payments. In particular, content providers have an incentive to reduce the volume of
time-sensitive traffic that they send.
Summarizing the state of the net neutrality debate is beyond the scope of this article
(for a recent survey on the academic analysis of net neutrality, see
Kra
¨
mer et al., 2013).
The important lesson emerging from the debate is that regulatory decisions affect the rent
distribution between content providers and ISPs. This, in turn, may affect the strategy of
media platforms. In particular, if congestion is not priced, content providers may add a lot
of traffic stemming from advertising (e.g., advertising preceding videos). If congestion is
priced, media platforms may obtain a larger fraction of revenues from charging con-
sumers directly.
23
10.4. USERS CHOOSING MEDIA CONTENT
In traditional audiovisual media, such as television, consumers usually need to choose
which content to consume at any given point in time and which to dismiss. Consider
a television viewer who is interested in two different movies and one sports game broad-
cast by different stations on the same evening. This viewer has to choose one program to
watch but, by making this choice, misses the other programs. This problem of linearly
progressing content of TV or radio is absent in online media offers.
Content provided by Internet platforms can be quite durable. For example, media
libraries allow users to access content at any time. Thus, the Internet is a nonlinear
medium in which each consumer can choose her preferred time and order of content
consumption.
In this respect, Internet media offer content at the individually preferred time, in con-
trast to the predefined time slots of traditional audiovisual media. So it shares important
features of (digital) VCRs and on-demand content with respect to time shifting, but
unlike these other forms of on-demand content, the cost to obtain the same content
at different points in time on the Internet is almost negligible. In fact, another website
is “just one click away,” and accessing it does not require costly hardware. For traditional
media, time shifting requires special hardware devices such as VCR or PVR setup boxes
(e.g., offered by TiVo or DirecTV). In addition, accessing multiple websites is often an
almost mindless activity, while deciding which content to videotape is a more conscious
decision.
Consequently, online media consumers are usually multi-homers, whereas in tradi-
tional media markets, consumers are more likely to choose one outlet and stick to it.
Consider, for example, the market for newspapers: Most consumers read only one daily
23
Moreover, the pricing of congestion may affect the choice of format of media content. For instance, video
consumes more bandwidth than text, and with video, a higher resolution requires more transmission
capacity.
470
Handbook of Media Economics
newspaper (if any) due to time constraints and often stick to this choice for a long time.
24
Also, during the course of an evening, TV viewers who want to watch a movie usually
choose one and single-home on the channel showing the movie. For the other side of the
market, this implies that an advertiser can reach a particular consumer only by placing ads
in the particular newspaper that this consumer is reading, or by placing commercials dur-
ing the movie that the viewer is watching. To inform a large number of consumers,
advertisers need to buy ads on multiple outlets due to consumers’ single-homing behav-
ior. This problem is captured by the seminal competitive bottleneck model of
Anderson
and Coate (2005)
and follow-ups.
If, instead, consumers choose multiple outlets, an advertiser can reach a consumer not
only on a single platform but also on multiple ones. In this respect, platforms lose their
monopoly power of delivering consumers’ attention to advertisers. In the competitive
bottleneck model, in order to attain such a monopoly position, platforms fight for the
exclusive turf of consumers, thereby capturing rents on the advertiser side but dissipating
parts of these rents to consumers. In a market with multi-homing on both sides, this is no
longer necessarily true. This implies that the well-known force that competition intensity
is determined by the strength of business stealing on the consumer side is less relevant in
online media markets. Since consumers are active on multiple platforms, new forces
come into play and old ones are probably disabled. This can affect platforms’ content
choice.
In
Section 10.4.1, we provide different formalizations of multi-homing on the con-
sumer side. We focus particularly on implications for competition and distinguish them
from models with single-homing consumers. Before doing so, we note that if advertisers
could perfectly coordinate their messages, then multi-homing would be equivalent to
single-homing. Consider the situation in which advertising has decreasing returns-to-
scale—that is, the first impression is very valuable, but further impressions are less
valuable because there is a probability that the consumer has already noticed the ad
on another platform. If advertisers can perfectly coordinate their messages, they can
prevent a consumer from being exposed to the ad multiple times. For example, on
TV, this requires that an advertiser choose the same time slot for its ads on each station.
Therefore, competition in this model is equivalent to competition in a model with
single-homing consumers.
Anderson and Peitz (2014a) use the formulation to study
advertising congestion. We discuss this paper in more detail in
Section 10.6.2.
24
However, Gentzkow et al. (2014), using historical data on newspaper readership, find that even in the
newspaper market, multiple readership is quantitatively important. In particular, in their data, 15% of
households that read a daily newspaper read two or more newspapers.
471
The Economics of Internet Media
10.4.1 Consumer Choice with Multi-Homing and Its Implications
on Content
One of the first attempts to allow consumers to combine consumption of multiple prod-
ucts was done in the
Hotelling (1929) framework in a one-sided market. Suppose that
there are two platforms, 1 and 2. The content provided by each platform is interpreted
as its location on the Hotelling line. Platform 1 offers content α and platform 2 offers
content 1 β. Most of the literature works under the assumption that a consumer sub-
scribes to only one platform. This implies that the disutility of a consumer located at x
from not consuming the preferred content is g(jxαj)org(j1β xj), depending on
which platform the consumer is active, where g is an increasing function.
Anderson
and Neven (1989)
extend this formulation by allowing a consumer to consume any
mix of the two contents of
ωα +1ωðÞ1 βðÞ;
with 0 ω 1. The disutility incurred by a consumer at x under the assumption of qua-
dratic disutility is ωα +1ωðÞ1 βðÞxðÞ
2
. Therefore, the consumer can obtain her
optimal content by combining the existing content in the right way. The Hotelling
model with content mixing can be straightforwardly interpreted in the media market
context. Suppose that each consumer has some amount of time that she can allocate
between the two platforms. Then, the consumer spends a share ω(x) of this amount
on platform 1 and 1ω(x) on platform 2.
Suppose that consumers are uniformly distributed on the interval between 0 and 1.
Although the utility formulation is very different from the standard single-homing
Hotelling model, the resulting aggregate demand function is exactly the same if con-
sumers pay for the amount of time they spend on a platform.
25
To see this, recall, first,
that in a traditional Hotelling model, the aggregate demand of firm 1 is
D
1
¼
1+α β
2
+
p
2
p
1
21α βðÞ
: (10.1)
Let us now briefly derive the aggregate demand in the mixing model. The utility function
of a consumer located at x is
vp
1
, p
2
, α, β, x
ðÞ
¼u
0
ωα +1ω
ðÞ
1 β
ðÞ
x
ðÞ
2
ωp
1
1 ω
ðÞ
p
2
;
where u
0
is the gross utility from using the platform. Maximizing with respect to ω,we
obtain
25
Anderson and Neven (1989) consider such linear pricing and find that the model gives similar results as the
standard Hotelling model.
Hoernig and Valletti (2007) analyze two-part tariffs and find that consumers
then do not necessarily choose their preferred product mix because they need to pay fixed fees to both
platforms.
472
Handbook of Media Economics
ω xðÞ¼
21α βðÞ1 β xðÞp
1
+ p
2
21α βðÞ
2
for
α +
p
2
p
1
21α β
ðÞ
< x < 1 β +
p
2
p
1
21α β
ðÞ
;
while ω ¼1 for x α + p
2
p
1
ðÞ= 21α βðÞðÞand ω ¼0 for x 1 β + p
2
p
1
ðÞ=
21α β
ðÞðÞ
. Therefore, consumers whose preference is close to the content of one
of the platforms do not mix, while those located at less extreme positions choose to
mix the content. Determining the aggregate demand of firm 1, we obtain
D
1
¼α +
p
2
p
1
21α βðÞ
+
ð
1 β +
p
2
p
1
21αβðÞ
α +
p
2
p
1
21αβðÞ
ω xðÞdx ¼
1+α β
2
+
p
2
p
1
21α βðÞ
;
which equals
(10.1), implying that the two formulations are equivalent. In other words,
single-homing in the Hotelling model can also be interpreted as multi-homing of con-
sumers who mix content. As a consequence, competition plays out in exactly the same
way in the two models.
Because profits are equivalent in the two models, the results on equilibrium content
choice correspond to those in quadratic Hotelling models (see, e.g.,
d’Aspremont et al.,
1979
). In particular, if α and β are restricted to be positive, firms in a two-stage location-
cum-price game choose maximal differentiation in equilibrium. That is, α ¼β ¼0, and
firm 1 is located at 0 whereas firm 2 is located at 1.
26
If α and β can be negative, in equi-
librium α ¼β ¼1=4, implying that firms choose content outside the unit interval. In
media markets, this means that platforms may polarize content even if it does not match
the heterogeneity of tastes.
Although the positive results of content mixing in the Hotelling model coincide with
the ones of the standard analysis, the normative economics are different. For example, as
shown by
Anderson and Neven (1989), the socially optimal locations in the mixing
model are indeed 0 and 1. Hence, if firms are bounded to choose locations within the
consumer taste space, they choose the welfare-optimal ones. By contrast, in the tradi-
tional Hotelling model, welfare-optimal locations are in the interior of the taste space
(at α ¼β ¼1=4), leading to excessive differentiation.
We now extend the framework of mixing content to a two-sided ad-financed media
model—i.e., platforms obtain their revenues from advertisers instead of consumers. That
is, prices p
1
and p
2
are equal to zero, but consumers view advertising levels a
1
and a
2
on
the platforms as a nuisance. Therefore, the utility function of a consumer located at x is
26
Peitz and Valletti (2008) show that this result holds true also in the context of advertising-financed media,
in which platforms receive revenues from consumers and advertisers.
473
The Economics of Internet Media
va
1
, a
2
, α, β, xðÞ¼u
0
ωα +1ωðÞ1 βðÞxðÞ
2
γωa
1
γ 1 ωðÞa
2
;
where γ represents the nuisance parameter of advertising.
Gal-Or and Dukes (2003) use this framework to analyze content choice in media
markets. They find that platforms choose the same location on the Hotelling line, a result
in stark contrast to the one obtained in the traditional framework. In the model of
Gal-Or
and Dukes (2003)
, consumers obtain content for free but incur a disutility from adver-
tising. Advertisers compete in the product market and inform consumers about their
products via advertising. A lower advertising intensity leads to less-intense product mar-
ket competition, implying that advertisers’ prices and profits are higher. Each platform
and advertiser negotiate about the payment made by the advertiser in return for adver-
tising on the platform. In this negotiation, the two parties maximize their joint surplus
and share it equally. By choosing minimal differentiation, platforms reduce the amount of
advertising in equilibrium because advertising is a nuisance to consumers. Hence, intense
competition for consumers in the media market results in low advertising levels. Via min-
imal differentiation platforms commit to a low advertising intensity, thereby reducing
product market competition. This, in turn, allows advertisers to reap higher profits. Since
platforms do not set advertising prices but negotiate with producers, minimal differen-
tiation does not lead to zero advertising prices but increases the surplus in the
negotiation.
27
Gabszewicz et al. (2004) also consider multi-homing consumers but do not consider
advertiser competition in the product market. Instead, they assume that the disutility of
consumers is convex in the advertising level—that is, the disutility from advertising is a
i
θ
,
with θ 1. As they show, in equilibrium, platforms may choose a location in the interior
range of the Hotelling line; that is, the content is relatively similar.
28
In fact, the equi-
librium locations are closer to each other, the larger is θ. That is, the program diversity
is smaller, the larger is the advertising aversion of consumers (measured by increasing
marginal disutility of advertising).
These papers are based on the idea that consumers mix the time that they spend on
different platforms, keeping the total amount of time fixed. However, in most markets,
the availability of content increases consumption. These features have been incorporated
27
Exclusive advertising contracts are a different means to mitigate competition between advertisers. These
contracts are standard practice, e.g., in the US television industry. By offering single-category advertising
rights, a platform guarantees not to sell another slot in the same advertising break to any close competitor.
Therefore, consumers are less informed about competing products, yielding higher profits for advertising
firms. For a detailed analysis, see
Dukes and Gal-Or (2003).
28
Peitz and Valletti (2008) also find that platforms do not choose “maximal” differentiation to obtain higher
surplus from advertisers in case consumers obtain content for free. Their model is cast in a framework in
which all consumers single-home and advertisers multi-home. However, as they note, their results carry
over to a setting in which consumers mix content.
474
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.