However, in the pay-TV regime the broadcaster in most cases can do better than this
by bundling its content together as a package.
66
(Clearly, in the free-to-air context, chan-
nel bundling has no role to play. For simplicity, suppose in the following discussion that
the broadcaster does not use advertising at all.) To illustrate most transparently, suppose
instead of retailing its various content separately that the broadcaster sells its full range of
content as a single bundle in return for a single price.
67
The key point is that a viewer’s valuation for the whole bundle is often more predict-
able (i.e., less idiosyncratic) than her values for individual pieces of content (
Bakos and
Brynjolfsson, 1999; Crawford, 2008
). This might be because of negative correlation in
values for different types of content, so that those viewers who particularly like content i
place less value on content j, and vice versa.
68
A deeper reason is that, even without neg-
ative correlation, idiosyncrasies in valuations tend to get “averaged out” when bundling is
used. (The phenomenon is similar to the insight that a diversified portfolio is less risky for
an investor than holding a single asset.) If a viewer’s valuation of one kind of content is
independently distributed from her tastes for other kinds of content, and if there are many
pieces of content on offer, then the law of large numbers implies that her value for the
bundle is highly predictable. The result may be that the broadcaster can extract a large
fraction of viewer surplus with bundling, and relative to non-bundled pricing viewers
may be harmed and total welfare may rise.
69
As with the comparison between advertising-funded and pay-TV regimes, the fact
that a broadcaster typically makes more profit when bundling is used implies that some
content can be profitably supplied only if bundling is used. As a result, even if bundling
often reduces consumer surplus for a given range of content, the fact that more content
might be offered with bundling could lead to consumer gains from the practice.
70
For
66
The pioneering articles on bundling are Stigler (1968) and Adams and Yellen (1976). The many-product
discussion which follows is taken from
Armstrong (1999b) and Bakos and Brynjolfsson (1999).
67
More ornate schemes would allow viewers to pick and choose between different content, but the basic
insight is seen most clearly with this “pure bundling” format.
68
For example, suppose half the population value a sports channel at $10 and a news channel at $2, and
remaining viewers have the reverse preferences. If the broadcaster had to set a separate price for each chan-
nel, it would charge $10 for either channel and viewers would see only their preferred channel. However,
if it set a price of $12 for the bundle of both channels, all viewers would just be willing to pay this, and
profit and welfare both rise.
69
Continuing with the example in the previous footnote, suppose for any given piece of content half the
viewers have value $10 and the rest have value $2, and valuations are independently distributed across
products. As before, with per-channel pricing, the broadcaster would set the price at $10 per channel,
viewers would see only the content they value highly and average revenue per subscriber per channel
is $5. However, if there are many channels, most viewers place a high value on around 50% of these chan-
nels, and so most people are willing to pay around $6 per channel for the bundle. The result is that the
broadcaster can extract most of the surplus from most of the viewers by giving them the whole bundle, and
welfare increases since most people see all available content.
70
See Crawford and Cullen (2007) for an investigation of this tradeoff.
317
The Economics of Television and Online Video Markets
similar reasons, an incumbent broadcaster with an existing portfolio of content may be
willing to pay more for new content than a new entrant, since the latter is less able to
achieve extra revenue obtained with bundling.
71
The result may support a tendency
toward concentration in the market for content discussed in the previous subsection,
due here to demand-side economies of scope rather than any supply-side economies
of scope.
7.4.4.2 Bundling Empirics
A sizeable empirical literature has analyzed consumer demand for bundles of television
channels.
Mayo and Otsuka (1991) and Rubinovitz (1993) estimate demand for cable
bundles and attempt to measure the impact of regulation on cable prices.
Crandall and
Furchtgott-Roth (1996, Chapter 3)
and Crawford (2000) also estimate bundle demand
and calculate the welfare effects of changes in prices and product offerings.
Goolsbee and Petrin (2004), Chu (2010), and Crawford and Yurukoglu (2012) are
the most recent and comprehensive empirical papers analyzing demand for pay-television
bundles.
72
I summarize here CY, as it builds demand for bundles from heterogeneous
preferences for individual channels.
A challenge when estimating demand for bundles is to determine the relative impor-
tance of each channel in the purchase of the bundle when the bundle contains as many
channels (50+) as is common in the pay-television industry. Variation in the contents of
bundles across cable markets, or across bundles of different size within cable markets,
helps to trace out the demand for each component channel, but generally is not rich
enough to recover the full distribution of preferences for individual channels. Further-
more, these distributions are critical to understanding the core issues of pricing, content
choice, and welfare both in existing television markets and how they may differ in
different economic environments.
CY resolve this issue by pairing data on bundle composition and price with comple-
mentary data on individuals’ viewing habits. The latter, both in the form of average ratings
for channels across markets as well as individual households’ viewing behavior, provide
rich information at the level of individual channels, but do not have price information.
It is only the combination of viewing data that allows estimation of the relative utility
of alternative channels, and bundle purchase data that allows the translation of channel
utility into bundle WTP, that allow them to recover demand curves for individual
channels.
71
In the running example, an incumbent broadcaster which already possesses many pieces of content
would be willing to pay up to $6 per subscriber for an extra channel, while a stand-alone entrant could
pay only $5.
72
Other recent work includes Rennhoff and Serfes (2008), Byzalov (2010) , and Crawford et al. (2012b) .
318
Handbook of Media Economics
Supported by evidence from viewing patternsintheirdata,CYalsoaccommodate
the “long-tail” feature of preferences for medi a products.
73
They do so by assuming
that with some probability depending on demographics, households have a zero pref-
erence for c able channels. If positive, they further assume that the marginal distribu-
tion of preferences across households i s distributed as an exponential. T hey then
estimate the zero-taste probabilities and exponential parameters for each channel.
They also estimate distributions for preferences allowing for pos itive or negative cor-
relations in tastes for pairs of channels, an important consideration in the bundling
literature.
74
Figure 7.18 reports both the share of p ositive WTP and, among thos e tha t are pos-
itive, the estimated WTP for each of nine popular cable channels arising from their
analysis. Preference heterogeneity is evident: some people (32%) do not value the
cable news channel CNN at all , while othe rs va lue i t at more than $20/month.
75
Furthermore, preferences for each channel a re estimated to have a long tail, with
many va luing channels at or around $2/month and only few at values at or above
$10/month.
7.4.4.3 Welfare Effects of à la Carte
Crawford and Yurukoglu (2012) use the results of their analysis to evaluate the implica-
tions of the bundling theory summarized earlier as well as the policy of forcing cable
channels to be offered on an à la carte basis, a policy sometimes proposed by policymakers
in the pay-television industry.
76
The results both confirm the bundling theory summa-
rized above and refute its application to the specific case of television channels due to
bargaining between channels and distributors.
73
See Anderson (2006) for a general exposition of this issue and Shiller and Waldfogel (2011) for evidence of
long-tail preferences for music.
74
In subsequent research analyzing demand for both national cable television channels and Regional Sports
Networks,
Crawford et al. (2015) extend this demand model to allow for heterogeneous value of time
between sports and non-sports channels, finding this to be an important extension which is necessary
to explain the relatively large affiliate fees sports networks are able to obtain in negotiations with
distributors.
75
Furthermore, table 4 in Crawford and Yurukoglu (2012) lists, for each of the 50 cable channels in their
analysis, the second channel for which households had most positively correlated preferences. For the nine
channels listed in
Figure 7.18, these were TV Land (ABC Family), MTV2 (BET), Fox News (CNN),
MTV (Comedy Central), Nickelodeon (Disney Channel), ESPN2 (ESPN), VH1 (MTV), USA
(TNT), and TNT (USA).
76
They also analyze the impact of channels being offered as a part of Theme Tiers as well as the Bundle-Sized
Pricing strategy proposed in
Chu et al. (2011).
319
The Economics of Television and Online Video Markets
In their counterfactual analysis, CY simulate market outcomes in one large and one
small cable market, both competing with a “national” satellite provider. Outcomes are
compared for two scenarios. In the first, baseline, scenario, distributors are assumed to set
a fixed (bundle) fee for the 49 cable channels included in their analysis. In the second,
“full à la carte” scenario, distributors are assumed to set a fixed fee for access to any chan-
nels and then individual channel prices for each of these 49 channels. Competition is
between the single cable operator serving each market and the national satellite operator
setting a common price in both markets.
Their baseline results confirm the predictions of the discriminatory theory of
bundling summarized above: in an à la ca rte world, house holds would cho ose o nly
22 of 49 channels, expenditure on cable would fall by an estimated 23.8%, and con-
sumer surplus would rise by 19.2%. Total industry profits would fall by 12.7%, but
0 2 4 6 8
0
500
1000
1500
2000
ABC Family Channel
Pos: 0.49
0 5 10
0
500
1000
1500
2000
BET Networks
Pos: 0.34
0 10 20
0
1000
2000
3000
CNN
Pos: 0.68
0 2 4 6 8
0
2000
4000
6000
Comedy Central
Pos: 0.61
0 5 10
0
2000
4000
6000
8000
Disney Channel
Pos: 0.65
0 5 10 15
0
1000
2000
3000
4000
ESPN
Pos: 0.64
2 4
0
2000
4000
6000
8000
MTV
Pos: 0.59
1068050
0
2000
4000
6000
TNT
Pos: 0.72
0 5 10
0
2000
4000
6000
USA Network
Pos: 0.51
Figure 7.18 Estimated WTP for a subset of television channels. Notes: Reported from Crawford and
Yurukoglu (2012)
, the estimated share of 20,000 simulated households that value each of nine
large pay-television networks positively (Pos) and the estimated distribution of their willingness-to-
pay (WTP) for each network are shown. In each figure, the y-axis reports households and the x-axis
reports WTP in 2000 dollars.
320
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.