These results come from an “inverted pyramid” structure of preferences. On each level
are genres, these getting finer from the bottom (the LCD) to the top (where each taste
type is represented). The idea is that most people will tune in if their most preferred
option is available, fewer if a broader-based option is all that is available. For illustration,
suppose that 80% of the population will listen to the lowest-level (LCD) program. If two
middle-level programs are available, 45% of the population will listen to each. At the top
level, all listen and the split is equal. So, let us trace how the market develops as we
decrease fixed costs (equivalently, as we increase the number of consumers across the
board). This is the market size effect. With a small market, just the LCD programming
is provided. Then, as the market expands, the two mid-level ones are offered. This arises
because, rather than sharing the smaller LCD base, stations do better going higher for a
base-extension effect as they offer better-matched content. And, if the rival is doing so, it
is better for a second station too, since the LCD loses half its potential audience when the
rival “upgrades.” A further market size rise will double the number of offerings again, for
similar reasons. Notice though that the “doubling” at the last stage here was an artifact of
the assumed symmetry in the preference divisions. If instead the middle level were split
say 55% to 40%, then the first market expansion effect above the LCD is to offer the two
more specialty genres, but then a further across-the-board increase will impact first the
55% who will get upgrades, and the 40% will be temporarily left behind until a further
expansion makes them worth dividing.
The important take-aways from these models are therefore mainly for markets served
by few platforms. A market served only by a single platform will tend to serve up an LCD
offering. With two platforms, the LCD type might be duplicated, or a second popular
genre (or another LCD) might be broached. As the number of platforms rises, more
diverse preference types will be served, and pure LCD types will tend to be surpassed
(although they may indeed represent the first preferences of some viewers, in which case
they will prevail). However, duplication will pile up in the most popular formats.
Moreover, there is bias toward those viewers whom the advertisers most want to reach.
Notice finally the positive preference externalities in the examples above. As own-group
size expands, it becomes more likely one’s higher preferences get catered to. Moreover,
by taking away some of the clientele of the erstwhile LCD, there is a greater likelihood
that the other clients on the LCD base get an upgrade. Thus, we expect positive pref-
erence externalities with respect to own types, with weaker spillovers to similar types.
While the principles described above in the Beebe and Steiner analyses resonate, the
models are too sparse. To gain more depth, we apply their insights into first a spatial
model and then into a logit model. In both cases we explicitly introduce different con-
sumer types so as to be able to track the effects of population composition on product
selection (positioning and variety). The former approach is well configured to deal with
markets with very few firms, while the latter deals better with markets with larger
numbers.
15Preference Externalities in Media Markets
Our objectives in this section are to incorporate the fundamental features of media
markets discussed above—fixed costs, product differentiation, tastes that vary across
groups, and advertiser finance—into models of product differentiation. We discuss the
ensuing empirical implications for the relationship between the size and mix of the num-
ber of potential consumers and the number and targeting of products, as well as the ensu-
ing welfare of various kinds of consumers (i.e., the within-group and across-group
preference externalities).
1.3.2 Spatial Models
1.3.2.1 Negative Preference Externalities Under Monopoly
We begin with a monopoly model with two types of consumers, illustrated by reference
to daily newspaper markets. The two groups, whom we term “whites” and “blacks,”
have different product preferences, with their ideal products represented along a one-
dimensional spectrum. Assume there is a continuum of W-type agents uniformly distrib-
uted on [0,1] with density f
w
and a continuum of B-type agents uniformly distributed on
[z,1 + z] with density f
b
<f
w
. Let z E [0,1] so that the degree of overlap is 1 z. The prob-
ability of any type buying the newspaper, and hence getting exposed to any ads in the
paper, is d(p + tjxx
n
j), which is decreasing in its argument. Here p is the price of the
newspaper (if any), x
n
is its location in the content space, x is the “ideal” content location
of an agent of type x, and t is the “transport” cost rate from not getting the ideal news-
paper content. Furthermore, assume there is a mass A of advertisers. Each is willing to pay
w to contact a W-type reader and b to contact a B-type one. Let w > b so that the W-types
are more attractive to advertisers. Therefore, we have a majority group which is also
worth more to advertisers, and its preferences overlap with the minority one: apart from
the assumption that the majority is worth more, there is no loss in generality assuming
they are on the left of the spectrum. The assumptions are illustrated in
Figure 1.1.
We consider in turn the preference externalities under pure advertising finance, pure
subscription pricing, and the mixed business model: see
Chapter 2 for more details, along
with a consideration of ad nuisance to consumers, which has here been suppressed.
Loosely, the first case transpires if advertiser demand is strong enough.
The equilibrium has the (profit-maximizing) newspaper choose its content location,
and price where appropriate. Each consumer is therefore worth p plus either w or b
(depending on its type) times A, the mass of advertisers (and A ads will be placed in
the paper, charging advertisers the full value of their surplus). In what follows, we eschew
0 z 11+z
Figure 1.1 Preference structure (consumer densities) producing a negative preference externality
with monopoly.
16
Handbook of Media Economics
a full equilibrium description because the factors on which we focus are quite immediate
in their impact on the price/location solution: the properties claimed are readily derived.
Notice first that if all consumers are equally valuable to the advertisers, and their densities
are the same, then the monopolist’s profit-maximizing location (under each of the three
business models) is the center of the full market, i.e., at (1 + z)/2. This benchmark allows
us to compare the externalities exerted under asymmetric advertiser preference to con-
tacting the two types, or different densities of them.
Under pure advertising finance (e.g., commercial TV or terrestrial radio), the B-types
are less valuable. If indeed b ¼0, then they have no commercial value and their prefer-
ences are irrelevant. The content format chosen is then x
n
¼½, which is the mean of the
W-type consumer distribution. As b rises, the chosen content location rises, reaching the
benchmark point (1+ z)/2 only when b gets as high as w and even then only when f
b
<f
w
.
Equivalently, we can say that the larger is f
w
or w, then the more is the content tilted
toward the W-types, so that the preference externality of catering to the type of greater
mass or greater value is detrimental to the other type (but is advantageous to the majority).
The impact of the preference externality to the B-types is more severe the greater is z,
corresponding to a greater preference divergence.
For pure subscription pricing, the difference in advertiser valuations is taken off the
table, and the preference externality depends solely on the single source of differentiation
between the two types, their different densities. The preference externality effect is there-
fore weaker. But still, the more W-types there are, the closer is the content location to
their mean at ½, and the further away from the B-types’ desiderata.
24
With a mixed-finance model, the solution for the content location choice lies
between the two cases noted above. Again, more W-types move the location away from
the B-types, as does a higher advertiser preference for contacting the W-types. Moreover,
the larger is A, the more reliance on advertising finance, and the lower the copy-price so
as to attract more consumers to deliver to the advertisers. Then again, the closer is the
location to the mean of the W-types. The problem is more severe the more tastes differ
across groups.
To summarize, the bias against the B-types is greater the more W-types there are, the
larger the discrepancy in their values to advertisers, the more different their tastes from the
mainstream, and the greater the weight given to advertisers in the business model.
1.3.2.2 Preference Externalities Under Duopoly
When demand is sufficiently strong, there will be more than one product. There remain
some markets in the US with more than one paper, though this was quite common some
decades ago. Accordingly, we next allow for two competing media outlets. To do so, we
24
Another possible driver impinging on locations would be differential purchase probabilities. They have
been suppressed here but are readily introduced.
17
Preference Externalities in Media Markets
assume that subscription prices are fixed, or indeed zero (as per commercial TV or radio),
so we take price competition off the table.
25
We continue to assume there is no ad nui-
sance impinging on reader choices. These missing features are analyzed in detail in
Chapter 2, but our emphasis here is on preference externalities between different groups.
We simplify the spatial model above slightly by assuming that z ¼1, so that the
B-types “start” where the W-types end. The total market length is now 2. Denote
the media firms’ locations as x
1
<x
2
, so that Firm 1 serves predominantly the W-types.
We first characterize these via the first-order conditions for equilibrium locations. The
key condition is that a move in of δ by Firm 1 (located at x
1
) expands its loyal base by δ,
while picking up only an extra base of δ/2 from the rival’s consumers. The change in
profits is thus the value gained on the LHS minus the value lost on the RHS (where
its market length has decreased by δ/2). Thus, its profit increases by corresponding values
of the consumers gained and lost. In equilibrium, the marginal consumer is a W-type.
Thus, the local profit increment from moving in for Firm 1 is
δd p + tx
1
ðÞðÞδ=2ðÞd p +1=2tx
1
+ x
2
ðÞðÞ½wf
w
,
which is the value of demand gained on its LHS minus the loss on the RHS.
The corresponding effect for the other firm from moving left is analogously
δbf
b
d p + t 2 x
2
ðÞðÞδ=2ðÞwf
w
d p +1=2tx
1
+ x
2
ðÞðÞ:
For this firm, each gained consumer is worth less because they are B-types.
Setting both of these first derivatives to zero yields the equilibrium solution when
firms are not back-to-back. Because the left sides of these expressions are the same (these
are transfers at the interior margin), the equilibrium conditions imply that wf
w
d(p + t(x
1
)) ¼bf
b
d(p + t(2 x
2
)). Then, because wf
w
>bf
b
(by the assumption that Ws are
worth more than Bs) the implication is that x
1
<2 x
2
. That is, Firm 1 is closer to
the median W-type than Firm 2 is to the median B-type. Moreover, both are closer
to the market center (here set to 1) than to the edges. The reason is that each firm picks
25
There is a voluminous literature on equilibrium existence in prices and locations. Anderson et al. (1992,
Ch. 8)
summarize the state of the art two decades back, and the field has not developed much more in the
interim, at least as relates to the current application. In particular,
Hotelling’s (1929) claimed “minimum
differentiation” result was shown to be incorrect by
d’Aspremont et al. (1979). Osborne and Pitchik
(1987)
solve his original two-stage location-then-price model with linear transport costs, by analyzing
the mixed strategy equilibrium to the price sub-game and engaging this to the location stage. They find
(pure strategy) equilibrium duopoly locations just inside the quartiles, at which locations the price equi-
librium is in (non-degenerate) mixed strategies.
d’Aspremont et al. (1979) propose the “fix” used by most
subsequent authors, by replacing linear transport costs with quadratic ones. This ensures tractable price
sub-game solutions in pure strategies, but with a radical change in the equilibrium locations. Instead
of “minimum differentiation,” the locations are at the extremes. Therefore, fixing prices as we do here
is hardly innocuous.
18
Handbook of Media Economics
up more demand on the interior boundary between them, and that demand is worth rel-
atively more to the firm with the predominantly minority readership.
The implication is that the Ws are better served than the Bs, although each group
would be better off in aggregate if the other were not there.
26
Note too that the presence
of the B-types draws Firm 1’s location inward even if few B-types actually read it. The
induced bias toward middle-of-the-road coverage is greater for the minority paper than
for the majority one.
The bias toward the center—and the minimum differentiation result associated with
Hotelling (1929)—is tempered by the elasticity of demand in the model above. The more
inelastic is demand, then the closer together will the firms locate.
27
We can now describe preference externalities at the differentiated outcome. The
mechanism is quite interesting. Suppose the population of Ws rises across the board
(an increase in f
w
) or, equivalently, if the value of the Ws rises to advertisers (an increase
in w). Then the first-order condition of outlet 1 remains unchanged, because the
relative value of readers at each margin is unchanged. However, the other outlet’s profit
is now higher when it moves in because of the increased value of serving the marginal
W-type. Thus, outlet 2 moves left; this in turn causes the first outlet to cut left too. So the
upshot is that both outlets are further left in the new equilibrium. The implication for the
well-being of the two groups (in aggregate) is that the Bs are worse off on average. The
Ws are better off for two reasons (although those just right of the erstwhile location of
outlet 1 are worse off ). First, those served by outlet 1 are better off on average because 1’s
location is more central to them. Second, those Ws whose preferences are more extreme
so that they chose before the outlet catering predominantly to B-types (the B-leaning Ws)
are also better off because their outlet now delivers content closer to their ideal.
Now consider the effects on locations of a merger (here to monopoly), and assume
that the two-outlet firm keeps publishing both media. Then we can view the location
choice for each outlet as internalizing the effect of its location choice on the demand
of the other. The upshot is that equilibrium locations are further apart because they avoid
cannibalization of the sibling outlet’s readers. Therefore, the prediction is that mergers
lead to more diversification, moving away from the excessive tendency to centralize that
is epitomized in the Hotelling model.
26
Modulo the possibility that the B-types might not be served at all if they were alone: the existence of the
W-types raises the profitability of the B-type paper and may thus enable the fixed costs to be covered. This
form of “cross-subsidization” is discussed further below when we explicitly consider the model with
entry.
27
The term “Principle of Minimum Differentiation” is due to Boulding (1966). In the setup above, there
is back-to-back location if demands are too inelastic, i.e., (from the location derivatives) if firms do not
want to move outward from a common location x
m
satisfying d(p + tx
m
) >½d( p) and bf
b
d(p +
t(2 x
m
)) >½wf
w
d( p)). Note that x
m
should further satisfy the condition that firms’ profits are equal,
so that neither firm wishes to flip its position with its rival.
19
Preference Externalities in Media Markets
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.