two-sided market balance when media consumers choose only one platform (single-
homing consumers—SHCs). The competitive bottleneck induced from single-homing
begets several puzzles for the positive analysis and strong conclusions for the normative
analysis. These are addressed through considering the possibility that (some) consumers
multi-home (
Section 2.4). Multi-homing gives rise to incremental pricing of ads
whereby platforms can only charge for the extra value they deliver to (multi-homing)
advertisers. This leads to reluctance for platforms to deliver multi-homing consumers
(MHCs) due to their lower value.
Section 2.5 draws some implications for platform
content choice in both product specifications (short run) and overall product diversity
(long run).
Section 2.6 concludes with some outstanding research directions.
2.2. CAST OF CHARACTERS
The essence of two-sided markets is that the interaction between two groups of actors is
mediated by platforms. In media markets, the two groups are advertisers and consumers,
and the platforms are the media themselves. The general theory of such markets was first
propounded by
Caillaud and Jullien (2001, 2003) and Rochet and Tirole (2003), and has
been further elaborated in a voluminous literature. Key milestones and surveys are
Armstrong (2006) and Rochet and Tirole (2006).
This literature traditionally distinguishes between usage externalities and participation
externalities. Usage externalities arise (for example) for credit cards or click-through
advertising, where the platform charges its members per interaction, so that individual
interaction behavior matters. Participation externalities arise (for instance) for club mem-
bership, where platforms charge their members for access, so that potential members care
about the level of participation on the other side. The distinction, although a useful oper-
ational tool, is somewhat artificial as it depends as much on tariff structure choices as on
technologies (see
Rochet and Tirole, 2006).
1
As with many two-sided markets, typical
media markets (and models thereof ) involve elements of both.
The theoretical literature on platform economics has traditionally concentrated on cases
where there are no own-side network effects—participation has no direct effect on the
well-being of other members on one’s own side of the market—but there are cross-side
network effects, which can be either positive or negative.
2
In the media context, consumers
typically do not care about how many other consumers are engaged on a medium (modulo
fashion and water-cooler effects), but advertisers might well care whether competitors are
also airing ads. Advertisers want an audience, the larger the better, so there is a positive
1
For instance, credit cards charge a fee per transaction to sellers but an annual fee to buyers (along with
rebating bonus points), implying that buyers care about overall seller adoption (see
Bedre-Defolie and
Calvano, 2013
).
2
Exceptions include Nocke et al. (2007) and Belleflamme and Toulemonde (2009).
43
Two-Sided Media Markets
network effect of consumer-side size on advertiser benefits. The relation in the other direc-
tion can be positive or negative (or indeed can vary across consumers). Typically, one
thinks of television and radio advertising as a net nuisance to consumers insofar as any con-
sumer surplus enabled from the advertising (in terms of information about better purchase
options, or enhanced product satisfaction) is outweighed by intrusive interruption of the
program content. Specialty magazines may involve positive net benefits,
3
especially insofar
as ads in magazines (and newspapers too) are more easily skipped over, and readers may
want to find out more about products related to a hobby (sailing or golf mags) or purchase
opportunities (classified ads in newspapers).
To be sure, some media are not financed by advertising at all. Such cases (HBO, Sirius
radio, and Consumer Reports—which has a mandate not to carry ads) are easily treated as
standard one-sided markets, whereby media firms set prices and consumers choose
among options in a standard manner, although such cases are rather rare. Instead, when-
ever consumers are paying attention (even subconsciously, as with billboards), then there
is a latent demand to send them advertising messages. Witness the sponsors’ emblems on
soccer players’ shirts and the billboards around the soccer field. Thus the common form of
business model is either joint finance with both advertisers and subscribers footing the bill
(magazines) or advertisers only paying ultimately for the programming (“free-to-air” or
“commercial” television and radio). The business model is then as follows. The platforms
want to attract consumers in order to sell their attention to advertisers. The program con-
tent is the bait, or lure, which in turn is either denigrated by the ads piggy-backed upon it
(when ads are a nuisance) or indeed part of the attraction (when ads have a positive value).
The program is thus a conduit for the ads to reach prospective customers, who are in turn
not attracted primarily by the ads (infomercials aside!) but by the entertainment content.
In this context, the platforms’ problem is to balance between extracting revenue from
advertisers, while delivering consumers who might be put off by the ads, and switch over,
or switch off.
4
Viewed in this light, one might anticipate a marginal condition for the
equilibrium at which the elasticity of revenue per viewer is equal consumer participation
elasticity, and that is exactly what we deliver formally below.
We next give some notation, and discuss more the three legs of the market, continu-
ing to mix our metaphors somewhat between the various media applications.
2.2.1 Consumers
The media consumers are the readers, viewers, listeners, or (web-)surfers. They choose
whether or not to subscribe to a particular channel (if there is a subscription fee) or buy a
magazine, and how much time and attention to pay to it (depending in part on the quality
3
See Chapter 9 (this volume) for more details on the empirical evidence for positive benefits.
4
This is one instance of a more general trade-off between third-party financing and consumers participation
that is analyzed in depth by
Hagiu and Jullien (2011).
44
Handbook of Media Economics
of the publication, how it matches with the consumer’s tastes, and the number and types
of ads carried). They may indeed consume several channels, although with most media
they can only devote attention to one at a time.
5
While engaged with a channel, a con-
sumer may register some of the ads on it, and she might buy something she otherwise
would not have.
Given the complexities of modeling the effects in the previous paragraph in terms of
the mapping from subscription fees and ad levels to ultimate purchases, it is not too sur-
prising that the literature has taken some drastic simplifications. One of the least egregious
is the assumption that the ad nuisance (or desire for ad exposure) can be monetized into
dollar terms. Frequently it is assumed that all consumers face the same valuation/cost per
ad, and that it is moreover a linear function of ad volume on the channel. Thus, the full
price from watching channel i is
f
i
¼s
i
+ γa
i
, i ¼1, …, n;
where s
i
0 is the subscription price, a
i
is the ad volume carried on the channel, and γ is
the (net) nuisance per ad (which may be negative if ads are enjoyable).
6
With these prices in mind, consumer choice of what to watch is determined by utility
maximization. The standard assumption has been that consumers make a discrete choice
of which (single) channel to engage with, and so the apparatus of discrete-choice models
(or spatial competition models) has frequently been deployed. This reflects that a con-
sumer can typically only engage a single channel at a time, and the import of this
“single-homing” assumption was only recently recognized. Indeed, the modeling of
“multi-homing” consumer choice is quite elaborate, and still in its infancy (more details
are given in
Section 2.4), even if the practice is perhaps becoming even more prevalent
with increased Internet penetration (see
Chapter 10, this volume). In broad terms then,
consumer choice delivers a demand system of substitute products whereby the number of
consumers choosing channel i is (denoting f
i
the vector of other platforms’ prices)
N
i
f
i
; f
i
ðÞ;
which is decreasing in the first argument and increasing in each element of the vector of
all other channels’ full prices.
While many contributions rely on discrete-choice models such as the Hotelling
duopoly model, the Vickrey–Salop circle model, or the logit model, an alternative inter-
pretation of N
i
is that it measures the total time or attention devoted by consumers to the
media platform (see
Dukes, 2006; Gal-Or and Dukes, 2003). Some contributions use a
representative consumer model (e.g.,
Kind et al., 2007, 2009). These are discussed in
5
Multi-tasking across media is becoming more common with the advent of the Internet, although even with
traditional media it is sometimes possible (reading the newspaper while listening to the radio).
6
Thus the nuisance might allow for netting off expected consumer surplus from buying advertised products.
45
Two-Sided Media Markets
detail below, as are extensions of the Hotelling-type models, to allow for individual con-
sumer demand for multiple platforms (multi-homing).
Consumer surplus (from the channel decisions) is then measured in standard fashion
given the consumer’s optimization problem and the full prices faced.
2.2.2 Advertisers
Advertisers are assumed to derive some benefit from reaching consumers. This should
realistically depend on the number and types of other advertisers reaching them, and also
the types and numbers of consumers reached. Both of these heterogeneities are typically
set aside. That is, first, the value to a particular advertiser from reaching a consumer is
usually assumed independent of the specific platform via which she is reached (thus ignor-
ing the matching problem that readers of a motorcycling magazine are more likely to be
interested in chain-lube than those of a sailing magazine).
7
Second, the value per con-
sumer is independent of the number of consumers reached,
8
so there are constant returns
to advertising (this would not be the case if there were non-constant marginal production
costs for advertisers’ goods, for example).
9
Third, competition in the product market is suppressed (so GM’s returns from adver-
tising are independent of whether Ford also advertises). This assumption is most tenable
when ads are from different sectors and there are negligible income effects (so that the
chance the consumer buys the steak-knife is independent of whether she accepts the
mortgage refinance).
Dukes and Gal-Or (2003), Gal-Or and Dukes (2003, 2006), and
Dukes (2004, 2006) analyze media with advertising for competing products.
Except where explicitly noted to the contrary, we assume away limited attention and
congestion (
Anderson and de Palma, 2009; Van Zandt, 2004) so that the return from an
ad does not depend on how many other advertisers reach the same individual.
Under these assumptions, we can rank the advertiser willingness-to-pay per consumer
in standard fashion, from high to low, to generate the advertiser demand curve for
impressions on a per-consumer basis. Moreover, when each consumer can be reached
through one platform (single-homing), the decision to buy advertising space on any plat-
form is independent of the decision for other platforms. Hence the single-homing
assumption for consumers implies that advertisers put ads on multiple platforms (they
multi-home).
Let then v(a) denote the willingness-to-pay per consumer for the ath highest adver-
tiser. If a platform gets the top a advertisers, its price per ad per consumer is v(a) and its
revenue per ad per consumer is
7
For exceptions, see the literature on targeting, such as Athey et al. (2014), discussed in the last section.
8
In models where demand is measured by the time spent on the platform, the unit of demand is consumer
time and the assumption is that the value per unit of time is constant.
9
As Rysman (2004, p. 491) puts it, advertiser profit per look is constant. An exception is Crampes et al. (2009).
46
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.