just with the public one could count these as effectively exclusives, and even those
shared with the public and one another broadcaster could be charged for at βv.After
allowing ads, advertise rs can reach such consumers through the public broadcaster
and thi s demotes their market value. If the value of βv is low, or there are a lot of shared
links that are thus devalued, the toughe r competition in the ad market will more than
offset any demand diversion effect.
2.4.2 MHC Demand with Observed Ad Levels
We now extend the model to consider the effect of advertising on demand when con-
sumers observe ad levels before choosing a platform (or can change after observation).
Assume that there is some heterogeneity in the population so that some individuals
may choose to consume from only one platform (single-homing) while others choose
to consume from both platforms. Some contributions assume more than two platforms,
but the main intuitions can be explained with n ¼2, which we assume here.
For a given vector of advertising levels a ¼ a
0
, a
1
ðÞon platforms 0 and 1, we denote by
N
i
E
(a) the mass of exclusive customers and by N
i
S
(a) the mass of consumers who multi-
home and thus are shared by the platforms. The total demand addressed to a platform is
thus N
i
aðÞ¼N
E
i
aðÞ+ N
S
i
aðÞ.
Consumer multi-homing results from the possibility of consuming contents from
both platforms, when joining one platform does not exhaust all consumption possibilities.
Ambrus and Reisinger (2006) assume that the two platforms are not rivals in the market
for consumers so that their demands are independent. Specifically, they consider a
Hotelling model with each platform at one extreme of a line and they assume that con-
sumers located on the line consume from each platform for which their utility is positive.
From the consumer side alone (i.e., in the absence of the advertising side), this would be
uninteresting because the platforms would just face monopoly problems: it is the inclu-
sion of advertisers that renders interaction. The demand structure implies that the MHCs
will be “in the middle” of the interval. If the platform at 0 sets ad level a
0
, its demand is the
set of consumers with positive surplus, thus all consumer locations x such that
V tx γa
0
> 0, where V is base consumption utility and t is the consumer
“transport” cost, here the disutility from distance in the program characteristics of the
platform at 0. Normalizing the consumer density to one, this yields a demand
N
0
a
0
ðÞ¼
V γa
0
t
(2.5)
(when this is below 1), and likewise for the platform at 1. Consumers overlap (in the
middle) if N
0
+ N
1
> 1, and the number of such overlappers (on both platforms) is
N
S
aðÞ¼max N
0
a
0
ðÞ+ N
1
a
1
ðÞ1,0
fg
:
71Two-Sided Media Markets
Consequently, the number of exclusive viewers on platform i is
N
E
i
aðÞ¼N
i
a
i
ðÞN
S
aðÞ, i, j ¼0, 1, i 6¼j; (2.6)
which is simply the number not served by the rival.
Note that this demand implies that when the market is covered, platform i does not
control its mass of exclusive customers, which depends solely on the level of advertising
of the other platform. One could easily extend the model to obtained more complex
patterns (e.g., as in
Gentzkow, 2007). As Ambrus et al. (2014) emphasize, (2.6) applies
more generally when the valuations for the two platforms are correlated (but remain not
rivals: a consumer listens/reads/views platform i if his utility is positive irrespective of the
utility at the other platform, but correlation between values for the two platforms implies
correlation between individuals’ demands for the two platforms). The multi-homing
demand then depends on the correlation between individual demands for the two plat-
forms. In particular the share of multi-homers increases with the correlation between the
utilities.
With the forces discussed above in mind, we can now think about unilateral
changes that would support a particular equilibrium with advertisers multi-homing
(a
0
+ a
1
> A).
45
An ad on both platforms is still worth v + βvN
S
a
ðÞ
, but an ad on platform
j alone is just worth vN
j
(a
j
). Differencing yields the market-clearing total ad price for plat-
form i, in terms of N values, as
P
i
¼vN
E
i
aðÞ+ βvN
S
aðÞ;
and duopoly platform i’s profit is
π
i
¼vN
i
a
i
ðÞa
i
1 βðÞvN
S
aðÞa
i
: (2.7)
The first part, vN
i
(a
i
)a
i
, is the profit of a (non discriminating) monopoly platform. The
presence of competitors induces a reduction in profits as multi-homing reduces the
attractiveness of the platform. In particular, a media platform sharing demand with others
would choose a lower level a
i
of advertising than a monopoly if the demand from multi-
homers is more elastic with respect to advertising than the demand from single-homers,
i.e., if
N
0
i
a
i
ðÞa
i
N
i
a
i
ðÞ
<
@N
S
aðÞ
@a
i
a
i
N
S
aðÞ
:
For a given total demand of a platform, the media platform would benefit from reducing
the share of multi-homers among its customers so as to raise its incremental value per
consumer. This means that at the margin of the monopoly level of advertising (at equal
45
If a
0
+ a
1
< A, then each advertiser patronizes one platform or none and the prices are P
i
¼vN
i
a
i
ðÞ: In this
case, each platform acts as a monopoly on both sides.
72
Handbook of Media Economics
demand for its product), a platform serving some MHCs would raise the advertising level
if this reduces the share of multi-homers. Thus entry may raise or reduce the level of
advertising per platform.
One implication of incremental pricing is that when the equilibrium involves MHAs,
platforms will need to leave some surplus even to homogeneous advertisers. By contrast, a
discriminating monopolist owning the two platforms could price-discriminate by charg-
ing different prices for ads depending on the whether the advertiser puts an ad on one or
both platforms. In doing so, the monopoly would extract the full advertiser surplus.
Denoting by A
S
¼a
0
+ a
1
A the mass of shared advertisers, the total monopolist’s
profit is
Π ¼
X
i
vN
i
a
i
ðÞa
i
v 1 βðÞN
S
aðÞA
S
:
It follows that the marginal profit for a monopoly is
@Π
@a
i
¼
@π
i
@a
i
+ v 1 β
ðÞ
@N
S
aðÞ
@a
i
A a
i
ðÞ
;
where π
i
is the duopoly profit defined in (2.7) and A a
i
is the mass of advertisers exclu-
sively on platform i. Thus, in the Ambrus and Reisinger setup, a monopoly would engi-
neer a lower number of multi-homers, N
S
, than under duopoly. In particular, this implies
that a merger reduces advertising if the consumer multi-homing demand decreases with
advertising.
Total welfare under discriminating monopoly is the sum of total vertical surplus Π and
consumer surplus. Clearly, such a monopoly would set an excessive level of advertising as
it would not internalize the negative effect of advertising on consumer surplus. The same
occurs for a duopoly if N
S
decreases with advertising. These conclusions, however, hold
only for homogeneous advertisers (where a monopoly achieves perfect discrimination),
as we have seen that with heterogeneous advertisers a monopoly would reduce the supply
of advertising and advertiser surplus.
Ambrus et al. (2014) propose an alternative approach by allowing homogeneous
advertisers to buy multiple ads on each platform. An advertiser which airs m
i
ads on plat-
form i gets a return of ϕ(m
i
) for each consumer who is not exposed to the advertiser’s ads
on the other platform. The return on a consumer who is exposed to m
j
of the advertiser’s
ads on the other platform is ϕ
2
(m
0
, m
1
). With a unit mass of advertisers, all advertisers will
advertise on both platforms and m
i
¼a
i
. For given intensities of advertising a
0
and a
1
, the
lump-sum price charged to an advertiser for placing a
i
ads on platform i is
P
i
¼ϕ a
i
ðÞN
E
i
aðÞ+ ϕ
2
aðÞϕ a
i
ðÞðÞN
S
aðÞ, which is also the profit of platform i.
In the same context of non-rival content as Ambrus and Reisinger, they confirm the
result that a media platform facing competition will air more ads than a monopoly if the
demand from multi-homers is sufficiently more elastic to advertising than the demand
73Two-Sided Media Markets
from single-homers.
46
They also show that, for a bivariate normal distribution of utility,
this condition holds when the two platforms deliver negatively correlated utilities.
2.4.3 MHCs and Heterogeneous Advertisers
One drawback of the two previous models is the simplifying assumption that advertisers
are homogeneous. The
Anderson and Coate (2005) framework, and subsequent models
in that vein, allowed for heterogeneity in the willingness-to-pay by advertisers. In the
presence of MHCs, different advertisers may choose different portfolios of platform pres-
ence, and do so even if there is no correlation between advertiser demand and program
choice (which remains a major outstanding research problem).
As pointed out by
Doganoglu and Wright (2006), when customers of two-sided plat-
forms are heterogeneous in their valuation of externalities, they will differ in their con-
sumption patterns. In the context of media and advertising, this means that some
advertisers will patronize only one platform while others will patronize two platforms.
Indeed, a natural framework for extending the advertiser demand is to deploy models
of vertical differentiation, which describe competition between firms selling different
quality products to consumers who have heterogeneous values over quality. These con-
sumers translate naturally in the media context to advertisers which have different values
for making an impression, and the “qualities” have a natural analog in the numbers of con-
sumers attending each platform. Notice that the standard models of vertical differentia-
tion (
Gabszewicz and Thisse, 1979; Mussa and Rosen, 1978; Shaked and Sutton, 1983 for
duopoly) have consumers making a single choice of option. In the oligopoly equilibrium,
a firm with a higher quality sells at a higher price to those consumers with high valuations.
The standard analysis was extended by
Gabszewicz and Wauthy (2004) to allow for con-
sumer multi-purchase. The consumers who most value quality are those who will multi-
purchase in equilibrium. Their model can then be directly transposed into the advertiser
demand side in a fuller-fledged two-sided market platform context with “qualities”
endogenously determined via the viewer demand side. In the spirit of the multi-homing
models above, overlapping consumers denigrate the “quality” of buying the bundle of
both platforms (an ad on each platform), so it is not worth the sum of its parts. That
is, advertisers who multi-home have to “pay twice” for overlapped viewers, so only those
advertisers with high willingness-to-pay will multi-home if there are many MHCs (and
second impressions are not worth much).
To see how this works, consider the duopoly model with fixed consumer demands N
i
and N
S
¼N
1
+ N
2
1. Suppose that v is heterogeneous. Faced with ads priced at P
1
and
46
Their condition compares the ratio of ads–elasticities of the two demands with the ratio of elasticities of
advertising returns per consumer for multi-homers (ϕ
2
) and single-homers (ϕ). For instance, if
ϕ
2
¼ϕ m
0
ðÞ+ ϕ m
1
ðÞϕ m
0
ðÞϕ m
1
ðÞ, the condition of Ambrus and Reisinger is sufficient for entry to raise
ad levels.
74
Handbook of Media Economics
P
2
, an advertiser will multi-home if the incremental value of each platform exceeds the
price, thus if:
v > v
m
¼ max
i
P
i
N
i
1 βðÞN
S
:
Advertisers with a lower v will choose to single-home or stay out of the market. This
means that while the incremental value is the relevant one for high-value advertisers,
competition with single-homing prevails for low-value advertisers. In this setup, an
advertiser which single-homes opts for platform i if vN
i
P
i
is larger than 0 and
vN
j
P
j
. If the return to ads is a linear function of the consumers reached, then a price
competition game for advertisers may fail to have a pure strategy equilibrium.
47
Anderson et al. (2013b) engage such a (vertically differentiated) advertiser demand
side with a specific consumer demand side that allows for consumer multi-homing. They
use the non-existence of a price equilibrium that we just noted for some consumer allo-
cations to rule out some types of configurations in advertiser and consumer homing.
Their consumer model is a horizontal differentiation model quite similar to the one
used by
Ambrus and Reisinger (2006) that was described above.
48
The equilibrium
concept is the same as that at the end of
Section 2.4.1: consumers do not observe ad
levels but rationally anticipate them. A priori, four regime combinations can arise: each
side can fully single-home or have some multi-homers.
Consider first any regime with SHCs. This is the “competitive bottleneck” case so
that all platforms set the monopoly ad level a
m
(and are expected to do so). It is an equi-
librium as long as no consumer wants to multi-home given the monopoly ad levels on all
platforms. Note that all active advertisers multi-home. Hence single-homing on both
sides cannot happen.
49
Another combination that cannot happen for a wide range of
parameter specifications (including a uniform distribution of advertiser valuations) is par-
tial MHC with SHAs.
50
Taken together, these results then mean that the relevant market
47
For example, suppose that N
i
¼N
j
. Then an outlet serves all the multi-homers if it sets the higher price but
it serves all multi-homers and all single-homers if it sets the lower price. The advertising demand discon-
tinuity implies that the price game has only mixed strategy equilibria.
48
These models are described in more detail in Chapter 10 (this volume). The Anderson–Foros–Kind model
(see Anderson et al., 2013b) has MHCs who value quality increments of their second-choice (further)
product differently from their first choice. However, for the current purpose, with symmetric media prod-
uct “qualities,” the model is the same as that of Ambrus–Reisinger.
49
The SHC–single-homing advertiser (SHA) combination can arise in the Ambrus–Reisinger model when
the length of advertisers is large enough that the sum of each platform’s advertiser level is below the total
mass A.
50
This is ruled out by the first-order conditions on the advertiser side. However, SHAs can arise with full
MHC. If all consumers are multi-homing, then platforms become perfect substitutes. Equilibrium ad
prices are zero, and advertisers are indifferent as to which platform to place an ad upon.
75
Two-Sided Media Markets
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.