Competition for Viewers and Advertisers in a TV Oligopoly

Abstract

This study considers a model of a TV oligopoly where TV channels transmit advertising and viewers dislike such commercials. It is shown that advertisers make a lower profit the larger the number of TV channels. If TV channels are sufficiently close substitutes, there will be underprovision of advertising relative to social optimum. This study also finds that the more viewers dislike ads, the more likely it is that welfare is increasing in the number of advertising-financed TV channels. A publicly owned TV channel can partly correct market distortions, in some cases, by having a larger amount of advertising than private TV channels. It may even have advertising in cases where advertising is wasteful per se.

ACKNOWLEDGMENTS

We are grateful to Steve Wildman, three anonymous referees, and seminar participants in Antwerp, Helsinki, and Milan for helpful comments. We thank the Research Council of Norway (the KIM program) for its financial support through Samfunns- og næ ringslivsforskning AS (SNF)–Institute for Research in Economics and Business Administration. Hans Jarle Kind thanks CESifo in Munich for excellent working conditions while revising this work.

Notes

In 1995, the average adult male American spent 17.3 hr watching TV each week (Robinson & Godbey, 1999). In 2002, TV advertising in the United States amounted to approximately $50 billion, out of a total of approximately $115 billion spent on advertising (see http://www.adage.com/images/random/lna03.pdf).

It is documented that viewers try to avoid advertising breaks (e.g., see Danaher, 1995; Moriarty & Everett, 1994). See also Wilbur (2006), who estimated a model of TV competition and found viewers' disutility from advertising to be significant and positive.

To be precise, advertisers impose negative externalities on viewers, whereas viewers impose positive externalities on advertisers. For general introductions to the theory of two-sided markets, see Armstrong (2006) and Rochet and Tirole (2006).

For a survey of the economics literature on advertising, see Bagwell (2007).

See also the work by Anderson and Coate (2005), who did a welfare analysis in such a Hotelling-style setting. Other contributions in the recent literature on the economic analysis of media industries include Armstrong (2006); Crampes, Haritchabalet, and Jullien (2005); Cunningham and Alexander (2004); Dukes (2004); and Nilssen and S⊘rgard (2003). The seminal work is Steiner (1952). For a review of the early literature, see Owen and Wildman (1992). The more recent literature is reviewed by Anderson and Gabszewicz (2006).

In Hotelling models it is typically assumed that each consumer watches one and only one channel. From the advertisers' point of view, each channel will thereby have monopoly power over its viewers. In our framework, on the other hand, advertisers can reach an individual viewer through several channels (although not at exactly the same time). This seems to be consistent with observations that a given TV viewer is likely to see an ad for Coca-Cola, for instance, at several TV channels over the course of an evening or a week.

See Motta and Polo (1997) for a survey of the media industry in Europe. See Armstrong (2005) and Armstrong and Weeds (2007) for some recent discussions on public service broadcasting.

Note that this is in contrast to the standard quadratic utility function, where one and the same parameter measures both product differentiation and market size. See Motta (2004) for details.

An exception is Gal-Or and Dukes (2003), who model pair-wise negotiations between TV channels and advertisers.

When transmitting newscasts or sport events, the TV channels are quite flexible with respect to how much advertising to show. Moreover, to accommodate a low amount of advertising a TV channel can fill in with advertising for its own programs (“tune-ins”). For details concerning tune-ins, see Shachar and Anand (1998).

Barros, Kind, Nilssen, and S⊘rgard (2004) formulated a model where media firms set prices of advertising slots rather than quantities. The equilibrium outcomes they find are analogous to the ones we report here. For a more detailed discussion of price versus quantity competition in the market for TV advertising, see Nilssen and S⊘rgard (2003).

Equation 7 shows that A ^M _i → 0 as s → 1. In the profit function (Equation 3) we have abstracted from, for example, TV channels' programming costs. It should be noted, however, that the TV channels will clearly not be operative in the long run unless they raise enough advertising revenue to cover costs.

See also Anderson and Coate (2005).

When viewers dislike ads, advertising is like an indirect price of watching TV. Collusion enables TV channels to increase this indirect price and thus have more ad time. Our analysis relates to the informal discussion in Wildman (1998), who pointed out that introducing viewer disutility of advertising provides a tendency for advertising to be higher when TV channels collude than when they compete. The question is also discussed in Masson, Mudambi, and Reynolds (1990), but their analysis disregards any externalities between viewers and advertisers.

As proposed by a referee, we might also interpret γ as measuring the viewers' ad aversion net of the expected utility of their product-market transactions.

Recall that we have abstracted from fixed and variable costs for the TV channels (and also for the advertisers, except for the price they have to pay for advertising slots).

It is easiest to use a numerical example to show that could be positive or negative, depending on parameter values. To this end, note that has the same sign as the term inside square brackets in the numerator of Equation 21. Suppose that n = 1, m = 2, and s = ½. The term inside square brackets is then equal to minus one for γ = ½ and equal to plus one for γ = 1. We then have dW ^M /dm < 0 in the former case and > 0 in the latter.

The European Union restricts TV advertising to 9 min on average, with a maximum of 12 min in any given hour, although some of the member states have stricter limits. See details in Anderson (2007) and Motta and Polo (1997). In the United States, the National Association of Broadcasters at one time set an upper limit. In 1981, this was found to violate antitrust laws (see Owen & Wildman, 1992, chap. 5; and Hull, 1990). No restrictions (except for advertising on children's programs) exist in the United States today.

One example of this comes from Norway, where there are restrictions on allowed advertising levels. Although these restrictions have become less severe over time, the private station TV3, owned by Modern Times Group AB, has chosen to broadcast from the United Kingdom to the Norwegian market to avoid the Norwegian restrictions on ad time. Thus, it is not merely an empty threat when TV stations argue that they will serve the market from abroad if there is a strict regulation of advertising levels.

There are several studies of mixed oligopoly (e.g., see Cremer, Marchand, & Thisse, 1991; De Fraja & Delbono, 1989). Nilssen and S⊘rgard (2002) presented, as far as we know, the only mixed-oligopoly study relating to the media industry. However, their study is a Hotelling model, not capturing consumers' disutility from advertising. In a related study, Nilssen (2000) discussed mixed oligopoly in a payments market where, not unlike the present media context, there are negative externalities among firms in addition to the traditional oligopoly externality.

The only instance where not all TV channels have positive advertising levels for s < 1 is when the expression in Equation 23 is negative so that the public channel's equilibrium level is zero (so that it must be completely financed by, e.g., public funding). This happens when γ > , defined later.

Note that there is no reason to set A ₁ > 0 if γ > 1 and s = 0. The reason for this is that the TV channels' programs now are completely independent, so that the ad time on TV1 does not have any indirect effect on the profit levels of the other channels.

The curve for A ^P ₂ when A ^P ₁ = 0 (γ > ) is, in the present two-channel case, given by A ^P ₂ = . This can be found by using Equations 3 and 6.