contextually targeted ad, such as a banner ad for a new computer displayed on a site
devoted to computing and technology. The other is a highly visible ad, which users
might consider obtrusive. For example, an ad is considered obtrusive if the ad is part
of an in-stream audio or video, if it is a pop-up window, or if it automatically (non-user
initiated) plays audio or video, among other characteristics.
Goldfarb and Tucker (2011b)
find that the effect of targeted ads alone (without obtrusiveness) and the effect of obtru-
siveness alone (without targeting) have a positive influence on the effectiveness of adver-
tising. However, both strategies in combination nullify this effect and are ineffective. An
explanation for this can be that consumers perceive themselves to be manipulated, which
reduces their purchase intentions. In particular, when exposed to targeted ads that are
obtrusive, consumers may have privacy concerns.
Goldfarb and Tucker (2011b) find evi-
dence that supports this view.
More generally, advertising across different types of media and consumer online
behavior are connected.
Joo et al. (2014) study how advertising in an offline medium
(television) affects consumers’ online searches and thus search engine advertising. They
consider TV advertisements of financial services and analyze how these commercials
affect consumer search behavior. They find a significantly positive effect. For example,
a few hours after being exposed to a TV ad for a particular brand, searchers have a stronger
tendency to enter branded keywords instead of generic keywords.
62
Overall, this suggests that, while online and offline advertising are substitutes, offline
advertising stimulates online product search. Further research, both theoretical and
empirical, could be fruitful to establish a solid pattern that links the influences of adver-
tising in one medium to consumer behavior in another.
10.6. MEDIA PLATFORMS MATCHING ADVERTISING TO USERS
Internet media facilitate the targeting of ads to specific consumers. Traditional media pro-
vide tailored offers such that consumers self-select into particular programming and con-
tent. Advertisers then benefit from the correlation of consumer tastes with media content
and with advertised products. Clearly, such tailoring strategies are also available on the
Internet and were analyzed in the previous section. A novel feature of advertising on
the Internet is the wealth of personal data available to data providers, which allows
the matching of advertising to consumer tastes on media platforms irrespective of the
media content that is consumed.
63
While this wealth of data raises serious privacy and
data protection issues (not analyzed in this chapter), it also affects the way media platforms
62
Rutz and Bucklin (2012) find a similar result for online advertising and online search.
63
We are not claiming that the tailoring of advertising is a completely new phenomenon. For instance,
advertisers may use personal information when sending out coupons by mail.
511
The Economics of Internet Media
operate.
64
In addition, since consumers mostly visit multiple sites, excessive advertising
beyond what consumers can digest also arises on the Internet.
10.6.1 Tracking and Personalized Offers
The Internet has opened new ways to track consumers by placing cookies. Cookies are
small pieces of data sent from a website, which track the user’s activities. To the extent
that previous user behavior allows inferences on users’ current tastes, it becomes possible
to, at least partly, avoid wasteful impressions. Google explicitly writes in its information
to users: “We use cookies to make advertising more engaging to users and more valuable
to publishers and advertisers.” Google then provides a more detailed explanation on the
use of cookies: “Some common applications of cookies are to select advertising based on
what’s relevant to a user; to improve reporting of campaign performance; and to avoid
showing ads the user has already seen.” While perfect public tracking would, in partic-
ular, allow websites to best match advertising to users, in many markets media platforms
may not share tracking information. Thus, tracking is often imperfect.
Tracking may allow for the segmentation of consumers according to some broad
categories without fully personalizing the targeting of ads. This segmenting of the consumer
pool may be based on past purchases (
Malthouse and Elsner, 2006).
65
This information helps
to increase the likelihood that the advertiser’s product and the consumer’s taste match.
Targeting can be based on personal characteristics. Some of the theoretical models
presented below are based on this idea. Advertising has been shown to be more effective
when it is targeted to particular consumers using consumer browsing behavior (
Chen
et al., 2009
) or using inferred or observed demographics as consumer characteristics
(
Joshi et al., 2011).
66
Thus, the empirical literature indicates that tracking can increase
ad effectiveness. A more challenging question is to uncover the impact of tracking on
industry outcomes.
Beyond the collection of information from cookies, the matching of advertising to
users may rely on information provided by database marketing companies.
Marwick
(2014)
provides some information on the second largest company in the industry,
Acxiom. According to
Marwick (2014), Acxiom has 23,000 computer servers and pro-
cesses more than 50 trillion data transactions per year, keeping records on hundreds of
million US residents. Data include 200 million mobile profiles, information gathered
from publicly available records (such as home valuations and vehicle ownership), infor-
mation about online behavior (1.1 billion browser cookies, information on browser
advertising, and other information), as well as data from customer surveys and offline
64
Privacy issues are discussed in Chapter 11 of the Handbook.
65
Other empirical work segments consumers according to their cognitive style (Hauser et al., 2009).
66
This empirical work combines tailoring and tracking, as it combines consumer characteristics with content
matching.
512
Handbook of Media Economics
buying behavior. On average, for each US resident, Acxiom keeps about 1500 pieces of
data. Thus, Acxiom has a wealth of information that it can sell to interested parties, in
particular with the aim to better match advertising or services to user tastes.
While on traditional ad-financed media, the user pays with her attention, on Internet
media, the user pays not only with her attention, but also with her personal data. Thus,
websites including Internet media may make revenues even if they neither charge users
nor carry any advertising. They can accomplish this by opening a third source of
revenues—selling user information.
A number of theoretical efforts help in understanding the forces at play when media
platforms track users or rely on third-party information in their effort to best match
advertising to users. The model presented at the end of this subsection explicitly includes
the sale of user data for the purpose of targeting.
A media platform may provide tracking information about consumers to advertisers.
Doing so allows advertisers to bid for ads conditional on the information they receive.
When advertising space is scarce, advertisers operating in such an environment internal-
ize that in case of tracking their bids will only be successful if they provide better matches
to consumers than other advertisers; absent tracking advertisers offer similar expected
match quality. As a consequence, advertisers set higher retail prices with tracking infor-
mation than without. While tracking improves average match quality, leading to higher
retail prices and thus larger industry profits, it also reduces the share of industry profits that
can be extracted by the platform, as advertisers set prices prior to learning consumer
types.
67
Thus, it is not obvious whether the platform benefits from tracking.
de Cornie
`
re and De Nijs (2014) formalize this tradeoff and investigate the platform’s
incentives to install a tracking technology. Here, through a second-price auction,
a monopoly media platform sells a single advertising slot to n advertisers.
68
This slot gives
exclusive access to the consumer. Thus, sellers act as monopolists in the product market.
The timing of the model is as follows: First, the platform decides whether to install a
tracking technology. Second, advertisers simultaneously set the product price p
i
,
i ¼1,…,n. Third, the consumer type is revealed to advertisers under tracking; it remains
unknown otherwise. Fourth, advertisers simultaneously place bids for the advertising slot
conditional on the information they received. The consumer is matched to the winning
advertiser.
67
If prices were set conditional on consumer types and thus advertisers customize retail prices, they would
extract a larger fraction of consumer surplus. However, this would drive up bids. Advertisers would be
worse off since the difference between valuation of the winning bid and the valuation of the second
highest bidder (and thus the price in the auction) shrinks when advertisers can customize the retail price
compared to the setting where they cannot.
68
Selling a single slot is perhaps the simplest setting and avoids the need to consider alternative multi-unit
auctions. Suppose that advertisers are potential competitors in the market place. Then, it is optimal to sell a
single slot if monopoly profits exceed industry profits with two or more firms.
513
The Economics of Internet Media
A consumer is of type (θ
1
, …, θ
n
), where θ
i
is i.i.d. across products and distributed
according to F with density function f on
θ,
θ
hi
. Type θ
i
for product i gives rise to a
conditional demand function D(p
i
; θ
i
). A higher type is assumed to be associated with
larger demand for the respective product (e.g., a larger probability to buy the product);
i.e., Dp
i
;θ
i
ðÞ> Dp
i
;θ
0
i
if and only if θ
i
> θ
0
i
. The profit of an advertiser selling to a con-
sumer is π
i
p
i
, θ
i
ðÞ¼p
i
cðÞDp
i
; θ
i
ðÞ. Absent tracking, if an advertiser’s bid is successful,
its expected profit gross of the advertising cost is
Ð
θ
θ
π
i
p
i
, θ
i
ðÞf θ
i
ðÞdθ
i
. The profit-
maximizing product price p
NT
solves
ð
θ
θ
@π
i
p
i
, θ
i
ðÞ
@p
i
f θ
i
ðÞdθ
i
¼0
in p
i
. Since advertisers are homogeneous at the bidding stage, the media platform can
extract the full expected industry profit
Ð
θ
θ
π
i
p
NT
, θ
i
f θ
i
ðÞ
dθ
i
.
With tracking, the profit-maximizing product price p
T
solves
ð
θ
θ
@π
i
p
i
, θ
i
ðÞ
@p
i
F
n1
θ
i
ðÞf θ
i
ðÞdθ
i
¼0
in p
i
, since advertiser i wins the auction if and only if θ
i
is larger than θ
j
, j 6¼i, which occurs
with probability F
n1
(θ
i
). Compared to the case without tracking, this extra term captures
that the firm with the largest θ
i
wins the auction. As a consequence, it will set a higher
price at the pricing stage than without tracking, p
T
> p
NT
. The price under tracking is
increasing in the number of advertisers. Consequently, tracking results in a better match
between advertiser and consumer and increases industry profits. With tracking, adver-
tisers obtain a positive information rent and thus a strictly positive share of industry profits.
When deciding whether to install the tracking technology, the media platform faces
the tradeoff between increasing efficiency (and industry profits) and rent extraction; such
a tradeoff also obtains in
Ganuza (2004).
69
As the number of advertisers turns to infinity,
the product price p
T
turns to the monopoly price of a firm facing a consumer with type
θ
and thus the information rent of advertisers disappears. Hence, for a number of advertisers
sufficiently large, the media platform installs the tracking technology and shares the con-
sumer information with advertisers.
Suppose now that the platform sells multiple advertising slots and advertisers offer
independent products to consumers. Advertising slots are sold through a uniform price
auction such that all advertisers with the highest bids pay the price equal to the highest bid
among losing advertisers. Then, the equilibrium product price is shown to be decreasing
in the number of advertising slots. In addition to the standard price–quantity tradeoff, an
69
The objective function of the platform is to maximize its profit given user participation.
514
Handbook of Media Economics
increase in the number of advertising slots renders winning a slot in the auction less infor-
mative about the expected elasticity of demand. When the number of slots is sufficiently
large,
de Cornie
`
re and De Nijs (2014) show that the platform chooses not to install the
tracking technology since the losing bidder who determines the ad price in the auction
tends to receive a rather bad signal with tracking.
Johnson (2013) also explores the effects of the tracking technology on advertiser
profits and consumer surplus. While he does not include media platforms in his model,
his analysis is useful in obtaining insights about the role of tracking when consumers can
block advertising.
In his model, advertising creates an opportunity for advertisers and consumers to form
a match. There is a mass 1 of advertisers and a mass 1 of consumers. For each advertiser–
consumer pair, the probability of such a match is ϕ, which is distributed i.i.d. across all
pairs according to some distribution function F with positive density of f on [0, 1].
A match generates a surplus Λ for the advertiser and 1 Λ for the consumer. Advertisers
offer totally differentiated products.
The advertiser learns about the match probability with probability ψ and does not
learn otherwise. In this model, improved tracking corresponds to a larger value of ψ.
Thus, the probability of a match is ψϕ+1ψðÞEϕ if the consumer sees the ad.
Advertisers incur a cost of κ > 0 for sending an ad. Consumers have the possibility of
blocking an ad with probability ς. Hence, an advertiser decides to advertise to a consumer
with signal ϕ if
κ 1 ςðÞΛψϕ+1ψðÞE ϕ½½:
A firm that sends an ad to consumers with signal ϕ will send an ad to all consumers with
larger signals. If h is the mass of ads sent by an advertiser, we have ϕ hðÞ¼F
1
1 hðÞas the
signal of the marginal consumer. An advertiser’s profit is, then,
κh +1ςðÞΛψ
ð
1
ϕ hðÞ
xdFxðÞ+1ψðÞhE ϕ½
"#
:
All consumers are exposed to the same number of ads if they decide not to block them.
Each ad they receive generates a nuisance γ from being exposed to it. However, adver-
tising leads to consumption, which enters the consumer’s utility as well. Thus, the
expected utility from receiving h ads is
v ¼γh +1ΛðÞψ
ð
1
ϕ hðÞ
xdFxðÞ+1ψðÞhE ϕ½
"#
:
If consumers block ads, they receive an outside utility u
0
, which is distributed according
to G, a continuously differentiable distribution function with support 1,0ð. Taking h
as given, each consumer compares her expected utility from receiving ads to the outside
515The Economics of Internet Media
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.