice called cookie matching—advertisers can avoid such multiple exposures.
Athey et al. (2014) consider a market with two platforms that perfectly track con-
sumers on their own platform (i.e., a consumer gets an ad at most once on this platform),
but may not observe the exposure of consumers to advertising on the other platform.
72
They investigate the impact of this lack of cookie matching on market outcomes and
70
The results of van Zandt (2004) and Anderson and de Palma (2009) can also be interpreted as showing the
possibility of socially excessive advertising. In their models, socially excessive advertising may arise because
lower-value ads crowd out higher-value ads, an issue we return to in
Section 10.6.2. In contrast, in
Johnson (2013), a larger ad level encourages consumers to block ads.
71
If advertisers have to pay a media website to place ads (an issue not considered by Johnson, 2013), the
website can manage the ad level of advertisers. Suppose that the media website charges a price per ad.
This price enters the advertiser’s profit function as part of its cost κ. A welfare-maximizing single media
website (which cannot directly control blocking) will then implement the second-best optimal advertising
level. A profit-maximizing, monopolistic ad-financed media website will also internalize the effect of ad
levels on blocking. However, it will typically not implement the second best. The effect of the tracking
technology on total and consumer surplus in such a media market has yet to be explored.
72
Ghosh et al. (2012) provide a related analysis, which also shows that in some cases platforms prefer to share
cookie information, whereas in others they do not.
516
Handbook of Media Economics
media platforms’ profits. In their model, as, e.g., in Ambrus et al. (2015), some consumers
exclusively consume media content of one platform, while the others consume both (see
Section 10.4.1). Thus, there are exclusive consumers and overlapping consumers.
Similar to previous models, media platforms provide access to consumers. To focus on
advertisers’ behavior, suppose first that media platforms do not make any decision (i.e., ad
levels are fixed and exogenous) and that ad prices clear the market for ads. Consider a
continuum of advertisers who are heterogeneous with respect to the profit per consumer
they derive when successfully contacting a consumer. Their behavior determines the
demand for advertising. The advertiser value per consumer is distributed between
0 and some upper bound. Advertisers with a high value per consumer have a stronger
incentive to contact consumers than advertisers with lower value. If there were no over-
lapping consumers (and perfect tracking on each platform), there would exist a marginal
advertiser such that every advertiser with a lower value per consumer would not adver-
tise, whereas all advertisers above this threshold would deliver each impression to a dis-
tinct consumer. Thus, no impression would be wasted and advertising would be
delivered efficiently.
Demand for advertising depends on the tracking technology. If tracking is perfect on
each platform but there is no cookie matching, multi-homing advertisers waste some
impressions as they sometimes show the same ad to switchers twice. This waste, together
with advertiser heterogeneity, implies sorting of advertisers: low-type advertisers single-
home (and miss some consumers), while high-type advertisers (who have a higher oppor-
tunity cost of not informing consumers) multi-home.
This establishes the main insight. With perfect tracking across platforms, the number
of impressions would map one-to-one into the number of consumers reached by an
advertiser. On the contrary, with the above imperfect tracking, some overlapping con-
sumers will see the same ad twice and so their attention is wasted. By increasing the num-
ber of overlapping consumers waste becomes more prevalent under imperfect tracking
and the value of the advertising inventory is further degraded.
The analysis can be extended to allow for the platforms simultaneous choosing adver-
tising inventory. In the presence of overlapping consumers, platforms become essentially
Cournot competitors.
Athey et al. (2014) show that, in equilibrium, an increase in the
fraction of overlapping consumers leads to higher advertising inventories and, in turn,
lower equilibrium advertising prices.
73
Taylor (2013b) provides a different perspective on the role of tracking. In his model,
media platforms choose content quality, taking into account that it increases the likeli-
hood that a consumer does not switch to another platform. The incentive to invest in
content quality is affected by the tracking technology available to the platforms.
73
Due to the heterogeneity of advertisers, the analysis is rather intricate and best-response functions are non-
monotone.
517
The Economics of Internet Media
Taylor’s (2013b) model has a number of features different from those of the three pre-
vious models. First, a key ingredient is product market competition among advertisers;
74
second, consumers are uninformed about content quality before visiting a website and
therefore access media platforms at random. Thus,
Taylor (2013b) focuses on how con-
tent quality affects consumer behavior after consumers have clicked on the media website.
Key to his results is the interaction between product market competition among adver-
tisers and the consumer decision of whether to switch to another media platform (endog-
enous multi-homing).
In
Taylor’s (2013b) model, two ad-financed media websites provide content to a large
number of consumers of measure D. Consumers enjoy media consumption but incur a
cost c for visiting a website. The two media platforms choose content quality q
i
2 0, 1½
and incur cost k(q
i
).
75
Here, content quality measures the probability that a consumer is
satisfied with the content and hence does not need to move to the other website for this
topic. Content quality is a search good; thus, consumers cannot assess content quality
before actually visiting the website. Consumers visit websites sequentially. If a consumer
is satisfied by the content at the first website, she does not visit the second. If she is not
satisfied, she visits the second website if the expected quality is larger than c. Consumers
recall ads and make consumption choices after their media consumption.
Websites will, nevertheless, choose positive quality so as to retain the consumer’s
attention. Why? Websites bundle advertising to content and sell ads to advertisers.
76
More specifically, the website places an ad together with the content offer. Each con-
sumer has a particular consumption interest (e.g., interest in a particular product cate-
gory), and an advertiser matches this interest if it offers a product in this category.
The surplus generated by a successful match between advertiser and consumer is normal-
ized to 1. Advertisers are assumed to be Bertrand competitors if a consumer happens to
see an ad from two advertisers within the same category. As it wants to raise revenues
from selling ads, a website will carry only one ad of a given product category. Each
website extracts the full advertiser profit.
The tracking technology determines the probability of a successful match; it allows
the website to identify the product category of interest with probability ϕ and delivers
any of the other categories with the remaining probability. If a consumer visits both
74
Most theoretical papers on media economics postulate that advertisers have monopoly power over con-
sumers in the product market. As discussed in
Section 10.4.1, an exception is Gal-Or and Dukes (2003).In
their model, as well as in
Taylor’s (2013b) model, intense product market competition creates incentives
for media platforms and advertisers to sign exclusivity contracts of the form that the media platform agrees
not to carry ads from competing advertisers.
75
The cost function satisfies the appropriate properties so as to ensure an interior solution characterized by a
first-order condition.
76
In contrast to consumers, advertisers can assess a platform’s quality because they need to be reassured that
the platform makes retention efforts.
518
Handbook of Media Economics
websites, she is exposed to the advertising on the other website. For simplicity, suppose
that this other website draws on a different set of advertisers and therefore that a consumer
will never see the same ad twice. If the consumer sees another ad in the product category
of interest, she can choose between the two offers and will choose the offer at the
lower price.
As long as some, but not all, consumers visit both websites (i.e., in any interior solu-
tion), there are consumers who observe one successful match and others who observe
two. Thus, advertisers randomize over price. Each advertiser’s equilibrium profit is equal
to its monopoly price 1 the likelihood that it provides the only match. Here, content
quality helps, as it increases the probability that consumers do not visit multiple websites
and thus prevents multiple exposures of advertising.
As
Taylor (2013b) shows, equilibrium content quality can be calculated by equating
marginal benefit from quality investment with marginal costs. This yields
q
¼ k
0
ðÞ
1
ϕ
2
D
2
:
The intuition is as follows: Since consumers do not observe quality, they visit platforms at
random. Therefore, each platform gets D/2 first visits. The benefit of retaining a con-
sumer is the security profit of that consumer, which is measured by ϕ.
77
Better tracking
(i.e., an increase in ϕ) is valuable even in a monopoly context, as it increases the prob-
ability of a match between advertiser and consumer. Everything else being equal, this
increases the incentive to invest in content quality. The tracking technology matters also
for product market competition since better tracking makes it more likely that a con-
sumer who visits both websites encounters two matches. Hence, better tracking makes
product market competition more intense. In this case, the website would extract lower
rents from advertisers. To reduce the probability of multiple exposures to ads within the
same product category, a website has to increase its content quality. Due to this compe-
tition effect, websites may actually overinvest in content quality compared to the welfare-
maximizing solution. More specifically, the result arises because of externalities between
websites: If a website invests in content quality, it does not internalize that its rival’s
opportunity to provide impressions to consumers via ads gets lower. This is detrimental
for welfare because it eliminates trade between some pairs of consumers and advertisers.
This effect is particularly pronounced if ϕ is large because then many of the foregone
impressions would have resulted in a match. As
Taylor (2013b) illustrates, websites
77
In the model, a consumer is retained (i) if she is satisfied with the content of the platform, and (ii) if she is
not satisfied with the content on both platforms and the second platform could not identify her product
category of interest. Therefore, the profit function of a website is D=2ðÞϕ q
i
+1q
i
ðÞ+ðð
1 q
i
ðÞÞ1 ϕðÞÞ. Taking the derivative with respect to q
i
and using platform symmetry yields q*.
519
The Economics of Internet Media
may actually suffer from improved tracking, as they are compelled to invest (excessively)
in content quality.
Tracking is made feasible by data providers that handle large amounts of data that they
obtain from placing cookies.
Bergemann and Bonatti (2015) explore the interaction
between data providers and advertisers. Advertisers face consumers with heterogeneous
match value v 2V . Through cookies, they obtain precise information on consumers’
value. This allows advertisers to segment the consumer side into two groups: a target
group about which it collects detailed information and to which it makes personalized
advertising offers; and an anonymous outsider group in which all consumers receive
the same level of advertising as everybody else in this group.
If an advertiser were fully informed, he would reach a consumer with probability ϕ and
then extract the full surplus v of the match between his product and the consumer. Thus, he
would make revenue vϕ. To reach consumers with probability ϕ, he has to place a(ϕ) ads
on the publisher’s platform, which is assumed to charge p per ad. Hence, a fully informed
advertiser makes profit vϕ pa(ϕ). The advertiser’s problem is that he knows only about
some prior distribution of v and thus may not be able to appropriate the full surplus.
To better extract surplus, the advertiser may acquire data about the individual con-
sumer. Such data provide a signal about the consumer’s match value. This is where cook-
ies come into play. A cookie, bought at a price P
c
for a subset of types W V , allows the
advertiser to identify a consumer v. Hence, for all v 2W , the demand for advertising
space is v ¼pa
0
ϕ
v
ðÞðÞ
, where ϕ
(v) is the full information demand for advertising space.
For all v62W ; the advertiser updates beliefs and acquires ad space according to
Evjv 62 W½¼pa
0
ϕðÞ. Clearly,
ϕ ¼ϕ
Evjv 62 W½ðÞ. Figure 10.9 illustrates the market
environment.
A fully informed advertiser makes profit vϕ
(v) pa(ϕ
(v)), which is convex in v.By
contrast, absent any information, the advertiser’s profit would be vϕ
(E[v]) pa(ϕ
(E[v])),
which is linear in v. An uninformed advertiser advertises too much to low-value consumers
and too little to high-value consumers. Thus, the advertiser has an incentive to buy cookies.
Figure 10.9 Internet advertising based on cookies in Bergemann and Bonatti (2015).
520
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.