Archive
Archive |
|
THE IMPACT OF ADVERTISING ON MEDIA BIAS
Preliminary 2011 Draft
ESTHER GAL-OR
TANSEV GEYLANI
PINAR YILDIRIM*
Economics, Katz Graduate School of Business, University of Pittsburgh, 222 Mervis Hall, Pittsburgh,
PA 15260 (email: esther@katz.pitt.edu , tel: 412 648-1722); Tansev Geylani: Associate Professor of
Business Administration, Katz Graduate School of Business, University of Pittsburgh, 320 Mervis Hall,
Pittsburgh, PA 15260 (e-mail: tgeylani@katz.pitt.edu, tel: 412 383-7411); Pinar Yildirim, Wharton
School (e-mail: pyild@wharton.upenn.edu).
THE IMPACT OF ADVERTISING ON MEDIA BIAS
Abstract
In this study, the authors investigate the role of advertising in affecting the extent of bias in the
media. When making advertising choices, advertisers evaluate both the size and the composition
of the readership of the different outlets. The profile of the readers matters since advertisers wish
to target readers who are likely to be receptive to their advertising messages. It is demonstrated
that when advertising supplements subscription fees, it may serve as a polarizing or moderating
force, contingent upon the extent of heterogeneity among advertisers in appealing to readers
having different political preferences. When heterogeneity is large, each advertiser chooses a
single outlet for placing ads (Single-Homing), and greater polarization arises in comparison to
the case that media relies on subscription fees only for revenues. In contrast, when heterogeneity
is small, each advertiser chooses to place ads in multiple outlets (Multi-Homing), and reduced
polarization results.
Keywords: Media Competition; Bias in News; Advertising; Two-Sided Markets
1
Bias in news media is well known (e.g., Groseclose and Milyo 2005, and Hamilton 2004) and
can be defined as selective omission, choice of words and varying credibility ascribed to the
primary source (Gentzkow and Shapiro 2006). In a recent paper by Mullainathan and Shleifer
(MS 2005), a link is established between subscription fees and media bias. By assuming that
readers prefer news consistent with their political opinions and that newspapers can slant toward
these opinions, MS (2005) show that when the papers sole source of revenue is from
subscription fees (i.e., price for news), they slant news toward extreme positions.
For many media outlets, however, 60% to 80% of total revenue stems from advertising
(Strmberg 2004), as opposed to subscription. Thus, in this study, we aim to complement the
work of MS (2005) by recognizing that newspapers rely on revenues that accrue both from
subscription fees paid by readers and advertising fees paid by advertisers. We investigate how
the existence of these two sources of revenue affect the extent of bias in reporting that is selected
by the media.
In order to understand the role of advertising in determining the nature of competition
between newspapers, we specify in the model the effectiveness of advertisements to enhance
consumers probability of purchase. We argue that this effectiveness, for some products, may
depend upon the political opinions of readers of the ads. It has been long established in the
Consumer Behavior literature that products reflect a persons self-concept (Belk 1988). They
provide a way for a person to express her self-image, which may be strongly correlated with her
political opinions. We introduce, therefore, a product specific variable that measures the extent to
which political preferences play a role in enhancing consumers probability of purchase of the
product. While for some products this measure is significant, for others it is trivial. For example,
2
while green products, such as Toyota Prius, or Apples Mac computer may appeal more to
liberals, American products, such as the Chevy Truck, may appeal more to conservative
consumers. However, there are many products, such as automobile tires or insurance policies,
for which political opinions do not affect consumers choices to a large extent.1 When political
preferences play an important role in consumers purchase decisions, advertising the product can
be effective if it targets the correct consumers. An advertisement that reminds the consumers that
the product is consistent with their political opinions may increase the likelihood that they
purchase the product.
Heterogeneity among advertisers with respect to the appeal of their products to
consumers having different preferences is distributed in our model over a bounded interval. The
length of this interval captures the extent of heterogeneity among advertisers, with longer
intervals indicating significant differences in the appeal of products to liberal vs. conservative
readers. In our model we show that the degree of heterogeneity among advertisers plays a role in
determining whether advertisers choose to place ads with a single newspaper or with both
newspapers. The literature on two-sided markets has referred to these two possible outcomes as
Single and Double-Homing by advertisers, respectively (See Armstrong (2006), for instance.)
While Single-Homing arises as the unique equilibrium when the extent of heterogeneity is large,
Double-Homing arises when it is small.
We further investigate the manner in which the advertisers choice between the
newspapers affects the slanting strategies of media outlets. We show that when newspapers rely
both on advertising and subscription fees, advertising can serve as a polarizing or moderating
force in affecting the reporting of newspapers through two effects. First, adding the advertising
market implies that newspapers reduce their reliance on subscribers in favor of advertisers. As a
3
result, they may choose less slanting in their reporting strategies to improve their appeal to
moderate readers, and by doing so, offer a bigger readership to advertisers. This readership
effect enables the newspapers to charge higher advertising fees.
However, in seeking to lure advertisers a second, counter effect may arise when
advertisers choose to Single-Home. Specifically, when downward pressures on subscription fees
arise due to reduced slanting of the newspapers, similar downward pressures on advertising fees
appear, as well, as each newspaper attempts to defend its market share among advertisers. Hence
newspapers may have stronger incentives to polarize in order to alleviate price competition in
both markets. This incremental pricing effect to polarize is above and beyond the traditional
attempt of companies to introduce product differentiation in order to soften price competition in
a given market. Due to the two-sided markets we consider, polarization serves to soften price
competition in both markets.
We demonstrate that at the Single-Homing equilibrium, the incremental pricing effect
is stronger than the readership effect, thus leading to intensified bias in reporting. In contrast,
at the equilibrium with Double-Homing the readership effect is the only force present, thus
giving rise to reduced bias at the equilibrium.
There is a growing body of literature on media bias as implied by the medias attempt to
appeal to readers beliefs. In addition to MS (2005), Gentzkow and Shapiro (2006) and Xiang
and Sarvary (2007) also investigate this kind of bias. In Gentzkow and Shapiro (2006) readers
who are uncertain about the quality of an information source infer that the source is of higher
quality if its reports are consistent with their prior expectations. Xiang and Sarvary assume that
there are two types of consumers, those who enjoy reading news consistent with their political
opinions and conscientious consumers who care only about the truth. This assumption is
4
different from MS (2005) or our paper, where each consumer values both some consistency with
political opinions and accuracy. The reporting strategy of the newspapers depends then on the
relative weights consumers assign to consistency with their political opinions vs. accuracy. In
addition, these earlier studies on bias assume that the medias sole source of revenue stems from
selling news. In contrast, in the present study we allow the papers to earn revenues from
advertising fees as well.
There are two recent papers that consider, like us, a media market with both advertising
and subscription fees as sources of revenue. In Gabszewicz, Laussel and Sonnac (2002) and
Ellman and Germano (2009), advertisers care only about the size and not the profile of the
readership of each newspaper. This assumption is different from our setting, where advertisers
wish to target audiences that are receptive to their advertising messages. This targeting objective
of advertisers is pursued in Bergemann and Bonatti (2010) in an environment where the sole
source of revenues of media outlets is from advertising. In this recent study, the authors
investigate how improvements in the targeting technology that is facilitated by online advertising
affects the allocation of advertisements across different media and the equilibrium prices of
advertising messages. The topic of targeted advertising is also investigated in Iyer, Soberman,
and Villas-Boas (2005) in an environment where the firms themselves and not media outlets
possess the targeting technology.
Another strand of literature related to our study deals with consumers who may choose
one or two of competing products. In Sarvary and Parker (1997) consumers decide whether to
rely on a single information source or to diversify their purchases to include competing sources.
They show that the segmentation of consumers between those who purchase one or two sources
of information depends upon the relative importance consumers assign to obtaining precise
5
information. In Guo (2006), a similar diversification of the consumption bundle may arise when
there is uncertainty about future preferences. Buying competing products simultaneously serves
as insurance against such uncertainty. The main difference between our study and the previous
two is our focus on competition between media outlets in two-sided markets instead of the onesided framework considered in these studies.
Consider a market with two newspapers, i=1, 2, a mass of A advertisers and a mass of M
consumers, where M1 of these consumers are subscribers to one of these two papers and M2 are
nonsubscribers. Newspapers provide news and print advertisements. By simultaneously
operating in these two markets, newspapers have two potential sources of revenue: subscription
fees (Pi) and advertising fees (Ki).
Each of the M1 consumers reads either Newspaper 1 or 2 (but not both), and may buy
products from the advertisers. We adapt the model developed by MS (2005) to capture the
interaction between subscribers and newspapers. Specifically, when reading the newspaper, a
subscriber receives information about a certain news item t, which is distributed according to
N(0,
). Each consumer has some belief about the news item that is affected by her political
opinion. We designate this political preference by , and assume that the consumer believes the
news item to be distributed according to N(b,
). In comparison to the true distribution of the
news item, the consumers belief is biased. The political opinion parameter
measures the extent
and direction of this bias. It is uniformly distributed in the population of readers between b0 and
b0. For example, readers with beliefs closer to b0 can be considered liberals, and those in the
proximity of b0 can be considered conservatives.
6
Newspapers report news about . They receive some data
variable is independently distributed of t according to
by the newspapers may be different since and
, where the random
. Note that the data received
are random variables. Hence,
,
where
.2 Newspapers may choose to report the data with slant
news is
. Readers incur disutility when reading news inconsistent with their political
, so the reported
opinions, as measured by the distance between the reported news and the readers opinions: (
-
b)2. Holding constant the extent of inconsistency with their opinions, they also prefer less
slanting in the news. As in MS (2005), the overall utility of a reader is:
(1)
Where is the reservation price of the reader,
0,
calibrates her preference for reduced slant, and
calibrates the readers preference for hearing news consistent with her political opinion. Note
that the utility of the reader increases the smaller the slant
between the readers opinion
and the reported news
, and the smaller the discrepancy
.
Similar to MS (2005) we also focus on the characterization of the equilibrium with full
coverage of the market and linear slanting strategies of the newspapers in the form
with
interpreted as a choice of location of newspaper i.3 This location choice of
the newspaper can be a point inside or outside of the interval [b0, b0] and reflect the newspapers
political preference. Using
, the paper slants data toward its preference
news. Notice that the extent of slanting is an increasing function of
when reporting
and a decreasing function
of . Hence, as readers derive higher utility from hearing news consistent with their political
opinions and reduce the importance placed on obtaining accurate information, newspapers
choose greater slanting in their reporting. Without loss of generality, we assume that Newspaper
7
2 is located to the right of Newspaper 1 (B1 left, Newspaper 2 slants more to the right. Substituting the linear slanting strategies for and into Equation 1 and using the distributional properties of the random variable d (specifically, that and ), yields the expected payoff of a consumer having opinion b at the time she chooses between the two newspapers. Note that at this time, the realizations of and are yet to be determined due to the fluctuations of the data supporting news stories. At the time of the choice, the reader is aware only of the locations and fees chosen by the newspapers ( and ) as well as her own political opinion . Since the actual news may fluctuate depending upon the realization of , in evaluating the utility she derives from subscribing to the papers the reader calculates expectation over all possible realizations in Equation 1. For Newspaper and reader of type this yields the following expected utility. . The consumer who is indifferent between the two newspapers satisfies the equation . Solving this equation for (2) yields: = + . the papers subscription revenues are: Given the expression derived for (3) and . The population of advertisers is distributed according to the appeal of their products to consumers having conservative opinions, namely those situated in the positive segment of the distribution of opinions. We designate this appeal parameter by and assume it is uniformly distributed on the interval [ indicate products , ],
8
unappealing to conservative consumers with opinions in the range [ ,
], with more negative
values indicating increased appeal to liberal consumers with opinions in the range [
Positive values of
.
indicate products having the opposite characteristics, with bigger positive
values indicating increased appeal to conservatives. Products whose attractiveness to the
consumer is unlikely to be determined by political opinions assume an
neighborhood of zero. Given the above specification, the parameter
value in the
can be interpreted as
reflecting the extent of heterogeneity of the appeal of different products to consumers with
different political opinions.
We assume that in the absence of advertising each consumer has a certain probability of
purchasing a product. This probability can be modified with advertising. The change in purchase
probability for a given reader depends on the extent of compatibility between the political
opinion of the reader (her location b) and the type of the product advertised (its appeal ). When
an ad is successfully targeted to enhance compatibility, the readers purchase probability of the
advertised product increases. However, with lack of compatibility, her purchase probability
might actually decrease. We designate by
the incremental probability (positive or
negative) when a reader of political preference
is exposed to an ad related to product , and
specify it as:
(
(4)
),
where
.
Hence, the effectiveness of advertising is higher when political opinions are more
consistent with the appeal parameter of the advertised product, measured by the term
Equation 4. Note that the product
in
is positive for both liberal consumers of products having a
negative measure of appeal
and conservative consumers of products having a positive measure
of appeal. The parameter
is a measure of the basic effectiveness of advertising to increase
9
consumers purchase probabilities. The change in the probability of purchase
depends
also upon the extent of compatibility between the variables b and . For example, when a liberal
consumer is exposed to an advertisement of a green product, this will cause an increase in her
probability of purchasing this product that is above
, which is the basic increase in purchase
probability when the consumer becomes aware of the product due to the advertisement.
However, an extremely conservative consumer can respond very negatively to this product in
which case the change in her purchase probability due to the advertisement
might even
become negative.4, 5
The specification in (4) implies that an advertiser is likely to pursue two objectives in
designing its advertising strategy: to obtain a large audience for its ads and to target an audience
that is receptive to its advertising message. The first component of the advertising response
function motivates the large audience objective and the second motivates the targeting objective.
Finally, for simplicity, we assume that advertising has the same effect on a subscriber and
nonsubscribers with whom she shares information about advertised products. This assumption is
reasonable since subscribers tend to communicate with friends and relatives who normally hold
similar political opinions.
The payoff of an advertiser is measured by the average increase in the number of
consumers likely to buy its product (average incremental probability times the mass of
consumers
) net of the advertising fees paid to the newspapers. Hence, when an advertiser of
appeal parameter
chooses to advertise only in Newspaper 1, its expected payoff as derived
from the subscribers of Newspaper 1 is given as:
(5)
(
)
if it chooses to advertise only in Newspaper 2 its expected payoff is:
,
10
(
(6)
)
,
and if it chooses to advertise in both papers its expected payoff is:
(7)
.
By choosing to advertise only in Newspaper 1, an advertiser recognizes that subscribers to this
newspaper tend to have left leaning political opinions, lying in the interval
at the symmetric equilibrium( when
green product
to the newspaper (
where
). For instance, if it advertises a
in Newspaper 1, it can expect a positive payoff if the advertising fee paid
) is not too large, given that the average change in these readers purchase
probability due the advertisement is positive (i.e.,
(
)
). In contrast, by
choosing to advertise only in Newspaper 2, the advertiser draws readers who have more right
leaning opinions, in the interval
aware of its product (
(
when
. In this case, even though these readers become
), their political preferences are inconsistent with the product
and
), thus possibly leading to a negative expected payoff.
When advertising in both newspapers, an advertiser draws the entire population of
readers. An advertiser chooses to advertise in a single newspaper if
From Equations 5-7 it follows that for this advertiser
and
for
namely the
added benefit from advertising in the second newspaper falls short of the fee newspaper
charges. This may happen if the advertisers product appeals mostly to readers having extreme
political opinions. Advertising in a newspaper whose readership consists mostly of readers with
opposing opinions in the political spectrum may not be worthwhile to the advertiser in this case.
In contrast, an advertiser whose products appeal is not highly correlated with political
preferences (having an appeal parameter in the neighborhood of zero) may advertise in both
11
newspapers since the added benefit from advertising in each paper is likely to be positive for this
advertiser, implying that
. The above discussion indicates that the
population of advertisers can be segmented into at most three intervals as described in Figure 1.
[Insert Figure 1 about here]
Advertised products with appeal parameter less than are advertised only in Newspaper
1 since the advertisers of these products try to target mostly liberals (from (6)
increasing function of , thus if
,
is an
). In contrast, those with
for all
appeal parameter bigger than are advertised only in Newspaper 2, since advertisers wish to
reach only conservative readers for such high values of appeal parameter (from (5)
decreasing function of , thus if
values of
,
is a
). For intermediate
for all
, advertisers choose to advertise in both newspapers (since both
and
are positive in this range). The number of segments in Figure 1 can be smaller than three.
If
, no advertiser chooses to advertise in both newspapers (referred to in the literature on
two-sided markets as Single-Homing) and if
and
all advertisers choose to
advertise in both newspapers, (Double-Homing). Note, in particular that when
of
advertisers is located at
, the mass
, and in this case, advertisers do not care about targeting. At
the symmetric equilibrium, from Equations 5 and 6 each advertiser derives the net benefit of
when placing an ad with either one of the newspapers. Double-Homing is obviously
implied, given that both newspapers offer the same net benefit to each advertiser.
From Equations 5-7 we can derive the expressions for and as functions of the
locations and advertising fees chosen by the newspapers as follows:
(8)
{ (
)
},
{
(
)
}.
12
The appeal parameter ( ) characterizes an advertiser who is indifferent between advertising
in Newspaper 1(2) and advertising in both newspapers (i.e.,
and
).
In the Single-Homing equilibrium, the interior segment of Figure 1 disappears and the
advertiser who is indifferent between Newspaper 1 and 2 can be derived from Equations 5 and 6
by solving for
in the equation
(9)
:
(
)
(
)
.
From Equation 9 we obtain the advertising revenues that accrue to the newspapers in the
equilibrium with Single-Homing as follows:
(10)
and
.
When some advertisers Double-Home, the segment of the market covered by Newspaper
1 is
and that covered by Newspaper 2 is
. As a result, the
advertising revenues of the newspapers are:
(11)
and
.
In what follows we will derive symmetric equilibria with the market of advertisers fully
covered. At such equilibria,
, and
. We will focus on two possible
cases: equilibrium with Single-Homing, where each advertiser chooses to advertise in a single
newspaper (
Double-Home (
in Figure 1); and Double-Homing, where all advertisers choose to
).
We formulate the decision process of the newspapers as a two stage game. In the first
stage, each newspaper simultaneously announces a strategy si (d) of how to report the news (its
location
. In the second stage, the papers choose their prices Pi and Ki simultaneously.
Subsequent to those two stages, advertisers choose where to advertise and readers decide to
13
which newspaper to subscribe. Next, papers receive data d and report news d +si (d). Finally,
consumers read the news, get exposed to the advertisements, and form new impressions of the
advertised products.
Using this framework but with no advertising, MS (2005) show that the equilibrium
locations of the newspapers are
and
. Hence, with subscription
fees being the only source of revenues of newspapers, extreme bias in reporting, to the right by
Newspaper 2 and to the left by Newspaper 1, are chosen at the equilibrium. Such extreme
differentiation in reporting alleviates the extent of competition on subscription fees. In what
follows, we investigate how these equilibrium locations change if newspapers earn revenues
from advertising as well.
It may be interesting to point out how bias in reporting as a vehicle to introduce
differentiation between newspapers is different from other product features aimed at achieving
horizontal differentiation. First, the utility of readers depends upon two different attributes of
news reports, accuracy and consistency with political opinions, thus introducing potentially
opportunities for both vertical and horizontal differentiation. While the location choice of each
newspaper ( ) is the vehicle to introduce horizontal differentiation, the weight assigned to this
location in designing the slanting strategy (i.e.,
) captures the relative importance of the
vertical versus the horizontal attributes (i.e., accuracy vs. consistency with political opinions) in
the utility function of the consumers. In particular, if the consumers appreciation for accuracy
(the vertical attribute) is infinite, the papers stop slanting the news and dont use reporting bias
for horizontal differentiation. Another aspect that distinguishes bias from traditional models of
horizontal differentiation is that newspapers attempt to appeal to two different audiences, readers
and advertisers. Hence, the positioning of each newspaper has implications for price competition
14
in both markets. This contrasts with most models of product differentiation, where features are
chosen by taking into account competition in a single consumer market.
When both subscription and advertising revenues are available, the objectives of the newspapers
are:
Single-Homing (
)
(12)
where
,
and
are given in Equations 2 and 9, respectively.
Double-Homing (
)
(13)
where
;
,
;
is given by Equation 2.
The newspapers choose subscription and advertising fees in the second stage to maximize
Objectives 12-13. When the newspapers locate symmetrically so that
, the
solution to the maximization is as follows:
Single-Homing (
)
(14)
Double-Homing (
(15)
,
.
)
,
.
Hence, for a fixed symmetric choice of locations, subscription fees are higher if
subscribers have greater preference for reports that are consistent with their political opinions
15
(bigger ), smaller preference for accurate reporting (smaller ), and are more heterogeneous
(bigger
). Subscription fees are also higher when the advertising market is smaller (smaller A),
the relative size of the population of subscribers is bigger (bigger
of advertising declines (smaller
), and the effectiveness
). In general, the more important advertising revenues in
comparison to subscription revenues, the lower the fees newspapers charge to subscribers at the
symmetric equilibrium.
Substituting the equilibrium advertising fees derived in Equations 14 and 15 back into
Equation 8 implies different types of homing depending on the extent of heterogeneity among
the advertisers (value of
). While for large values (
equilibrium, for small values (
), Single-Homing is the unique
, Double-Homing is the unique equilibrium.6 As
explained earlier, advertisers in our environment care both about the number and profile of
readers who are exposed to their ads. When heterogeneity among advertisers is significant,
targeting readers who are compatible with advertised products is very important to the
advertisers. Single-Homing is more successful than Double-Homing in achieving such targeting.
In the absence of targeting, ads might reach consumers with extreme political opinions
incompatible with the products advertised. When heterogeneity is large, such lack of targeting is
especially costly for advertisers since the product
might assume very large negative values in
Equation 4.
To obtain the equilibrium locations chosen by the newspapers in the first stage, one has to
solve first for the second stage fees,
and
, as functions of arbitrary location
choices selected in the first stage (not necessarily symmetric locations only). The second stage
equilibrium strategies have to be substituted back into Equations 12-13 to obtain the first stage
payoff functions of the newspapers.
16
Assuming the existence of an interior equilibrium, next we compare the locations
selected at the symmetric equilibrium (designated by
obtain revenues from subscribers only (denoted as
heterogeneity among advertisers, namely when
) to those derived when newspapers
). When there is no
, advertisers Double-Home and
, meaning that bias remains unaffected when advertising is added as a source of
revenue. However, when
, adding advertising to supplement subscription fees may
moderate or intensify bias. In Lemma 1, we first derive restrictions on the parameters of the
model to guarantee that those regimes can be supported with positive streams of revenues from
subscribers (namely that
and
). For ease of presentation, we introduce a measure
for the importance of advertising relative to subscription as a source of revenue for the papers,
(
), where
relative to the subscription market and
represents the size of the advertising market
is a measure of the importance consumers
attach to accuracy relative to consistency with their political opinions. If consumers attach great
importance to accurate reporting (i.e.,
is large), the papers cannot charge high
subscription fees. Hence, if either one of the two components of T increases, the subscription
market loses its importance as a source of revenues relative to the advertising market.
LEMMA 1. To ensure positive subscription prices and strict differentiation between
newspapers (i.e.,
and
):
(i) At the Single-Homing equilibrium:
, and
(ii) At the Double-Homing equilibrium:
, and
.
.
Restricting attention to the regions specified in Lemma 1, we derive the optimal locations
chosen by the newspapers at the symmetric equilibrium in Equations 16 and 17.
17
Single-Homing
(
(16)
)
(
(
(
))
)
.
Double-Homing
(
(17)
)
.
Proposition 1 follows from the expressions derived in Equations 16 and 17.
PROPOSITION 1. With both advertising and subscription fees contributing to the newspapers
revenues,
(i) When heterogeneity among advertisers is sufficiently large (
):
Each advertiser chooses a single newspaper for placing its ads (Single-Homing), and
newspapers introduce more bias in their reporting (
). This bias increases as the
importance of advertising as a source of revenue increases (
).
(ii) When heterogeneity among advertisers is sufficiently small (
):
Each advertiser chooses both newspapers for placing its ads (Double-Homing), and
newspapers introduce less bias in their reporting (
importance of advertising as a source of revenue increases (
). This bias decreases as the
).
To understand the results reported in Proposition 1, it is important to highlight the new
effects influencing the location choice of the newspapers that arise when advertising is added as
a source of revenues to supplement subscription fees. The first readership effect relates to the
intensified incentives of each newspaper to increase its readership (for Newspaper 1 this means
increasing
, and for Newspaper 2 decreasing it). Note that at the symmetric equilibrium
18
(when
)
and
.7 Hence, irrespective of the type of
homing, a newspaper that delivers a bigger readership can command a higher advertising fee
from advertisers. This implies that each newspaper has extra incentives to move closer to its
competitors location in order to increase its market share among readers (e.g.,
symmetry, when
at
).
Adding advertising as a source of revenue introduces, though, a second counter force
when advertisers Single-Home. We refer to this force as the incremental pricing effect to
capture the idea that a change in a newspapers location does not only have a direct effect on the
intensity of price competition in the subscription market but may also have an indirect,
incremental effect on the intensity of price competition in the advertising market.8 When a
newspaper modifies its location and advertisers Single-Home, the competing newspaper may
have to adjust its advertising fee in order to defend its market share among advertisers. For
instance, when Newspaper 1 increases
, it moves closer to the location of Newspaper 2, and
due to reduced differentiation, Newspaper 2 is forced to cut subscription fees . In addition, since
the new, moderated location of Newspaper 1 offers a larger readership to advertisers, Newspaper
2 has to cut its advertising fee as well in order to defend its market share in the advertising
market.9 The existence of this incremental pricing effect introduces, therefore, incentives for
Newspaper 1 to polarize in order to discourage aggressive pricing by Newspaper 2. These
incentives are stronger than in an environment where newspapers compete in a single, subscriber
market because Newspaper 2 is forced to cut both its advertising and subscription fees.
According to part (i) of Proposition 1, the incremental pricing effect present at the
Single-Homing equilibrium more than outweighs the objective of increasing readership, thus
19
leading to intensified bias at the equilibrium when advertising is added as a source of revenues to
augment subscription fees. Moreover, this bias increases as the importance of advertising as a
source of revenue ( ) increases. In contrast, according to part (ii) of the Proposition, at the
equilibrium with Double-Homing, bias in reporting the news is reduced when advertising
supplements subscription fees. At this type of equilibrium, the only additional effect that
advertising introduces is the added objective of newspapers to offer bigger readerships to
advertisers. Since the market share of each newspaper in the advertising market is fixed at 100%
and the newspapers dont need to defend their market shares among advertisers, the incremental
pricing effect is non-existent in the Double-Homing environment. Note that the readership
effect intensifies, in this case, when advertising is a more important source of revenue (large ).
Figure 2 depicts the relationship between the equilibrium locations of the newspapers and the
importance of advertising as a source of revenue to the newspapers, as reported in Proposition
1.10
[Insert Figure 2 about Here]
We can use the results reported in Proposition 1 to conjecture how the equilibrium is
likely to change in case of less than full coverage of readers. At the Single-Homing equilibrium
(when
is big) bias in reporting is significant. Hence, it is sensible that when the market is less
than fully covered, it is consumers with moderate opinions in the neighborhood of b=0 who
choose to drop out of the market
for such consumers). As a result, the subscribers of
each newspaper are fewer in number and have more extreme beliefs in comparison to a fully
covered market. This new composition of subscribers reduces even further the benefit from
Double-Homing. In the Web Appendix, we demonstrate that newspapers may have reduced
incentives to polarize as a result of incomplete coverage of the subscriber market. In fact, when
20
the reservation price of readers is relatively low and their valuation of accurate reporting is high,
bias is more moderate than that derived in MS (i.e., smaller than
Single-Home. At the Double-Homing equilibrium (when
) even though advertisers
is small) bias is moderate. It is now
consumers with very extreme opinions who are likely to drop out of the market. The population
of subscribers becomes less heterogeneous, as a result, thus enhancing the benefit from DoubleHoming. In the Web Appendix, we demonstrate, that in this case as well, incomplete coverage
may moderate the extent of bias selected by the newspapers if the reservation price of readers
(and their valuation of accuracy) is low (high), respectively.
In this paper we extend the work of MS (2005) by investigating media bias when advertising is
added as a source of revenue to supplement subscription fees. We show that the additional
advertising market introduces two counteracting effects on the behavior of newspapers. First, as
newspapers attempt to increase their readership in order to attract advertisers, they moderate
slanting in order to appeal to readers having moderate opinions. Second, when advertisers choose
to Single-Home a second effect arises that may lead to greater polarization in news reporting. If
newspapers moderate bias in this case they are forced to compete more aggressively not only for
subscribers, but for advertisers as well. Downward pressure on subscription as well as
advertising fees follows. To avoid such intensified price competition, newspapers may choose to
increase polarization. We demonstrate that when the heterogeneity among advertisers in
appealing to consumers with different political preferences is significant, the attempt to alleviate
price competition dominates, thus leading to greater polarization. When this heterogeneity is
negligible, reduced polarization is predicted.
21
REFERENCES
Armstrong, Mark (2006), Competition in Two-Sided Markets, RAND Journal of Economics,
37 (3), 668-91.
Belk, Russell W. (1988), Posessions and the Extended Self, Journal of Consumer Research,
15 (2), 139-68.
Bergemann, Dirk and Alessandro Bonatti (2010), Targeting in Advertising Markets:
Implications for Offline vs. Online Media, Cowles Foundation Discussion Paper 1758,Yale
University
Ellman, Matthew and Fabrizio Germano (2009), What do the Papers Sell? A Model of
Advertising and Media Bias, Economic Journal, 119 (537), 680-704.
Gabszewicz, Jean J., Didier Laussel and Nathalie Sonnac (2002), Press Advertising and the
Political Differentiation of Newspapers, Journal of Public Economic Theory, 4 (3), 317-34.
Gentzkow, Matthew and Jesse M. Shapiro (2006), Media Bias and Reputation, Journal of
Political Economy, 114 (2), 280-316.
Groseclose, Tim and Jeffrey Milyo (2005), A Measure of Media Bias, Quarterly Journal of
Economics, 120 (4), 1191-1237.
Guo, Liang (2006), Consumption Flexibility, Product Configuration, and Market Competition,
Marketing Science, 25 (2), 116-130.
Iyer, Ganesh, David Soberman and J. Miguel Villas-Boas (2005), The Targeting of
Advertising, Marketing Science, 24 (3), 461-76.
Mullainathan, Sendhil and Andrei Shleifer (2005), The Market for News, American Economic
Review, 95 (4), 1031-53.
Sarvary, Miklos and Philip Parker (1997), Marketing Information: A Competitive Analysis,
Marketing Science, 16 (1), 24-38.
Strmberg, David (2004) Mass Media Competition, Political Competition, and Public Policy,
Review of Economic Studies, 71 (1), 265-84.
Xiang, Yi and Miklos Sarvary (2007), News Consumption and Media Bias, Marketing
Science, 26 (5), 611-28.
22
FOOTNOTES
1
The appeal parameter of the advertised product that we introduce in the model assumes a value in the vicinity of
zero if political preferences do not play an important role in consumers purchase decisions.
2
Notice that there is no vertical differentiation between the newspapers in this setting (i.e., the accuracy of the data
received by both newspapers is identical:
). In the Web Appendix, we demonstrate that our
utility specification may also give rise to a tradeoff between vertical and horizontal differentiation. Specifically,
when
, | | | |.
3
In the Web Appendix we show the optimality of linear slanting strategies when the newspapers sole source of
revenue is from subscription fees. However, in our analysis, in which both advertising and subscription fees are
sources of revenue, we implicitly assume that the linearity of slanting strategies is still valid.
4
According to Equation 4, the change in the purchase probability for extreme products and consumers is larger than
that for moderate products and consumers. As we mention later in footnote 10, when this feature of our model is not
valid, some of our results may change, even though the strategic effects we identify will continue to operate.
5
Let
denote the initial probability of purchase in the absence of advertising by an individual with opinion and
denote the probability of purchase after advertising such that
. In order to guarantee that
we assume that
and
. Note that these parameter restrictions
do not conflict with those given in Lemma 1.
6
Note that between
and
there is an equilibrium in which while some advertisers Single-Home (place
their ads in a single newspaper), others Double-Home (place ads in both outlets). As well, multiple equilibria may
arise in this range. (See Web Appendix for derivations.)
7
(
The solution for the advertising fees as functions of the locations are:
(
(
)
) and
(
)
(
),
(
(
)
)
),
(
)
8
Note that this effect does not exist in standard models of horizontal differentiation in which a change in location
has implications on price competition in only one market.
9
As Newspaper 1 increases its readership by increasing
, Newspaper 2 loses market share among advertisers
since at the symmetric equilibrium from Equation 2 and from Equation 9
has an incentive to cut its advertising fee since
10
. Thus, Newspaper 2
at symmetry.
Note that with a different advertising response function, which implies that the change in purchase probability for
moderate products and consumers is larger than that for extreme products and consumers, the readership effect will
be stronger, since in this case, the moderate readers will be more valuable for the advertisers, and therefore the
newspapers. We predict that while the results for Double-Homing reported in Proposition 1 will continue to hold in
such an environment, the results for Single-Homing may change as the readership effect may outweigh the
incremental pricing effect.
23
FIGURES
Only Newspaper 1
Both Newspapers
Only Newspaper 2
Figure 1: Segmentation of the Advertising Market
Figure 2: Equilibrium Locations as a Function of T
1
WEB APPENDIX
Derivations of Equations 16-17 and Proof of Lemma 1
(i) Single-Homing: Second stage prices are obtained by optimizing (12) with respect to
and
as follows:
(A1)
[
(A2)
[
(A3)
[
(A4)
[
]
[(
]
)
[(
(
)
)]
(
]
,
]
,
and
)]
.
From (9):
(A5)
(
)
(A6)
(
(
(A7)
)
(
)
)
.
, and
.
From (2):
(A8)
and
.
Substituting (2), (9), (A5), (A6) and (A7), into the first order conditions (A1)-(A4),
evaluating them at symmetry (
), and solving for
and
, we get
and
given in (14). And substituting (A6), (A7) and (9) into (A3) and (A4) and solving for
and
as
,
one can get equilibrium advertising fees as a function of the locations:
(A9)
(
(
)
),
(
(
)
)
To obtain the equilibrium locations chosen by the newspapers in the first stage, one has to
solve first for the second stage fees,
and
as functions of arbitrary location
choices (not only symmetric). Substituting the equilibrium strategies back into (12), we obtain
the first stage payoff functions designated as
locations yields from the Envelope Theorem that:
. Differentiating with respect to the
2
(A10)
.
To illustrate the derivation of the first stage equilibrium, we focus on the optimization of
Newspaper 1. For this newspaper, the terms of (A10) can be derived as follows:
(A11)
(
=
)
(
(A12)
(
)
(A13)
(
)
While the expression for
expression from
)
,
(
)
and
.
in (A13) can be directly derived from (A9), to obtain the
in (A12), we need to utilize the Implicit Function Approach by totally
differentiating the first order conditions (A1) and (A2) that determine subscription fees (
and
= 0). We obtain:
(A14)
(
(A15)
(
)
(
)
(
)
(
)
(
)
and
)
(
)
(
)
.
From (A14) and (A15):
(A16)
[
]
[
]
[
].
Using (A1) and (A2) in evaluating (A16) at the symmetric equilibrium yields:
(A17)
where
[
]
[
]
,
[
];
and
.
For second order condition, the determinant of the inverted matrix on the RHS of (A17)
should be positive implying that Z< 1.5. From (A17), therefore:
(A18)
(
).
3
We can now complete the characterization of the optimal location choice of Newspaper
from (A9) and
from (A18) in (A10),
we obtain at the symmetric equilibrium:
(A19)
, where
At the symmetric equilibrium when
is given by (A18).
, we obtain from (A19) a quadratic
equation as follows:
(
(A20)
(
) )
(
)
(
))
.
The two roots of this quadratic equation are:
(
)
; where
(
(
Only the bigger root guarantees stability of reaction functions (i.e.
).
< 0.) As a result,
the optimal location at the Single-Homing equilibrium is given in (16). Note that if
quadratic expression (A20) is positive for all values of . Hence,
optimal location is the corner solution
(
(A21)
(
. Hence,
))
the
for all B and the
if:
(
)
.
Inequality (A21) holds if:
(A22)
.
We next investigate the conditions under which
(A23)
=
implies
(A24)
is positive. From (14):
(
.
or equivalently from (16):
(
))
(
)
(
).
Given that the LHS is positive, there are two cases where this inequality can hold: when
RHS is negative (Case 1) and when both sides are positive but the LHS is bigger (Case 2). Case
4
1 implies that
. For Case 2, squaring both sides of (A24) and solving for T yields
. Combining the two cases, yields that
(A25)
if:
.
Combining (A25) and (A22) yields the condition of part (i) of Lemma 1.
(ii) Double-Homing: Using a very similar approach to that developed when advertisers SingleHome, we obtain the following first order condition for the choice of location in the first stage.
[
(A26)
]
where the expression for
, which follows from the maximization of (11) is:
(
(A27)
,
)
(
(
),
)
(
).
At the symmetric equilibrium (A26) reduces to:
(A28)
[
]
To derive the expression for
, we have to use, once again, the Implicit Function Approach, by
.
totally differentiating the first order condition for the subscription fees
(A29)
. Those conditions are:
(
)
{
(
)
}
,
(
)
{
(
)
}
.
Total differentiation of the first order conditions yields the following system of equations for
and
:
[
][
order conditions
or
]
[
] where
. Solving for
and for second
, we obtain:
]. Substituting back into (A28), yields a quadratic equation in B as follows:
(A30)
(
)
.
There are two roots to this equation. However, only one satisfies also the condition for stability
of reaction function. It is given in equation (17). The discriminant of the solution in (17) is
5
positive if
. As well, to guarantee that
, it follows from (15) that
the expression for B from (17) in the last inequality, yields
. Using
. This is a more
demanding constraint than the one necessary to insure that the discriminant is positive, thus
yielding part (ii) of Lemma 1.
Proof of Proposition 1
First note from (16) and (17) that
expressions of
and
when
with respect to
type of equilibrium yields
and
Regions of the Parameter
. Differentiating the
for the range of parameters that support each
. Hence,
and
.
that Support Single and Double-Homing
(i) Single-Homing: To ensure that Single-Homing is an equilibrium, we use
from (14) in (8)
to obtain
and
and evaluate (8) at the symmetric equilibrium (i.e.,
. To guarantee that the interior interval in Figure 1 disappears, we impose the
restriction that
, which happens when
.
(ii) Partial Double-Homing: In Partial Double-Homing equilibrium, in which some advertisers
Double-Home and some Single-Home (i.e.,
), the newspapers choose
subscription and advertising fees in the second stage to maximize the objectives:
,
.
When the newspapers locate symmetrically so that
, the solution to this
maximization can be obtained as:
and
Using
we obtain
in (8) and evaluating it at the symmetry (i.e.,
. Hence,
if
, and
if
(iii) Double-Homing: All advertisers will Double-Home when
that this happens if
and
.
. It follows from part (ii)
.
From (i), (ii) and (iii) we conclude that while for
equilibrium, for
.
Single-Homing is the unique
Double-Homing is the unique equilibrium. It also follows from the
6
above that between
and
Single-Homing and Partial Double-Homing equilibria may co-
exist.
Optimality of Linear Decision Rules
We show the optimality of linear slanting strategies when the newspapers sole source of revenue
is subscription fees. To this end we first derive the first order conditions that follow from the
newspapers first stage location choices without restricting the functional form of the slanting
strategies
. Then we show that a linear slanting rule satisfies these first order
conditions.
The consumer who is indifferent between the two newspapers satisfies the equation
where
is given by (1). Solving this equation and using the distributional
properties of the random variable
(
) and the uniform distribution of the parameter
, yields:
(A31)
In the second stage, newspapers set their prices
and
to maximize (3). First order
conditions for this maximization are:
(A32)
((
)
)
,
(A33)
((
)
)
.
At the symmetric equilibrium (i.e.,
(A34)
) the solution to (A32) and (A33) is:
.
To obtain the equilibrium locations chosen by the newspapers in the first stage, one has to
solve first for the second stage fees,
as functions of arbitrary location choices (not only
symmetric). Substituting the equilibrium strategies back into (3), we obtain the first stage payoff
functions
. Differentiating with respect to the locations yields from the Envelope
Theorem that:
(A35)
It follows from (A35) and (3) that,
.
7
(A36)
,
(A37)
.
From (A31):
(A38)
,
(A39)
{
To obtain
}
, where
.
, we utilize, once again, the Implicit Function approach by totally differentiating the
first order conditions (A32) and (A33), and solving the resulting system of equations using
from (A34). This procedure yields
(A40)
.
Suppose the firms use a linear decision rule:
(A41)
, then:
(A42)
,
(A43)
,
(A44)
.
Substituting (A42)-(A44) into (A38)-(A40), and using these in (A37) at the symmetric
equilibrium yields:
(A45)
.
From (A45):
(A46)
.
Thus, the linear slanting rule (A41) allows us to solve the first order condition in (A37) and find
an optimal location as given in (A46). Since the newspapers are symmetric, the same rule also
satisfies the first order condition for Newspaper 1(i.e., (A36)) as well.
We will now show that this rule can be expressed as
for
newspaper i. We will assume that Newspaper 1 follows such a decision rule and demonstrate that
the best response of Newspaper 2 is to follow such a rule, as well. Specifically, assuming that
8
we designate the linear rule followed by Newspaper 2 as
, where
.
Given the assumed behavior of Newspaper 1, a consumer who chooses to subscribe to it
derives the expected payoff:
. Her payoff
when choosing Newspaper 2 is:
(
)
.
To find the consumer who is indifferent between the two newspapers we solve the equation
for b as follows:
(
(A47)
(
Assuming without loss of generality that
)
.
)
, in the second stage the newspapers
choose their subscription fees to maximize:
and
. Note that
simply guarantees that Newspaper 2 serves the upper end of subscribers above
and Newspaper 1 serves the lower end. Once the coefficients are derived this assumption
is indeed satisfied as
. Optimizing with respect to
prices as a function of
,
(A48)
and
and
, yields the second stage
as follows:
(
)
(
)
[
(
) ]
,
[
(
) ]
.
It follows, therefore, that:
(
)
,
When Newspaper 2 chooses its slanting strategy rule in stage 1, namely
its payoff function
for
.
and
, it optimizes
, given the prices established subsequently in the second stage. Substituting
and
function for Newspaper 2,
back into the payoff functions, yields the first stage payoff
. Using the Envelope Theorem:
9
(A49)
[
]
[
]
[
[
[
(
)]
]
when
decision rule of Newspaper 2
,
]
[
From the second equation of (A49), it follows that
where
]
.
. Hence, the
can be written as
has been normalized to
. To find the value of
, we
further restrict our attention to symmetric Bayesian equilibria, which implies that
Substituting into the first equation of (A49), implies that
symmetry assumption,
, thus,
.
, and by the
.
Incomplete Coverage of the Subscribers Market
Single-Homing:
In this section, we demonstrate that when advertisers Single-Home and when the subscriber
market is not covered as in Figure (A1), there exist conditions under which
Buy Newspaper 1
Do not Buy
Buy Newspaper 2
Figure A1: Segmentation of the Subscriber Market
The reader who is indifferent between buying Newspaper 1 and not buying at all satisfies the
equation:
(A50)
where
The above equation has two roots:
from (1).
10
(A51)
.
Notice that for the segmentation given in Figure A1 to hold we need
at
which implies:
(A52)
.
Further at
(A53)
, we should have
and at
,
, thus :
, and
.
The root in (A51) that satisfies (A52) is:
(A54)
(
)
(
).
Similarly, the location of the reader who is indifferent between buying Newspaper 2 and not
buying at all can be calculated as:
(A55)
and
(A56)
(
),
if
, and
.
When an advertiser of appeal parameter
chooses to advertise in Newspaper 1, its
expected payoff is given as:
(A57)
(
)
.
If it chooses to advertise in Newspaper 2 its expected payoff is:
(A58)
(
)
.
The advertiser who is indifferent between Newspaper 1 and 2 can be derived from (A57) and
(A58) by solving for
in
:
(A59)
.
In the last stage the newspapers set their subscription and advertising fees to maximize their
profits:
(A60)
which yields the following first order conditions:
,
,
11
(A61)
[(
)
(A62)
[(
)
(A63)
(
(A64)
(
(
)
(
)
(
)
(
)
]
]
(
)
)
(
) (
(
)
,
)
)
(
) (
.
)
Simultaneously solving (A61) and (A62) we get:
(A65)
(
(
)
),
(A66)
(
(
)
).
Thus, at the symmetric equilibrium (i.e.,
(
(A67)
)
):
.
Using (A67) in (A64) at the symmetric equilibrium yields:
(
(A68)
)
.
To obtain the equilibrium locations with incomplete coverage we differentiate the first
stage payoff functions
with respect to the locations and use the Envelope Theorem:
(A69)
.
.
For Newspaper 2, it follows from (A69) that,
(A70)
Using the Implicit Function approach by totally differentiating the first order conditions (A63)
and (A64) and solving the resulting system of equations using
from (A67) and
from
(A68) at the symmetric equilibrium we obtain:
(A71)
where
[
(
)(
)
(
)
(
)
].
(
)
From (12), (A59), and (A71), at the symmetric equilibrium (A70) becomes:
(
)
12
[(
(A72)
)
(
)]
(
From (A55),
)
. Notice that the first term inside the brackets in (A72) is always positive
and the second term is positive if
Therefore, if
,
and if
. From (A65)
if
.
, and the newspaper will continue to reduce bias. Since
it follows that bias will be reduced until
with complete coverage
. Because at the Single-Homing equilibrium
it follows that
Hence, incomplete coverage of the type
depicted in Figure A1, will never lead to an equilibrium where each newspaper covers less than
half of the segment of readers who prefer its location best (for 1 this segment is
this segment is
, and for 2
).
We now derive the conditions on the parameters of the model to guarantee that less than
full coverage moderates the extent of bias at the equilibrium in comparison to
. To
obtain the conditions we substitute the equilibrium price from (A68) back into (A55) and solve
for
in terms of
as follows:
(A73)
(
It is very easy to show that
)
in the above expression is positive when
right-hand side of (A73) at
(A74)
(
Hence, as long as
).
if:
.
is sufficiently small (e.g., when the reservation price of readers is low and
their valuation of accurate reporting
bias below
. Evaluating the
yields therefore, that:
The right-hand side of (A74) is smaller than
(A75)
.
is high), incomplete coverage may yield moderation of
. Note that condition (A75) does not necessarily contradict (A53) and (A56),
conditions necessary to support the type of incomplete coverage we consider.
Double-Homing:
13
Suppose that the newspapers, this time, cover only the middle of the market (i.e, between
and
in Figure A2) and advertisers Double-Home.
Do not Buy
Buy Newspaper 1
Buy Newspaper 2
Do not Buy
Figure A2: Segmentation of the Subscriber Market
We demonstrate that at the limit as
and
, the newspapers may have
incentives to moderate their bias in reporting in comparison to the full market coverage case that
is analyzed in Section 4.
Again, the reader who is indifferent between buying Newspaper 1 and not buying at all
satisfies equation (A50). For the segmentation given in Figure A2 to hold we need:
at
(A76)
Further at
which implies:
.
, we should have
and at
,
. The root in (A51) that
satisfies (A76) is:
(A77)
(
)
(
).
Similarly, the location of the reader who is indifferent between buying Newspaper 2 and not
buying at all can be calculated as:
(A78)
(
).
One can show that in this case as well, the newspapers will choose to cover at least one half of
the readers who prefer their locations best, namely
(A79)
{
and
. Note from (A78) that:
14
Equation (A79) implies that newspapers may have incentives to moderate their bias, once again.
(
, and
)
(
The benefit to advertisers in this case can be derived as:
)
.
When advertisers Double-Home, advertising fees are determined by the requirement that
and
, thus yielding:
(A80)
(
),
(A81)
(
).
Newspapers profits when they cover only the middle of the market (and when
advertisers Double-Home) are:
(A82)
where
,
, and
.
are as given in (A80) and (A81). Let
and
denote Newspaper
2s profits when it covers and does not cover the market as given in (13) and (A82), respectively.
Then, when
and
:
(
(A83)
)
.
Recall that at the Double-Homing equilibrium
. Hence, the sign of the above difference
depends only on the sign of
if
. From (A79),
. This inequality is more likely
when the variable , defined in (A75), is relatively small (e.g., when the reservation price of
readers is low and their valuation of accurate reporting
is high). If the sign of (A83) is
negative, newspapers have incentives to moderate when
when
equal to
and
. Note that even
, in which case the newspapers have incentives to polarize, bias will at most be
, because,
when
and
.
Asymmetric Accuracy
We investigate the impact of asymmetry in newspapers data accuracy on reporting bias when
the papers sole source of revenue is subscription fees. Specifically, we assume that Newspaper 1
15
has access to more accurate data than Newspaper 2:
. In this case location of
reader who is indifferent between the two newspapers is:
(A84)
=
+
.
For the newspapers second stage pricing decisions, we optimize (3) with respect to
and
which yields the following first order conditions:
(A85)
((
)
)
,
(A86)
((
)
)
.
Substituting (A84) in (A85) and (A86) and simultaneously solving for
and
we get
equilibrium subscription fees as functions of the newspapers locations:
(
(A87)
)
(
(A88)
(
)
),
(
).
Substituting these second stage equilibrium strategies
first stage payoff functions
into (3) we obtain the
. Differentiating with respect to the locations yields from
the Envelope Theorem:
(A89)
.
It follows from (A89) that,
(A90)
,
(A91)
.
Using (A87), (A88) and (A84) to find the terms of (A90) and (A91) and solving them
simultaneously for the newspapers first stage location choices yields:
(
(A92)
(A93)
(
)
,
)
.
The equilibrium locations derived illustrate that Newspaper 1, which has access to more
accurate data, introduces less bias in reporting and the opposite is true about Newspaper 2, which
has less precise data available. Hence, the utility formulation introduces some tradeoff between
16
vertical differentiation (in our case precision of data) and horizontal differentiation (in our case
bias in reporting). There may be different reasons why a newspaper has access to more accurate
data, including lower cost of conducting investigations due to greater experience in investigative
reporting. Hence, if
can be chosen endogenously, the newspaper facing lower cost will
likely choose greater accuracy in gathering information.