The main purpose of this chapter is to examine the demand side of the radio broad-casting industry in the Netherlands with a special interest in the degree of market expansion or business stealing. Several specifications of the nested logit have been estimated and the results indicate that a two-level nested logit model in which radio stations are grouped on the basis of their format is best suited for representing listen-ers’ choices. So, listeners first decide to which type of music to listen to or not to listen to radio and conditional on listening to a specific format they decide to which radio station to listen to.
Overall, the estimations show that σ is quite high for the different specifications used in this chapter. For the preferred model σ is close to 0.8 which implies that the ra-dio broadcasting market in the Netherlands can be characterized by business stealing within formats. So, an increase in the number of radio stations within a format does not have an impact on the decision whether or not to listen to other formats or to lis-ten to radio. Therefore the “pie of lislis-teners” hardly increases. However, if an entering radio station is able to differentiate itself from other radio stations in a such a way that it also creates a new nest then the “pie of listeners” will increase. The entrant will then attract listeners from the other formats as well as from the outside option.
From a policy-making perspective, the main conclusion is that more is not always preferred to less. The (potential) increase in the number of radio stations is mostly seen as a great advantage of digital radio by offering listeners more choice. However, the utility of listeners will only increase if the increased number of radio stations also results in increased variety. Otherwise the additional costs of the new radio stations will likely not outweigh the additional benefits of listeners.
Previous empirical evidence on radio broadcasting industries has shown that there is excessive entry when the market can be characterized by business stealing. The business-stealing effect is driven by the fact that listeners perceive radio stations to be close substitutes. Moreover, the experience in the UK shows that more radio stations does not lead to more listening in terms of time spent on listening to radio. Unless radio stations are able to differentiate themselves which might be facilitated by the other features that digital radio offers, there will be fiercer competition for listeners among radio stations and this is not expected to benefit any side of the market.
2.A Description of variables and data sources
Table 2.A1 Data description and data sources
Name Description Source
Advertising price Price for an advertising spot. www.radionieuws.nl
of 20 seconds Individual websites of radio stations
Contacting radio stations Amount of Number of advertisements based on www.radionieuws.nl advertising an average length of 20 seconds. Nielsen Media Research
Listen shares: AQH rating and AQH share. Intomart GfK
sj, s0, sj|g
North, East, Dummy variables equal to 1 if the South, West if the radio station is located in
the North, East, South, or West.
National, Public, Dummy variables equal to 1 if the Dutch Radiocommunications Agency No FM, Inside market radio station is a national, public,
or no-FM station, or operating inside the market.
Pop, Hits, News, Dummy variables equal to 1 if the www.rab.fm Rock, and Classical radio station has a pop, hits, news,
rock or classical format.
Number of Total number of stations or Intomart GfK
competing stations number of stations within a format.
Population 10+ Total population in each market Intomart GfK aged above 10 years.
Economic climate Variable measuring the economic Statistics Netherlands situation. This index is based on
the economic situation over the past 12 months and the future 12 months on a scale of -100 to +100.
2.B Derivation of three-level nested logit
In a three-level nesting structure an individual first decides to listen to radio or not, then he or she decides to which format to listen to, and finally he or she chooses a radio station. The utility of individual i from radio station j is given by:
Uij=δj+ϕij
where ϕ is given by:
ϕij=ζiL+ (1−σL)ζiF+ (1−σL)(1−σF)εij.
The terms ζiL, ζiF, and εij denote individual i’s idiosyncratic taste for listening to radio(L), listening to format F, and listening to radio station j. The parameters σLand σF are between zero and one. If σFapproaches one, radio stations within a format are considered to be closer substitutes. If σLapproaches one, formats are considered to be closer substitutes given that an individual listens to radio.
The market share for radio station j is the product of the conditional probability of choosing radio station j given that an individual listens to format F(sj|F), the con-ditional probability of choosing format F given that an individual listens to radio L (sF|L), and the marginal probability of listening to radio(sL). These market shares are given by:
The overall market share for radio station j is the product of these three market shares, that is, sj=sL·sF|L·sj|Fand can be written as:
sj= e
(δj/(1−γ))
DF
· D
1−σF F
DL
· D
1−σL L
D
=e(δj/(1−γ))·D1σFF·D−σL L·s0
Taking logs and rewriting gives the listeners’ demand equation:
ln(sj) −ln(s0) = δj
1−γ−σFln(DF) −σLln(DL)
= δj
1−γ−σFln
"
e(δj/(1−γ)) sj|F
#
−σLln
[e(δj/(1−γ))]1−σF s1−σj|F FsF|L
=σFln(sj|F) −σL(1−σF) δj
1−γ +σL(1−σF)ln(sj|F) +σLln(sF|L)
= [1−σF−σL+σFσL] δj
1−γ+ [σF+σL−σFσL]ln(sj|F) +σLln(sF|L)
=δj+γln sj|F+σLln(sF|L)
2.C First-stage regressions
Table2.C1First-stageresultsforlisteners’demand (a)(b)(c) ln(sjmt|g)Ajtln(sjmt|g)Ajtln(sjmt|g)ln(sgmt|L)Ajt Economicclimate−0.0010.008∗∗∗−0.0000.008∗∗∗−0.000−0.0001∗∗0.008∗∗∗ (0.001)(0.002)(0.001)(0.002)(0.001)(0.000)(0.002) Population10+−0.078−0.002−0.064−0.033−0.0860.0110.007 (0.131)(0.246)(0.127)(0.241)(0.130)(0.028)(0.246) Numberofstations−0.014−0.069∗0.021−0.014∗∗∗−0.037 (0.021)(0.040)(0.023)(0.005)(0.046) Numberofwithin-formatstations−0.145∗∗∗−0.163∗∗−0.155∗∗∗0.061∗∗∗−0.144∗ (0.039)(0.074)(0.043)(0.008)(0.083) Constant−1.6421.5151.3260.8101.2161.450∗∗∗1.004 (1.849)(3.480)(1.833)(3.472)(1.836)(0.368)(3.478) North−0.2100.190−0.1070.147−0.135−0.0730.198 (0.209)(0.396)(0.207)(0.393)(0.208)(0.051)(0.396) East−0.1350.112−0.0640.065−0.010−0.0370.128 (0.170)(0.324)(0.166)(0.318)(0.169)(0.029)(0.324) South−0.166−0.058−0.057−0.036−0.065−0.079∗∗−0.022 (0.173)(0.303)(0.169)(0.300)(0.170)(0.037)(0.301) National0.2573.400∗∗∗0.2723.358∗∗∗0.269−0.0353.362∗∗∗ (0.238)(0.378)(0.232)(0.377)(0.231)(0.054)(0.377) Public0.193−1.981∗∗∗0.188−1.995∗∗∗0.187−0.003−1.994∗∗∗ (0.194)(0.344)(0.192)(0.345)(0.192)(0.028)(0.345) NoFM−0.393∗−2.894∗∗∗−0.379∗−2.890∗∗∗−0.379∗−0.010−2.889∗∗∗ (0.205)(0.297)(0.202)(0.296)(0.202)(0.036)(0.296) In-market3.021∗∗∗0.1852.962∗∗∗0.1712.966∗∗∗0.0500.163 (0.238)(0.296)(0.230)(0.296)(0.230)(0.062)(0.297) Pop0.835∗∗∗2.426∗∗∗−0.671∗∗∗2.883∗∗∗−0.642∗∗∗1.743∗∗∗2.832∗∗∗ (0.222)(0.384)(0.228)(0.435)(0.231)(0.052)(0.446) News1.870∗∗∗2.922∗∗∗0.3423.293∗∗∗0.3671.722∗∗∗3.249∗∗∗ (0.213)(0.381)(0.232)(0.422)(0.237)(0.075)(0.432) Hits1.219∗∗∗2.374∗∗∗−0.405∗3.088∗∗∗−0.3581.989∗∗∗3.005∗∗∗ (0.194)(0.337)(0.238)(0.469)(0.249)(0.059)(0.500) Rock0.685∗∗∗3.984∗∗∗0.995∗∗∗3.785∗∗∗0.983∗∗∗−0.411∗∗∗3.807∗∗∗ (0.223)(0.420)(0.229)(0.434)(0.230)(0.079)(0.436) F-valueforboldvariables0.557.054.857.453.6417.516.16 R20.490.460.560.460.560.930.46 Note:Forexpositionalpurposes,thedependentvariableAjtisinthousandsofadvertisements.Robuststandarderrorsclusteredbystation- marketcombinationaregiveninparentheses.∗∗∗p<0.01,∗∗p<0.05,∗p<0.1.
Advertising, Market Power, and Network Effects in Radio
Broadcasting
3.1 Introduction
Two-sided markets can be characterized by two distinct types of users that interact on a common platform. An important aspect of two-sided markets is that utility obtained by one side of the market is affected by the number of participants on the other side of the market. Platforms need to take into account these cross-group externalities when making strategic decisions to maximize their profits. Some common examples of two-sided markets are credit-card markets, shopping malls, dating agencies, and media markets.
An important question is how platforms make these strategic decisions with re-spect to both sides of the market. Theoretical work by Rochet and Tirole (2003), Cail-laud and Jullien (2003), and Armstrong (2006) shows that one side of the market is often priced below marginal cost and the other side acts as the profit segment. The profit-maximizing outcome might even have negative prices or free services. This im-plies that the pricing strategies are more complex for two-sided markets and that the conventional tools for market power cannot be applied. The standard Lerner pricing formula, which is the markup over price, will give an incorrect indication of market power because it does not consider the cross-group externalities. This is also pointed out by Evans and Noel (2008) and Wright (2004). For example, Rochet and Tirole (2003) show that the two-sided version of the Lerner index depends on both price elasticities for the case with positive externalities between both sides of the market.
The purpose of this chapter is to estimate a two-sided Lerner index that takes into account the cross-group externalities for the Dutch radio broadcasting industry. The radio broadcasting industry can be characterized as a two-sided market because the two types of users, listeners and advertisers, interact on the common platform, a ra-dio station. Rara-dio stations provide broadcasting of programmes and/or music to lis-teners and provide advertisers with a channel to market their products to potential consumers.
The main challenge is to derive a Lerner index under the assumption of price-setting behavior of radio stations that incorporates the network effect between adver-tisers and listeners. Suppose that a radio station increases its price for advertisements.
The direct effect is a decrease in the number of advertisements broadcasted. The
in-direct effect is that, due to less advertising, the radio station attracts more listeners which, in turn, positively affects advertisers’ willingness to advertise. So, the Lerner index will be underestimated if this feedback loop is not taken into account.
This chapter quantifies this indirect effect by using a structural two-sided model for the radio broadcasting industry. Our model includes listeners’ demand and adver-tising demand as well as profit-maximizing behavior of radio stations from which our two-sided Lerner index is derived. The two demand models provide the link between the estimated demand parameters concerning the network effect and the estimation of the Lerner index. Hence, the model allows listeners to dislike advertising and adver-tisers to value listeners. Moreover, adveradver-tisers are not only assumed to be interested in the total number of listeners but also in the demographic composition of the audience.
This chapter is closely related to empirical work on the pricing structure in media markets. Kaiser and Wright (2006) focus on the market for magazines in Germany and show that the cover price for readers is subsidized and the profit comes from adver-tisers by using a Hotelling demand specification. Song (2011) finds a similar pricing structure for TV magazines in Germany. Moreover, Song addresses the issue of market power and finds that advertising aversion increases the advertising markup because magazines want to charge advertisers more heavily to compensate their readers. Ar-gentesi and Filistrucchi (2007) employ a similar approach to measure the degree of market power in the Italian newspaper market. They find evidence of joint profit max-imization for the cover price and competition in the advertising market by comparing the estimated and observed markups. Van Cayseele and Vanormelingen (2009) show that newspapers publishers in the Belgian newspaper market make negative profits on the readers’ side but this is more than compensated by the profits from the advertisers side.
Although Song (2011) considers single-homing and multi-homing by advertisers, Argentesi and Filistrucchi (2007) only estimate a logit model for readers’ demand and advertising demand. A disadvantage of the logit model is the assumption of single-homing implying that advertisers and readers can only use one platform. Although this assumption is not problematic for listeners’ demand, it is somewhat unrealistic for advertising demand in radio broadcasting. Moreover, the logit model has quite restrictive substitution patterns which are primarily based on market shares rather
than station characteristics.
In contrast to the magazine and newspaper market in which not only advertisers are charged but readers also pay a cover price, radio stations do not charge both sides of the market. Listeners do not pay a subscription fee to listen to radio whereas ad-vertisers do pay an advertising price to broadcast their advertisements. Although lis-teners can freely listen to radio, they implicitly pay a price by hearing advertisements which interrupts the music and/or programming of the radio station.
There is some empirical evidence that the degree of advertising nuisance differs by media market. For example, Wilbur (2008) and Tyler Mooney (2007) show that viewers and listeners dislike advertising in the television and radio broadcasting market in the US, respectively. However, Rysman (2004) finds that consumers value advertising in the market for Yellow Pages in the US. De Smet and Van Cayseele (2010) distinguish between small and large advertisement in the market for Yellow Pages in five Euro-pean countries. Their results indicate that readers value small advertisements and place no value at large advertisements. Kaiser and Song (2009) find mixed results for the German magazine market. Whereas their results indicate that readers appreciate advertising in the women’s, business/politics, and cars magazines, readers dislike ad-vertising in adult magazines. These outcomes are related to the informational content of the advertisement which are high for magazine segments in which readers appreci-ate advertising and low for segments in which readers dislike them.
We extend this work on network effects, pricing issues, and market power by ex-amining a two-sided market in which a platform only charges one side of the market and in which media consumers dislike advertising. Hence, the network effect is ex-pected to be negative for radio broadcasting. If a radio station attracts more listeners it will also attract more advertisers. Yet, more advertisements are likely to have a negative effect on the number of listeners. We allow for somewhat more flexible sub-stitution patterns among radio stations by using the nested logit model for listeners’
demand. Advertising demand closely follows the model of Rysman (2004) to allow for multi-homing and is adjusted to apply it to radio broadcasting.
The parameters are estimated using unique firm-level data on listener ratings in 11 regions for 23 two-monthly periods from 2005 to 2006. In addition, we have infor-mation on advertising quantities, advertising prices, radio stations characteristics, and
listener demographics. This unique dataset allows us to include more detailed infor-mation and to follow a novel approach to estimate a two-sided model for the Dutch radio broadcasting industry. The estimates will indicate whether there is a network effect and the size of the network effect. Moreover, the degree of advertising aver-sion of listeners is estimated. The degree of competitiveness of the radio broadcasting market is examined by calculating the two-sided Lerner indices for the radio stations.
This approach has not yet been followed for an European media market in which only one side of the market is being charged and therefore contributes to the literature on market power issues in two-sided markets.
The remainder of this chapter is organized as follows. The next section gives a de-scription of the Dutch radio broadcasting market and the data. The two-sided model is presented in Section 3.3 together with the supply side and the corresponding Lerner indices. Section 3.4 presents the estimation strategy and the empirical results are dis-cussed in Section 3.5. Finally, Section 3.6 concludes.