Available online at www.sciencedirect.

com
INFORMATION
ECONOMICS
AND POLICY

Information Economics and Policy 20 (2008) 107–119

www.elsevier.com/locate/iep

Advertising and the screening role of mass media

Susanne Kremhelmer a, Hans Zenger b,*
a
University of Munich, Department of Economics, Ludwigstr. 28, 80539 München, Germany
b
European Commission, Chief Economist Team, DG Competition, 1000 Bruxelles, Belgium

Received 26 June 2006; received in revised form 23 November 2007; accepted 24 November 2007
Available online 25 January 2008

Abstract

We argue that there exists a problem of adverse selection in the provision of advertising which makes it impossible to
establish direct markets for it. The media are regarded as intermediaries that can channel advertising and allocate it effi-
ciently by screening consumers. This role of media may result in excessive quality and prices of media products. These dis-
tortions will be more severe if consumers’ quality preferences are relatively homogenous.
Ó 2007 Elsevier B.V. All rights reserved.

JEL classification: D82; L11; L82

Keywords: Adverse selection; Advertising; Media economics

1. Introduction

Both in the US and in Europe numerous professional websites offer consumers money for reading E-mail
advertisements, viewing advertising banners in their browsers, and the like. Those websites are financed by the
firms that book the advertising. What they offer is something that is a rare phenomenon; they are direct sup-
pliers of the product advertising, which is typically provided indirectly, as a bundle with some media product.
The question arises why there are so few markets on which advertising is traded directly. We argue that this
is mainly due to adverse selection. Clearly, firms are willing to pay different amounts of money to different
consumers depending on unobservable characteristics like income, age, or interest in the advertised product.
Hence, every consumer has an incentive to claim that he has profitable characteristics. Following Akerlof’s
(1970) classical paper, this incentive may lead to a market unraveling so that only consumers with bad char-
acteristics can trade. The market for rich consumers (who require a relatively high compensation since they
usually have higher opportunity costs of time) breaks down because it is not possible to distinguish them from
consumers who are less attractive as advertising targets.

*
Corresponding author.
E-mail address: hans.zenger@ec.europa.eu (H. Zenger).

0167-6245/$ - see front matter Ó 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.infoecopol.2007.11.003
108 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119

This phenomenon is precisely what can be observed at the websites that were mentioned above. All of the
major advertising providers are visibly plagued by adverse selection. Most of them pay extremely low rates –
usually around one cent for viewing an on-line advertisement for a significant amount of time and then fol-
lowing a confirmation procedure. Also, the advertising that they feature is obviously targeting low income
consumers. Major advertisers are bargain-websites, loan-sharks, financial institutions offering credit cards
without solvency check, and dubious internet business opportunities. Clearly, the average payout rate is
too low for high income consumers, who drop out of the market, reducing the quality of the pool; this implies
that the payout rate has deteriorated even more, and the process repeats until only the lowest income type
remains.1,2
As will be demonstrated below, the same kind of adverse selection problem exists when people are not
directly paid for consuming advertising – that is, when advertising is provided through the media. We show,
however, that distributing advertising through media platforms can at least mitigate the problem of adverse
selection from which direct markets suffer so severely. There are some obvious ways in which this happens.
For instance, tennis rackets are advertised in tennis magazines. Also, high income types can be targeted by
placing ads in golf magazines. But there is a limit to this kind of targeting. In particular, most companies want
to reach broader audiences of high income consumers than the very small subset of those happening to read
golf magazines. In addition, a great many products are not as target-group specific as tennis rackets. Hence,
most advertisers have to rely on mass media like TV broadcasting in order to get their messages across.
TV broadcasters sell a bundle of products consisting of a primary product (the content) and a secondary
product (the advertising). Agents have to pay for consuming programs and they receive a reimbursement for
consuming the advertising in the form of a lower (possibly even zero) subscription fee. In trading the adver-
tising, the broadcaster acts as an intermediary on behalf of the advertising companies. Since it offers both
products as a bundle, the broadcaster can tackle the adverse selection problem; by distorting the market
for its primary product (for example by altering its price or quality), it can mitigate the distortion in the sec-
ondary market. Thus, the two-sided market nature of media firms allows more efficiency in the provision of
advertising.
As an illustration of what we have in mind, consider a broadcaster that offers two movie channels in an
attempt to discriminate between rich and poor viewers. If the channel that targets high income viewers con-
tains less advertising, that channel will attract low income viewers, counteracting the broadcaster’s intent to
show more advertising to low types (who generate lower advertising revenues). Therefore, the broadcaster may
increase the subscription fee or alter the content of the high type channel in order to discourage low income
viewers, leading to lower profits than would be possible if it could identify viewers without having to rely on
self-selection. In competitive environments, this adverse selection problem will entail a softening of price com-
petition; price reductions exert a negative externality on the profitability of the advertising side of the market
as they attract consumers who are less valuable to advertising clients than inframarginal consumers. Hence,
the incentive to decrease prices is reduced and competition will take place in other dimensions. It will be shown
below that one such dimension is quality, leading to the overall result that media products will tend to have
higher prices and more quality than would be the case if firms could identify consumer types.3
Our results are well in line with empirical evidence on media markets. In particular, Thompson’s (1989) and
Kaiser’s (2003, 2005) analyses of newspaper and magazine markets show that media firms face a severe trade-
off between increased sales and deteriorating advertising rates when considering a cover price cut. Such a
trade-off is generated by adverse selection in our model. Lowering the price not only attracts clients from com-

1
For instance, the media data of Bonimail, the largest German provider (http://www.bonimail.de/) show that only around one third of
their customers stated that they were currently employed or self-employed.
2
Obviously, the webpages also have to handle a problem of moral hazard, as it is difficult to guarantee that consumers pay sufficient
attention to the advertisements. However, software technologies tackle this problem quite efficiently. For instance, websites that pay for
viewing banners only award credits if the user clicks a confirmation button in regular intervals in order to prevent that the ads are run
while she is absent from the computer.
3
If broadcasters were able to identify consumers’ identities, it would be possible to directly discriminate between different types of
consumers and to offer them customized products which would depend on their attractiveness as advertising targets. Given that this is not
possible, media markets will bring about a differentiated product space that guarantees self-selection of consumers towards the efficient
product.
 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119 109

petitors but also low income consumers who are unattractive as advertising targets. In order to prevent this
outcome, advertisers distort prices upwards.
Recently, media markets have been analyzed extensively in the literature on two-sided markets.4 Models in
this spirit are Anderson and Coate (2005), Crampes et al. (2005), Ferrando et al. (2004), Gabszewicz et al.
(2001), Gal-or and Dukes (2003), Haeckner and Nyberg (2000) and Nilssen and Sorgard (2001), among
others. These papers offer interesting insights into the nature of competition in media markets, but do not
address the question of adverse selection we are concerned with here.
The monopoly case in this paper is closely related to the literature on monopolistic price discrimination
with endogenous quality choice (most notably Mussa and Rosen, 1978; Srinagesh and Bradburd, 1989). As
should become clear below, however, the basic intuitions that were derived in this literature are often reversed
in the case of media markets. Monopolistic screening has been applied to broadcasting markets in Chae (1992)
and Crampes and Hollander (2005). In these papers, TV stations are fully subscription-based, however, so the
issue of advertising is not addressed.
The remainder of this paper is organized as follows. In Section 2, we present the formal model. Section 3
characterizes the equilibrium in a monopolistic TV market. Section 4 considers extensions of the basic model,
among others to the case of a competitive TV market. Finally, Section 5 concludes by discussing the results
and relating them to empirical work.

2. The model

Consider a market for a homogeneous media product with a monopoly media firm. To fix ideas, let the firm
be a TV broadcaster that can offer different channels. There are two types of viewers: high types (denoted by
subscript H) with high income or a large interest in advertised products and low types (denoted by subscript
L), who are less interesting for advertising firms. Since the broadcaster cannot directly observe individual
viewers’ types, there is asymmetric information between broadcaster and consumers. There is a continuum
of viewers of measure 1. A proportion c (known by the broadcaster) is of type H, while a proportion 1  c
is of type L. Consumers have unit demand and watch the television program that provides them with the high-
est utility (if above zero).
The broadcaster can offer channels which are characterized by the vector (p, a, q), where p P 0 denotes the
subscription fee, a P 0 denotes the amount of advertising and q 2 R denotes the quality of the content of the
channel. ‘‘Quality” refers to some unambiguously measurable characteristic of the station, e.g., the amount of
money it spends on movie rights.5 We assume that the consumers’ utility from watching a channel increases in
q and decreases in p and a. That is, there are nuisance costs of advertising.6
For simplicity, we take the utility function to be linear. The (indirect) utility of a consumer of type i from
the consumption of a given media product is thus
V i ðp; a; qÞ ¼ ui  bi a  p þ vi q: ð1Þ
The parameter ui > 0 is a constant measuring the utility of consuming a free, base-quality TV program con-
taining no advertising. bi > 0 measures type i’s nuisance costs of advertising. Let bH > bL, implying that rich
consumers suffer more from advertising than poor consumers. Above all, this is due to higher opportunity
costs of time.7 The parameter vi P 0 is an indicator of type i’s willingness to pay for quality. Let vH P vL. With

4
See Armstrong (2006) and Rochet and Tirole (2003, 2006) for some of the general insights from this literature. They also contain some
analyses of media markets.
5
This notion of quality is chosen for expositional purposes only. Section 4 contains a discussion of the case where product
characteristics can not be easily measured along a vertical dimension. We discuss there how the results of the paper can be reinterpreted
when the characterization of quality is more ambiguous.
6
It is generally accepted that TV advertising typically annoys viewers. Note, however, that print advertising may be desired by some
readers (see Rosse, 1967; Sonnac, 2000). The arguments in this paper apply only if consumers dislike advertising at least at the margin.
Otherwise, consumers have no incentive to misrepresent their type in order to be exposed to less commercials.
7
As Vi() has been normalized with respect to p, bi can also be interpreted as the marginal rate of substitution between watching
advertising and making a monetary payment to finance the station. Therefore, bH > bL can alternatively be interpreted as stating that low
type consumers suffer more from an increase in the price than high type consumers.
110 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119

this assumption we do not want to suggest that low types have a lower appreciation for quality, but rather that
their lack of wealth (or involvement) induces a lower willingness to pay for it. We normalize vL = 0, which has
no impact on the quality of our results but greatly simplifies calculations.
Situations in which our assumptions on demand parameters do not necessarily hold, are conceivable. For
instance, there may be cases where low income consumers have a higher willingness to pay for quality
improvements than high types. Therefore, Section 4 contains a discussion that presents the results of the model
if different assumptions are made on bH, bL, vH and vL. For reasons of space, however, the main body of the
paper will focus on the case described above which in our view is the most prominent one.
Let di denote the market price for one unit of advertising that is consumed by a consumer of type i. That is,
di is the price that brings demand and supply for type i advertising into equilibrium across media markets. It is
natural to have dH > dL. This assumption is supported by Fisher et al. (1980), who find that US broadcasting
stations with high income viewers receive much more favorable advertising rates than otherwise equal stations.
Consumers can remunerate broadcasters for obtaining the content they desire by paying a subscription fee
or by accepting advertising. Which form of obligation they prefer, of course, depends on their preferences.8
Note also that the knife-edge preference for advertising only or subscription fee only is a consequence of
the linearity of the utility function. From (1) we have that consumers of type i strictly prefer advertising-based
broadcasting to subscription-based broadcasting if di > bi. If di = bi, they are indifferent between the two and
if di < bi, they strictly prefer paying in money rather than consuming advertisements. As this paper is con-
cerned with the provision of advertising, it is natural to assume that advertising is the efficient form of financ-
ing the broadcaster, that is di > bi for i 2 {H, L}. We forgo the analysis of the other possible cases. Note,
however, that the adverse selection problem may be less severe in alternative settings, because if some consum-
ers prefer subscription-based television while others prefer advertising-based television, screening is
facilitated.9
The broadcaster has fixed production costs of F > 0 and marginal costs of c P 0. For simplicity, (fixed)
quality costs are assumed to be 12 q2 , where q is the highest quality the broadcaster offers in any channel.10
Another possible extension is to include channel-dependent fixed costs. The results in this case are also
equivalent, unless these costs are so high that the broadcaster wants to pool both types in a single channel,
in which case the adverse selection problem of course disappears. A market only exists, of course, if the con-
sumers’ willingness to pay for the media product is high enough in order to cover the costs of production,
which we assume throughout the paper.

3. A monopoly broadcaster

We will now analyze the market outcome. It turns out that the equilibrium in a monopoly market depends
on the agents’ maximum willingness to pay and to accept advertising. Here, we will restrict the analysis to the
case that seems most plausible to us: the one where high types have a higher willingness to pay for a basic TV
channel (uH P uL), but low types have a higher willingness to accept advertising to finance it (uL/bL P uH/bH).

3.1. Full information

As a benchmark, consider first the market equilibrium under full information. For the moment, we main-
tain the idealized assumption that the broadcaster can perfectly identify types. Therefore, we can look at the
H- and the L-market separately.

8
Note that broadcasters’ preferences concerning the form of payment are perfectly in line with consumers’ in a full information world:
letting consumers choose the form of payment maximizes the sum that can be extracted from them – directly via a subscription fee or
indirectly via advertising revenue.
9
In the real world, it must be that consumers desire some positive level of advertising. Otherwise, media firms could increase their profits
by having less advertising and increasing the media price such that consumers’ utility levels stay constant. As will become clear below, this
argument does not hold for subscription fees. Asymmetric information might guarantee their existence even in a world where every
consumer would prefer all types of media to be fully advertising-financed.
10
It is straightforward to extend the model to include marginal costs of quality. This case is analyzed in the working paper version of this
paper, where we show that the results are qualitatively equivalent.
 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119 111

In order to make sure that consumers purchase, the monopolist has to bear in mind the participation
constraints
u H  bH a H  p H þ v H q H P 0 ð2Þ
for high types and
u L  bL a L  p L P 0 ð3Þ
for low types, which follow directly from (1). Also, the non-negativity constraints
ai P 0 for i 2 fH; Lg ð4Þ
and
pi P 0 for i 2 fH; Lg ð5Þ
have to be met. The monopolist maximizes the following expression:
 
1
max cðaH dH þ pH  cÞ þ ð1  cÞðaL dL þ pL  cÞ  F þ ðmax fqL ; qH gÞ2
fðpH ;aH ;qH Þ;ðpL ;aL ;qL Þg 2 ðP1Þ
subject to ð2Þ–ð5Þ:
2
The solution to program (P1) is straightforward to find; pFI FI 2
H ¼ 0, aH ¼ uH =bH þ cdH vH =bH , and
qFI
H
FI FI FI
¼ cdH vH =bH for the high type and pL ¼ 0, aL ¼ uL =bL , and qL ¼ 0 for the low type. These results are
quite intuitive. As advertising is the efficient way of paying for the channels, prices are zero for both types
and the programs are fully financed by advertising. This way, the monopolist is able to extract most rent.
Low types receive zero quality, as they do not care about it. High types, on the other hand, receive a positive
amount of quality, which is increasing in their valuation of quality vH, in the proportion of high types c (more
high types mean that quality costs can be shared among more viewers), and in advertising receipts dH, but
decreasing in the nuisance parameter bH.
The monopolist provides both types with the efficient amount of quality, as this maximizes the amount of
rent the firm can extract from them. Which type has to accept more advertising depends on the quality pref-
erences of H-types, which determine the maximum amount of advertising they are willing to bear. If their
desire for quality is relatively small, high types have to endure less advertising, while they will be exposed
to more advertising if their preferred product is of very high quality. Both consumers groups are held to zero
utility. It is instructive to note that this solution may not be incentive compatible in the limited information
setting. The reason is that – as long as high types have to endure less advertising than low types – there is
always an incentive for low types to consume the high type product, which costs the same (zero) but has less
advertising and more quality.

3.2. Limited information

We now turn to the more realistic case of asymmetric information with respect to consumers’ types, where
perfect discrimination is not feasible. The monopolist has two possibilities. Either it can engage in (second
degree) price discrimination and thereby sell to the whole market. Or it can offer a single bundle only in order
to extract the whole rent from one type alone, while rationing the other one. Which of these strategies is opti-
mal typically depends on the relative frequency of types (Salant, 1989).
As is usual in the literature on price discrimination, we first consider the case of price discrimination with-
out worrying about the possibility of rationing for the moment. In this case the monopolist has to take the
following incentive constraints into account; these constraints guarantee that both types actually prefer the
bundle that is designed for them. The high type’s incentive constraint can be easily constructed from (1) as
bH ðaL  aH Þ þ pL  pH  vH ðqL  qH Þ P 0: ð6Þ
Similarly, the low type’s incentive constraint is
bL ðaH  aL Þ þ pH  pL P 0: ð7Þ
The broadcaster again maximizes profit given participation constraints of the two types, but now also takes
the incentive constraints (6) and (7) into account.
112 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119
 
1 2
max cðaH dH þ pH  cÞ þ ð1  cÞðaL dL þ pL  cÞ  F þ ðmax fqL ; qH gÞ
fðpH ;aH ;qH Þ;ðpL ;aL ;qL Þg 2 ðP2Þ
subject to ð2Þ–ð7Þ:
It turns out that the solution to this program crucially depends on the size of c. Let us first consider the case
where there are relatively many low types so that c 6 c. Proposition 1 formally states the result in this case.
Proposition 1. In the solution to program (P2) for the case c 6 c, the low type always receives his full information
bundle. The high type’s bundle is as follows.11

(i) If vH 2 [0, vH), then pLI

H ¼
bH uL bL uH
bH bL
 c bðbL ðdb
H bL Þ 2
v , aLI
Þ2 H
uH uL dH bL 2 LI dH bL
H ¼ bH bL þ c ðb b Þ2 vH and qH ¼ c bH bL vH .
H L H L
bH uL bL uH
(ii) If vH 2 ½vH ; vH Þ, then pLI LI LI
H ¼ 0, aH ¼ uL =bL and qH ¼ b L vH
.
(iii) If vH 2 ½vH ; 1Þ, then the high type receives his full information bundle.

It is interesting to note that the low types receive their full information bundle, while the high types’ bundle
is distorted. This happens because, as we have argued, the low types have to be deterred from opting for the
high type station. It is convenient to present the solution graphically as we do here in Fig. 1, which depicts
both the full and the limited information bundle for high types depending on the valuation vH. As can be seen,
the optimal price, advertising and quality largely depend on the size of the H-type’s preference for quality vH.
Fig. 1 shows that an upward distortion of both quality and price is used to achieve incentive compatibility. As a
price increase hurts low types relatively more, the distortion in pH can be used to scare away L-types from the H-
bundle. Quality is a useful screening instrument as high types have a higher willingness to pay for it than low types.
While increasing qH beyond its full information level is costly for the broadcaster as the rents that can be extracted
from high types decrease, this is still worthwhile, since low types are deterred. This deterrence occurs because the
increased level of high type quality allows the broadcaster to air more commercials in the high type channel.
If the difference in quality preferences of the two types is low (vH < v0H ), both price and quality distortions
are used by the monopolist. An increase in vH then has two effects. The first effect is that the increasingly dif-
ferent preferences for quality make it more profitable to use qH as a screening instrument. Hence, the high
type’s quality moves further away from its full information level. At the same time, the price distortion can
be used less, until pH reaches zero at vH ¼ v0H . Therefore, the first effect of a widening gap between quality pref-
erences is an exchange of screening instruments.
The second notable effect is that screening becomes easier overall, as the products the two types would
like to consume are getting more differentiated. This effect drives the use of both screening instruments
down. This process is completed if vH P v0H , where the full information outcome is attained. Thus, the more
differentiated the products are that high and low types would ideally like to consume, the easier it is for
media firms to efficiently sort types. If tastes are sufficiently heterogeneous, media firms can fully solve
the adverse selection problem that is inherent in the provision of advertising.12 Note that this property does
not depend on the assumption of vertical product differentiation; screening also becomes easier if high and
low types’ tastes are differentiated along other (horizontal) dimensions. If different types want to watch dif-
ferent programs, the monopolist will find it easier to extract surplus as viewers are less tempted to switch
channels.13
The quality distortion that potentially occurs in monopoly media markets is quite different to the quality dis-
tortion observed in regular goods markets with price discrimination (see Mussa and Rosen, 1978). While quality
is distorted downwards there, here we have an upwards distortion. The distortion is reversed because in media
markets, the low types may have to be deterred from consuming the high type bundle, while this is the other way
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
11
The threshold levels are defined by vH :¼ ðbH bcbL Þðb H uL bL uH Þ
, vH :¼ bbH bH uLcdb L uH dL ðbH bL Þ
, and c :¼ b ðdH b . All proofs are
L ðdH bL Þ L H L H ÞþdL ðbH bL Þ
contained in Appendix A.1.
12
This is akin to a result from insurance economics. Bond and Crocker (1991) show that differential consumption of hazardous goods
may allow insurance companies to reach the first best in insurance markets with adverse selection.
13
Fully efficient sorting of types, however, is an artefact of the two type model: while more differentiated valuations also relax incentive
constraints in a continuous type model, this will not generally imply that the full information outcome is attainable at some point.
 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119 113

Fig. 1. The monopoly case for c 6 c.

around in regular goods market. This outcome occurs even though the high types are usually the more profitable
clientele for the monopolist because they generate higher advertising receipts than do low types.
A second major difference to regular goods markets is the absence of information rents to consumers. While
in the canonical model of second degree price discrimination, the type whose incentive constraint is binding
receives an information rent, here the monopolist is able to extract all surplus, holding viewers down to their
reservation utility as in Crampes and Hollander (2005).
The level of advertising, unlike price and quality, is not a screening instrument but adapts to these distor-
tions in such a way that consumers are just willing to buy the product. The amount of advertising for high
types, aH, is below the full information level for small values of vH. There, the broadcaster must show less
advertising because it demands a positive price. As vH increases, two forces drive aH up: the reduction in
the subscription fee makes it possible to increase aH and in addition to that, the rapid increase in qH allows
more advertising. As a result, aH increases above its full information level beyond some point. Finally, aLI
H and
aFI
H approach each other again as all distortions go down.
Next we turn to the case where there are many high types (c > c). Proposition 2 states the equilibrium in
this case.
Proposition 2. In the solution to program (P2) for the case c > c, the low type always receives his full information
cdH þð1cÞdL 2
bundle, except that aLI uH
L ¼ b þ b2
vH if vH 2 ½0; v0H Þ. The high type’s bundle is as follows.14
H H

uH cdH þð1cÞdL 2 cdH þð1cÞdL

(i) If vH 2 ½0; v0H Þ, then pLI LI
H ¼ 0, aH ¼ bH þ b2
vH and qLI
H ¼ bH
vH .
H
bH uL bL uH
(ii) IfvH 2 ½v0H ; vH Þ, then pLI LI LI
H ¼ 0, aH ¼ uL =bL and qH ¼ bL vH
.
(iii) If vH 2 ½vH ; 1Þ, then the high type receives his full information bundle.

Fig. 2 represents this solution graphically. The figure reflects a sudden change in the monopolist’s
screening policy if c gets larger than c. While the screening of the high type quality remains unchanged,
the monopolist now refrains from using the price instrument. This allows it to increase the advertising
of H-types. To guarantee incentive compatibility it now becomes necessary to decrease aL below full
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
14 bH bH uL bL uH
v0H is defined as v0H :¼ bL ½cdH þð1cÞdL .
114 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119

Fig. 2. The monopoly case for c > c.

information levels. This policy is intuitive. The more high types there are, the more costly it becomes for
the monopolist to distort the high type bundle in order to achieve incentive compatibility. Beyond c, there
are so few low types that it is profitable to distort the low type bundle instead by lowering its advertising
below its full information level. As the low type product otherwise remains unchanged, this policy generates
an information rent for L-types.
Fig. 2 shows that the two effects of an increase in vH that we observed before are at work here, too. First of all,
the quality instrument becomes more effective, so the advertising distortion is exchanged for the quality distor-
tion until aL reaches its full information level. Second, screening becomes easier, which decreases the use of both
instruments. Again, for vH P v0H the two types demand such different goods that self-selection is without costs.
It turns out that if c increases beyond c, the remaining quality distortion of high types gets still smaller and
is gradually exchanged for a further distortion of the low types’ advertising. Therefore, we can conclude that a
media firm will alter its screening policy if the proportion of high types increases beyond some threshold.
As was mentioned above, the solution to program (P2) only determines the optimal behavior of the
monopolist if it is not preferable to ration one type of customer while extracting the whole rent from the other.
Proposition 3 presents the surprising result that we do not have to worry about this issue here.
Proposition 3. Independent of the size of c, there will be no rationing of consumers. That is, the monopolist always
makes his pricing decision according to Propositions 1 and 2.
This is an important observation. Quite contrary to standard models of monopolistic price discrimination,
rationing in media markets is never profitable for a monopolist – irrespective of the relative frequencies of
types. This means that the efficiency loss from monopolization may be smaller in media markets than in stan-
dard goods markets.15 This stems from the fact that the monopolist can demand different forms of payment
from different types, which gives it greater leeway in profitably separating types.

15
Of course, monopolization in media markets may bring about other inefficiencies than purely economic ones. For instance, it may lead
to political biases and disinformation. This alone may be enough in order to worry about media monopolies. On the other hand,
competition may also lead to biased news by making it necessary to exaggerate stories in order to attract readers or viewers. See
Mullainathan and Shleifer (2005) for a model that analyzes both types of media bias.
 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119 115

4. Extensions

4.1. Competition

One natural extension of the basic model is to introduce competition. The working paper version of this
paper contains a full characterization of this case.16 What is shown there is that the equilibrium distortions
closely resemble the monopoly case. While the full information outcome entails fully advertising-financed
TV stations, this is no stable equilibrium with limited information as long as high type channels show less
advertising than do low type channels. As programs contain just enough advertising to recover their produc-
tion costs, broadcasters would make losses when low types consume the high type product since the advertis-
ing revenue they generate is too low. In order to prevent this from happening, broadcasters again use an
upwards distortion of price and quality to separate types.
One significant difference is that this will be the broadcasters’ strategy independent of the proportion of
high types. Distorting the low type bundle in order to achieve incentive compatibility is not feasible in the
competitive scheme. This is because a reduction of low type advertising below full information levels would
entail losses with this channel. With competition, cross-subsidization through profits from a high type channel
is not possible as perfect competition precludes strictly positive profits with any type of channel. Therefore,
even if there are almost only high types in the market, the high type channels will have more quality and a
higher subscription fee than is efficient.
The main intuition of the competitive case therefore is that despite fierce competition prices will not be dri-
ven down for high type channels. Instead, competitive pressure reduces the amount of advertising shown and
increases the quality level of these programs. A particularly striking real world example of this effect in the
newspaper industry is described by Mander (1978, p. 75), Deputy Chief Executive of the Times in the late
1960s:
‘‘From 1967 to 1969 the Times . . . sale(s) shot up from 270,000 to 450,000 – a remarkable achievement.
But its higher sales made it no more attractive as an advertisement medium . . . adding to the readership
just watered down the essential target group and increased the cost of reaching it. A reversal of policy
changed the situation with a consequent dramatic improvement in profitability. The circulation is back
down to 300,000.”

4.2. Incomplete market coverage

One assumption that we have made is that consumers’ basic valuations, uL and uH, are sufficiently high
compared to production costs, so that trade of both high and low type media products is efficient. Alterna-
tively, consider the situation in which the H-type’s valuation is high enough to make production worthwhile,
while production of the low type product is inefficient (uL < c). It is tempting to suspect that in such a case with
no low type media products, the adverse selection problem vanishes and we return to the full information
ideal. This, however, turns out to be wrong. Even though low types are not prepared to consume a potential
low type bundle, they may still have to be deterred from buying the high type bundle. This is because the high
type bundle is very attractive due to its relatively low amount of advertising. It is straightforward to show in
our basic model that whenever uL decreases below some threshold level, production for low types becomes
inefficient. As uL decreases from this point on, the necessary distortion of the high types becomes smaller
and smaller, because buying the high type bundle becomes less and less attractive to L-types. But until uL
has decreased below some second threshold level, the H-type bundle remains distorted.
The case where there are no low-type bundles may be the relevant one in many newspaper markets. Here,
the distinguishing characteristic of readers is probably the average reading time more than income. It is well
known that usually only a small percentage of newspaper subscribers actually read their paper on a given day.

16
As all TV programs of a given quality are homogenous in the basic model, Bertrand competition would ensue, implying that no more
than one firm would enter if there are fixed costs. Therefore, we consider a cost structure with positive marginal costs and zero fixed costs
in the competitive case, which makes entry of competitors feasible.
116 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119

Still, if the price is low enough, most people want to subscribe to a newspaper, even if they do not read it reg-
ularly. Naturally, advertisers pay lower rates for ‘‘readers” who do not actually see their advertisements. It is
quite conceivable that in a full information world, it would be efficient to finance newspapers entirely by adver-
tising. But given that free papers would attract subscriptions of readers who merely use the paper for checking
the lottery numbers and the cinema program, newspapers are forced to distort their pricing policy by charging
a cover price that discourages low interest readers.

4.3. Parameter variations

Generally, it can be said that different assumptions on the main parameters of the model will yield different
results. For instance, if we have bL P bH, implying that low types find a price increase less annoying than do
high types, then only the quality instrument can be used to screen consumers – not the price instrument. This is
because the only price change that relaxes the low types’ incentive constraint is a price reduction of the high
type bundle, which is not possible when we already have pH = 0. The outcome with bL P bH is therefore that
both products are fully advertising financed and that the high types’ bundle has a higher quality than in the
full information case (unless quality valuations are sufficiently different, in which case the full information out-
come can again be attained).
If vL > vH, that is if low types have a higher willingness to pay for quality than high types, the equilibrium
distortion in quality will be opposite to the case analyzed in the paper; under limited information, the high
type bundle will feature less quality than in the full information case (and of course also less quality than
the low type bundle).
Finally, if uL/bL < uH/bH, so that low types have a lower ‘‘willingness to pay” in terms of advertising than
high types, a quality distortion of the low type is used. In this situation the low type bundle features less qual-
ity than under full information. Furthermore, the price instrument is not used anymore. Therefore, this case is
equivalent to the standard price discrimination model with the only modification that the product is not
financed by a subscription fee but by advertising. This is because in such a situation, high types have to be
deterred from consuming the low type bundle as in the usual framework of second degree price discrimination.

4.4. Product heterogeneity

The main model has considered a TV market in which channels were only differentiated along a vertical
quality dimension, where more quality was costly to produce for the broadcaster. However, consumer tastes
are often differentiated along other (horizontal) dimensions. What one person perceives as ‘‘good quality” may
be considered as poor by someone else. Also, good quality is not necessarily more costly than bad quality, as a
cursory look at any given night’s TV program reveals.
It is important to note, however, that product characteristics are a useful screening instrument for the
monopolist even when product differentiation is horizontal and the costs of providing a TV service do not
change with the nature of the content. Consider a situation where TV stations’ characteristics are defined hor-
izontally, for instance the choice of whether they show action movies or documentaries (with identical costs).
If low types have a larger preference for documentaries, the monopolist will again use the ‘‘quality” instrument
to attain separation: by putting more emphasis on action content in the high type bundle than high types
themselves would prefer, low types can be successfully deterred as they suffer a larger marginal disutility from
the distorted content.
In summary, the vertical differentiation described in the main model is best regarded as an example of a
content distortion. Other dimensions of product heterogeneity are clearly conceivable. When they play a role,
screening will take a different form. What remains unchanged, however, is that the high type bundle is dis-
torted away from the full information outcome in the direction that low types dislike.

5. Conclusion

We have argued that direct markets for advertising, in which consumers are compensated for being exposed
to advertisements, are prone to market failure. Due to asymmetric information, consumers with preferable
 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119 117

traits for advertisers cannot be distinguished from consumers who are less valuable advertising targets. Hence,
payments cannot be made contingent on consumer characteristics, which is likely to be necessary to convince
high types to consume advertising.
The provision of advertising through media platforms like TV broadcasting was shown to mitigate this
problem by the broadcasters offering different channels in order to show targeted advertising. While there
may still exist a selection problem (low types will prefer the high type channel if it contains less advertising),
sorting can be achieved by distorting the high type program. This results in excessive prices and quality levels
of channels targeting high income consumers. Even if several broadcasters compete, price competition will be
soft as lower subscription fees would induce low types to switch to high type channels, deteriorating the adver-
tising rate that can be demanded from advertisers.
It is interesting to note that the classic papers in media economics already hint at problems of adverse selec-
tion in media markets. Reddaway (1963, p. 214) observes the dangers of using cover price cuts as a competitive
instrument in the newspaper industry:
‘‘Even if the price-cut succeeded, however, there is a real risk that the new readers would be concentrated
in low-income groups, and so lower A’s status as an advertising medium.”
In a similar vein, Corden (1953, p. 186) explains that advertising revenue increases less than proportionally
with increases in circulation (caused, say, by a cover price cut):
‘‘[. . .] as circulation increases the average income of readers usually falls; hence the quality of the adver-
tising space to advertisers is decreased, and to some extent the increase in quality resulting from the rise
in circulation itself is offset.”
A testable implication of our theory along the lines of these quotes is that audience expansions through
price cuts will result in lower CPM advertising rates. Indeed, Thompson (1989) finds that price setting in
the British newspaper industry involves a trade-off between circulation and advertising revenue. If producers
decide to reduce the cover price of their newspaper, this increases circulation, but substantially deteriorates
advertising rates.17
Similarly, Kaiser (2003, 2005) empirical results for women’s magazines show that increasing circulation
by one percent increases total advertising revenues by far less than one percent. The result implies that price
cuts dramatically worsen advertising rates per reader. The effect is particularly strong for magazines that are
aimed at high income women (as proposed by this paper); a one percent increase in circulation results in a
less than 0.25% increase in advertising revenue for high priced magazines. That is, an inframarginal reader
generates an advertising rate that is more than four times higher than the corresponding rate for a marginal
reader.

Acknowledgements

We would particularly like to thank Simon Anderson, Klaus Schmidt, the editor David Waterman, and two
anonymous referees for very helpful comments. Furthermore, we are grateful to Tobias Böhm, Bruno Jullien,
Marco Ottaviani, Martin Peitz, Ray Rees, and seminar participants at Hamburg (3rd Workshop on Media
Economics 2005), Madrid (Annual Meeting of the European Economic Association 2004), Munich, Stras-
bourg, Vienna (Annual Meeting of the Austrian Economic Association 2004), and Warsaw (Spring Meeting
of Young Economists 2004) for helpful discussions. The second author gratefully acknowledges financial sup-
port from the German Research Foundation.

17
In principle ad rate reductions following an increase in circulation could also be due to other factors than adverse selection (e.g.,
because economies of scale allow lower rates). However, Thompson’s simultaneous equation framework shows that new readers indeed
have lower incomes on average; this in turn is demonstrated to deteriorate advertising rates.
118 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119

Appendix A

A.1. Proofs

Proof of Proposition 1

(i) First conjecture that (2), (3) (5) for low types, and (7) are binding, whereas (4) and (5) for high types, and
(6) are slack. This gives us the Lagrangian:
1
L ¼ cðaH dH þ pH  cÞ þ ð1  cÞðaL dL þ pL  cÞ  F  q2H þ l2 ½uH  bH aH  pH þ vH qH 
2
þ l3 ½uL  bL aL  pL  þ l5l pL þ l7 ½bL ðaH  aL Þ þ pH  pL :
Using c 6 c, vH < vH, and the assumptions from Section 2, we find that all multipliers are larger than
zero and we get the equilibrium values proposed in the proposition for the case in which vH 2 [0, vH].
Substituting the solutions back in the constraints and rearranging shows that none of the constraints
are violated.
(ii) Next conjecture that (2), (3), (5) and (7) are binding, whereas (4) and (6) are slack. The Lagrangian then
becomes
1
L ¼ cðaH dH þ pH  cÞ þ ð1  cÞðaL dL þ pL  cÞ  F  q2H þ l2 ½uH  bH aH  pH þ vH qH 
2
þ l3 ½uL  bL aL  pL  þ l5h pH þ l5l pL þ l7 ½bL ðaH  aL Þ þ pH  pL :
Using c 6 c, vH 6 vH < vH , and the assumptions from Section 2, we again find the multipliers to be lar-
ger than zero and get the equilibrium values proposed in the proposition for the case vH 2 ½vH ; vH . As
before, substitution of these values into the constraints and rearranging confirms that all constraints
are fulfilled by the solution.
(iii) Finally, using c 6 c, vH P vH , and the assumptions from Section 2, simple substitution of the full infor-
mation values into the constraints and rearranging confirms that the full information outcome can be
achieved for vH P vH . h

Proof of Proposition 2

(i) First conjecture that (2), (5) and (7) are binding, whereas (3), (4) and (6) are slack. This gives us the
Lagrangian:
1
L ¼ cðaH dH þ pH  cÞ þ ð1  cÞðaL dL þ pL  cÞ  F  q2H þ l2 ½uH  bH aH  pH þ vH qH 
2
þ l5h pH þ l5l pL þ l7 ½bL ðaH  aL Þ þ pH  pL :
Using c > c, vH < v0H , and the assumptions from Section 2 we find that all multipliers are larger than zero
and we get the equilibrium values proposed in the proposition for the case where vH 2 ½0; v0H . Substitut-
ing the solutions back into the constraints and rearranging shows that none of the constraints are vio-
lated. (ii) and (iii) See the proof for parts (ii) and (iii) of Proposition 3. h

Proof of Proposition 3. First note that in the case vH P vH we are in full information. Hence, rationing can
only be profitable if vH < v0H . Let us assume this to be the case. Denote by M FI LI
i (M i ) the full information (lim-
ited information) contract offered to type i according to program (P2); and denote a no-trade situation by N.
As for low types M FI LI
L ¼ M L , rationing high types can never improve the monopolist’s profit. The most prof-
itable strategy that possibly rations low types is offering M FI H only.
As the low type’s incentive constraint is binding if vH < vH , we must have M FI FI
H L M L . Furthermore,
M FI LI FI LI
L L M L as M L ¼ M L . Finally, since the low type’s participation constraint is fulfilled, we must have
 S. Kremhelmer, H. Zenger / Information Economics and Policy 20 (2008) 107–119 119

M LI FI
L L N . Hence, by transitivity of L’s preferences, M H L N . This means that the most profitable candidate
for rationing actually does not involve any rationing if offered alone. Therefore, there exists no rationing
contract that generates a higher profit for the monopolist than the solution to program (P2). h

References

Akerlof, G., 1970. The market for lemons: quality uncertainty and the market mechanism. Quarterly Journal of Economics 89, 488–500.
Anderson, S.P., Coate, S., 2005. Market provision of broadcasting: a welfare analysis. Review of Economic Studies 72, 947–972.
Armstrong, M., 2006. Competition in two-sided markets. RAND Journal of Economics 37, 668–691.
Bond, E.W., Crocker, K.J., 1991. Smoking, skydiving, and knitting: the endogenous categorization of risks in insurance markets with
asymmetric information. Journal of Political Economy 99, 177–200.
Chae, S., 1992. Bundling subscription TV channels. International Journal of Industrial Organization 10, 213–230.
Corden, W.M., 1953. The maximization of profit by a newspaper. Review of Economic Studies 20, 181–190.
Crampes, C., Haritchabalet, C., Jullien, B., 2005. Advertising, Competition and Entry in Media Industries. Universit é de Toulouse I,
Mimeo.
Crampes, C., Hollander, A., 2005. Product specification, multi-product screening and bundling: the case of pay TV. Information
Economics and Policy 17, 35–59.
Ferrando, J., Gabszewicz, J.J., Laussel, D., N. Sonnac, 2004. Two-Sided Network Effects and Competition: An Application to Media
Industries. CORE, Mimeo.
Fisher, F.M., McGowan, J.J., Evans, D.S., 1980. The audience–revenue relationship for local television stations. Bell Journal of
Economics 11, 694–708.
Gabszewicz, J.J., Laussel, D., Sonnac, N. 2001. TV-broadcasting competition and advertising, CORE Discussion Paper 2000/6.
Gal-or, E., Dukes, A., 2003. Minimum differentiation in commercial media markets. Journal of Economics & Management Strategy 12,
291–325.
Haeckner, J., Nyberg, S., 2000. Price Competition, Advertising and Media Market Concentration. University of Stockholm, Mimeo.
Kaiser, U., 2003. The effects of website provision on the demand for German’s women’s magazines. ZEW Discussion Paper 02-54.
Kaiser, U., 2005. Price Setting Behavior and Returns to Scale in the German Magazine Industry. University of Southern Denmark at
Odense, Mimeo.
Mander, M., 1978. The integration of advertising and circulation sales policies. In: Henry, Harry (Ed.), Behind the Headlines: The
Business of the British Press–Readings in the Economics of the Press. Associated Business Press, London.
Mullainathan, S., Shleifer, A., 2005. The market for news. American Economic Review 95, 1031–1053.
Mussa, M., Rosen, S., 1978. Monopoly and product quality. Journal of Economic Theory 18, 301–317.
Nilssen, T., Sorgard, L., 2001. The TV Industry: Advertising and Programming. University of Oslo, Mimeo.
Reddaway, W.B., 1963. The economics of newspapers. Economic Journal 73, 201–218.
Rochet, J.-C., Tirole, J., 2003. Platform competition in two-sided markets. Journal of the European Economic Association 1, 990–1029.
Rochet, J.-C., Tirole, J., 2006. Two-sided markets: a progress report. RAND Journal of Economics 37, 645–667.
Rosse, J.N., 1967. Daily newspapers, monopolistic competition, and economies of scale. American Economic Review 57, 522–533.
Salant, S., 1989. When is inducing self-selection suboptimal for a monopolist? Quarterly Journal of Economics 104, 391–397.
Sonnac, N., 2000. Readers’ attitudes toward press advertising: are they ad-lovers or ad-averse? Journal of Media Economics 13, 249–259.
Srinagesh, P., Bradburd, R.M., 1989. Quality distortion by a discriminating monopolist. American Economic Review 79, 96–105.
Thompson, R.S., 1989. Circulation versus advertiser appeal in the newspaper industry: an empirical investigation. Journal of Industrial
Economics 37, 259–271.