Essays On The Economics of Media Platforms

FACULTEIT ECONOMIE EN BEDRIJFSWETENSCHAPPEN
KATHOLIEKE UNIVERSITEIT LEUVEN
ESSAYS ON THE ECONOMICS OF MEDIA PLATFORMS
Proefschrift voorgedragen tot het behalen van de graad van Doctor in de Economische Wetenschappen door Dries De Smet
Nummer 367
2011
FACULTEIT ECONOMIE EN BEDRIJFSWETENSCHAPPEN
KATHOLIEKE UNIVERSITEIT LEUVEN
ESSAYS ON THE ECONOMICS OF MEDIA PLATFORMS
Proefschrift voorgedragen tot het behalen van de graad van Doctor in de Economische Wetenschappen door Dries De Smet
Nummer 367
2011
ii
Daar de proefschriften in de reeks van de Faculteit economie en bedrijfswetenschappen het persoonlijk werk zijn van hun auteurs, zijn alleen deze laatsten daarvoor verantwoordelijk. Since the dissertations in the series published by the Faculty of Business and Economics are the personal work of their authors, only the latter bear full responsibility for them.
Doctoral Commission
Chairman: Advisor: Co-Advisor: Other Members: Wim Moesen (KU Leuven) Patrick Van Cayseele (KU Leuven) Jo Swinnen (KU Leuven) Simon Anderson (University of Virginia) Fabrizio Germano (Universitat Pompeu Fabra) Frank Verboven (KU Leuven)
iv
Preface
Before starting this PhD, I attended an introductory evening on doing a PhD in economics. The big dierence between an undergraduate thesis and a PhD, one of the speakers explained, is that an undergraduate thesis has a clear target. You get a bow and arrow and you know what you have to do to hit the bulls eye. The question is clear, and so is the answer, or you know at least in which direction or within which boundaries you have to search. For a PhD youre much better equipped, with graduate courses and the help of senior academics, but the questions and answers are far less obvious. A PhD, and by extension any academic research, should be state of the art, trying to answer new and detailed questions and use newly discovered and advanced methodologies. In sum: it is repositioning the frontiers of knowledge. If writing a thesis was a journey, an undergraduate thesis would probably resemble a city trip: you know where the highlights are, and the only thing you have to do is reach your destinations and hit the camera button. Writing a PhD is much more like starting an explorative expedition. It is hard to know what to expect and to predict which road will lead to success. After the rst year in Leuven, PhD students are required to present a draft proposal of their PhD. Comparing the initial road map with the ultimate results would prove that such an initial road map is not very useful, at least to nd your way in the PhD. None of the questions answered in this PhD emerged as one of the original questions at the start of my PhD. There are many questions which look interesting at rst sight, but turn out to be trivial, answered by someone else, not relevant anymore or way too dicult to solve adequately. New questions emerge during research and new courses are taken. What
vi
Preface
ended up in this PhD is only a fraction of the work done; just as important are the time and eort put in ideas and drafts that never made it to the nal version. The absence of clear questions and answers makes academic research often a frustrating task, even for experienced researchers. As a result, more than 50 % of all starters at UGent and KU Leuven quit early and never nish their PhD (Verlinden et al. 2005). However, nding answers on well-dened questions, allowed me to present my work on workshops and conferences, which are inspiring environments to exchange ideas and to catch up new ways to dene and answer interesting questions. With the deadline in sight, I also became more focused and more productive and Im glad to have brought this PhD to a good conclusion. In this dissertation, I was able to combine two of my biggest passions: media and economics. Since my childhood, I am passionate about media, news and writing. The constant changes in the media sector, due to the rise of the internet and the loss of advertising revenues following the nancial crisis, made it extremely interesting to explore and explain this sector from an economic viewpoint. Whether I will stay in academia and/or in economics remains to be seen. Whats for sure, is that I will remain an economist with an academic background. Whatever the problem, I catch myself nding out how agents interact, what the incentives are and which trade-os emerge. Economics is, contrary to what many people believe, not the science of money but a powerful tool to explain reality. I hope to have explained part of this complex reality in this PhD.
Dries De Smet Mannheim, July 2011
Contents
Preface Contents 1 General Introduction 2 Newspapers 2.1 Introduction . . . . . . . . . . . . . . 2.2 Theoretical Model . . . . . . . . . . 2.2.1 Monopoly model . . . . . . . 2.2.1.1 Set-up . . . . . . . . 2.2.1.2 Results . . . . . . . 2.2.2 Competition Model . . . . . . 2.2.2.1 Set-Up . . . . . . . 2.2.2.2 Results . . . . . . . 2.2.3 Welfare Analysis . . . . . . . 2.2.4 Conclusion of the Model . . . 2.3 Empirical Evidence . . . . . . . . . . 2.3.1 Data Description . . . . . . . 2.3.2 Empirical Strategy . . . . . . 2.3.3 Results . . . . . . . . . . . . 2.3.4 Concerns . . . . . . . . . . . 2.4 Survey of Belgian Journalists . . . . 2.4.1 Related Literature . . . . . . 2.4.2 Methodology . . . . . . . . . 2.4.3 Results . . . . . . . . . . . . 2.5 Advertising Bias Channels . . . . . . 2.5.1 Supply and demand channels v vii 1 7 8 11 12 12 15 16 16 18 21 23 23 23 26 27 31 35 35 37 38 45 45
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
viii 2.5.2 Inside channels . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . Monopoly prots . . . . . . . . . . . . . . Figures & Tables . . . . . . . . . . . . . . Overview Surveys on Inuence Advertisers Advertising Bias Channels . . . . . . . . . D1 Supply and demand channels . . . D2 Inside channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CONTENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 46 49 50 53 58 58 62 67 68 72 76 76 80 86 89 94 94 95 96 99 101 101 104 104 105 107 108 110 112 117 118 120 124 124 125 128
2.6 A B C D
3 Directories 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 3.2 Set-Up of the Theoretical Model . . . . . . . . . . . 3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 The Availability Restriction . . . . . . . . . . 3.3.2 The Comprehensive Directory Restriction . . 3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . A Proofs . . . . . . . . . . . . . . . . . . . . . . . . . . B Empirical Investigation of the Yellow Pages Industry B1 Industry Characteristics . . . . . . . . . . . . B2 Data . . . . . . . . . . . . . . . . . . . . . . . B3 Identication . . . . . . . . . . . . . . . . . . B4 Results . . . . . . . . . . . . . . . . . . . . . B5 Robustness Check: SURE . . . . . . . . . . . B6 Reconciling Theory with Empirical Findings . C Extensions to the Model . . . . . . . . . . . . . . . . C1 Negative Utility from Advertisements . . . . C2 Endogenizing r (as a function of s1 and s2 ) . C3 Price discrimination . . . . . . . . . . . . . . C4 Results for a Ramsey Planner . . . . . . . . . C5 Competition . . . . . . . . . . . . . . . . . . D Proofs of Extensions . . . . . . . . . . . . . . . . . . 4 Aggregators 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . 4.2 Related Literature . . . . . . . . . . . . . . . . . . 4.3 Theoretical Model . . . . . . . . . . . . . . . . . . 4.3.1 Set-Up of the Model . . . . . . . . . . . . . 4.3.1.1 Framework without metaplatform 4.3.1.2 Framework with metaplatform . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
CONTENTS 4.3.1.3 Timing of the Game . . . . . . Basic Results . . . . . . . . . . . . . . . 4.3.2.1 Stage 2: Seller Quantities . . . 4.3.2.2 Stage 1: Allowance Decision . 4.3.3 Extensions . . . . . . . . . . . . . . . . 4.3.4 Future research . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . Equilibrium quantities . . . . . . . . . . . . . . A Legal Discussion on Metaplatforms . . . . . B1 Arguments in favor of the Platform . . . B2 Arguments in favor of the Metaplatform 4.3.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix 130 130 130 132 135 139 144 146 146 146 148 151 157 159 160 171
4.4 A B
5 General Conclusion Acknowledgements About the Author Bibliography Dissertations
CONTENTS
List of Figures
2.1 2.2 2.3 B1 B2 3.1 3.2 A1 C1 C2 C3 C4 D1 D2 D3 4.1 4.2 4.3 Questions on Relationship Advertisers Newspapers Suggestion to Write about Advertiser . . . . . . . . Suggestion to Adjust Article about Advertiser . . . Total Advertising Expenditure over Time . . . . . Total Coverage over Time . . . . . . . . . . . . . . Dierent regimes given availability . . . . . . Dierent regimes given comprehensiveness . . Platform Prot . . . . . . . . . . . . . . . . . Dierent Solutions in the r-s-space . . . . . . Dierent solutions in the v-s-space . . . . . . Relative prices (p = p1 ) . . . . . . . . . . . . p2 Optimal welfare solutions in the r-s-space. . . Dierent Solutions in the r-s-space with r < 0 Dierent solutions in the r-s-space. . . . . . . Price Ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 41 42 50 50 77 82 90 105 106 107 109 113 115 116
Set-Up of the Game . . . . . . . . . . . . . . . . . . . . . . 124 Equilibrium Seller Quantities for All Cases . . . . . . . . . 132 Equilibrium Zones for Allowance Decisions . . . . . . . . . . 134
xii
LIST OF FIGURES
List of Tables
2.1 2.2 2.3 2.4 2.5 2.6 2.7 B1 B2 B3 C1 B1 B2 B3 Summary Statistics of Advertising and Coverage Coverage explained by advertising and dummies Robustness Checks . . . . . . . . . . . . . . . . . Granger Causality Test . . . . . . . . . . . . . . Coverage explained by adv. of last 12 months . . Summaries Survey . . . . . . . . . . . . . . . . . Survey Results per Group . . . . . . . . . . . . . Coverage explained by logged adv. and dummies Coverage explained by lagged advertising . . . . Advertising explained by lagged coverage . . . . Overview Surveys on Inuence Advertisers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 28 31 33 34 37 44 51 52 52 53
Variables: summary statistics . . . . . . . . . . . . . . . . . 95 Results of the OLS regression . . . . . . . . . . . . . . . . . 100 Results of OLS and SURE . . . . . . . . . . . . . . . . . . . 103
xiv
LIST OF TABLES
Chapter 1
General Introduction
When the global payments technology company Visa came under attack from the antitrust authorities in the 1990s, the company hired the economists Richard Schmalensee and David Evans (Warsh 2007). In 1999, they wrote Paying with Plastic in which they defended the debit and credit cards industry. As a side product, they also analyzed how the market for plastic cards was functioning. They stated that standard economic models failed to observe the chicken-egg problem that confronted card issuers: if consumers dont possess a certain card, then merchants wont accept it, and vice versa, if merchants dont accept it, then consumers dont want this card. When Schmalensee asked Jean Tirole and Jean-Charles Rochet to have a closer look at the problem, these French economists gured out that these chicken-egg problems are not limited to credit cards, but are present in many markets. Every market where an intermediary is active that connects two sides of the market, serves as a candidate to be examined as a two-sided market. The list of applications seems endless: real estate brokers, auction houses, internet platforms, dating services, employment agencies, magazines, radio, television, payment systems such as debit and credit cards, operating systems for personal computers, video games, mp3 players... (Evans & Schmalensee 2008) The three topics discussed in this PhD thesis, newspapers, telephone directories and buyer/seller internet platforms, might seem at rst sight unrelated, but a closer look learns that they are all two-sided markets, essentially bringing together readers and advertisers. That immediately explains why this PhD is titled Essays on the Economics of Media Platforms. To model these markets, I rely on the fast evolving literature on twosided markets, started by the seminal papers of Rochet & Tirole (2003) and Armstrong (2006)1 . Two-sided markets can be loosely dened as markets where one or more platforms serve two distinctly dierent types of agents. There is a certain kind of interdependence between both groups of agents, i.e. they generate an externality on each other. A key feature of two-sided markets is that these externalities (or also: indirect network eects) are not internalized by the dierent groups. As a result, an intermediary is needed to internalize these externalities. As Rysman (2009) argues, this denition might be overly inclusive. While many markets can
1 Other seminal papers are Caillaud & Jullien (2001), Caillaud & Jullien (2003), Evans (2004), Parker & Van Alstyne (2005) and Rochet & Tirole (2006).
3 be considered as two- or multi-sided markets, these externalities are not always important for the analysis. Rysman (2009) gives the example of cars. While the existence of car dealers and gasoline stations was very important for car drivers by the introduction of cars, today it is much less important given the abundance of dealers and gasoline stations. But these networks might become important again when new car brands or new fuel sources are introduced. Nearly all papers in the two-sided market literature place the platform or intermediary at the foreground. It are in particular the key decisions of platforms that are investigated, given the decisions of both agent groups to join or use the platform. Most of these articles look at price setting platforms, though some consider quantity setting platforms. A big divide in the literature has been the question whether agents pay a membership fee (e.g. Armstrong (2006)) or a transaction fee (e.g. Rochet & Tirole (2003)). Weyl (2010) tried to unite both models by introducing two-part taris which are combinations of both. He stresses more the heterogeneity of agents: are they heterogeneous in their valuation for interaction (as in Rochet & Tirole (2003)), for membership (as in Armstrong (2006)) or for both? In the models I have build in the next chapters, I have not chosen in an outspoken way for one type of payment (e.g. membership or transaction fee), or for one type of user heterogeneity. I believe that a consistent approach is important to derive general results, as has been done by Armstrong (2006), Rochet & Tirole (2003) and Weyl (2010). Though, for the questions I try to answer in the next chapters, it is more important to model the markets under investigation in a logical, sensible way. In sum, I combine price and quantity setting platforms, between membership and transaction price and between membership and transaction level heterogeneity. Structure The main text of this dissertation consists of three chapters. In chapter 2, I investigate the potential favorable treatment of advertisers by newspapers. Chapter 3 questions whether the Universal Service Obligation for telephone directories is ecient. In chapter 4, I discuss the eect of aggregators on traditional platforms such as real estate websites. In chapter 2, the main question is: do newspapers write more about their advertisers? I try to answer this question on three levels. In my theoretical model, I show that the answer is yes, if readers do not mind
bias too much and if bias is relatively eective for advertisers. In a survey of Belgian journalists, 21 percent answers this question positively, but direct pressure is almost non-existent. In the empirical part, I nd that advertisers in Belgian Dutch-language newspapers receive a signicantly higher coverage. This chapter is part of a larger and quickly emerging eld of the economic literature, known as forensic economics (Zitzewitz (2009) and Fisman (2010)). This eld is aimed at uncovering the hidden behavior of dierent types of agents. These empirical studies use measurement inconsistencies, violations of the ecient market hypothesis, discontinuities, variation in incentives and eld experiments to discover all types of corruption. One of the rst forensic economists avant la lettre was the Belgian statistician Adolphe Quetelet. Quetelet (1846) compared the chest size of French conscripts with the expected values, assuming that chest size was normally distributed over the population. He observed a puzzling shortage of men with a chest-size slightly above 1.57 meter and an excess number of men below 1.57 meter. It might not surprise the reader that men below 1.57 meter were excused for the Imperial army. Zitzewitz (2009) provides numerous examples of cheating and corruption on taxes, education, government procurement, sports, weather forecasts and yes, also academic performances. Advertising bias is also a hidden action because newspapers will never admit that they write more about advertisers because the reputation loss might be devastating. In chapter 3, I examine the universal service obligation for telephone directories. In most countries, this obligation consists of two parts. First, publishers must oer a comprehensive directory, including all telephone subscribers. Second, this directory should be available to all users. Building a simple theoretical model, inspired by the literature on two-sided markets, I nd that one of these two obligations is redundant. Imposing availability to all users is sucient to induce publishers to oer a comprehensive directory. The reverse however does not hold. In chapter 4, the eect of metaplatforms on traditional platforms is investigated. Aggregators, or metaplatforms as I call them, are both an opportunity and a threat for traditional platforms. The metaplatform redirects customers that would otherwise not have visited the traditional platform. But the metaplatform scores o some of the advertiser profits generated through customers. In a theoretical model, I show that traditional platforms might allow the metaplatform, even when it is not
5 protable, due to a coordination failure. In this chapter, I shed my light on the so-called New Economy that was created by the rise of the Internet. In chapters 2 and 3, I remained silent on the Internet though it goes without saying that it had and has a big impact on newspapers and telephone directories. Especially during the nancial crisis, when the fast declining advertisement revenues accelerated the damage done by the Internet, many newspapers kicked the bucket. Printed telephone directories survived, but is is unlikely that they will survive once everyone is connected to the Internet. Note that the focus of chapters 2 and 3 on the traditional media does not degrade my results, since Im condent that the basic mechanisms in print also hold for online newspapers and telephone directories.
Chapter 2
The Advertiser is Mentioned Twice. Media Bias in Belgian Newspapers

This chapter is based on joint work with Stijn Vanormelingen.
Newspapers
2.1
Introduction
Companies are indispensable for newspapers. They are not only an important source of news, e.g. by the products they release or by the uctuations in their stock prices, but are also an important source of revenues through advertisements. While this is not unusual in itself, it might become worrisome when a newspaper mixes up this double role of companies. When an advertiser has a higher probability of being covered in the newspaper, because the newspaper wants to please the advertiser, then we label this as advertising bias. The advertiser is mentioned twice: once in the advertisement and once in advertising bias, which is wrapped as news. Allegations that editors and journalists are inuenced by advertisers are probably as old as advertisements in newspapers1 . One of the oldest documented cases of advertising bias is found in an editorial in the New York Herald. When other pill makers complained about the favorable treatment of dr. Brandreth cure-all pills, the editor of the New York Herald wrote in 1835 in his newspaper: Send us more advertisement than dr. Brandreth does and well cut dr. Brandreth to dead (Bagdikian 2004). A more recent example is the newspaper Chicago Sun-Times, who mentioned in an article on the sales after the death of Versace only department stores that also advertised in the newspaper. The editor defended the approach with we have to take care of our customers. The same newspaper later declared that theaters advertising get preferential editorial treatment. Similarly, according to a leading journalist, it was quite normal in the 1950s that articles on womens and fashion pages reected the relative advertising strength. The empirical research on advertising bias is relatively new. Reuter & Zitzewitz (2006) nd that mutual funds that advertise in the US nancial press receive more as well as more positive coverage. Rinallo & Basuroy (2009) document a strong inuence of advertisements by Italian fashion companies on the editorial content of newspapers and magazines. A one percentage increase in the number of pages advertised generates 0.4 percent additional coverage (also measured in pages). Similar results are found by Bignon & Miscio (2009) for the French nancial press reporting on nancial assets in 1907, Gambaro & Puglisi (2009) in their study about Italian newspapers writing about major listed companies and
1 For a historical view on advertisements in newspapers, see Baldasty (1992) and DeLorme & Fedler (2005).
2.1 Introduction
Gurun & Butler (2010) investigating coverage of local advertisers by US local media. Di Tella & Franceschelli (2009) nd that advertisements by the incumbent government (which can be perceived as subsidies) have a strong negative impact on the coverage of scandals in Argentinia.2 These studies conrm surveys on adverting bias. Soley & Craig (1992) report that 90% of the US editors in their sample have been pressured by advertisers. More than one third of the respondents answered that advertisers succeeded in inuencing the news. In a survey of Hays & Reisner (1990), 37% of farm magazine writers report that advertisers attempts to inuence what stories appear harmed the profession. An equal amount criticized pressure from their editor or publisher to please advertisers.3 In a study of the American Society of Newspaper Editors, 50% of the readers believed that newspapers allow advertisers interests to inuence news decisions (ASNE 1999). In the literature, many dierent denitions have been used to describe media bias. There seems to be no accepted denition of bias, but most authors follow the slanting denition of Hayakawa (1940): Slanting is the process of selecting details that are favorable or unfavorable to the subject being described.4 We dene advertising bias as the existence of an implicit or explicit bundle of advertisement and favorable coverage. Suppose news stories can be ranked on newsworthiness. Since a newspaper has only limited space, there exists a newsworthiness threshold. An unbiased newspaper will publish all news above the threshold and ignore all news below the threshold. Positive bias will occur when a certain news story appears in the newspaper, despite being under the newsworthiness threshold. Negative bias occurs when a news story is not in the newspaper, despite being above the newsworthiness threshold.5
of non-existence of advertising bias is scarce, perhaps because of a publication bias towards statistically signicant results. Reuter (2009) does not nd any statistal dierence in the ratings of a wine magazine that accepts advertisements versus a magazine that doesnt. 3 For network television correspondents, this gure appears to be much lower. Price (2003) nds that only 7% reported some advertiser pressure. 4 For an overview of bias denitions, see McCluskey & Swinnen (2010). 5 Note that for political bias, the construct of newsworthiness is seldom used, because it is dicult to operationalise. In the context of political media bias, bias is often expressed as the relative position towards two extreme viewpoints. A medium that balances the two sides is seen as unbiased. In empirical studies, these two sides are left vs. right (Groseclose & Milyo (2005) and Gentzkow & Shapiro (2010)) and government vs. opposition (Di Tella & Franceschelli (2009) and Durante & Knight (2009)). Theo2 Evidence
10
Newspapers
In their theoretical models, Ellman & Germano (2009) and Germano & Meier (2010) consider only negative bias, i.e. the suppression of negative news. Bias is considered as something negative for news consumers and something positive for advertisers, creating a clear trade-o. In our theoretical model, bias is presented as the bundle of an advertisement with favorable coverage. This coverage can be favorable both at a general level, i.e. by promoting business or consumption, and at the company level. At the general level, the media outlet is inclined to create a positive climate for all businesses, or to choose content that stimulates consumption. At the company level, an outlet pleases advertisers by inserting more positive news and downplaying negative news. This concept is very similar to stealth marketing or covert marketing, where advertisers try to invade editorial content with promotional messages. A dierence is that this invasion mostly happens through a sponsorship, and not as a complement to advertising6 . The empirical exercise exploits the fact that advertisement levels should have no inuence on the coverage of companies. We consider a newspaper as unbiased if advertisement expenditures have no inuence on the newspapers reporting of companies. A statistically signicant eect of advertisement on coverage, is an indication of advertising bias. This is similar to other empirical studies on advertising bias, such as Reuter & Zitzewitz (2006), Rinallo & Basuroy (2009) and Gambaro & Puglisi (2009). In this chapter, we attempt to answer the question whether advertising bias is present by applying several approaches. First, we build a theoretical model. Second we combine two data sets holding information about advertising and coverage in dierent Belgian newspapers and nally, we conduct a survey among Belgian journalists. The remainder of this chapter is structured as follows. In section 4.3, we build up a theoretical model to explain advertising bias. In this simple theoretical model, we investigate the incentives of a monopolist newspaper to insert positive advertising bias. We argue that the amount of bias if any depends on the relative eectiveness of bias and advertisement and the relative aversion of readers towards bias and advertisement. Whether bias will prevail is ultimately an empirical question.
retical studies often consider a binary state of the world (Baron (2006), Gentzkow & Shapiro (2006), Sobbrio (2009) and Suen (2004)). 6 Roy & Chattopadhyay (2010) give an overview of the dierent denitions of stealth marketing; Goodman (2006) provides a legal discussion.
2.2 Theoretical Model
11
In section 2.3, we measure advertising bias in the Dutch-language Belgian newspapers. For that purpose, we matched a dataset of advertisement expenditures in newspapers with the number of articles mentioning these advertisers. We nd that advertisers are mentioned twice as often as non-advertisers. Controlling for the fact that some newspapers are more business oriented, we document that an expenditure of 100.000 euro generates depending on the specication between 0.47 and 1.86 additional articles. Alternatively, every four to fourteen full page black and white advertisements generate one additional article in the same month. The question remains how this bias is implemented. To answer this question we conducted a survey of Belgian Dutch-speaking newspaper journalists in section 2.4. Based on the qualitative answers of this survey, and the many examples in the literature, we dene a number of explicit and implicit channels that might facilitate advertising bias. The channels are described in section 2.5. Section 4.4 concludes and discusses policy proposals to reduce advertising bias in the future.
2.2
The Newspapers Incentives for Advertising Bias
In this section, we sketch a simple theoretical model, in order to understand better the incentives and determinants of advertising bias. This model will help us to better understand under which conditions bias is likely, i.e. when we can expect advertiser bias is likely. We model a newspaper as a platform that serves both readers and advertisers7 . The newspaper has the option to introduce advertising bias. By doing so, the publisher takes into account two eects. On the one hand, advertising bias increases the willingness to pay by advertisers. On the other hand, advertising bias scares o readers. The trade-o between both eects determines the amount of advertising bias, if any. This mechanism is similar to the one in Ellman & Germano (2009) and Germano & Meier (2010). To keep the model tractable, we make several non-trivial assumptions.
7 For an overview of platforms and the two-sided markets literature, see Rochet & Tirole (2006) and Rysman (2009). For theoretical applications of two-sided markets to media markets, see Anderson & Coate (2005), Anderson & Gabszewicz (2006) and Armstrong (2006). Rysman (2004), Kaiser & Wright (2006) and Argentesi & Filistrucchi (2007) provide empirical investigations of media as a two-sided market.
12
Newspapers
First, we only look at one specic type of bias, i.e. positive news which is not relevant for readers, in return for advertisement. Other types of bias (e.g. political bias) are likely to produce other results. Second, we assume that readers dont pay. Third, we eliminate all marginal and xed costs, i.e. producing news (and advertisement) is free. Introducing a cost to produce news might, together with the introduction of a reader price, reduce the averseness of readers towards bias and ads, because they might help nance news. Finally, we consider at most 2 media outlets. Adding more media outlets might provoke dierent dynamics not captured by this model. Most of these assumptions can be relaxed, but this comes at the cost of losing tractability and/or obfuscating the focus of this chapter, i.e. the likeliness of advertising bias in equilibrium.
2.2.1
Monopoly model
In the rst subsection, we introduce the model with a monopoly newspaper publisher to make clear the underlying mechanism behind advertising bias. In section 2.2.2 we introduce competition in the model. 2.2.1.1 Set-up
The monopolist newspaper acts as a platform facing a double demand of readers and advertisers. The newspaper oers them a product that consists of three parts: news (n), advertisements (qA ) and bias (b). Bias is the mentioning of an advertiser in a news story, despite the fact that this mentioning is not newsworthy.8 The newspaper is a quantity setter and chooses the amount of news, bias and ads. The total available space for content in the newspaper is xed and normalized to 1. Therefore, the variables n, qA and b can be interpreted as percentages and add up to one. Choosing two of these variables automatically determines the third one, therefore we concentrate on b and qA . Readers Readers make a decision whether to read the newspaper or not. Readers like news but attribute a disutility to advertisements and bias. To simplify the model, we make abstraction of the reader price.The
8 If the advertiser is mentioned in a news story and this mentioning is newsworthy, then it is counted under news (n). Moreover, we only have two levels of newsworthiness: either it is newsworthy (news) or it is not (bias). We dont allow for more detailed levels.
2.2 Theoretical Model utility of reader k reads uRk = 0 n 1 b 2 qA wk
13
(2.1)
where i are the (dis)utilities gained from each type of content. We assume that there is a unit mass of readers who have an opportunity cost w to read or pick up a copy of the newspaper, uniformly distributed over 0 and 1. Note that we (implicitly) assume here that readers know the amount of bias in the newspaper, but cannot distinguish news from bias9 . To simplify the notation, we use n = 1 b qA , and introduce 0 0 , 1 0 + 1 and 2 0 + 2 . All potential readers whose utility is larger than zero (uRk 0), read the newspaper, and therefore the number of readers is equal to qR = 0 1 b 2 qA (2.2)
Advertisers As readers of newspapers act as consumers in the market, companies want to persuade them to buy their products. There are two channels through which a company can reach consumers10 . First, by advertisements in the newspaper and second, by bias published by the newspaper. Companies want to advertise in the newspaper because it enables them to persuade consumers to buy their products. Bias and advertisements dier in their propensity to persuade consumers, labeled 1 respectively 2 . We further assume that newspapers spread the total amount of bias equally over all advertisers, i.e. the amount of bias per
9 It might be possible to introduce a more complex model where news and bias can be veried ex-post, in the spirit of Milgrom & Roberts (1982). Gentzkow & Shapiro (2010) model ex post verication in the context of political bias; Blasco, Pin & Sobbrio (2010) do so for advertising bias. We dont introduce a micro model on verication, but our modeling might be seen as a long run eect of bias. Suppose that readers detect bias ex post; and that the choice of bias does not change over time, then readers know perfectly the amount of bias, but they dont know which articles are biased. Moreover, if ex post revelation is not perfect, then our parameter 1 will capture both the aversion against bias and the chance that bias is detected. 10 In the same spirit, one can also replace consumers by voters, donors or investors. Interestingly, DellaVigna & Gentzkow (2009) show in an overview of empirical literature on persuasion that these latter categories might be even more prone to persuasive communication than consumers.
14 advertiser is equal to
b 11 qA .
Newspapers Therefore, the persuasion per reader is = 1 b + 2 qA (2.3)
This persuasion is equal to the amount of products or services the average reader buys from the advertiser. If < 1, this can be interpreted as the chance that the average reader will buy the product or service. There is a mass 1 of advertisers which obtain a prot of Ak per product or service sold. The utility function of advertiser k reads uAk = Ak qR pA (2.4)
This parameter Ak is uniformly distributed over 0 and 1. We further assume that each company can place only one ad. All potential advertisers whose utility is larger than zero (uAk 0) will advertise in the newspaper. Therefore, the implicit price of the quantity setting monopolist is pA = (1 qA )qR (2.5)
Newspaper The newspaper is a prot maximizer. It chooses the optimal amount of bias b and advertisement qA . The remaining part of the newspaper consists of news. With a cost of c per copy, the prot function reads = (pA c)qA (2.6) In setting its optimal advertisement quantity, the newspaper takes into account the direct eect of this quantity on prot, but also its indirect eects through a lower equilibrium price and lower number of readers. Similar for bias, the newspaper will weigh the positive eect of increasing the willingness to pay of advertisers with the negative eect of chasing away some readers. For the sake of simplicity, we will assume that c = 0 for the remainder of the chapter. Further assumptions To obtain the results in the next section, newspaper prot is maximized with Kuhn-Tucker constraints. To simplify the
11 Note that we implicitly assume here that the newspaper knows the distribution of Ak , but that the Ak of each company is private information. If this information was public, then the marginal rms would receive more bias, since they drive the advertiser demand price (in the absence of price discrimination). In that case, bias would not be given to infra-marginal advertisers, since they are going to advertise anyway.
15
analysis, we impose that the parameters 1 and 2 are larger than 1 and 0 , respectively. These assumptions exclude two sets of corner solutions: rst, newspapers with only ads and bias, and second full coverage on the reader side no matter whats in the newspaper. We maintain these assumptions throughout the chapter. 2.2.1.2 Results
The monopolist newspaper maximizes his prot = pA qA subject to the constraints n, b, qA , qR 0. We solve this problem with Kuhn-Tucker conditions. Lemma 2.2.1 The optimal quantities are
0 b= qA =
2 21 2 1 2 1 +2 (2 1)2 2 1 31 1 (2 1 1 2 ) 1 21
if if if if if if
1+2
( 12 + 2 ) 1 < 1 (2 1)2 2 2 2 1 ( 12 +2 ) 1 (2 1)2 2 21 1 > 1+22 2 ( 12 + 2 ) 1 < 1 (2 1)2 2 2 2 1 ( 12 +2 ) 1 (2 1)2 2 21 1 > 1+22 2
21 1+22
2 12 +2 32
1 22 1 +21 2 3(1 2 2 1 )
21 1+22
The second solution is the unconstrained maximum. The rst and third solution are corner solutions where respectively b and qA are zero. Given the assumptions on parameters 0 , 1 and 2 , the constraints n, qR 0 are never binding in this problem. In the rst solution, the newspaper decides not to insert bias, either because it is not eective enough (1 too low or 2 too high) or because it is too annoying relative to advertising (1 too high or 2 too low). For the newspaper, it is more protable to earn its revenue directly through advertising than oering the bundle of advertisement and bias. If advertising is relatively uneective or disliked by consumers, the newspaper will have only one advertiser (qA 0). All the biased articles will be written about the only advertiser. The newspaper can then be seen as a company newspaper. Depending on the annoyance of bias (1 ), this amount of bias can increase up to one half of the newspaper content.
2 21 In the region where 1 (2 1)2 2 1 1+22 , the newspaper will 2 have both bias and advertisements. Each advertiser also gets a share ( qb ) A of the biased news. The higher 1 and 2 , the higher the bias and the lower the advertisement quantity. The reverse holds for 2 and 1 .
1 + 2 )
16
Newspapers We summarize these results in the following proposition.
Proposition 2.2.2 Bias appears if relative eectiveness of bias is high compared to the relative annoyance. The newspaper will implement advertisements if these are relatively eective compared with the relative annoyance. Only for intermediate eectiveness of bias (vs. ads), both bias and ads will be implemented. The newspaper prot12 increases the more eective bias and advertisement (1 and 2 high) and the lower the annoyance of bias and advertisement (1 and 2 low). Rather remarkable, for the second solution, the number of readers is increasing in 1 , but decreasing in 2 . This means that readership increases when bias becomes more annoying. The reason for this increased readership lies in the fact that the increased annoyance of bias induces the newspaper to lower the bias levels in equilibrium, which is partly replaced by ads. This reduction in bias overcompensates the negative eect of 1 on readership; the result is that readership increases13 . In sum, we can conclude that the monopolist newspaper will appear bias-free if bias is not eective for advertisers or highly disapproved by readers. There will be only one advertiser (and a substantial amount of bias) if an ad is not eective for advertisers or it is highly disapproved by readers. A newspaper will oer the bundle of advertisement and bias if the relative eectiveness ( 1 ) and annoyance ( 1 ) of bias are intermediate. 2 2
2.2.2
2.2.2.1
Competition Model
Set-Up
The set-up of the duopoly model is similar to the monopoly model. To introduce competition between newspapers, we assume newspapers are
prots corresponding with the three solutions can be found in Appendix A. that these results can also help to explain the impact of ad avoidance technologies, such as digital video recorders (e.g. TiVo) and whitelisting (e.g Ad Block Plus for Firefox and Chrome). These technologies make it possible to avoid ads, either by skipping or blocking them. As a result, advertisements become less annoying (2 decreases), but also less eective (2 decreases). In the extreme case, where 2 = 2 = 0 and as a result , , the newspaper implements the third regime, with bias and only one advertiser. An advertisement alone has no value for an advertiser and is therefore complemented with bias. Of course, when the model is altered and bias can be bought directly (without having to by it as a complement to advertising), it is likely that bias will be oered to various companies. See Anderson & Gans (2010), Wilbur (2008a) and Wilbur (2008b) for a theoretical and empirical discussion of ad avoidance on advertisement levels.
13 Note 12 The
17
horizontally dierentiated in the style of Hotelling competition. This competition model is commonly used in media economics models (Armstrong & Weeds (2005), Anderson & Coate (2005)). Readers In contrast to the monopoly model, where readers decide whether to read the newspaper or not, readers now decide which of both newspapers theyll read. Readers read one and only one newspaper, so we assume that the market is covered. The Hotelling model implies that the newspapers are horizontally dierentiated. One example to explain this dierentiation is political dierentiation. Suppose that one newspaper is politically left-leaning and that the other one is politically right-leaning. In the absence of price and quality dierences, readers prefer the newspaper that is closest to their political opinion. It is important to notice that the two newspapers are at the end of the Hotelling line, i.e. newspapers take the most extreme position, or in other words, there is a maximal dierentiation between both newspapers. The utility functions reads: uRik = 0 ni 1 bi 2 qAi twik (2.7)
where wik is the distance of reader k towards the position of the newspaper i; t is the parameter that captures the travel cost, or the disutility from every unit that the newspaper diers in position with the own position. We assume the Hotelling line has a length of one and that readers are uniformly distributed between 0 and 1. Introducing uRi as the utility without the travel cost (i.e. uRi = 0 ni 1 bi 2 qAi ), the quantity of readers of both newspapers reads: 1 uR1 uR2 + 2 2t 1 uR2 uR1 + 2 2t
qR1 qR2
= =
(2.8) (2.9)
Similar to the monopoly case, we also introduce 0 , 1 and 2 to simplify the computational output. Advertisers While readers are reading one and only one newspaper, advertisers might advertise in one, two or no newspaper. In the two-sided markets terminology, this means that readers are singlehoming and advertisers are multihoming. As such, the newspapers compete for readers, but act as monopolists with regards to advertisers. Since the readers of
18
Newspapers
both newspapers do not overlap, the decision for companies to advertise in newspaper 1 is independent from the decision to advertise in newspaper 2. This is similar to the set-up of Armstrong (2006). The utility and demand functions for advertisers are equal to those in the monopoly case and therefore omitted. Newspapers The prots of the newspapers are similar to those in the monopoly case. Newspapers do charge for advertisements but not for bias. Readers also dont pay. The newspaper is a quantity setter, i.e. it choses how many place is reserved for news, bias and ads. Since the total place in the newspaper is normalized to one, we only look at the decisions for bias (bi ) and ads (qAi ). The prot function reads: i = (pAi c)qAi which is maximized with respect to bi and qAi . Further assumptions As discussed above, we assume that the market is covered on the reader-side, i.e. all potential readers read one and only one newspaper. It suces to assume that 0 3 . In the comparison 2 between monopoly and duopoly, we set 0 equal to 3 , but this assump2 tion does not aect the competition results, since 0 does not appear in the equilibrium quantities. To reduce the notational burden, we set t = 1, though we relax that assumption below to investigate the eect of a strengthening or lessening of competition. 2.2.2.2 Results (2.10)
The competing newspapers maximize their prot i with respect to bi and qAi , subject to the constraints 0 n, b, qA , qR 1. We solve this problem with Kuhn-Tucker conditions. Lemma 2.2.3 The optimal quantities are symmetric:
0 if if if if if if bi =
2 22 1 +2 2 1 2 1 (2+2 ) 2 1 1 1 (1 2 2 1 ) 1 1 2 2+2 4+2 22 2 (2+2 4+2 ) 1 < 1 2 2 22 1 (2+2 4+2 ) 1 22 2 1 1 > 1+2 2
1 1+2
qAi =
1 2 1 +1 2 1 2 2 1
2 (2+2 4+2 ) 1 < 1 2 2 22 1 (2+2 4+2 ) 1 22 2 1 1 > 1+2 2
1 1+2
19
As can be seen, the results are qualitatively similar to the ones obtained in the monopoly situation. We also obtain three solutions, where the second solution is the unconstrained maximum and the rst and third solutions are the corner solutions where bi and qAi , respectively, are equal to zero. Similar to the monopoly situation, newspapers will oer the 1 bundle of advertisement and bias if the relative eectiveness ( 2 ) and annoyance ( 1 ) of bias are intermediate. 2 Intensity of competition To investigate the eect of the strength of competition on the equilibrium results, we relax the assumption t = 1 and consider t as an additional parameter. When this parameter t is low, competing newspapers are close substitutes. Based on a comparative statics analysis, we can state the following. Proposition 2.2.4 If newspapers are closer substitutes for the reader, the newspaper lowers the levels of bias and advertisement. The only exception is when the newspaper is oering both bias and advertisement: more competition decreases bias but increases the ad level. These results show that if the competition for the reader becomes more intense, newspapers react to this by increasing the utility of the readers. Even in the case where the ad level is increased, reader utility increases due to the decrease in bias. These results conrm our intuition that more intense competition will drive out some (but not all) bias, and the amount of news will certainly increase. As can be expected: increased competition diminishes newspapers prot. Comparison monopoly and competition It is interesting to compare our competition results with those obtained under monopoly. We summarize the main ndings in the following proposition. Proposition 2.2.5 Bias is (weakly) higher under competition. The number of ads is higher under competition if the relative eectiveness of bias (vs. ads) is low and/or the relative aversion against bias (vs. ads) is high. The prot of a monopolist is always higher than the prot of a duopolist. The nding that bias is always higher under competition is rather remarkable. One would have intuitively expected that the bias level would be higher under monopoly. The result might be an artefact of our modeling strategy where the monopoly is located at one side of the Hoteling line,
20
Newspapers
competing with an outside good (with utility 0) whereas duopoly newspapers are competing with each other, and unaected by the height of 0 , given that 0 is large enough such that the market is covered. The dierence in ad levels are determined by the parameters for eectiveness and aversion of bias and ads. Competitive newspaper have lower levels of ads when ads are relatively annoying and/or relatively ineective. The reverse holds when bias is more annoying and more ineective than ads. Prot is always lower under competition, which is in line with the expectations. In sum, we can conclude that bias does not disappear under competition. Our results show that more intense competition for the reader reduces advertising bias, though switching from monopoly to competition increases the amount of bias and decreases the amount of (real) news. These results dier from those obtained in the literature. Ellman & Germano (2009) and Germano & Meier (2010) nd that in duopoly, at least one newspaper eliminates all bias. In their models, bias is bad news about advertisers that is suppressed. Competitive newspapers are likelier to tell the truth about advertisers, because they do not fully internalize the reduction in advertisement spending (in all newspapers) due to the bad news. If a newspaper is the only newspaper revealing the news, it gains on the reader side. In models of political bias, it is not uncommon that bias increases if competition is introduced (Mullainathan & Shleifer 2005). The mechanism behind political bias is quite dierent from our modeling. Political bias models assume that readers have a (positive) preference for bias, and assume that people like to read or hear the opinion that is closest to their prior belief. Switching from monopoly to duopoly or oligopoly, means that newspapers can better cater to the preferences of the reader, by introducing bias. Note however, that bias and the eect of bias are quite dierent in this setting. Since these biases are in the opposing direction (e.g. politically left-leaning for one newspaper, right-leaning for the other), the amount of (aggregated) misinformation in society is limited. In the case of advertising bias, it is unlikely that a newspaper will enter with reverse advertising bias, i.e. blaming its advertisers, since this strategy is unlikely to be successful14 (Reuter & Zitzewitz 2006).
14 In theory, reverse advertising bias is not impossible. Criticizing advertisers might be perceived as a strong signal for the trustworthiness of the newspaper. As a consequence, it increases the willingness to pay of readers. Even if readers dont pay for the newspaper, a better reputation increases the readership, indirectly increasing revenue. If this increase from the reader side is large enough, it overcomes the direct revenue
21
2.2.3
Welfare Analysis
Journalism researchers tend to hold the view that only one level of bias is acceptable for the society: when bias is zero. From an economic point of view, though, it is not guaranteed that the absence of bias is welfare maximizing. In this section, we analyze the results of a monopolistic welfare maximizers (i.e. a Ramsey planner). For clarity, we split the analysis in reader welfare and advertiser welfare. For simplicity, we only look at the case where the Ramsey planner publishes one newspaper. Reader Welfare Readers gain utility from reading the news in the newspaper, but value bias, advertisement and a travel cost negatively. Since readers dont pay to read a newspaper, the consumer surplus of the readers is equal to uR qR WR = (2.11) 2 where uR stands for the utility without the travel cost. Written in full, 2 WR = (0 1 b2 qA ) . Since the newspaper doesnt occur costs in our 2 model, reader welfare is feasibly maximized when both the amount of bias and advertisement is zero. Proposition 2.2.6 Reader welfare is maximized when the newspaper contains neither bias nor advertisements. Note that this resembles a public newspaper where, similar to some public broadcasters, no bias and ads appear. Advertiser Welfare In this two-sided market setting, advertisers are a demand side too, and therefore advertiser welfare is also computed as a consumer welfare. Advertisers gain from every reader, but have to pay a price pA to appear in the newspaper. The advertiser welfare reads: WA =
2 uA qA 2
(2.12)
Maximizing advertiser welfare with respect to bias and advertisement, generates the following results.
decrease from the advertiser side.
22
Newspapers
Proposition 2.2.7 The quantities that maximize welfare are b= 0

0 (22 1 1 2 ) 31 (2 1 1 2 ) 20 32
if if if if
1 2 1 2 1 2 1 2
< <
21 2 21 2 21 2 21 2
qA =
1 0 3(1 2 2 1 )
These results learn that an advertiser welfare maximizer will always implement advertisements, as could have been expected. Bias will appear in the newspaper if the relative eectiveness of bias (vs. ads) is relatively high, or the relative aversion against bias (vs. ads) is relatively low. Note also the welfare maximizer never brings the advertisement to the maximum level; the highest amount of qA is 20 , which is, given our assumption 32 that 2 > 0 , at most 2 . The reason is that advertisers only gain utility 3 if there are also readers. Implementing a newspaper with only advertisements, or only advertisements and bias, would leave the advertisers with a newspaper without readers. Due to this two-sided nature, maximizing advertiser welfare always ensures that a substantial amount of readers still reads the newspaper. The next step in this welfare analysis is a Ramsey planner that maximizes the sum of reader and advertiser welfare. Though equilibrium results are not hard to compute, we omit them here because they dont add much insight. As could have been expected, the Ramsey planner weights the interests of readers and advertisers. As the advertiser welfare also contains part of the reader welfare, the Ramsey planner has a strong tendency to abolish all bias. Bias might however still occur, e.g. when bias is relatively eective (1 is high) and/or bias is relatively harmless for readers (1 is low). We conclude that the optimal amount of bias might be positive, at least when the welfare denition also includes the advertisers consumer surplus. Comparison Welfare under Monopoly and Competition To conclude this section on welfare, we compare the welfare levels of the protmaximizing solutions that were discussed in sections 2.2.1 and 2.2.2. Our computations are summarized in the following proposition. Proposition 2.2.8 The total welfare of readers and advertisers is higher under competition, if the relative eectiveness of bias over advertisements is low and/or relative annoyance of bias is high.
2.3 Empirical Evidence
23
While the exact parameter values where welfare is higher under competition might dier for reader welfare, advertiser welfare or for both, the results are qualitatively the same. As we have seen in the comparison between monopoly and competition regime, competitive newspapers always implement more bias than their monopolistic counterparts; this is likely to reduce welfare, and therefore welfare is only higher under parameter values that urge to implement low amounts of bias (because it is ineective or annoyant).
2.2.4
Conclusion of the Model
In sum, we can conclude that a newspaper will appear bias-free if bias is not eective for advertisers or highly disapproved by readers, whether the newspaper faces competition or not. There will be only one advertiser (and a substantial amount of bias) if an ad is not eective for advertisers or it is highly disapproved by readers. A newspaper will oer the bundle of advertisement and bias if the relative eectiveness ( 1 ) and annoyance 2 1 ( 2 ) of bias are intermediate. A comparison between monopoly and duopoly newspapers learns that these results are qualitatively the same, but that newspapers under duopoly set more bias and make less prot than their monopolistic counterparts. If duopoly newspapers are closer substitutes for readers, then newspapers set less bias; while the eect on advertisements is ambiguous.
2.3
Empirical Evidence of Advertising Bias in Belgian Newspapers
The previously presented theoretical model predicts that the combination of bias and advertisement will only exist for certain parameter values. It is therefore an empirical question whether advertising bias will prevail. We test the existence of advertising bias in the Belgian Dutch-language newspapers for the period 2001-2005.
2.3.1
Data Description
Clearly, in order to measure advertising bias, one needs rich data about advertising spending at the advertiser/newspaper level as well as data on media coverage of these advertisers in the dierent newspapers.
24
Newspapers
Advertisers Data on advertising expenditures are obtained from Aegis Media (see also Van Cayseele & Vanormelingen (2009)). The data set contains monthly expenditures by companies in each newspaper separately, from January 2001 till June 2005. From this set, we selected the large companies that concentrated their ads over newspapers. First, we focused on rms with a total advertising spending of over 300000 euros over the whole sample period. Moreover, we computed for each advertiser the share of each newspaper in total advertising spending of the advertiser. We used these shares to compute for each advertiser a Hirschman-Herndahl index of advertising spending and selected all companies with an index above 3000. We deleted two categories of companies from this list. First, we eliminated company names which are also common Dutch words or common family names, in order to avoid too many false positives. Second, we deleted the companies closely linked to a newspaper, such as the own advertising agency15 . Finally, 57 advertisers are withheld in the dataset and we observe for each of them monthly ad spending in each dierent newspaper. Coverage The second part of our dataset contains data about newspaper coverage of advertisers and we obtain a measure of how often an (potential) advertiser is mentioned in the newspaper. To be precise, we count for each newspaper edition, published on a given day, the number of articles in which the advertiser is mentioned. To this end, we ran keyword searches on the company name or the major company product on the Mediargus database, which is an online archive of all Belgian Dutch-language newspaper articles. Note that we counted several mentions in one article only as one observation and did not distinguish between positive and negative news16 .
15 The reason for focusing on the companies that typically concentrate their advertising spending in a limited number of newspapers is to optimize the time consuming collection of data about coverage. In our empirical strategy we are going to exploit the variation of advertising spending by a specic advertiser across dierent newspapers in a given time period and link this with advertiser coverage in that particular newspaper. High advertising spending by an advertiser in all newspapers simultaneously will be picked up by control variables and would not help in identifying the advertising bias parameter. Therefore we opt to focus on those advertisers concentrating their spending in a limited number of newspapers. 16 Obviously, it would be interesting to also include a content analysis. However, given the size of the dataset on coverage, this would be very time consuming. Moreoever, it is not easy to dene some objective measures of positive versus negative news
25
Table 2.1: Summary Statistics of Advertising and Coverage

Variable Advertising (euros) Coverage (nr. of articles per month) Sample All Non-Zero All Non-Zero Advertisers Non-Advertisers Mean 2249 28558 1.58 3.43 3.39 1.43 S.D. 20039 65200 3.25 4.06 5.66 2.89 max 842108 842108 96 96 94 96 obs 24624 1978 24624 11389 1978 22624
Our nal dataset contains information about coverage and advertising spending in all newspapers active in the Belgian-Dutch language newspaper market17 . The market consisted of 3 elite newspapers and 5 popular newspapers18 during the period of study. Coverage data is aggregated at the newspaper-advertiser-month level as we observe advertising spending at the monthly level instead of the daily level. We present some summary statistics in table 2.1. In total there are 24624 observations19 . From this table, we can infer that only 1978 of the 24626 observations contain positive advertisement expenditures, i.e. the average advertiser in our sample advertises 34 newspaper months. The average monthly expenditure is 28558 euro for non-zero expenditures, which equals roughly two black and white full pages ads. The last two rows indicate an advertiser is mentioned twice as much as a non-advertiser. This is a very rough indication that advertising bias might exist.
against which the article could be screened. Therefore, we refrain from doing so and leave this issue to further research. 17 The dataset excludes the free newspaper Metro as we do not have any content information about this newspaper. 18 These newspapers are: De Morgen, De Standaard, De Tijd (formerly FinancieelEconomische Tijd), Gazet van Antwerpen, Het Belang van Limburg, Het Laatste Nieuws, Het Nieuwsblad and Het Volk. The rst three newspapers are elite newspapers that focus more on political and international news compared to the popular newspapers who typically spend more articles on showbusiness and criminal news. These are all the Dutch-language national newspapers, except Metro, which is a free newspaper that is not available for subscription. 19 We have data on 57 rms advertising in 8 newspapers over a period of 54 months and 57*8*54 = 24624.
26
Newspapers
2.3.2
Empirical Strategy
In order to test for advertising bias, we investigate the eect of advertisement expenditure on news coverage. The summary statistics already showed how coverage of rms that advertised in the newspaper in a given month was already higher compared to the coverage of a rm that did not advertise. However, there could be various mechanisms going on here besides advertising bias to explain this positive correlation. First, larger rms would tend to spend more on advertising and would also be more likely to be covered in newspapers. Second, some characteristics of the newspaper will make it more likely to publish articles about companies as well as receiving more advertisements. For example, the average mention in the only business newspaper in our sample, De Tijd, is with 2.27 articles higher than the average of the other newspapers, although the dierence is not statistically signicant. Third, there could be some underlying common trends in both advertising spending and newspaper coverage. Moreover, these trends could be newspaper specic. Fourth, due to some unobserved characteristics of the readership of a particular newspaper, rms could be more interested in that newspaper compared to other newspapers and that particular readership could be more interested in the products of the advertiser as well. Given that newspapers tend to choose articles based on what their readers interests, a positive correlation between coverage and advertising would arise. Fifth, rms could be more likely to advertise when there has happened something newsworhty at the rm (e.g. a new product launch) and again a positive correlation between advertising spending and coverage would be found. In our empirical specication, we will control for these factors. If the newspaper is unbiased, one would expect the positive relation between coverage and advertising spending to fade away when controlling for the above mentioned confounding eects. More precisely, we will estimate the following equation: Cijt = Aijt + i + j + t + ij + jt + it + ijt (2.13)
where Cijt represents coverage of advertiser i by newspaper j in month t. Coverage is measured by the number of articles in which the advertiser or its main product is mentioned. Aijt is advertising spending of advertiser i in newspaper j in month t, and advertising expenditure is measured in thousands of euros. As discussed above, there could be some advertiser specic eects i leading to higher coverage. Newspaper time and
27
newspaper-time xed eects are represented by j , t and jt respectively and ij are advertiser-newspaper idiosyncrasies in coverage. Finally, it are advertiser-time specic xed eects on coverage. By controlling for all these factors, we solve for most of the concerns in identifying a causal impact of advertising spending on coverage mentioned in the literature. Moreover, we provide some Granger causality tests as well as include lagged advertising spending to address additional concerns about the interpretation of the correlation between advertising and coverage.
2.3.3
Results
The results of estimating equation 2.13 are displayed in table 2.2.
28
Table 2.2: Coverage explained by advertising and dummies

(3) 0.0147*** [0.00379] 0.00946*** [0.00263] 0.00891*** [0.00244] 0.00472*** [0.00156] (4) (5) (6) (7)
(1)
(2)
Advertising
0.0186*** [0.00352]
0.0155*** [0.00384]
Total Ads Adv, Newsp Month 24624 0.402 Adv X Newsp Month Adv X Newsp Newsp X Month Adv X Newsp Newsp X Month
0.00657*** [0.00242] 0.00185*** [0.000702]
Dummies
Adv
Adv X Newsp Newsp X Month Adv X Month 24624 0.628
N 24624 24624 24624 24624 24624 R2 0.013 0.175 0.195 0.383 0.401 Standard Errors in brackets. * p < 0.10, ** p < 0.05, *** p < 0.01 Number of articles as dependent variable Total Ads is the advertising spending of the same advertiser in other newspapers
Newspapers
29
As a rst indication, we regress coverage on advertising. The coecient of advertising is statistically dierent from zero. The magnitude is surpris1 ing. An additional 54000 euro (=1000 0.0186 ) advertising expenditure generates one additional article in the same month. Note that only a very small part of the variation is explained (R2 =0.013). Again, this is just a correlation between coverage and advertising spending and all above confounding eects are also at play. To account for these eects, we include in each specication more and more controls. For example moving from the rst to the second column, we include advertising dummies to take into account the above decribed advertiser specic eects. As expected the coecient on advertising spending goes down, but remains positive and signicant. From columns (4) and (5), it can be seen that controlling for advertiser-newspaper eects, has a substantial impact on our estimate for the impact of advertising on coverage. Finally, we also want to control for the advertiser specic changes in coverage over time. First, we include advertising spending in other newspapers by the particular advertiser as a control variable and second we include interactions between time and advertiser dummies in the regression. The second strategy is the most robust one, but has the disadvantage of slicing away much potentially useful variation in both coverage and advertising. The rst strategy is computationally less intensive and preserves more variation in the dependent and independent variable but rests on the assumption that changes in newsworthiness of advertisers over time can be represented by a linear function in advertising spending in other newspapers. Even with these controls, the eect of advertising on coverage remains statistically signicant and positive as displayed in columns (6) and (7). The coecients imply that respectively an extra 152000 or 212000 euro of advertising spending will generate an additional article in the newspaper. To put these results in perspective, remember that the average (nonzero) advertiser spends 28858 euro in one newspaper in one month. This corresponds with less than one seventh of 212000 euro, the additional spending necessary to generate one additional article. As such, the average advertiser has a 14% chance of being mentioned. The average full page black and white advertisement cost 14450 euro (in the sample period). Stated dierently, an advertiser has to buy over 14 full page black and white advertisements to receive one additional article. Whether this is a good deal for newspapers or advertisers, depends on the reputation loss for newspapers (think of the reader averseness 1 in our theoretical model)
30
Newspapers
and the eectiveness of bias (1 in our theoretical model). Based on these results, we nd that if an advertisers stops advertising, his coverage drops on average from 1.58 to between 1.05 and 1.45 articles per month per newspaper20 . These results also allow to determine the total amount of newspaper articles that are due to advertiser bias. The average amount of advertising money spend in newspaper between 2001 and 2005 is 240 million per year. This means that, on average, 1132 up to 4464 articles per year are related to advertiser bias. There are many caveats with this simple extrapolation from our sample to the total population of advertisers, though these results give at least an indication of the order of magnitude. The nding of advertising bias in general newspapers is remarkable. Reuter & Zitzewitz (2006) document that the coverage of nancial magazines is aected by mutual funds advertisement, but general newspapers (New York Times and Wall Street Journal) remain unaected. Gambaro & Puglisi (2009) nd a positive eect of advertising expenditure on the daily coverage in general newspapers, but the eect disappears when newspapers times company xed eects are included. It is not so straightforward to compare the magnitude of our coecients with previous studies, since advertisement prices also depend on the readership; some studies are also in logs. The results of Reuter & Zitzewitz (2006) show that spending 1 million dollar on advertising increases the probability of a mention with 0.1 to 0.2 percent. Stated dierently, of all mutual funds mentions roughly 8 up to 23 percent are due to advertiser bias. Rinallo & Basuroy (2009) measure bias in pages and nd that advertising one percentage more pages generates 0.4 percent more articles in fashion magazines. Robustness Checks We carry out a number of robustness checks in table 2.3. We exclude the rst eight months, because the aggregate data show a peak in coverage in these months (see Appendix B, graph B2). Next, we exclude the business newspaper De Tijd because this newspaper reports a lot more on companies. We also exclude the high coverage advertisers and nally include advertising squared. Our results are unaltered. The quadratic function seems to better t the data, though the coecient for the quadratic term is rather small. Furthermore, we also put adver20 The dierence is the average amount spent, divided by 1000 and multiplied by the coecient of advertising, e.g. 28558 0.00472 = 0.135 for specication (7). 1000
31
Table 2.3: Robustness Checks

(1) Advertising Total Ads Advertising2 000532*** [0.00196] 0.00129** [0.000529] (2) 0.00547*** [0.00203] 0.00189** [0.000770] (3) 0.00286*** [00.000862 0.000569 [0.000393] (4) 0.0157*** [0.00464] 0.00175** [0.000750] -0.0000219* [0.00000642] Adv X Newsp Newsp X Month 21546 0.369
Dummies
Adv X Newsp Newsp X Month 20976 0.406
N R2
Standard Errors in brackets. *p<0.10, ** p<0.05, *** p < 0.01 Number of articles as dependent variable Total Ads is the advertising spending of the same advertiser in other newspapers Model (1) First eight months excluded Model (2) De Tijd excluded Model (3) Observations for which coverage > 10 excluded Model (4) Advertising squared included
tising in logs. Since most of our observations included zero advertisement expenditure, more than 90 percent of our observations is dropped. We still obtain a positive eect but signicance decreases to a 95% condence interval or disappears in the specication with newspaper times month interaction dummies. These results are represented in Appendix B, table B1.
2.3.4
Concerns
Killing negative news In our empirical identication, we did not distinguish between good and bad news. We adopted the assumption that any news is good news. In section 2.5, we provide some anecdotal evidence on newspapers and magazines suppressing negative news on advertisers. This might aect our coverage variable. We refrained from distinguishing positive and negative news because it is hard to nd an objective measure. Some authors have attempted to separate positive from negative news; though only in specic contexts. Reuter & Zitzewitz (2006) investigate mutual funds and classify recommendations as positive, dissuasions as negative. Gambaro & Puglisi (2009) analyse changes in stock prices news following an increase is considered as positive, and vice versa. Gurun & Butler (2010) count the number of negative words used in articles about companies. Lott Jr. & Hassett (2004) look at the coverage of economic news. Based on the economic variable discussed in the title of a news article, and the adjacent verb (e.g. increase or decrease) or adjective (e.g.
32
Newspapers
strong or weak), they coded the news. A similar approach is followed by Doms & Morin (2004). These two latter papers are more sophisticated versions of the R-word index used by the Economist21 . Hamilton (2004) uses the DICTION software to analyse hard vs. soft news. This software analyses word use and word length. For general news on companies, it is much harder to reach agreement how to classify articles. Content analysis might prove helpful (see Fico, Lacy & Rie (2008) for an overview), but it is very hard to set up an automated procedure. Wood, Nelson, Cho & Yaros (2004) manually coded the appearance of brands or companies in television news or news shows as negative, neutral and positive. They claim a high percentage match amongst independent coders. Upshaw, Chernov & Koranda (2007) conduct a similar analysis. Williams (2009) uses a combination of automated evaluation and coding by humans in a case study evaluating the tone of the news on a nancial institution during the nancial crisis. If suppressing negative news is an important force, then our test is inconclusive about advertising bias. A statistically positive coecient can then be explained by more a higher positive coverage (i.e., more advertising bias) or by less killing of stories (i.e., less advertising bias). However, previous empirical studies that distinguish between positive and negative news, nd that advertising has an inuence on positive news but not on negative news (Reuter & Zitzewitz (2006) and Gambaro & Puglisi (2009)). Timing In this article, we suggest that advertising generates additional coverage. One might argue that the causality can also be the other way around. It might be the case that a newspaper acts over-friendly towards some companies in order to attract them as potential advertisers. To test for this, we perform a Granger causality test. We regress coverage on lagged ads and lagged coverage; and we also regress ads on lagged coverage and lagged ads. If the coecient of the lagged ads are jointly signicant in the rst specication while the coecients of lagged coverage arent in the second, then we can conclude that advertising Granger causes coverage but not the other way around. The results in table 2.4 indeed conrms our hypothesis. We also ran regressions with lags, showing a similar picture (see tables
21 The R-word index is a simple count of the number of times the word recession appears in newspapers. It was proposed in 1992 as an alternative indicator of economic activity. (Rrrrrrrecession?, The Economist, 16-07-1998.
33
Table 2.4: Granger Causality Test

Coveraget Coveraget1 0.359*** [0.00663] 0.138*** [0.00691] 0.132*** [0.00682] 0.102*** [0.00628] 0.00188 [0.00148] 0.00381** [0.00172] -0.00176 [0.00173] 0.00145 [0.00156] 22800 0.413 10.58 0.000 Advertisingt -0.0476 [0.0300] 0.0110 [0.0313] 0.0105 [0.0308] 0.0397 [0.0284] 0.564*** [0.00668] 0.218*** [0.00777] 0.0972*** [0.00781] 0.103*** [0.00704] 22800 0.746 16429.05 0.000 1.14 0.3373
Coveraget2
Coveraget3
Coveraget4
Advertisingt1
Advertisingt2
Advertisingt3
Advertisingt4 N R2 Joint signicance (F) Lagged Ads (Prob > F)
Joint signicance (F) 3859.97 Lagged Coverage (Prob > F) 0.000 Standard errors in brackets * p < 0.10, ** p < 0.05, *** p < 0.01
34
Newspapers
Table 2.5: Coverage explained by advertising of last 12 months

(1) Ads (lag year) 0.0237*** [0.00412] (2) 0.00925*** [0.00158] (3) 0.00795*** [0.00167] 0.000867 [0.000686] Adv Adv X Newsp Newsp X Month 19152 0.404 Adv X Month (4) 0.0202*** [0.00476]
Total Ads (lag year)
Dummies
N 19152 19152 R2 0.019 0.404 Standard errors in brackets Number of articles as dependent variable * p < 0.10, ** p < 0.05, *** p < 0.01
19152 0.462
B2 and B3 in Appendix B). In the specication with lagged advertising on coverage, coecients are still signicantly (at least jointly) dierent from zero. In the reverse regression, the signicance disappears. This can also be interpreted as weak evidence that advertising inuences coverage, but not the other way around. Reuter & Zitzewitz (2006) show that the reverse eect (an increase in coverage followed by an increase in advertisement) is much stronger for those companies that already advertise. This also suggests a causal advertisement-coverage relation. Targeted Advertising It might be that our results do not pick up only bias, but also targeted advertising. Suppose that a certain rm releases a new product. If the readers of a particular newspapers are very interested in this type of product, it is not illogical that the rm will advertise in this newspaper (and not in others); and that the newspaper will report on the product (while others dont). Our specications, that exploit the dierences between newspapers and dierences over time, will pick up this eect as bias, while it is not. However, note that in our most demanding specication, we include not only interaction between advertiser and newspaper dummies but also between advertiser and month dummies. So we control for the average targeted advertising in a certain newspaper as well as the dierent advertising intensities by an advertiser over time. Although we control to a large extent for the targeted advertising problem, there could still be an eect present, i.e. advertisement and coverage increase due to a product launch (see above). To circumvent this problem, we also run a regression with all the lagged advertisement expenditure in
2.4 Survey of Belgian Journalists
35
the specic newspaper over the last year. The coecients, shown in table 2.5, remain signicant throughout these specications.
2.4
Survey of Belgian Journalists
Conditional on the fact that our sample is representative for other Belgian advertisers, we nd that advertiser bias counts for 1000 up to 5000 additional articles per year. An important question is how these articles end up in a newspaper and how the practice of advertiser bias is perceived by journalists. In this section, we present a survey conducted among journalists working for Dutch-language Belgian newspapers. First we provide an overview of other surveys on journalists, editors and related parties. Then we describe the methodology and the most important results.
2.4.1
Related Literature
Given the potential impact of mass media on society, the people working in this sector have been repeatedly surveyed. Some of these surveys question the perceived inuence of advertisers on coverage. Not surprisingly, the perceived advertising bias strongly depends on what you ask and whom you ask. Outside parties, such as marketeers and the audience, are more concerned about advertiser pressure than inside parties, such as journalists and editors. In this section, we present an overview of the existing survey results on advertiser inuence and news. Results are dicult to compare, because the methodology diers in nearly every study. Besides, media actors interviewed work in dierent countries and in dierent media branches. A summary of survey results can be found in Appendix C. Journalists Journalists are at the heart of the news publishing process and are often surveyed on their views of their own profession. All studies presented here, except for Price (2003), question the advertising bias issue as a part of a longer questionnaire. One remarkable conclusion is that the percentage of journalists that perceives advertiser inuence varies between a mere 7% and a worrisome 85%. Note however that this last result might be biased upward because the survey questioned whether there was inuence from advertisers, owners and marketeers. If we look at results of questions asking for advertiser inuence only, then the percentage varies between 7% and 38%. This means that a majority of journalists is not
36
Newspapers
pressured or inuenced by advertisers. However, this does not rule out advertising bias since on average 25% feels the heat from the advertisers. Editors A similar variance in answers is found when we look at studies that question editors. Pew (2008) nds that only 15% of newspaper and television editors perceive that advertiser concerns inuence news coverage decisions, while in Soley & Craig (1992) 90% of the newspaper editors report that advertisers try to inuence the news. Interestingly, when asked whether their news organization has engaged in a new revenue experiment that has raised concerns about editorial independence, 36% of journalism executives answers yes (Pew 2010). Detailed answers included ad sponsorships of specic content and blurring the lines between ads and news. These large dierences might be explained by the way the questions are asked. Soley & Craig (1992) ask whether they know of one case where an advertiser tried to inuence while the Pew studies (Pew (2000), Pew (2004) and Pew (2008)) and ASNE (1999) ask whether advertiser inuence exists in general. Other Parties There are only a limited number of studies on other parties involved in advertising bias, such as marketing and sales inside and outside the newspaper, readers and owners. Managers and owners consider the amount of advertiser inuence as very limited (10 to 16 %) while half of the readers in a study by ASNE (1999) think that newspapers allow advertisers interests to inuence the news. An & Bergen (2007) question advertising directors at US newspapers. One third nds it acceptable that a salesperson asks to favor advertisers as news source, while more than 90% nd it unacceptable to suppress a negative story because an advertiser asked it. Roughly 60% of marketing experts and PR people believe that Polish newspapers insert advertisement bias (Tsetsura 2005). The overview provided above showed that dierences on advertising bias perception are large, but one should be cautious to compare these studies, given the dierences in methodology, professions, countries and media sectors. In general, some respondents perceive inuence of advertisers on editorial decisions but this concerns only a minority of the respondents.
37
Age 20-30 31-40 41-50 50+
Table 2.6: Summaries Survey Number 91 21 19 23 28 90 26 14 50 63 26 21 16 89 23 11 25 30
Percentage 100% 23.1% 20.8% 25.2% 30.8% 100% 28.9% 15.6% 55.6% 100% 41.3% 33.3% 25.4% 100% 25.8% 12.4% 28.1% 33.7%
Experience 0-5 years 6-10 years +10 years Media Group Persgroep Corelio Concentra Editorial Oce Politics/Domestic/Foreign Business Editor (in chief / graphical) Other
2.4.2
Methodology
In order to measure advertisers pressure as perceived by journalists, we conducted an anonymous survey among journalists working for Dutch language newspapers in Belgium. To our knowledge, there is only one study that questions Belgian journalists on advertiser pressure. Ppress (2010) reports that 85% of the journalists think that Objectivity of journalism is under increasing inuence from advertisers, marketeers and owners. Their study is not comparable to ours, since this is the only question on advertiser pressure and since the respondents are magazine writers, compared to newspaper journalists in ours. The survey was created on a secured online web page. The invitation
38
Newspapers
to participate in the survey with the link to the website was sent by email to the population of around 750 newspaper journalists working for Dutch language newspapers in Belgium. We retrieved the e-mail addresses from the Association of Belgian Journalists (VVJ), which also encouraged to its members to participate in the survey. Note that only contractual journalists were contacted, so freelancers were excluded. In total, 131 journalists lled out the survey, which was reduced to 101 after dropping the respondents that spent less than 2 minutes on the survey. This implies a response rate of around 13.5%. In the introduction to the survey it was clearly mentioned that the survey was about the relation between advertisers and newspapers. Table 2.6 displays some summary characteristics of the respondents. The journalists included in the survey are relatively equally spread out over the dierent age groups, although there are slighltly more journalists older than 50 years. Concerning the experience at their current employer, the majority has been working for over 10 years at the same newspaper and 30% of the respondents has less than 5 years experience. In accordance with their market shares, most journalists are working for Persgroep, followed by Corelio and Concentra, although a substantial number of journalists opted not to respond to this question. Finally, we grouped the journalists according the editorial oce they are working for. Around one fourth of the respondents writes about politics or brings more general domestic or foreign news and 12% works for the business (economics) editorial oce. Around 30% works as editor (in chief) or as graphical editor. The rest (33%) writes regional, lifestyle, sports or other news.
2.4.3
Results
The rst part of the questionnaire asked the respondents after their general opinion about the inuence of advertisers on newspaper content. The respondents were presented several statements which they had to judge on a 5-point Likert scale, namely 1 (Strongly disagree) 2 (Disagree) 3 (Neither Agree nor Disagree) 4 (Agree) 5 (Strongly Agree). The responses to a selection of these questions are reported in gure 2.1, where for expositional purposes we have pooled Strongly (dis)Agree and (dis)Agree together. First we asked whether advertisers received as much attention in the newspaper as non-advertisers. Over 60% of the journalists thinks that there exists no advertising bias as advertisers are mentioned as much as non-advertisers, while 20.8% (strongly) disagrees with the statement.
39
Figure 2.1: Questions on Relationship Advertisers Newspapers Almost one fourth of journalists agree with the statement that advertisers are trying to exercise some inuence on the content of the newspapers, although only 12% of them believes they succeed in doing so. According to results, not reported here for brevitys sake, over 90% of the journalists strongly dislikes advertising bias as they state it would be inappropriate or detrimental for newspaper quality when advertisers manage to inuence the editorial content. Next, the correspondents were asked directly how often they are approached by somebody, either the editor in chief, direct editor, marketing department or advertiser, to write an article about an advertiser. Again, the responses were measured on a ve-point scale, namely 1 (Very Often) 2 (Often) 3 (Sometimes) 4 (Seldom) 5 (Never)22 . The results are reported in gure 2.2. Pressure to write an article about an advertiser comes mostly from the marketing department. Over 40% of the journalist are at least seldom contacted by the marketing department to write an article about
22 In an explanatory note, we indicated that Very Often means approximately on a daily basis, Often means approximately once a week, Sometimes means approximately once a month and Seldom means approximately once a year.
40
Newspapers
an advertiser and for over 20% this happens more or less frequently (at least once a month). Around 35% journalists approached by the marketing department say they never follow the advice to write about an advertiser (which leaves of course 65% of the journalists following at least seldom this advice)23 . Also advertisers themselves try to put some pressure on journalists as 20% of the journalists report that they are at least sometimes contacted by rms to write an article about them. However, 60% of the journalists say they never follow this suggestion. Suggestions to write an article about an advertiser by the editors occur much less frequently, but if they occur, they are, not surprisingly, more often followed. When we combine the dierent channels through which the advertisers can inuence the editorial content, we nd that 13% of the journalists gets at least often approached by at least one of the four dierent agents (the editor in chief, direct editor, marketing department or advertiser) to write an article. 36 % gets at least sometimes approached and almost 60% gets at least seldom approached. These numbers seem to indicate that there is some advertisers pressure, although limited to a subset of journalists and only a part of them gets approached regularly. We perform this exercise for dierent groups seperately and report the results in the rst three columns of Table 2.7. First we look at whether there exist dierences in advertiser pressure across dierent experience categories. Although it is dicult to do draw rm conclusions due to the low number of observations, it appears that, in line with our priors, less experienced workers get approached more often compared to more experienced workers. Splitting the sample by the editorial oce the journalist is working for, displays no clear picture, Looking at the percentage of journalists that are approached at least sometimes, it appears that mostly journalists working at Other editorial oces are more prone to experience some advertisers pressure. The reason that we do not nd a clear eect could be the relatively low number of observations which obliges us to divide the sample into coarsely dened subgroups. Finally, we investigate whether there exists dierences across the dierent newspaper groups in Belgium. Again, the results should be interpreted with caution, but advertiser pressure seems somewhat larger for the Concentra newspaper group.
23 These results come from a second question where we asked whether the journalists followed the advice. The full results are not reported in the paper, but are available on request.
41
Figure 2.2: Suggestion to Write about Advertiser
Advertising bias can not only appear through the simple writing of an article about the rm in question, but also through the adjustment of articles about the advertiser, bringing it more in line with advertisers expectations and wishes. We asked our respondents about the occurrence of this kind of pressure and results are reported in Figure 2.3. Clearly, the prevalence of pressure to adjust an article about an advertiser is lower than the pressure to write an article about an advertiser. Only 10% of the journalists report that they receive sometimes or often the suggestion from the marketing department to rewrite an article. Similar gures are achieved for advertisers, editors in chief and direct editors. This low gure can also be due to the fact that the marketing department, and certainly advertisers, normally do not preview the articles before they are published. Computing the percentage of journalists that is approached by at least one of the four dierent agents, shows 5.2% is at least often approached, 17.5% is at least sometimes approached and 34% is at least seldom approached to adjust the content of an article about an advertiser. When we split up the sample by dierent categories of journalists, we obtain similar results as for the part where we asked whether they were approached or not. Besides these quantitative questions, we also asked qualitative ques-
42
Newspapers
Figure 2.3: Suggestion to Adjust Article about Advertiser tions where journalists could comment or give examples. These comments and examples are incorporated in the next section. To conclude one could say that apparently there is some pressure of advertisers, either directly or indirectly, to steer the content of newspapers. However, only a limited number of journalists is aected. Unfortunately, our sample of respondents is too small to split up the sample by the dierent characteristics of the journalists and perform regressions. The nding that only a small minority is often approached, does not conict with the results found in section 2.3. We found that advertiser bias accounts for roughly 1000 up to 5000 additional articles per year. Given that Flanders counts 755 newspaper journalists (freelancers not included) and that the average journalist in our sample writes 10 articles per week, it is not illogical that journalists are approached only seldom, and that many journalists are not approached at all. Moreover, advertiser bias might be implemented without the journalists consent and therefore not be perceived as bias. In the next chapter, we shed light on how this bias might be implemented. Moreover, note that the survey measures the perception of journalists about advertising bias, which is not necessarily equal to actual pressure of
43
advertisers to inuence newspaper content. As is typically the case with surveys, respondents might not have the right incentives to reveal the truth. In this survey, one might argue that journalists have no incentive to reveal advertisers pressure or bias, since this would hurt the reputation of their profession. Moreover, the - legally not binding - Belgian Code of the Council of Journalism (Raad voor de Journalistiek 2010) stipulates in article 11 that the journalist does not engage in advertising or propaganda, and does not give in to pressure from advertisers or stakeholders of information. However, we hope to have mitigated these eects by making our survey anonymous.
44
Table 2.7: Survey Results per Group

Often 12.9% 23% 14% 10% 17% 9% 12% 17% 19% 0% 13% 27% 33% 50% 54% 52% 75% 8% 0% 13% 35% 18% 28% 47% 56% 55% 64% 57% 0% 9% 4% 10% 17% 18% 13% 25% 15% 10% 27% 54% 36% 24% 65% 64% 54% 4% 7% 6% 19% 21% 16% 42% 21% 36% 30% 36% 28% 40% 31% 14% 56% Approached Sometimes Seldom 35.6% 59.4% Often 5.2% Asked to change Sometimes Seldom 17.5% 33.7%
Total (N=101)
Experience (N=90) 0-5 years(N=26) 6-10 years (N=14) +10 years (N=50)
Editorial (N=89) Politics/Dom./For. (N=23) Business (N=11) Editor (chief / graph.) (N=25) Other (N=30)
Media Group (N=63) Persgroep (N=26) Corelio (N=21) Concentra (N=16)
Newspapers
The numbers in this table are the percentage of journalists that are approached at least often, sometimes or seldom by advertisers, marketing department, direct or indirect chief.
2.5 Advertising Bias Channels
45
2.5
Advertising Bias Channels
In the previous sections, we showed that it is likely that the bundle of advertisements and coverage exists. It remains an open question however, how this bundle might be implemented. Based on our survey and anecdotal evidence worldwide, we try to uncover these channels. We determine four channels. We label the rst two as supply and demand channels, because they uncover how the bundle might be oered (supply) or asked (demand). We complement them with inside channels which explain how the journalistic ethics might be circumvented inside the news rm. We discuss these channels only briey here, a more elaborate version with examples can be found in Appendix D.
2.5.1
Supply and demand channels
In this subsection, we discuss how bundles of advertisements and coverage might realize, by investigating the supply (i.e. the newspaper) and the demand (i.e. the advertisers).
Supply channel The clearest way how the advertising bias is implemented, is when the media outlet oers news for sale. This can be very direct, i.e. by selling news space directly; or more indirectly by oering a bundle of coverage and advertisement. This bundled coverage might contain additional news about the advertiser, or more in general positive news that encourages consumption.
Demand channel It might also be the case that the bundle is not offered by the media outlet, but exacted by the advertiser. Advertisers might explicitly negotiate such a bundle; or withhold future advertisements if it is not supplemented by favorable coverage. Some advertisers have successfully used withdrawal threats, especially to suppress negative news. Other apply a withdrawal rule, i.e. they automatically withdraw their advertisements if certain content is published. It goes without saying that media outlets are not bound by such rules, though they are nancially punished if they publish on sensitive topics.
46
Newspapers
2.5.2
Inside channels
While media outlets might oer bundles, or advertisers might command them, in many media rms, there is a strict wall of separation between the business side and the editorial sta. If the editorial sta can freely choose the topics they write about, then the scope for advertising bias is limited. That is exactly why such a wall of separation was constructed. There are a number of channels though, that allow the business side to circumvent this wall. We summarize them as crossing the wall and editorial sta composition. Crossing the wall If journalists dont want to implement this bundle, then the business side might do it themselves. In many newspapers, the business side engages in journalism, especially for the lighter sections such as lifestyle, travel, real estate, cars and food. But the business side might also invite journalists to cross the wall, by having them perform in advertisements, or organizing meetings between advertisers and journalists. Editorial sta composition If some journalists dont want to engage in advertising bias, then media outlets might hire simply other journalists. The most visible composition changes are lay-os. A more subtle mechanism is dedicating more journalists to the elds that are interesting to advertisers.
2.6
Conclusion
Many scholars have argued that there should be a rm wall between the advertisement department and the editorial sta of a newspaper. The advertising department should not exert pressure on the editorial sta to shape articles. And if they do the professional ethics of journalists should prevent any direct or indirect inuence of advertisers on the selection of news. The analogy with a church-state separation of journalism and commerce is never far away. Newspaper owners and editors might nevertheless have strong incentives to tear down the wall between the advertisement department and the editorial sta. Our theoretical model shows that it is protable for both monopoly and duopoly newspapers to oer a bundle of bias and advertisement if the eectiveness of bias is relatively high for advertisers and if
2.6 Conclusion
47
the annoyance of bias is relatively low for readers. If duopoly newspapers are closer substitutes for readers, then newspapers set less bias; while the eect on advertisements is ambiguous. But, as Davies (2008) argues, to the outsiders eye, [advertising bias] is very tempting. Advertisers have money, the media outlets need money, so they must be vulnerable to some kind of pressure from the advertisers. Its a ne theory, but its truth is very limited. Whether bias will prevail is indeed ultimately an empirical question. We have tried to answer this empirical question for the Belgian Dutchlanguage newspapers. We investigated to which extent advertisements aect the coverage of companies in these newspapers in the period 20012005. Depending on the specication, we nd that spending 54000 to 212000 euro in a newspaper in one month generates on average one additional article in this newspaper in the same month. Our results appear robust to alternative specications. Our results are unsurprising and surprising at the same time. On the one hand, newspapers are prot maximizing businesses. And as our theoretical model predicts, pleasing advertisers might generate higher prots. Therefore, it is not illogical that newspapers insert bias. But on the other hand, newspapers are no normal business. Correctness of information generates a lot of externalities on society. Many decisions, be it voting, consumption or investment, are based on information retrieved from newspapers. Therefore, newspaper owners and editors should not allow that advertisers or the advertisement department attempt to inuence editorial content.
Policy proposals If corporate and society interests on bias do not align, there might be some solutions to root out advertising bias. A rst and drastic solution is to eliminate all advertising. Without advertising, advertising bias becomes useless and will not be implemented. Though it is doubtful whether a complete ban on advertising is desirable. First, as argued by Baker (1994), less advertising also means less media outlets, which hurts dierentiation. Second, the absence of advertising revenues opens the door for inuence by other revenue sources, such as political inuence. Petrova (2009) shows for American newspapers in the 19th century that newspapers were more likely to be independent from political parties if advertising revenues were high. Third, the experience of the
48
Newspapers
womens magazine Ms. learns that a ban on advertisements24 also indirectly resulted in a lower audience. Two of the main reasons were that Ms. had to give up full color and reduce the frequency to a bi-monthly publication (Cunningham & Haley 2000)25 . Baker (1994) oers two interesting proposals to reduce the reliance on advertisers. A rst is his TA-SR proposal, which stands for Tax Advertising / Subsidize Readers. The aim is to redistribute advertising revenues according to the audience size. This tilts the trade-o of inserting advertising bias more in the direction of the readers26 . Another proposal is to randomize advertising over programs or content, such that advertisers cannot know to which content their advertisement is adjacent. This proposal is most relevant for television. This randomization already happens to some extent in newspapers, magazines and internet websites. Another solution is to install a government body that watches over the general interest. However, determining whether an article crosses the line between bias and news is not always easy to do. Note that we have only tried to estimate bias in general and not in specic articles. It is very dicult to determine which articles about advertisers are genuinely newsworthy, and which constitute bias. Moreover, a government might have its own incentives to inuence news. A last solution is to make the reader aware of the existence of advertising bias. Many books (e.g. Seldes (1943), Bagdikian (2004), Baker (1994)), citizen activists (e.g. Fairness and Accuracy in Reporting27 ) and scientic articles (e.g. Reuter & Zitzewitz (2006), Rinallo & Basuroy (2009) and other references in this article) revealed advertising bias. This increased awareness might reduce the eectiveness of bias and render it unprotable. This article is also a step in this direction.
24 For the rationale of this ban, see the editorial by Gloria Steinem in Ms. Magazine, July/August 1990. 25 Interestingly, Kaiser & Song (2009) show that German magazine readers do not dislike ads. Most readers like ads, especially in magazines where ads are informative, such as TV and women magazines. This runs contra the common nding that news consumers are adverse. 26 Some governments have already implicitly implemented such a scheme. The Belgian government subsidizes the reader side by giving a tax exemption for newspapers (VAT 0%), and applying a very advantageous postal tari (Cochez 2010). This is no pure TA-SR scheme as Baker (1994) proposed since this subsidization is not paid by a tax on advertising. 27 See www.fair.org
A Monopoly prots
49
Appendix A Monopoly prots

=
The monopoly prots corresponding with the three regimes in proposition 4.3.5, are:
2 (z322 )(622 +z)(42 3+z) 2 2162 (1 (22 3)22 1 )3 2161 1 (1 2 2 1 ) 91 161 2 9 62 + 42 .
if if if
1 2 1 (
<
1 (
2 12 +2 )
(2 1)2
2 12 +2 )
(2 1)2 21 1 > 1+2 2 2
1 2
21 1+22
with z =
50
Newspapers
Figures & Tables
Figure B1: Total Advertising Expenditure over Time
Figure B2: Total Coverage over Time
Table B1: Coverage explained by logged advertising and dummies

(3) 0.328** [0.160] 0.115** [0.0539] Adv,Newsp 1978 0.373 1978 0.410 1978 0.502 1978 0.610 Adv, Newsp Month Adv X Newsp Month Adv X Newsp Newsp X Month Adv X Newsp Newsp X Month 1693 0.635 0.362** [0.153] 0.319* [0.181] 0.177 [0.207] 0.0392 [0.237] (4) (5) (6) (7)
(1)
(2)
ln(Advertising)
0.466** [0.204]
0.363** [0.144]
B Figures & Tables
ln (Total Ads)
Dummies
Adv
N R2
1978 0.010
1978 0.353
Standard errors in brackets * p < 0.10, ** p < 0.05, *** p < 0.01
51
52
Newspapers
Table B2: Coverage explained by lagged advertising

(1) Advertisingt1 0.00501* [0.00294] 0.00615** [0.00254] 0.00250 [0.00208] 0.00812*** [0.00284] (2) 0.00421 [0.00327] 0.00549** [0.00248] 0.00153 [0.00198] 0.00663** [0.00256] Adv (3) 0.00249 [0.00341] 0.00357 [0.00279] -0.000735 [0.00181] 0.00391* [0.00229] Adv X Newsp Newsp X Month 22800 0.402 11.45 0.000 (4) 0.00314 [0.00196] 0.00666*** [0.00202] 0.00184 [0.00214] 0.00678*** [0.00234] Adv X Month
Advertisingt2
Advertisingt3
Advertisingt4 Dummies
N R2 Joint significance (F) Lagged Ads (Prob > F)
22800 0.016 15.99 0.000
22800 0.192 10.26 0.000
22800 0.432 11.30 0.000
Standard errors in brackets Number of articles as dependent variable * p < 0.10, ** p < 0.05, *** p < 0.01
Table B3: Advertising explained by lagged coverage

(1) Coveraget1 0.387* [0.214] 0.261 [0.163] 0.134 [0.109] 0.205* [0.108] (2) 0.344** [0.169] 0.225* [0.126] 0.137 [0.100] 0.218** [0.0978] Adv (3) 0.169 [0.112] 0.116 [0.0970] 0.0654 [0.0792] 0.141* [0.0727] Adv X Newsp Newsp X Month 22800 0.434 1.48 0.2083 (4) 0.353* [0.198] 0.258* [0.147] 0.165 [0.108] 0.216** [0.106] Adv X Month
Coveraget2
Coveraget3
Coveraget4 Dummies
N R2 Joint significance (F) Lagged Coverage (Prob > F)
22800 0.015 1.63 0.1666
22800 0.140 2.15 0.0738
22800 0.303 1.66 0.1585
Standard errors in brackets Advertising expenditure as dependent variable * p < 0.10, ** p < 0.05, *** p < 0.01
C
Table C1: Overview Surveys on Inuence Advertisers Pro Neutral Contra Research Year Country N
Overview Surveys on Inuence Advertisers
Proposition
Editors na 22% 0% 78% Pew (2000) 1999 US na 75% ASNE (1999) 1997 US 50 162
25%
6%
71%
Pew (2004)
2004
US
92
15%
1%
84%
Pew (2008)
2007
US
156
C Overview Surveys on Inuence Advertisers
15%
na
na
Soley & Craig (1992)
1991
US
147
37%
na
na
1991
US
147
55%
na
na
1991
US
147
71% 83% na
na
na na
Soley & Craig (1992) Soley & Craig (1992)
1991 1991
US US
147 147
Allow advertisers interests to inuence the news In your opinion, to what extent do advertising concerns inuence news organizations decisions about which stories to cover or emphasize? Have there been instances in which your newsroom was encouraged to do a story because it related to an owner, advertiser, or sponsor? In your opinion, to what extent do advertising concerns inuence news organizations decisions about which stories to cover or emphasize? Our newspaper seldom runs stories which our advertisers would nd critical or harmful. Has any advertiser succeeded in inuencing the news or features in your newspaper? Has there been pressure from within your paper to write or tailor news stories to please advertisers? Has any advertiser tried to kill a story at your newspaper? Has any advertiser threatened to withdraw advertisng from your paper
53
54
88%
na
na
1991
US
147
89%
na
na
1991
US
147
because of the content of the stories? Are you aware of an attempt by any advertiser to inuence what news or features apeared in your newspaper? Has any advertiser ever withdrawn advertising in response to the content of your newspaper? Has any advertiser tried to inuence the content of a news or feature story? 90% na na Soley & Craig (1992) 1991 US 147
Editors & Journalists 7% na na Williams (1992) 1990 US 41
32%
na
na
Williams (1992)
1990
US
41
44%
na
na
Williams (1992)
1990
US
41
Have you ever been pressured by a senior editor or their publisher to delay or kill a story a real estate advertiser might nd oensive? Do you know of one case where advertisers actually pulled ads because of a story? Our real estate sections do not include coverage that would oend a real estate advertiser as a matter of policy. Did real estate advertisers ever threaten to pull ads from the papers because of the coverage? 83% na na Williams (1992) 1990 US
41
Journalists 10% 55% 11% 33% 25% 65% ASNE (1999) Hollings et al (2007) 1997 2007 US New Zealand 1714 391
Allow advertisers interests to inuence the news Newsrooms have been pressured to do a story because it related to an advertiser, owner or sponsor. In your opinion, to what extent do advertising concerns inuence news organizations decisions about which stories to cover or emphasize? 28% 2% 70%
Newspapers
Pew (2000)
1999
US
228
36%
0%
64%
Pew (2000)
1999
US
124
26%
7%
67%
Pew (2004)
2004
US
96
35%
8%
57%
Pew (2004)
2004
US
45
34%
2%
65%
Pew (2008)
2007
US
232
85%
6%
10%
Ppress (2010)
2009
Belgium
140
7% 7% 38% 49% 13% na 93% Price (2003)
na
93%
Price (2003)
1999 1999 2005
US US Poland
131 131 99
(Traditional Media) In your opinion, to what extent do advertising concerns inuence news organizations decisions about which stories to cover or emphasize? (Internet) Have there been instances in which your newsroom was encouraged to do a story because it related to an owner, advertiser, or sponsor? (Traditional Media) Have there been instances in which your newsroom was encouraged to do a story because it related to an owner, advertiser, or sponsor? (Internet) In your opinion, to what extent do advertising concerns inuence news organizations decisions about which stories to cover or emphasize? Objectivity of journalism is under increasing inuence from advertisers, marketeers and owners Have you felt pressured to report a story because of advertisers Have you felt pressured not to report a story because of advertisers National media publish materials about a company or a product in exchange for a paid advertisement appearing elsewhere in the same medium. Tsetsura (2005)
Marketing & Sales (intern) 3% 6% 91% An & Bergen (2007) 2010 US 211
It is acceptable that a negative story on an advertiser is suppressed on the request of this advertiser. It is acceptable that a feature story on a restaurant is written by the 14% 24%
62%
An & Bergen (2007)
2009
US
211
55
56
23%
21%
56%
An & Bergen (2007)
2008
US
210
restaurant, which is also an advertiser in that section. It is acceptable that the advertising department encourages a photo editor to use photo with the logo of a shirt sponsor (and advertiser) prominently displayed. It is acceptable that a salesperson asks to favor advertisers as news source and avoid non-advertisers. 33% 12% 55% An & Bergen (2007) 2007 US 210
Marketing & Sales (extern) 58% 33% 9% Tsetsura (2005) 2005 Poland 98
National media publish materials about a company or a product in exchange for a paid advertisement appearing elsewhere in the same medium. (PR) National media publish materials about a company or a product in exchange for a paid advertisement appearing elsewhere in the same medium. (Marketing) 59% 32% 9% Tsetsura (2005) 2005
Poland
90
Readers 50% 32% 18% ASNE (1999) 1997 US 3000
Allow advertisers interests to inuence the news
Owners & Managers 14% 1% 85% Pew (2000) 1999 US 101
In your opinion, to what extent do advertising concerns inuence news organizations decisions about which stories to cover or emphasize? Have there been instances in which your newsroom was encouraged to do a story because it related to an owner, advertiser, or sponsor? 16% 7% 77%
Newspapers
Pew (2004)
2004
US
59
10%
0%
90%
Pew (2008)
2007
US
79
In your opinion, to what extent do advertising concerns inuence news organizations decisions about which stories to cover or emphasize? In the last two years, has your news organization tried or discussed a new revenue experiment that has raised concerns about editorial indepence or ethics? 36% 10% 53% Pew (2010) 2010 US 353
57
58
Newspapers
Advertising Bias Channels
In many theoretical models, the implementation of bias is fairly simple: the business side of the newspaper decides it and therefore it is implemented. In reality though, explicit money for bias schemes almost never occur, or at least not open, because this would reduce the reputation towards news consumers. Moreover, the business side does not make the news items themselves, therefore it is worthwhile to stress the role of the journalist. In this section, we expand the advertising bias channels that are discussed in section 2.5. We try to illustrate the channels with some examples, though advertising bias typically suers a tip-of-the-iceberg problem, where only the most visible cases come to the surface. As Fleetwood (1999) puts it, for every printed story that draws advertiser ire, there are thousands that never see the light of the day. Many cases, where the media outlet inserts self-censorship, are never known, perhaps not even by the journalists who implement it.
D1
Supply and demand channels
In this section, we discuss how the advertising bundle might look like in practice. Basically, this bundle of advertising and favorable coverage is an (implicit) agreement between the business side of the media outlet and the advertiser. Both parties can initiate this bundle. Supply channel The purest form of bias is where the media outlet sells news space to the advertiser. This practice was commonly accepted for US newspapers during the late 19th and beginning 20th century (Baker (1994), Baldasty (1992)). Companies wrote or commanded news articles or editorials, and these were published without any disclaimer. For readers, it was impossible to distinguish these so-called reader notices from other news, apart from its overoptimistic content. The practice steadily declined with the Post Oce Appropriation Act of 1912, prohibiting payment for news. Though milder forms of these reader notices still appear in the media. For example, in some news segments, Seattle television station Kiro TV featured only experts who had paid a fee of 1000 dollars to the station28 . Rush Limbaugh, a conservative American radio talk show host, allows companies to buy words that he then weaves into his monologues29 .
28 Seatle 29 New
Times, 07-06-2002, documented in Jackson, Hart & Coen (2003) York Times, 07-06-2008, documented in Hart (2009)
D Advertising Bias Channels
59
Similarly, AOL had an agreement with Innitys Radio to be mentioned at least six times a day in the programming30 . While these recent examples illustrate that selling news space directly still occurs, we focus on the bundle of selling an advertisement and favorable coverage. In a few cases, the news outlet makes this bundle explicit. Press-Enterprise announced in 2001 advertise your restaurant and get a free feature story31 . Richard Shortway, publisher of Vogue, said he cold, hard facts of magazine publishing mean that those who advertise get editorial coverage32 . This scheme was also applied in the Chicago Sun-Times, where theaters advertising get preferential treatment on the news pages33 . When other pill makers complained about the favorable treatise of Doctor Brandreth cure-all pills, James Bennett, editor of the New York Herald wrote in 1835 in his newspaper: Send us more advertisement than dr. Brandreth does [...] [and] well cut dr. Brandreth to dead34 . More often, and perhaps more importantly, advertising bias is an implicit bundle where the media outlet attracts or thanks advertisers by inserting favorable coverage, or by avoiding negative news. This selfcensorship by the media is proven by a quote of Helen Gurley Brown, editor of Cosmopolitan, who empathizes with advertisers: Who needs somebody youre paying thousands of dollars a year to come back and bite you on the ankle?35 . And further, on a question why she didnt criticize tobacco products: We just dont say rotten things about our advertisers. There is very little to warn our readers about, because advertisers products are so good36 . Seldes (1943), for example, investigated newspapers in 1940 and found that many newspapers did not report the Federal Trade Commissions reports on mistakes in advertisements, if they also ran these ads. According to Nan Robertson, a New York Times journalist and author, it was quite normal in the 1950s that articles on womans and fashion pages reected the relative advertising strength (Fleetwood 1999). Boston Herald ordered its consumer columnist to stop writing about a local bank merger. The reporter attributed this decision to the fact that one of the banks was also a big advertiser and lender
30 Metro
Times, 11-06-2003, documented in Hart & Hollar (2004) Journalism Review, 09/10-2001, documented in Jackson & Hart (2002) 32 Advertising Age, 17-04-1972, quoted in Bagdikian (2004) 33 New York Times, 03-06-2002, documented in Jackson, Hart & Coen (2003) 34 McClung Lee (1937), documented in Bagdikian (2004) 35 Quoted in Bogart (2000). 36 Quoted in Bogart (2000).
31 Columbia
60
Newspapers
to the newspaper37 . In the earliest edition of The Wall Street Journal October 3 edition in 2002, the newspaper put an advertisement of Bear Stearns next to an article about an error in a stock market order. Instead of moving the advertisement, the paper decided to eliminate the news article on Bear Stearns and place it as a subsection of another story38 . In a story about increased sales after the murder on Versace in 1997, the Chicago Sun Times mentioned only department store advertisers. We have to take care of our customers, said editor Larry Green39 . More subtle is the creation of additional reporting to attract advertisers. CBS admitted to reduce the number of prime-time war specials to please advertisers40 ; US News & World created in 2003 a war-free zone, called Second Front41 . Similarly, some respondents of our survey mentioned that there is disproportionate attention to soft news that sells advertisements, such as a winter special for the travel section and interior and design for the lifestyle section. Ironically, more advertiser-friendly soft news sections might indicate that the hard news section is perhaps less aected by advertiser inuence. Abraham Rosenthal, New York Times executive director (1977-1988) said on soft news: If they keep my A section pure, then I will let them have all those other sections. (Fleetwood 1999)42 . Demand channel Though the media outlet is the central player in advertising bias, advertisers play their role too. In the most basic form, some advertisers at least expect that such a bundle exists. The head of the Grocery Manufacturers Association, Paul Willis, wrote in 1962 to television stations that their editorial department and business department might better understand their interdependent relationships, and later claimed the publication of favorable articles in print magazines as a success of his action43 . One respondent of our survey mentioned that organizers of cultural events often expect an article if they buy an advertisement. Apart from creating expectations, bigger advertisers apply a more efPost, 01-05-2000, documented in Jackson & Hart (2001). 11-10-2002, documented in Jackson, Hart & Coen (2003) 39 Documented in Fleetwood (1999) 40 See Baker (1994). 41 MediaWeek, 24-02-2002, documented in Hart & Hollar (2004) 42 In the same spirit, Claussen (2002) relates the story of an older man asking a journalist What is the rst responsibility of the press? The young journalist answers enthousiastically: To print the truth. He is corrected by the older man. No, the rst responsibility is to make prot so that it can aord to print the truth. 43 Advertising Age, 17-04-1972, documented in Bagdikian (2004)
38 thestreet.com, 37 Washington
61
fective weapon to obtain favorable coverage: they withdraw their advertisements if coverage is unfavorable or not favorable enough. While this withdrawal does not undo the negative coverage, it might serve as a reputation device to impede negative coverage in the future. Baker (1994) and Bagdikian (2004) give many examples of withdrawals. Piano advertisers withdrew from Esquire in 1940, after it declared that the guitar is a better accompaniment to singing than the piano. Pharmaceutical companies threatened to withdraw all advertisements from Modern Medicine, after parent company New York Times ran a series of articles on medical malpractices in 1976. WDHD lost 4.5 million of advertising from the automobile industry in 1991; and 1 million from Proctor & Gamble in 1990 after it ran an article that criticized Proctor & Gamble. Cosmetics rm Revlon withdraw all its advertisements from Ms. because women on the cover were not wearing make-up, 60 Minutes lost advertising money because it alarmed for health issues on apples. But often the threat of withdrawing already suces to change the content. Time magazine was said to retreat Osama Bin Laden as person of the year in 2001 after pressure from Walmart44 . Georgia Governor Sonny Perdue threatened to withhold 500000 dollar; as a result an incident during the campaign was not aired on the CBS television station WGCL (Atlanta)45 . Nordstrom pulled its advertising from the Seattle Times and the Post-Intelligencer after they published stories on unfair labor practices (Bogart 2000). While the cases above might seem incidental, some bigger companies have an explicit withdrawal rule. They automatically withdraw advertisements if certain content is published. During the Federal Communications Commissions hearing in the 1960s, many of these rules were revealed. Proctor & Gamble refused to advertise in issues depicting business as cold or ruthless (besides other stipulations); Brown & Williamson tobacco directed that tobacco products should not be used in a derogatory or harmful way46 . General Motors announced in 2004 that they refused to advertise on TV programs about the atrocities in Iraq47 . BP and Morgan Stanley issued directives demanding their ads be pulled from any edition that included objectionable content48 . Google Adsense installed sensiYork Post, 2002, documented in Jackson, Hart & Coen (2003) Loang, 13-12-2006, documented in Jackson (2007) 46 See Bagdikian (2004) for a discussion of these withdrawal rules and for additional examples. 47 USA Today, 18-05-2004, documented in Hart & Hollar (2005) 48 AdAge.com, 24-05-2005, documented in Hollar, Jackson & Goldstein (2006)
45 Creative 44 New
62
Newspapers
tivity lters in 2003, to avoid that advertisements run next to damaging content. Many websites that ran stories on war (e.g. Iraq), sex (e.g. explicit content, rape) or business bashing (e.g. boycotting Sony) lost all their Google ads49 . Note that, even with these withdrawal rules, media companies can still write about sensitive topics, though they are punished by obtaining lower advertisement revenues. This business concern might inuence the news selecting process. To sum up, we argued that advertisement bias might be initiated by media outlets, advertisers or it might be a mutual agreement. Media outlets might oer an explicit bundle of advertisement and favorable coverage, but in many cases the oer is implicit, by self-censoring the news to avoid conicts. Advertisers can threaten to withdraw advertisement money in order to obtain favorable coverage, or they can constitute a withdrawal rule.
D2
Inside channels
The examples and mechanisms above only considered the business side of the newspaper, and the advertisers. This overlooked the role of the journalist. Journalists perceive themselves as independent from the business side. The ethics of the journalists determine that journalists should care about the public interest, by describing the news as truthfully as possible and by collecting information independently50 . The New York Times code explicitly stresses the wall of separation between the sales department and the editorial sta51 . Our company and our local units treat advertisers as fairly and openly as they treat our audiences and news sources. The relationship between the company and advertisers rests on the understanding that news and advertising are separate that those who deal with either one have distinct obligations and interests, and each group respects the others professional responsibilities.
Bay Express, 02-08-2006, documented in Jackson (2007) The Belgian Code of the Council of Journalism (Raad voor de Journalistiek 2010). For an extensive list of Codes of Ethics for journalists, see http://www.rjionline.org/mas/codes-of-ethics.php. 51 The New York Times Company Policy on Ethics in Journalism, article 80 (http://www.nytco.com/press/ethics.html)
50 49 East
63
Inserting advertising bias does not fulll these codes and therefore journalists might be reluctant to do so. Some respondents of our survey testied that they were asked to insert bias, but refused. We do not question the ethics of journalists here. On the contrary, the paragraph below describes how advertising bias can be implemented, despite the existence of ethical journalists. Pressure doesnt make journalists happier, write Hamilton & Krimsky (1996): Newsroom moral plummets when a publisher strolls into the editors oce and asks for special handling of a story relating to an advertiser. Direct pressure might even work contra-productive, they argue. Reporters balk at pung advertisers; and when they uncover something negative about one of them, they expect that the bad news will be published. Therefore, media outlets use more indirect forms of pressure. In general, we observe two ways in which media outlets can circumvent the independence of journalists. The rst is to tear down the wall between the business side and the editorial sta. The second way is to change the sta composition such that independence from the business goals is limited. Crossing the wall When the wall between the sales and editorial sta is torn down, then it is easier to cross the line between both, in both directions. This becomes very clear when the sales department is also involved in the activities of the editorial sta. In many newspapers, lighter news, such as real estate, travel and food is written by the ad department. As labeled by the editor of the Houston Chronicle, this is nothing controversial 52 . In 2003, the marketing department of the Pioneer Press wrote an article about a restaurant to compensate for another negative article on the same restaurant53 . To extend this habit further, the Fairbanks Daily News-Miner hired an advertorial writer54 . A more subtle form of the journalistic ambitions of the sales or marketing department are suggestions to the journalists. In fact, such suggestions are not per se wrong, as long as the journalists have the freedom to act or not. Though, as Bagdikian (2004) writes, these suggestions were known as business oce musts it was hard not to follow these suggestions, as recently as the 1980s. Some editors also provide a list of the biggest ad& Publisher, 31-03-1979, documented in Bagdikian (2004) Reader, 05-09-2003, documented in Hart & Hollar (2004) 54 Columbia Journalism Review, 9/10-2005, documented in Hollar, Jackson & Goldstein (2006)
53 Chicago 52 Editor
64
Newspapers
vertisers. One of our respondents mentions that the directors named the three biggest advertisers, directly followed by the statement that it was not the goal to adjust articles toward these advertisers. The statement was just to inform the sta. After some negative coverage, CNN informed its sta that Nasdaq is a major advertiser55 . A step further, television station KTVO urged its journalists always to go to station advertisers rst for expert opinion and industry comment56 . The business side can also facilitate direct contact between journalists and advertisers. Again, this is not wrong per se, but it might be considered as a further attempt to break down the wall between business and journalism. In 2009, Washington Post organized o-the-record salons with advertisers and journalists57 ; Des Moines Register asked advertisers to see what they can do together for the coverage of high prole events58 . Or, as summarized by Mackenzie Warren of News-Press: Keeping reporters away from the business side of the paper is old-school snobbery59 . But it is not always the business side that visits the journalistic side. Journalists also turn to the business side, or the business side applies journalistic practices in its advertisements. By doing so, the journalistic reputation is used to increase the eectiveness of the advertisement. Though this is not directly advertising bias (since there is no bundle and there is no biased coverage), the mechanism exploited here is the same. In both cases, the outlet uses the fact that reader can not fully distinguish news from bias; or here: news from advertisements with journalists. Either the ad looks like a news item (NBC ad in the LA Times60 , insurer ad on television channel Grin Communications61 ), or journalists are featured in ads or purchased items (KVVU, KLAS and KTNV used reporters to conduct interviews for purchased segments62 ). Some journalists are paid by companies, but they fail to disclose their interdependence. Similarly, some media rely heavily on business-paid experts. The technology editor of Child magazine, for example, was paid by Kodak to promote products, which he did on a local television channel and NBCs Today show63 . Tele55 Daily
News, 06-09-2001, documented in Jackson & Hart (2002) Journalism Review, 9/10-2004, documented in Hart & Hollar (2005) 57 Politico, 03-07-2009, documented in Hart (2010) 58 Forbes, 17-04-2008, documented in Hart (2009) 59 Washington Post, 04-12-2006, documented in Jackson (2007) 60 New York Times, 10-04-2009, documented in Hart (2010) 61 Tulsa World, 05-04-2009, documented in Hart (2010) 62 Las Vegas Review-Journal, 09-10-2009, documented in Hart (2010) 63 Wall Street Journal, 19-04-2005, documented in Hollar, Jackson & Goldstein
56 Columbia
65
com analyst Je Kagan was cited in various media, such as Kansas City Star, seemingly as an independent expert, but in reality paid by several telecom companies64 . Editorial sta composition A second mechanism to reduce the independence of the sta is to change the composition of this sta. The most visible composition change is ring unwilling journalists and replacing them by easier journalists or less ethical ones. These dismissals might serve as a signal to other journalists. Similarly, but less visible, is a media organisations decision not to advance unwilling journalists. There are many examples of journalists red due to pressure of advertisers. In 2009, Summit Daily News red a reporter due to a conict on ski resorts65 ; Hartford Courant did the same with a columnist after a negative consumer column on mattress company Sleepy66 . In 2006, Evening Sun red its long-serving columnist after calling Walmart products Made in China thingy-ma-bob67 . CityBusiness (New Orleans) red its editor after she objected to the introduction of an advertiser-sponsored news page68 . A more subtle mechanism is dedicating more journalists to the elds that are interesting to advertisers. As such, bias slips in without the consent of the journalists involved. By hiring relatively more freelance journalists, news outlets are also able to shift power away from journalists. Although freelancers are in principle more free to write about what they like, the market logic predicts that they are going to write articles that are bought by outlets.
(2006). 64 Pitch,28-07-2008, documented in Hart (2009). 65 CBS4Denver.com, 09-12-2009, documented in Hart (2010) 66 New York Times, 18-08-2009, documented in Hart (2010) 67 New York Times, 04-12-2006, documented in Jackson (2007) 68 Columbia Journalism Review, 09/10-02, documented in Jackson, Hart & Coen (2003)
66
Newspapers
Chapter 3
The Universal Service Obligation for Telephone Directories: Regulating the Redundant
This chapter is joint work with Patrick Van Cayseele and is adapted from De Smet & Van Cayseele (2010) and De Smet & Van Cayseele (2011).
68
Directories
3.1
Introduction
After the liberalization of the telecommunications sector in the 1990s, the member states of the European Union decided to implement a universal service obligation1 , to ensure that in every country the necessary services are available at determined quality and an aordable price, even if the market would not provide it2 . In many other countries, a similar framework was implemented3 . One of these obligations concerns the publishing of telephone directories. Article 5, paragraph 1(a), of the European Universal Service Directive species that member states shall ensure that at least one comprehensive directory is available to end-users. Article 25 adds that subscribers to publicly available telephone services have the right to have an entry in the publicly available directory. We can disentangle this obligation in two parts. First, the telephone directory should be comprehensive, i.e. it should list all the persons and companies that have a phone4 . Second, the directory should be available to all users. In this chapter, we develop a simple model to investigate whether the inclusion of both conditions is sensible. Is the imposition of both conditions necessary, or is it sucient to impose one condition, to obtain also the other? More specic, we check rst whether imposing complete availability implies that the directory will be comprehensive. Second, we investigate whether the reverse holds, i.e. whether imposing a comprehensive directory results in a free directory. A priori, it is not clear why one condition would imply the other. By providing a comprehensive list, the publisher gives away a valuable listing for which some persons or companies may want to pay. Similarly, certain users perceive the directory as a valuable search instrument for which they may be willing to pay. Therefore, it might be reasonable to charge not only (some of) the listed persons or companies, but also the users.
Service Directive, Directive 2002/22/EC http://ec.europa.eu/information_society/policy/ecomm/toda ys_framework/universal_service/index_en.htm, emphasis added 3 See for example the Telecommunications Act of 1996 for the United States (P.L. No. 104-104, 110 Stat. 56 (1996)) or Telecommunications (Consumer Protection and Service Standards) Act 1999 for Australia (http://www.austlii.edu.au/au/legis/cth/consol_act/tpassa1999620/). 4 Note that telephone directories will not be comprehensive in practice, since everyone has the right to opt-out. Furthermore, mobile numbers are, as a rule, not included in the directory, though subscribers have the right to opt in.
2 See 1 Universal
3.1 Introduction
69
Our model is particularly relevant in the policy discussion about the universal service obligations. Since the introduction of a universal service framework for telecommunication, postal services and electricity in the 1990s, the literature on universal service obligations has boomed. Scholars have been investigating the rationale of these obligations (Cremer, Gasmi, Grimaud & Laont (1998), Crew & Kleindorfer (2002), Graham, Cornford & Marvin (1996)), the costs and funding (Garbacz & Thompson (2005), Jaag, Koller & Trinkner (2008), Mirabel & Poudou (2004), Rodriguez & Storer (2000), Weller (1999)) and the eects on industrial organization topics like competition, entry and pricing (Armstrong (2008), De Donder (2006), Riordan (2001), Valletti, Hoernig & Barros (2002)). Due to technological changes, such as the Internet and mobile phones, a lot of research is devoted to the question whether the universal service should be adapted, extended or even abolished (Crmer (2000), Downes e & Greenstein (2007), Xavier (2003))5 . Theoretical contribution This article expands the understanding of the strategic decisions of telephone directories publishers in the context of two-sided markets. Inspired by this literature, started by Rochet & Tirole (2003), Parker & Van Alstyne (2005) and Armstrong (2006), the directory is modeled as a platform that connects two distinct sides of the market: receivers, also labeled as users, buyers or readers, and senders, also referred to as sellers or advertisers. While the roles of receivers and senders can overlap in reality, i.e. sellers can also be buyers and vice versa, we keep both sides of the market strictly separated. This claries the analysis in section 4.3, and makes the interpretation of telephone directories as a two-sided markets easier. Note that the framework of two-sided markets is important for the
5 At the end of this chapter, we come back to the policy debate on telephone directories. Note that this policy discussion is focused on White Pages. There are three generally accepted types of telephone directories: White Pages, Yellow Pages and Grey Pages. The rst type contains an alphabetical list of persons, with address and telephone number. The directory is divided in regions. Yellow Pages is synonymous with a business directory. It classies rms by their business type or goods or services provided. Grey Pages are less known. These are so-called reverse telephone directories where one can browse the numbers and nd the associated customer details. The latter was mostly used by emergency services, phone companies, law enforcement, and public libraries. Our model applies to both White Pages and Yellow Pages. In what follows, we use the more general term telephone directory which can both refer to White and Yellow Pages.
70
Directories
research question dealt with in this article. The universal service obligation for telephone directories simultaneously imposes a restriction on both sides of the market. The platform has to provide access to the receiver side (availability), and similarly, it has to provide access to the sender side (comprehensiveness). The focus of this article is on whether it is sucient to have a restriction on only one side to obtain a market outcome that satises the restrictions on both sides. Although dealing with a precise policy problem, we add to the theory of two-sided markets by including product dierentiation on the platform. The sender side types can self-select by choosing their listing format; the other side of the market can see the dierent formats on one single platform. In the seminal contributions, platforms can be dierentiated (e.g. through the Hotelling model), but this dierentiation only takes place between platforms. If the other side wants to see dierentiated content, it has to connect with multiple platforms (i.e. multihoming). As a result of this dierentiation, platforms have the possibility to charge dierent prices. This is quite unusual for the models used in the two-sided markets literature. As Weyl (2010) summarizes the literature, he emphasizes that platforms are price setters on both sides and typically set uniform prices as one of the key features of the current models. Weyl (2010) also adds that the possibility for platforms to oer multiple products is an important question for future research. We try to ll this gap in the literature, and by doing so we complement the small subsegment of two-sided markets articles that allow for multiple products on a platform. In Damiano & Li (2007), a monopoly matchmaker dierentiates both sides, men and women, in many types by setting a schedule of prices. The matchmaker uses price discrimination to allow men and women to select their optimal type, which generates an ecient matching process. The argumentation is that dierentiation is better than the uniform pricing used in online dating markets. A price that is used as a signal mitigates the misrepresentation in markets with uniform pricing. Our model is similar to their model to some extent since we also model the platform as a monopolist6 and we also allow the dierentiated side
6 Interestingly, in their accompanying paper Damiano & Li (2008), they compare the results of monopoly with competition. They nd that while monopolistic match makers can use prices to sort high types from low types, their duopolistic counterparts are involved too much in price competition and therefore they are less ecient. In our article, we choose for a monopoly model, for computational simplicity, but also because this resembles the reality of telephone directory publishing in most European countries.
3.1 Introduction
71
to self-select. Though, we do not model the potential inference of quality by a price mechanism, i.e. we do not allow to use the listing formats as a screening device for users and hence do not address potential quality distortion as in Mussa & Rosen (1978)7 . Viecens (2006) presents a duopoly model inspired by shopping malls. Similar to our model, for the buyer side, not only the number but also the type of shops present in the mall matters. The relative importance of quality vs. quantity determines whether the platforms will both engage in attracting low or high quality shops. This relates to our approach where we distinguish between the type and the volume of the sender side. Contrary to our model, shop keepers in Viecens (2006) cannot self-select their presence form in the mall. Other papers that implement non-uniform pricing are Gomes (2010) and Hagiu & Lee (2008). In Gomes (2010), the platform auctions its audience to advertisers. A platform charging dierent prices related to the action space of the demand side is introduced by Hagiu & Lee (2008). Final remark. The universal service obligation for telephone directories is at the heart of this article. This might mislead the reader to conclude that our contribution is old-fashioned and therefore irrelevant for two reasons. First, the universal service obligation is subject to debate, as we discuss in section 4.4 Second, it is beyond doubt that the paper telephone directories will be replaced by the internet. Though, we have at least three reason to rmly believe that these two trends do not devaluate the relevancy of this chapter. First, there is no qualitative dierence between paper and online directories. While paper directories are likely to vanish, online directories belong to most visited websites worldwide. Second, with minor modications, our model also applies to online markets such as search engines. Similar to directories, search engines apply dierentiation on the platform, by oering dierent formats: regular links which are free and sponsored links which are paid. The ranking of these results also introduces dierentiation. Therefore, our results might be interpreted also in the light of search engines. Third, though search results are not regulated, it is not unthinkable that some regulation will emerge in the future (New York Times (2010) and Mayer (2010)). The European Commission and the Fed7 An empirical investigation of price discrimination in the Yellow Pages industry is found in Busse & Rysman (2005).
72
Directories
eral Trade Commission (US) announced to scrutinize Googles search results for cross-promotion, i.e. favoring its own related services over rivals in both regular and paid results (European Commission (2010), Singhal (2011)). Even if the universal service obligation for telephone directories is abolished, our results remain relevant because they provide guidance too for potential regulation on search engines. The remainder of the chapter is organized as follows. In section 4.3, we introduce the model that subsequently is used to investigate the relevant policy issues: the consequences of imposing the two main restrictions of the universal service obligation for directories. In section 3.3, we analyze for each restriction the equilibrium outcome, conduct a comparative statics analysis and question whether it is necessary to supplement the restriction focussed on with the other restriction. Section 4.4 concludes and adds the policy debate on universal service. In the appendix of this chapter, we suggest an empirical method to calibrate the relevant parameters, based on the empirical work of Rysman (2004). We also work out some extensions of the model and suggest how our model can be applied to other industries, such as dating events, shopping streets, online search and online media.
3.2
Set-Up of the Theoretical Model
In this section, we set up a theoretical model that is tailored to the telephone directories industry. The goal of this model is twofold. First, this model is one of the rst attempts to model the important dynamics of the directories business. Second, we try to shed light on the two main features of the universal service obligation for directories: comprehensiveness and availability. The telephone directory industry is characterized by three players: receivers, senders and publishers. Receivers and senders can be both natural persons or companies. The reasons why receivers and senders want to connect with each other can be various. To clarify the analysis, we assume that the phone is used only for business reasons. Senders are retailers or companies that advertise in the directory to connect with users in order to sell their products or services. Receivers are consumers that use the directory to connect with senders in order to buy their products or services. This narrowing of the scope of telephone use allows to clearly distinct both sides of the market. An alternative interpretation is that the decisions of reading the directory as a receiver on the one hand and being listed in the
3.2 Set-Up of the Theoretical Model
73
directory as a sender on the other hand are made independently. Our analysis is focused on the platform. The platform is the publisher who connects receivers and senders. We assume that the platform is a quantity setter8 on the sender side, and a price setter on the receiver side. We further assume that there are two types of listings in the directory: small listings (indexed 1), and large listings (indexed 2). Small listings contain only limited information; large listings contain more information, but it is not guaranteed that receivers value this additional information. The platform sets the optimal quantities q1 and q2 and determines the optimal price for receivers pR . We investigate a monopolistic market situation9 . Receivers Receivers value both small and large listings. The valuation for small listings is given by r1 , the valuation for large listings by r2 . Small and large listings are valued dierently but it is not necessarily the case that readers appreciate large listings more than small listings. Large listings might contain more information, such as a website, e-mail address or map. Or it might contain pictures or just being printed in a larger font size. Some receivers might perceive this as a nuisance, hampering the browsability of the telephone directory. Utility is negatively aected by the price pR and the opportunity cost k. This opportunity cost can be related to the availability of outside options, such as going to a company one already knows, asking a friend for advice, or driving to a city and searching randomly for retailers. The net utility of using the directory is equal to: UR = max(r1 q1 + r2 q2 pR k, 0) (3.1)
8 While publishers have a tradition of announcing list prices each year, the practice learns that they adjust these prices with discounts to accommodate the number of senders, i.e. list prices can be seen as maximum prices but can dier substantially from the real prices. 9 In Europe, 70% of the countries has only one player in printed directories. Some countries have two publishers. These countries include the bigger countries (UK, Spain), Scandinavian countries (Sweden, Denmark, Finland) and alpine countries (Austria, Switzerland). For most countries, the market structure is not monopolistic if one considers the market for internet directories. We can expect further consolidation as was recently the case in the Netherlands (merger approved in 2008, NMa Decision 6246, case European Directories Truvo Nederland). (Data collected from the EADP website (EADP.org). EADP is the European Association of Directory and Database Publishers and coordinates most European directory publishers. We counted for each country the number of publishers in the category Telecommunication Directories that publish print directories.)
74
Directories
Receivers are homogenous in the valuations r1 and r2 but are heterogeneous in their opportunity cost. We assume there is a mass one of potential receivers and the opportunity cost k is uniformly distributed between 0 and 1, so the number of receivers qR is given as: qR = r1 q1 + r2 q2 pR (3.2)
Since qR is the number of potential receivers that reads the directory, we also refer to it as the readership. It is clear from expression 3.2 that a directory without listings will have no receivers. Senders Senders value looks at their listing since they generate prot from each look. We assume that this prot is related to the sales of services and goods. The number of looks for each listing depends on the size10 of the listing si and the number of receivers: L1 L2 = = s1 q R s2 q R (3.3)
We assume that large listings attract more receivers than small listings (s2 > s1 )11 . Further we assume that each sender buys at most one ad. Senders pay a price pi to use the platform. The net benet of a sender reads: Li pi (3.4)
with i = 1, 2 being the type of listings. Senders are of mass one and are heterogeneous in their prot per look which is uniformly distributed p2 p between 0 and 1. Dene L2 L1 to be the prot level where senders 1 p are indierent between both ad types, and 1 L1 is the lowest prot for 1 which the senders net benet is positive. Then q1 q2
10 Both
= 1 = 1 (3.5)
parameters s1 and s2 stand for the size of the listing format. Note that it can also be interpreted as salience, as in the framework of Haan & Moraga-Gonzalez (2009). Salience is the chance that a certain rm is remembered. 11 This assumption can be made without loss of generality. If it would be the case that small listings attract more attention than large listings, then small listings can be relabeled q2 and large listings q1 .
3.2 Set-Up of the Theoretical Model
75
Since we model the platform as a quantity setting monopolist on the sender side, we transform the demand functions to inverse demand functions: p1 p2 = (1 q1 q2 )L1 = (1 q2 )L2 q1 L1 (3.6)
Publisher The platform maximizes its prot: = p1 q1 + p2 q2 + pR qR (3.7)
While deciding on the quantities, a platform has to take into account the direct and indirect eects of changing these quantities. An increase in quantity directly aects prots by a decrease in prices, which is complicated by product dierentiation. Another, indirect eect is the eect of quantity on usage. An increase in the number of listings increases the number of receivers and therefore increases the willingness to pay of senders. This eect, the feedback loop eect, dampens the negative effect of a quantity increase on sender price. A similar mechanism exists along the receiver side: an increase in receiver price reduces the number of receivers, which further decreases the willingness to pay of senders. In the next sections, we adapt this set-up to shed light on the research questions. In order to present these optimal decisions more clearly, we dene three additional parameters: r = r1 , R = r1 + r2 and s = s1 . Thus r2 s2 r is the relative appreciation of q1 versus q2 , whereas s is the relative attention-attraction of q1 versus q2 . We replace r1 , r2 and s1 to obtain expressions that contain only r, R, s and s2 12 . To simplify the analysis, we assume that the platform has no costs. Including a xed cost would not change our results. Printing industries typically have large xed costs compared to marginal costs13 . Therefore, it seems reasonable to focus on xed costs only. As a consequence, there is no cost dierence between small and large listings.
rR R replace r1 by 1+r , r2 by 1+r and s1 by ss2 . digital media, one can argue that there is only a rst copy cost, since marginal costs are often nearly zero. 12 We 13 For
76
Directories
3.3
Investigation of the Universal Service Obligations for Telephone Directories
In this section, we try to answer the following questions. Will the platform provide a comprehensive telephone book, given that is available to all end-users? And will the book be available to all end-users given that the platform provides a comprehensive directory? Subsections 3.3.1 and 3.3.2 are structured as follows. First we explain the restriction imposed (i.e. availability in 3.3.1 and comprehensiveness in 3.3.2). Then we compute the optimal outcomes for the monopoly platform and perform a comparative statics analysis. Third, we extend the choice set of the platform with the relative size of the listing (s) We end with a discussion on the necessity of the other restriction, given the outcomes obtained.
3.3.1
The Availability Restriction
In this section, we examine whether the monopolist platform will provide a comprehensive directory, including all potential senders, given that the directory is available to all receivers. In order to publish such a directory, the platform must oer the small listings for free. Otherwise, some senders (i.e. those with a low willingness to pay) will opt out and prefer not to appear in the directory. With the restriction that the directory is available to all receivers, we assume that receivers dont have to pay for the directory, i.e. that pR = 0. Note that even with pR = 0, not all potential receivers use the directory. Therefore, the maximization problem of the platform reads: max = p1 q1 + p2 q2
q1 ,q2
(3.1)
s.t. q1 , q2 0 s.t. q1 + q2 1 Solving this problem allows to state the following results: Proposition 3.3.1 The optimal outcome for the monopoly platform is such that:
Regime 1 A large listings only platform (i.e. q1 = 0, q2 > 0) prevails if r s Regime 2 A small listings only platform (i.e. q1 > 0, q2 = 0) prevails if s r 4s3 3 4
< s < 1 and
Regime 3 A comprehensive platform (i.e. q1 + q2 = 1: everyone is listed) prevails if s4s2 +2 s2 (15s+4s2 ) s < 1 and r 4 14s
3.3 Results
77
Figure 3.1: Dierent regimes in the r-s-space, given availability
Regime 4 A dierentiated platform (i.e. q1 > 0, q2 > 0: small and large listings are oered) prevails otherwise
Note that all proofs are relegated to Appendix A. When s is large vis-`-vis r, as in regime 1, this means that the ata tractiveness of a small listing is quite high, though its contribution to readership is quite low. A high s makes a large listing more attractive for senders, and will increase the willingness to pay for large listings. This high s will also induce self-selection of senders, who trade in their large listing for a small listing. The platform incurs a double loss because the price paid by a sender will be lower (p1 p2 ), but also because the directory will attract less receivers (small r), reducing the general price level. Therefore, a monopolist will set q1 as low as possible even negative if it were possible. If r s, the monopolist will oer only large listings. The reverse occurs in regime 2 when both s and r are suciently large. Then the small listings q1 contribute (much) more to the readership than large listings q2 . Since both listings appear to be quite similar in terms of attractiveness when s approaches 1, they are equal the dierence
78
Directories
in willingness to pay for both types becomes smaller. Given this higher readership (and therefore the general price level), it becomes optimal to steer senders away from q2 to q1 . If s and r are suciently large, then the monopoly platform will oer only small listings. A third regime occurs when s is suciently small, and r is suciently large. With these parameter values, q1 has a relative high contribution to readership. But even in the case where this contribution is smaller than large listings (r < 1), only a limited amount of the more protable senders self-select into q1 (against q2 ) because the dierence in size is quite large. In this case, the platform oers more quantity space than there are potential senders, leading to a negative price for type 1. This is compensated by the prot made through q2 . Since negative prices are impossible in our model, we cap q1 + q2 to 1, i.e. the amount of potential senders. This results in an implicit price of zero for the lowest category. Figure 3.1 graphically summarizes these regimes in the r-s-space14 . When both s and r are intermediate, both listing formats gure next to each other in the directory. Regime 4 is the unconstrained optimum for this maximization problem. As shown above, this regime will not always be feasible, because for some values of s and r, this will violate the non-negativity conditions on q1 and q2 , or the upper limit q1 + q2 1. In this range, s is small enough to allow self-selection, and large enough to charge the small listings. Simultaneously, r is large enough such that readers value both listing formats. Performing comparative statics analysis allows us to formulate three interdependent propositions on the impact of R and s2 , r and s on the equilibrium quantities and prot. Proposition 3.3.2 R and s2 positively aect prot, but do not affect the optimal quantities. The proportion of r1 to r2 (= r) does aect the optimal quantities, but only in the comprehensive and dierentiated optimum. The bigger r1 relative to r2 , the higher the amount of small listings and the lower the amount of large listings. The proportion of s1 to s2 (= s) does aect the optimal quantities, but only in regime 4. If r1 is substantially larger than r2 , an increase
14 As s4s2 +2 s2 (15s+4s2 ) , 14s 1 and 3 . 4 4
can
be
seen,
the
boundaries
between
regimes
and
4,
and between regimes 4 and 2, r =
s , 4s3
go asymptotically
to
3.3 Results
79
in s increases q1 and decreases q2 . If r1 is smaller than r2 , an increase in s decreases q1 and increases q2 . The eect of r on the optimal quantities is straightforward. The more a listing type is appreciated by receivers, the more prominent this type will be on the directory. The eect of s is slightly more complicated. A higher s increases the willingness to pay for q1 , but it might also lure away senders from choosing a large ad. Therefore, the ultimate decision of the platform depends on the contribution of both types to readership. If q1 listings contribute suciently more to readership, then an increase in s results in an increase of q1 and a decrease of q2 . The reverse holds when q1 contributes less than q2 . An important remark on these comparative static ndings is that the eect of r and s is not limited to play a role only within these regimes, but also may aect which regime is optimal. A change in r or s can induce a regime shift. Note that regime 3 is the only regime where the platform oers a comprehensive directory. In this regime, the constraint q1 + q2 1 is binding, i.e. q1 + q2 = 1. This also means that q1 is oered for free to the potential senders. The platform only makes prot from the category of large listings, q2 . Endogenous Size Decision The results of the constrained problem with pR = 0, indicate that the platform will oer a comprehensive di2 1 s2 (15s+4s2 ) rectory if s < 4 and r s4s +2 14s . In all the other cases, the platform will either oer only one type (regime 1 and 2) or two types at a positive price. All the latter cases result in a non-comprehensive directory. This clearly shows how the choice to oer a comprehensive directory depends on the parameters r and s. In our analysis, we treat size as exogenous, which gave us a good insight in the interplay between r and s. In reality, however, platforms can choose the size of the listings. Since we assumed no costs to size in the model, platforms will set the level parameter s2 of the listings as high as possible. It is more interesting to investigate the choice of s (for a given s2 ). Proposition 3.3.3 If a prot maximizing monopolist can choose the size of the listings, it makes the large listing as large as possible and the small listing as small as possible. It oers receivers a directory with all potential senders.
80
Directories
Except for the special cases r = 0 or r = , the prot maximizing monopolist sets s = 0, i.e. it reduces the attractiveness of the small listings as much as possible. The prot maximizing monopolist implements regime 3, i.e. it oers receivers a directory with all potential senders (q1 +q2 = 1). It oers two possibilities to senders: a large listing that attracts many receivers, or a small listing that attracts virtually no receivers. The latter is oered for free15 . Casual observation of many European directories shows the prevalence of this outcome. It goes without saying that this proposition has an important corollary in the debate whether a government should impose a universal service constraint on comprehensiveness. Corollary 3.3.4 If receivers dont pay, the inclusion of a comprehensiveness constraint in the universal service obligation is not necessary, since a prot maximizing publisher will always choose to open the directory for all potential senders. If the receiver price is zero, and the platform can choose the relative size s, then it will oer a comprehensive directory. Note that the inclusion of costs, marginal or xed, might potentially change these results. When s is xed (i.e. by receiver preferences), our regimes show that oering a comprehensive directory is not always the prot maximizing regime for a platform. If we add more listing categories to our model a closer t with telephone directories in reality the likely result will be that the largest listing is as large as possible, the smallest as small as possible. This is similar to the quality degradation models of Mussa & Rosen (1978), Maskin & Riley (1984) and Besanko, Donnenfeld & White (1988). In these models, a monopolist deteriorates the quality oered to the groups with the lowest willingness to pay for quality. If a regulator rules that a mere listing is a product with too low quality, it can still implement regulatory corrections, such as minimum quality standards though this is not always welfare improving (Besanko, Donnenfeld & White 1988).
3.3.2
The Comprehensive Directory Restriction
In the previous section, we investigated whether the directory is comprehensive if the price for receivers was zero. In this section, we look at
15 Note
that in regime 3, q1 is always oered for free, even if s > 0.
3.3 Results
81
the reverse problem: will the platform oer the directory for free if it is imposed that the directory should be comprehensive. The maximization problem of the platform reads:
q1 ,q2 ,pR
max
= p1 q1 + p2 q2 + pR qR s.t. q1 , q2 , pR 0 s.t. q1 + q2 = 1
(3.2)
Solving this problem allows to state the following results: Proposition 3.3.5 There are four regimes for a monopoly platform:
Regime 1 A large listings only platform (i.e. q1 = 0, q2 = 1, p > 0) prevails if R s2 R s2 s2 R s2
s < 1 and r r1 s < 1 and r r2
R(1s)s2 R+(1s)s2 R+(1s)s2 R(1s)s2
Regime 2 A small listings only platform (i.e. q1 = 1, q2 = 0, p > 0) prevails if R
Regime 3 A dierentiated platform not charging users (i.e. q1 > 0, q2 > 0, p = 0) R s2 3R s2 3R s2 2R prevails if s s ; or s s s and 2 2 2 3R2 2R(1s)s2 +(1s)2 s2 2 3(R2 (1s)s2 ((1s)s2 2R)) 2 r3 r 3R2 +2R(1s)s2 (1s)2 s2 2 3R2 2R(1s)s2 +(1s)2 s2 +2 3(R2 (1s)s2 ((1s)s2 2R)) 2 r4 3R2 +2R(1s)s (1s)2 s2
2 2
Regime 4 A dierentiated platform charging users (i.e. q1 > 0, q2 > 0, p > 0) prevails R otherwise
Similar to the problem of section 3.3.1, regime 4 is the unconstrained regime (given the constraint q1 + q2 = 1). This regime is not always feasible. The other regimes are corner solutions where q1 , q2 and pR are equal to zero, respectively. When s and r are both suciently large (regime 2), then the same selfselection problem pops up as in the specication where pR = 0. Since both listings appear very similar, the platform gains more by steering senders to the listing type that maximizes readership (and therefore prot). The same holds when r is suciently small (regime 1), i.e. then it is optimal to oer only large listings. Note that it depends on both R and s2 to determine whether s is suciently large. If R is relatively large with respect to s2 , then even with a small s, the platform will choose to oer only one type if r is suciently large or small. If s is small and R is suciently small with respect to s2 , then the directory is oered for free to users (regime 3). If R is small, then the willingness to pay of receivers is low and therefore, it might be optimal to give away the directory for free. It might even be prot maximizing
82
Directories
Figure 3.2: Dierent regimes in the r-s-space, given comprehensiveness
to subsidize the use of the directory, i.e. a negative price (pR < 0). In principle, it is not strictly necessary to impose that pR 0. A platform could subsidize its receivers, although in practice it is not directly clear how. Note that a small R is not sucient, but should be accompanied with a small s as well. Since the restriction holds that the directory should be comprehensive, the platform will only make prots from large listings and receivers. The more interesting it is to extract money from large senders, the more the platform is inclined to charge the receivers no price. This happens when s is suciently small. Figure 3.2 shows these regimes graphically in the r-s-space. In this graph, R and s2 are equal to 0.25 and 1. These parameters are crucial in determining the boundaries of the regime. If 2R > s2 , then the platform will never set a receiver price pR equal to zero, as is the case in the third regime. Contrary to the problem in 3.3.1, the boundary conditions are also determined by the parameters (R and s2 ). Within these regimes, it is interesting to see what happens if the parameters change. The comparative statics for the regimes large listings
3.3 Results
83
only and small listings only are the easiest to interpret. These regimes are not aected by the size parameters s and s2 . For both regimes, an increase in R aects price, readership and prot positively. An increase in r also has a positive eect on price, readership and prot in the small listings only outcome, but a negative eect on all three in the large listings only outcome. For the outcome where pR = 0, only r aects the equilibrium quantities ( dq1 > 0, dq2 < 0). Prots are aected positively if r increases, but only if dr dr r > 1. The same holds for the eect of r on readership. An increase in s is negative for prots which anticipates already proposition 3.3.7. R and s2 aect prot positively. Comparative statics for the dierentiated paid platform are slightly more complicated. The relative contribution to readership r aects q1 positively and q2 negatively. An increase in r has a positive eect on price, readership and prot, if r > 1. The reverse holds if r < 1. Both s and R aect q1 positively and q2 negatively if r > 1. An increase in s2 aects q1 negatively if r > 1. The increased prot from senders allows the platform to lower receiver prices. As before, prot is aected positively by R and s2 , but negatively by s. These insights can be summarized as follows. Proposition 3.3.6 R aects prot positively. For regimes 3 and 4, s2 aect prot positively, an increase in s has a negative eect on prot. R, s2 and s only matter for the equilibrium quantities in regime 4 (the dierentiated paid platform). Their eect is crucially determined by whether r is bigger or smaller than 1. r increases q1 , but decreases q2 . It has a positive eect on prot if r is already quite high (above 1). The reverse holds if r < 1. Prot is higher if receivers have extreme preferences, i.e. highly favoring one ad type over the other. These comparative statics are very similar to those presented in proposition 3.3.2. Apart from some dierences, the general insights on the eects of changes in the corresponding parameters holds. We conclude this section with the observation that a platform prefers receivers with extreme preferences, i.e. preferences where the dierence between both types is large. The closer r to 1 (i.e. where tastes are equal), the lower the prot will be.
84
Directories
Endogenous Size Decision The results of the constrained problem with q1 + q2 = 1, indicate that the platform will oer the directory for free to receivers if s s2 3R ; or s2 3R s s2 2R and r3 r r4 . s2 s2 s2 In all the other cases, the platform will charge the receivers a positive price. Similar to the question on comprehensiveness, we allow now that the platform also determines the relative size s. Proposition 3.3.7 If a prot maximizing monopolist can choose the size of the listings, it makes the large listing as large as possible and the small listing as small as possible. Which regime will prevail, depends on the magnitude of R and s2 . The smaller R, the more likely that receivers dont pay.
2 2 If R < s3 , then the receiver price will be zero. In the range s3 R 2 the receiver price is zero if r3 |s = 0 r r4 |s = 0. If R > s2 , then receivers will always be charged to access the platform. We can summarize this in the following corollary.
s2 2 ,
Corollary 3.3.8 The comprehensiveness restriction is not sucient to guarantee that receivers dont have to pay. In sum, whereas our model sustains the claim that the directory will be comprehensive if the price is zero, the reverse is less likely. Receivers will only be oered a free directory if their total valuation for listings R is relatively small16 , or the gain from a large listing s2 is relatively high. In this case, the senders cross-subsidize the receivers. If R is relatively high, or s2 is relatively low, the reverse holds: receivers cross-subsidize senders. Before we conclude, we discuss two further issues, readability and uniform pricing.
16 In the past, it was often assumed that receivers were very price elastic. Given the two-sided markets models ` la Armstrong, a receiver price of zero is then a likely a outcome, at least if senders elasticity is (much) lower. Recent research shows that receivers are not that averse to pay a price. van Caspel, Moerman, Vermeer & Hermans (2002) show that a price increase of 1 euro causes a modest drop in demand of 1.75 tot 2 percent. A price around 5 euro might be optimal, although no calculations on prot maximization are given. A likely eect of a positive price is that people might keep their directory for a longer period. In most countries, publishers oer an annual update, because that is mandatory for White Pages in the European universal service directive. Of course, if directories are provided online, updates can occur permanently.
3.3 Results
85
Readability Our conclusion that availability implies comprehensiveness, but not the other way around, is based on the premise that it is feasible for the platform to implement a relative size s = 0. This implies that the size of a small listing is innitesimally small. As a result, the information might become unreadable. A drawback of our model is that this extremely small size gets accepted by the reader, because the valuation for small listings r1 is not related to size. It is reasonable to assume that a listing should have a minimum size to be readable. In that case, for a xed s2 , s also has a lower limit s. This implies that our claim on redundancy 1 of the comprehensiveness condition given availability only holds if s < 4 2 s4s2 +2 s (15s+4s2 ) and r . The reverse claim, whether availability 14s implies comprehensiveness, only partially relies on the minimum size s. If 2 R > s2 , then it will never be the case that receivers dont pay, no matter how small s is17 . In sum, the existence of a lower limit for s softens the basic claim in this article. Our results are only guaranteed if this lower limit is suciently small with regard to r. If not, then both conditions are necessary to implement and it does not suce to impose only the availability condition. Uniform Pricing In models on two-sided markets, nearly all contributions assume that platforms set uniform prices on both sides (Weyl 2010). Our model diverts from these models by allowing multiple products and hence dierent prices. Interestingly, our results show that in two of the four regimes (for both restrictions), the platform chooses not to use the possibility to oer multiple listing formats. As a result, the platform oers only small, or only large, listings. As such, to the casual observer, market outcomes for these regimes seem similar to the earlier models that allow only for one product and one price. The traditional explanation for not oering the full range of products, is that it is neither feasible nor desirable in the presence of returns to scale (Spence 1976). Further, the self-selection of consumers prevents that a seller can fully exploit discriminatory prices. We complement these explanations with a two-sided market logic: it also depends on the other side of
17 Another solution is to make the valuations r and r dependent on the size. We 1 2 are reluctant to do so, because the advantage of keeping the size and the informational value separate is that it does not determine a specic relationship between both. It might be that this relationship is non-monotic: an increase in the size when size is small is probably positively valued; an increase when size is already large is negatively valued.
86
Directories
the market (captured by parameter r) whether oering multiple products is optimal. We showed that platforms also oer a less protable format if receivers value this type suciently (and the size dierence between both listing formats is large enough).
3.4
Conclusion
In this article, we examined the main conditions of the universal service obligation for telephone directories. These conditions are comprehensiveness and availability. Telephone directories should contain all telephone subscribers and should be available to all end-users. We investigate whether one of these conditions is redundant, i.e. if imposing one condition also implies that one automatically gets the other. We nd that imposing a reader price equal to zero (and hence availability to all endusers) results in the publishing of a comprehensive directory. Conversely, it is not guaranteed that imposing a comprehensive directory results in availability to all end-users. An important caveat to our model is that costs are ignored and that an exogenously imposed minimum size could soften our redundancy result. These results might provide guidance in the discussion on regulating search engines. These engines share the same micro-economic mechanism with directories: paid listings (on the top of the ranking) feature next to algorithmic listings (with unknown ranking). The debate on the regulation of search engines is centered around a fair ranking; the arguments are similar to the rationale for a universal service obligation for directories. Search engines structure vital information for a society and the market might not provide valuable search results. Both conditions of the obligation, availability and comprehensiveness, are at the heart to achieve an informed society. Our results suggest that the availability condition might be favored over the comprehensiveness condition. We end with two remarks related to the current policy debate. First, is a universal service obligation for directories desirable? And second, is all data on subscribers readily available? Discussion on desirability Our model investigated the positive question whether one of the conditions of the universal service obligation for directories is redundant. The normative question, whether such an obligation is desirable, is beyond the scope of this article and can probably not
3.4 Conclusion
87
be solved by a simple theoretical model. As shown by Cremer, Gasmi, Grimaud & Laont (1998), many reasons might justify the implementation of a universal service regulatory framework. A universal service obligation might be the solution for the non-internalization of network externalities or the under-provision of a public good. Besides, it might serve a redistributive or regional purpose. It is currently under discussion whether a telephone directory ts these policy goals. In 2005, the European Commission asked For how long will there be a need to keep directories and directory enquiry services within the scope of universal service?18 According to the same document, the common view of the industry was that a universal service obligation in the retail market is unnecessary and harmful to competition. This was followed by a Commission sta working document in 2006 that proposed to remove the provision of directories and directory inquiry services from the scope of universal service and leave the market to meet demand for these services19 . This proposal was not implemented, and as a response to the Public Consultation on Universal Service Principles in E-Communications of March 2010, many telecom operators reiterated that this obligation was an anachronistic obligation (Telecom Italia20 ) and increasingly unjustied (Telefonica21 ). It seems likely that directories and directory assistance will be left out from the universal service directive in a future revision. Moreover, 9 of the 27 countries of the European Union have not implemented the part of
18 COM(2005) 203, On the Review of the Scope of Universal Service in accordance with Article 15 of Directive 2002/22/EC, http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=COM:2005:02 03:FIN:EN:PDF. 19 SEC(2006) 816, Annex to COM(2006) 334 nal, http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX:52006 SC0816:EN:NOT 20 Telecom Italia (2010), Telecom Italia response to the Public Consultation on Universal Service Principles in E-Communications, http://ec.europa.eu/information_society/policy/ecomm/doc/library/ public_consult/universal_service2010/comments/telecom_italia.pdf. 21 Telefonica (2010), Telefonica Response On The Public Consultation On Universal Service Principles In E-Communications, http://ec.europa.eu/information_society/policy/ecomm/doc/library/ public_consult/universal_service2010/comments/telefonica.pdf.Other telecom operators, such as British Telecom, Deutsche Telekom, Orange, Fonecta and Vodafone, joined these arguments.
88
Directories
the universal service directive on directory services22 . In 2003, it was also noted at an OECD round table that there is no evidence that universal directory services would not be provided in absence of a universal requirement23 . Besides, a participant noted that directory service is mostly used by business users and not general consumers and should therefore not be regarded as a universal service. In the United States, the ocial viewpoint remains that directory service is one of the core services included in the universal service funds. At the same OECD round table, the Federal Communications Commission noted that access to directory service is used by a substantial majority of residential customers, is widely available, is essential for education, public health, and safety, and is consistent with the public interest, convenience, and necessity. Discussion on availability of data Another point that has been put forward in the policy debate on directories, is the availability of information. In our model we assumed that all information on companies is readily available to the platform. In reality, this might not be the case. Therefore, many parties have pointed out that there is a dierence between the universal service obligation for directories at the wholesale and the retail level. The retail level concerns the comprehensiveness and availability of the directory to the end user. The wholesale level includes the obligation on network operators to make directory data available to third parties. The non-availability of contact data might constitute a serious barrier to entry. It is reasonable to impose the obligation on all market players to share contact information, at least for the purpose of composing a directory.
(2010), Vodafone comments on European Commission questionnaire for the public consultation on universal service principles in e-communications, http://ec.europa.eu/information_society/policy/ecomm/doc/library/ public_consult/universal_service2010/comments/vodafone_uk.pdf 23 DAFFE/COMP(2004)19, OECD Policy Round Table, Non-Commercial Service Obligations 2003, http://www.oecd.org/dataoecd/43/35/33691140.pdf
22 Vodafone
A Proofs
89
Appendix A Proofs
Proof of Proposition 3.3.1 If we solve equation 4.5 with respect to the constraints q1 , q2 0 and q1 + q2 1, we nd four optimal regimes:
regime 1 regime 2 regime 3 regime 4 q1 0 2 3 2r 1r+r 2 3(1r) r(s1)s+s2 s3 + z 6(1s)s(r 2 +s2rs) q2 2 3 0
12r+ 1r+r 2 3(1r) 3 (s1)8r 2 (s1)s+2s2 2s3 z+r(5s3 3s2 2s+z) 3r 6(sr)(1s)s(r 2 +s2rs)
where z = (rs)2 (s1)s(3r2 +6rs+s2 4s). Regime 4, the dierentiated optimum, is a local maximum, at least when r > s. The global maximum is reached when q1 and q2 approach , because in that case prices are also negative and prot approaches +. O course, this regime will not be feasible. Regime 4 will also be infeasible for some parameter values. Since we have solved this model with Karush-Kuhn-Tucker conditions, we should check whether and when these conditions hold. The only non-zero Lagrange operator related to the constraint q1 +q2 1, prevails in regime 3. Therefore, if the Lagrange operator, which is equal to 3 should be implemented and q1 + q2 will be equal to 1. The corresponding price p1 is equal to 0, i.e. small listings are not paid. The condition q1 0 is violated if r s. Moreover, the second order derivatives show that regime 4 is a maximum only when r > s. In this case regime 1 will be implemented where q1 = 0. The other non-negativity s condition q2 0 is violated if r > 4s3 , in this case regime 2 will be implemented where q2 = 0.
Rs2 (r 3 (14s)3(1+ 1r+r 2 )s+r 2 (4+ 1r+r 2 +(74 1r+r 2 )s)) 9(1+r)2 (1+r) Rs2 (r(1+ 1r+r 2 +(2+5 1r+r 2 )s)) + , is positive, then regime 2 (1+r) 9(1+r)
90
Directories
Figure A1: Platform prot. Left panel: r=0.5, right panel: r=1000
Proof of Proposition 3.3.2 To trace back the eects of R, s2 , r and s, we perform a comparative statics analysis on the equilibrium quantities of the dierent regimes and on the prots under the dierent regimes. First, we nd that the derivatives under all regimes of quantity with dq dq respect to R and s2 are zero: dq1 = dq2 = ds1 = ds2 = 0. If we derive the dR dR 2 2 i equilibrium prots (with index i for the regime) towards R and s2 , we obtain:
d1 dR d2 dR d3 dR d4 dR = d1 ds2 d2 ds2 d3 ds2 d4 ds2 = 4 27(1 + r) 4rs 27(1 + r) 1 2r + = 1 r + r2 1 + 2r 2 1 r + r 2 + 2r 1 + 1 r + r2 (1 + s)
27(1 + r)2 (1 + r) (9r 3 (1 + s)s + 3r 2 (9s2 + 9s3 + (r s)2 (1 + s)s(3r 2 + 6rs + (4 + s)s))
108(1 + r)(r s)(1 + s)s(r 2 + s 2rs) rs(8s 11s2 + 19s3 + 6 (r s)2 (1 + s)s(3r 2 + 6rs + (4 + s)s))
108(1 + r)(r s)(1 + s)s(r 2 + s 2rs) s(8s2 + 7s3 + s4 + 4 (r s)2 (1 + s)s(3r 2 + 6rs + (4 + s)s))
108(1 + r)(r s)(1 + s)s(r 2 + s 2rs) s2 (r s)2 (1 + s)s(3r 2 + 6rs + (4 + s)s))) 108(1 + r)(r s)(1 + s)s(r 2 + s 2rs)
d d with = s2 in the case of dR and = R in the case of ds2 . If r 0, R > 0, s2 > 0 and 0 s < 1, then these derivatives are unambiguously positive. An increase in R or s2 always increases prot.
A Proofs Second, if we derive the optimal quantities to r, we obtain:

1 dq1 dr 2 dq1 dr 3 dq1 dr 4 dq1 dr = 0
91
0 1 r + 2 6(1 + r)2 (r s) 1 r + r2 1 r + r2 (r s)2 (1 + s)s 3r 2 + 6rs + (4 + s)s 2
r(1 + s)s + s2 s3 +
3(1 + s)s r 2 + s 2rs 6r 3 + 18r 2 s + 4s2 + 2s3 2rs(2 + 7s) + 6 r 2 + s 2rs
(r s)2 (1 + s)s 3r 2 + 6rs + (4 + s)s
(r s)2 (1 + s)s 3r 2 6rs (4 + s)s
Results for
3 dq2
dq2 dr
are similar, and, as can be expected in the comprehensive

dq 3
case, dr is exactly the same as dr1 , but with the opposite sign. With the same restrictions as above, we can show that dq1 > 0 and dq2 < 0 in the dr dr dierentiated and comprehensive case. Third, we can derive the optimal quantities towards s. It is easy to 1 2 3 4 dq1 dq1 dq1 dq1 show that ds = ds = ds = 0. The derivative ds is somewhat more complicated; and crosses the horizontal axes once in the r-space, i.e. it can be both negative and positive, depending on the value of r.
92 We can summarize all the comparative statics as follows:

Comparative statics of the Free Directory Restriction regime 1 dq1 dq2 dr 0 0 ds 0 0 dR 0 0 ds2 0 0 regime 2 dr ds dR ds2 regime 3 dr ds dR ds2 regime 4 dr ds dR ds2 dq1 0 0 0 0 dq1 >0 0 0 0 dq1 >0 < 0 if r < 1 > 0 if r > 1 0 0 dq2 0 0 0 0 dq2 <0 0 0 0 dq2 <0 < 0 if r > 1 > 0 if r < 1 0 0 d <0 0 >0 >0 d >0 >0 >0 >0 d < 0 if r < 1 > 0 if r > 1 <0 >0 >0 d if r < if r > if r < if r > >0 >0
Directories
< > < >
0 0 0 0
2 2 3 3
Interpretation: each cell is the derivative of column with respect to row.
where 1 , 2 and 3 are parameter values above 1 where there is a 2 dq1 switch in the comparative statics, i.e. 1 is the value of r for which ds = 0 and mutatis mutandis the same holds for 2 and 3 . Proof of Proposition 3.3.3 To nd the solution of the private monopolist who decides on quantities and size (s), we maximize equation 4.5 with respect these variables. We nd two regimes for this problem, which are equivalent to regimes 2 and 3 of the problem in section 3.3.1, supplemented with a size choice of s = 1 in regime 2 and s = 0 in regime 3. Further we nd that regime 3 dominates for all values of r. The only exception is r ; then the platform is indierent between both regimes. In sum, one can say that s = 0 is optimal for all values of r. Moreover, it is the only solution if r2 > 0. Proof of Proposition 3.3.5 timal regimes. Solving equation 3.3, we obtain four op-
A Proofs
q1 0 1 q2 1 0
93
p R
regime 1 regime 2 regime 3 regime 4
2r 1r+r 2 3(1r) R(r1) 1 + 2(1+r)(1s)s 2 2
R 2(1+r) rR 2(1+r)
12r+ 1r+r 2 3(1r) R(1r) 1 + 2(1+r)(1s)s 2 2
0
3(r1)2 R2 +2(1+r)2 R(1s)s2 8(1+r)2 (1s)s2 (1+r)2 (1s)2 s2 - 8(1+r)2 (1s)s 2 2
Similar to the proof of proposition 3.3.1, we solved this problem under Karush-Kuhn-Tucker conditions. Checking for non-negativity of the key variables and computing the Hessian matrices for these regimes, we conclude that that these are the four regimes that might prevail. The boundaries where these regimes apply are in the main part of this article.
Proof of Proposition 3.3.6 In the table below, we summarize all signs of the rst order derivatives. The results in proposition 3.3.6 follow straightforward from these tables.
Comparative statics of the Comprehensive Directory Restriction regime 1 dq1 dq2 dpR d dr 0 0 < 0 < 0 ds 0 0 0 0 dR 0 0 > 0 > 0 ds2 0 0 0 0 regime 2 dr ds dR ds2 regime 3 dr ds dR ds2 regime 4 dr ds dR ds2 > < > < > < 0 0 0 0 0 0 dq1 0 0 0 0 dq1 > 0 0 0 0 dq1 > 0 if if if if if if r r r r r r > < > < < > 1 1 1 1 1 1 > < > < > < 0 0 0 0 0 0 dq2 0 0 0 0 dq2 < 0 0 0 0 dq2 < 0 if if if if if if r r r r r r < > < > > < 1 1 1 1 1 1 dpR > 0 0 > 0 0 dpR 0 0 0 0 dpR > 0 if r > 1 < 0 if r < 1 > 0 > 0 < 0 d > 0 0 > 0 0 d > 0 if r > 1 < 0 if r < 1 < 0 > 0 > 0 d > 0 if r > 1 < 0 if r < 1 < 0 > 0 > 0
dqR < 0 0 > 0 0 dqR > 0 0 > 0 0 dqR > 0 if r > 1 < 0 if r < 1 0 > 0 0 dqR > 0 if r > 1 < 0 if r < 1 < 0 > 0 > 0
Interpretation: each cell is the derivative of column with respect to row.
Proof of Proposition 3.3.7 The proof of proposition is similar to the proof of proposition 3.3.4 and therefore omitted. It can also be derived from the comparative statics results for d . ds
94
Directories
Empirical Investigation of the Yellow Pages Industry
As we stressed in the introduction of this dissertation, the interaction between theory and empirics is often lacking. In this section, we try to bridge the gap between theory and empirics. In propositions 3.3.1 and 3.3.5, we make the claim that it is possible to predict the dierent regimes based on the parameters r and s. O course, these predictions are conditional on the restrictive assumptions made, but at least, it gives an empirically testable hypothesis. An empirical exercise can exploit variation in both parameters to test whether these parameters can explain the equilibrium quantities of listings. A major concern of such an application is endogeneity. Since r and s are already determined by the equilibrium quantities, we dont obtain valuable results. Therefore, in this section, we propose only a rst move, i.e. by suggesting how r and s might be estimated. For this application, we collected data on the European Yellow Pages. We estimate the major parameters and reconcile our results with our theoretical model.
B1
Industry Characteristics
The total Yellow Pages revenues worldwide are estimated at 30.9 billion US dollar in 2008 (Kelsey 2009). Yellow Pages attract 5.2 % of the global advertising market and employ 82300 people, nearly half of them are sales representatives. Print media are in turmoil everywhere in the world, and also the print Yellow Pages do not escape the crisis. Yellow Pages companies hold out relatively well in 2008, with revenues declining only 2.3% vis-`-vis 2007, a but share prices nosedived at the end of the rst decennium of the third millennium. If we compare the shares of the beginning of 2010 with the beginning of 2005, Seat (Italy) lost 99.7% of its price. Yell (UK) lost 91% and Pages Jaunes (France) 56%. The reason for this decline, besides worldwide nancial problems, is clear cut: the Internet. In 2008, Yellow Pages companies got 85% of their revenues from their print division and 15% from online outlets. It is likely that the print edition will vanish, though the decline is slowed by two facts24 . First, Yellow Pages are still seen as more reliable and extensive
24 The
Economist, Dial I for internet, May 22nd 2008.
B Empirical Investigation of the Yellow Pages Industry
95
Table B1: Variables: summary statistics

Variable Quantity Persuasive Ad Quantity Informational Ad Quantity Free Ad Price Persuasive Ad Price Informational Ad Usage Penetration Internet Penetration Circulation Ination (Index=100 in 2000) Income per Capita Mean 20937 96805 657706 1622 310 0.40 0.29 5807407 101 26038 Std. 10955 67842 694509 1092 199 0.12 0.17 2423109 8 5511
than online search engines. Second, rms are loyal to Yellow Pages, helped by an extensive sales force. Below, we describe the data, discuss the relevance of the advertisement types and present the results of our empirical study on Yellow Pages.
B2
Data
To test the mutual eect of usage and advertisements in Yellow Pages, the rst thing we need are data on usage and ads. Data on usage and advertisements are rare, even for those who advertise. Therefore, we contacted several companies to collect data. We received book level data on advertising, i.e. quantities, list prices and real prices. The latter is important because it is common practice to give a discount o the price. This creates a gap between list prices and real prices. Since advertisers decide on real prices and not on list prices, we use real prices in the investigation. Considering usage, we have book level data on distribution, but surveys on book usage are only available at the country level. To approximate the number of users, we multiply the percentage of people frequently using the book with the circulation in a certain area. Because the lack of detailed usage data, we carry out our analysis on country level and not on book level. We also have data on country characteristics which are used as control variables. These data are obtained from international sources, such as IMF, OECD and Eurostat. The sample includes 5 European countries. Those countries are rela-
96
Directories
tively homogenous in population and market structure. It is dominated by countries with a single market player. In two countries, there is some competition, e.g. from local city directories, but we treat those countries as monopolies too. The earliest data point is 1995, the latest 2006. Since the data is compiled from dierent sources, we often lack data for the complete time span; i.e. the data set is an unbalanced panel. The statistics of the variables are summarized in table B1.
B3
Identication
Reader Side Most users pick up the Yellow Pages to nd a particular company or a particular good or service. If the directory is used to nd a company which is already known to the user, then this is labeled as known search. If the directory is used to nd new suppliers, then this is unknown search. The distinction between both types of look-ups can have a substantial impact on usage. At one extreme, one might argue that the users only value raw information: the name and phone number of each supplier, be it classied by the category of the good or service that is supplied. In such a setting, it even might be the case that the users would pay for receiving a well classied directory that alphabetically lists all the suppliers in a certain category. This would be the case when each consumer is in a satisfactory or even optimal relationship with a certain supplier. The directory then is an ideal instrument for retrieving the coordinates of the particular supplier a consumer wants to patronize. But it merely serves the administrative purpose of an organized inventory of ones business contacts. Large colored ads oating around then could disturb the user that is only interested in nding the coordinates of his trusted supplier. Hence these large ads might reduce the attractiveness of the directory. At the other extreme, the large ashy ads in the directory serve to persuade consumers that have no relationship with a supplier yet, or look for change. These ads aim at the starting of a relationship with the particular supplier that uses the directory for this purpose. In this setting, larger and ashy ads may convey information that this type of user is keen for. It could separate the good suppliers from the bad in a signaling environment: the most ecient suppliers who have substantial turnover can pay for the larger ads, while the less ecient or inexperienced suppliers can not. The nuisance then comes from the small entries that merely provide for contact data. Since they signal no quality, they are not looked
97
at and hence redundant. They could be dismissed entirely by the phone directory provider. In the case of the rst extreme, people only use the directory as an index: only the company name and number suces. In the other extreme, people use the directory to nd out the reputation of a company: only persuasive ads count. But perhaps there is also a category in between: informative ads. Users do not search only for the telephone number, but also for more information about a company, i.e. fax number, e-mail or web address, opening hours or mobile number. Besides, it can also be interesting to have additional information about the activities of the company. If one is are looking for a replacement of ones central heating, when looking at the category central heating, from a simple list of companies and telephone numbers, one can not nd out whether a supplier is specialized in gas heaters or oil-red central heating. An informative ad can provide more information. We can model Yellow Pages as a usage generating market. By choosing the right amounts of quantities, a platform manages to create look-ups for the advertisers. One potential production function is a Cobb-Douglas function. U = BQ0 Q1 Q2 0 1 2
i controli i
(B1)
If we take logs, then we obtain: ln(U ) = + 0 ln Q0 + 1 ln Q1 + 2 ln Q2 + i ln controli (B2)
We will estimate this equation in the results section. Usage is captured by the percentage of people answering yes on the survey question Have you used the printed Yellow Pages last month?. These are considered regular users of the directory. To count the number of persuasive ads, we add up the larger advertisements with logo or graphic. The informational ads are smaller ads that still t into one column. These ads are in the within category alphabetical list of companies. Large persuasive ads are always accompanied by a regular entry in the alphabetical list, with a reference to the ad (e.g. see also advertisement on previous page). We dont have data on the free listings, i.e. what companies get if they dont advertise, but we can approach them by subtracting the total number of ads of the total number of companies in a country.
98
Directories
Advertiser Side Rysman (2004) treats ads as a homogeneous service and explains the amount or volume of advertising chosen by an individual business, but not the type. Equilibrium is reached when two osetting network eects keep each other in balance. On the one hand, more advertising leads to increased usage of the directory, i.e. users like ads in Yellow Pages unlike what seems to be the case in other media. And since advertisers like eyeballs, they buy more advertising as usage increases. This is the positive feedback loop that links the two sides of the market. A countervailing force exists because of the negative network eect of congestion that takes place when too much advertising crowds a category: an overwhelming number of large ads for plumbers running over several pages of a directory is not likely the medium that still another plumber will choose to list his services in. The two eects taken together lead to directories in which not every business decides to buy ad space. As noticed above however, the type of ad that is requested by advertisers can be quite dierent and hence this can result in a dierent impact regarding the network eects just mentioned. Large ads may attract more potential customers, but are perhaps annoying to users. At the same time they trigger the possibility that another business ad becomes unnoticed or that trade is diverted rather than created. The small ads on the other hand probably convince fewer potential clients. But because they contain additional information in a condensed and surveyable way, they probably contribute more to the usage of the directory. By this, they might inict a positive externality on the other advertisers of the directory. At the same time they might steal less business from the larger ads. In this respect, they might be complementary to larger ads, and the directory that has one large and two small ads might be a better product than the directory of the same size that is composed of two large ads. It is not hard to understand that the providers of directories will take these dierences into account. The result will be inter alia a dierent pricing strategy for each type of ad. These pricing strategies will take into account that the contribution to usage of each type of ad is dierent and the eect of the ad on reaching potential customers. The congestion eect might be circumvented to some extent by including dierent types of ads that do not compete for the attention of the user. This leads to the claim that the dierent types of ads might be heterogeneous services. Following Rysman (2004), we model the advertising site as a Cobb-
99
Douglas function. Similar to the usage production function, it contains the major ad type and control variables. Besides, the price also depends on the number of users U .
pj = AQ0 0 Q1 Q2 U 3 1 2 i controli i
(B3)
If we take logs, then we obtain: ln pj = +0 ln Q0 +1 ln Q1 +2 ln Q2 +3 ln U + i ln controli (B4)
We will estimate this equation in the results section. In this price equation, the quantities are the same as in the usage equation. As explained above we prefer real prices above rate card prices. For usage, i.e. the number of consumers, we have no direct data. Therefore, we approach this number by multiplying the percentage of people regularly using the directory with the circulation. This may lead to an underestimation of the number of users, because directories are distributed to families which consist usually of more than one person. Robustness checks show though that multiplying this usage gure with average family size does not change the results below.
B4
Results
Reader side Do users value a directory mostly for its listings, small ads or large ads? In this section we test this question and present the results. Because we work with an unbalanced panel, consisting of 5 dierent countries, we use a least squares estimator with xed eects. This estimation technique allows to control possible characteristics of particular countries - even without measuring them, as long as those characteristics do not change over time. The number of observations is quite limited, though our results appear to be quite robust to other specications of the model (see table B3). Rysman (2004) proposed instrumental variables in his study on the Yellow Pages in the United States. In the usage equation, the advertisement level is instrumented by the number of people covered by a directory. In the advertisement equation, he instruments usage by the people that recently moved. For advertisements he uses the earnings level in a county, because this approximates the hourly wage and can be seen as a cost shifter. We do not apply these instruments because we have only a limited number of observations and we do not have the necessary data to
100
Directories
Table B2: Results of the OLS regression

Quantity Listings Quantity Small Ads Quantity Large Ads Internet Income per Capita Usage CPI Constant Observations R-squared -13.52 (1.21) 24 0.73 Usage 0.371 (2.98)*** 1.697 (1.82)* 0.207 (0.25) -0.109 (0.22) 0.18 (0.23) Price Small Ads -0.023 (0.45) -0.851 (2.26)** -0.155 (0.57) 0.373 (2.35)** 0.346 (1.07) 0.253 (3.07)*** -0.004 (0.64) 9.905 (2.63)** 24 0.96 Price Large Ads 0.137 (3.02)*** 1.744 (5.29)*** -1.352 (5.71)*** 0 0 -0.423 (1.5) 0.17 (2.35)** 0.027 (4.92)*** -2.52 (0.76) 24 0.94
Absolute value of t statistics in parentheses * signicant at 10%; ** signicant at 5%; *** signicant at 1%
instrument. In this section, we present the results for the OLS regression. The robustness checks with seemlingly unrelated regression estimation can be found in table B3. As can be seen from table B2, the coecients of the quantities of the small advertisements and the free listings are signicantly dierent from zero in the usage equation. The coecient of small ads is quite large: increasing the number of small ads with 1 percent increases the number of users with 1.7 percent. Large ads seem to have no eect on usage. These results indicate that readers value most the small ads in the book, but listings (=only name and telephone number) are also appreciated. Large advertisements do not aect usage. While large advertisements can be interesting for advertisers because they attract a lot of eyeballs, users are not interested in the additional (persuasive) elements. Advertiser side In table B2, the results for the price equation are given in the third and fourth column. Since price, quantity and usage are given in logs, we can easily interpret the coecients. As expected, both ad categories have a negative and signicant own elasticity. Higher usage increases the willingness to pay in both the small and
101
large ad category. In the large ads price equation, the quantity of small ads has a positive and signicant eect on the price. This means that more small ads increase the willingness to pay for large ads. The amount of large ads has no eect on the price of small ads. The quantity of free ads aects the price of large ads positively. The fact that internet penetration has a signicantly positive eect on the price of small ads, might be related to the fact that the internet is not a substitute, but a complement for additional information, e.g. it would be of no use to add a mail and website address if no users had internet access. As explained above, internet might rather be a substitute for large ads, because it serves the same needs: in depth search for a product or service.
B5
Robustness Check: SURE
To check for the robustness of our OLS results, we apply seemingly unrelated regression (SUR, Zellner (1962)). SURE allows for correlated errors between equations that are seemingly unrelated but based on the same data. As can be seen in table B3, the results of the seemingly unrelated regression are similar to the OLS regression.
B6
Reconciling the Theoretical Model with Empirical Findings
With this empirical results in mind, it is interesting to look back to the theoretical model. The most interesting application of these results comes from the usage equation. If we add a third advertisement category to the theoretical model, say q0 (which is associated with s0 < s1 ), then we get results for estimations of r0 , r1 and r2 . Point estimates for these parameters are 0.37, 1.70 and 0.21 respectively25 . If we ignore r0 for a moment, the estimates of r1 and r2 can be merged to a r estimate. Since r1 r = r2 , r = 8.2. One can wonder, if ratio is really that large, why is q2 still in the directory? The answer can be seen in gure 3.1. With a high r, the result depends on the value of s. If s is smaller than 0.25, then q1 is oered for free; if s is larger than 0.75, then only q1 is oered. Between
25 See
table B2 in this appendix.
102
Directories
0.25 and 0.75 both quantities are oered and a positive price occurs in the market. Even if no one likes the large advertisements, publishers will still oer them as long as the dierence in views is large enough. If our models extends to three categories and the listing is small enough, then it will be oered for free and the whole market is covered. If the large advertisement is substantially larger than the small advertisement, then all three advertisements will be oered. In most directories, the free listing is only one line, which ts with our prediction that, if platforms can choose the size dierence, they will make it as large as possible. We dont have size information for all observations, though rough estimations show that the size of the average small advertisement is between 0.04 and 0.11 of size of the average large advertisement. This is relatively small but still a lot larger than the listing. The average listing is between 0.002 and 0.006 of the size of a large advertisement. With the assumptions of our model, this size dierence would lead to a solution where all advertisers are included and the smallest advertisement is given away for free. Our model does not incorporate the possibility of a third advertisement type, which explains why small advertisements are paid in reality. It is harder to reconcile the estimation of the advertisement equation with the theoretical model. The reason is that we have not modeled the potential congestion eect in the directory. We have assumed that the number of rival advertisements have no inuence on the visibility of the 1 2 1 2 advertisement, or ds1 = ds1 = ds2 = ds2 = 0. What we do model is dq dq dq dq the self-selection eect. Therefore, an increase in the amount of small advertisement should not only decrease the price the small ads, but also of the large ads. We do nd the negative eect of an increased quantity on the own price, but we do not nd an eect on the cross price. A shift in large ads has no eect on small ads, while small ads have a positive eect on large ads. The theoretical model falls short in explaining this sign.
Table B3: Results of OLS and SURE
Quantity Listing
Quantity Small Ads
Quantity Large Ads
Internet
Income per Capita
Usage 0.371 (2.98)*** 1.697 (1.82)* 0.207 (0.25) -0.109 (0.22) 0.18 (0.23)
Usage 0.371 (3.44)*** 1.697 (2.10)** 0.207 (0.29) -0.109 (0.26) 0.18 (0.27)
Usage
CPI -13.52 (1.4) 24
Constant
Observations R-squared
-13.52 (1.21) 24 0.73
OLS Price Small Ads -0.023 (0.45) -0.851 (2.26)** -0.155 (0.57) 0.373 (2.35)** 0.346 (1.07) 0.253 (3.07)*** -0.004 (0.64) 9.905 (2.63)** 24 0.96 Price Large Ads 0.137 (3.02)*** 1.744 (5.29)*** -1.352 (5.71)*** 0 0 -0.423 (1.5) 0.17 (2.35)** 0.027 (4.92)*** -2.52 (0.76) 24 0.94
SURE Price Small Ads -0.023 (0.55) -0.851 (2.77)*** -0.155 (0.7) 0.373 (2.88)*** 0.346 (1.32) 0.253 (3.76)*** -0.004 (0.78) 9.905 (3.22)*** 24
Price Large Ads 0.137 (3.70)*** 1.744 (6.48)*** -1.352 (6.99)*** 0 0 -0.423 (1.84)* 0.17 (2.88)*** 0.027 (6.03)*** -2.52 (0.93) 24
Absolute value of z statistics in parentheses * signicant at 10%; ** signicant at 5%; *** signicant at 1%
103
104
Directories
Extensions to the Model
In this appendix, we enrich our initial model with some extensions. These extensions do not shed light on the main question of this article, i.e. whether a universal service obligation is necessary and desirable. Though, these extensions provide additional insight on the functioning of the Yellow Pages industry. All extensions are based on the model where pR = 0.
C1
Negative Utility from Advertisements
In section 4.3, we implicitly assumed that r1 and r2 are positive though it is likely that in some markets this assumption is violated. Especially in media markets, advertisements are often seen as a nuisance (Anderson & Coate (2005) and Peitz & Valletti (2008)). In telephone directories, this is less likely since advertisements contain relevant information for users, though it might be that some advertisement types do not contribute to usage, on the contrary, they decrease usage (see also the empirical application). In our model, it should hold that at least one of the listing formats contributes to usage, i.e. either r1 or r2 is positive. Otherwise there would be no usage on the platform. But the model can perfectly cope with one negative parameter (either r1 < 0 or r2 < 0). Proposition C.1 If r1 < 0 and r2 > 0, then the platform only oers large listings. If r2 < 0 and r1 > 0, then the platform always oers small listings. The platform will also oer large advertisements if the size dierence is large enough. This proposition is visualized in gure C1. In the left panel, below the x-axis, we plotted the case where r1 < 0; in the right panel the case where r2 < 0. As be seen from the graphs, the border r < s extends to negative values of r as well in the case where r1 < 0. Since small listings are less protable and are disliked by users, they do not appear anymore in the directory. In the case r2 < 0, things are somewhat more complicated. The platform will keep out the large listings only if s is large; more precisely if s 3 r > 4s3 (and for all r-values if r < 0 and 4 < s < 1). Small listings will 2 s4s2 2 s (15s+4s2 ) be oered for free if |r| . 14s
C Extensions to the Model
105
Figure C1: Dierent Solutions in the r-s-space. Left panel: if r < 0, then r1 < 0, right panel: if r < 0, then r2 < 0
C2
Endogenizing r (as a function of s1 and s2 )
The parameters r and s were treated independently in section 4.3, though one can argue that they are related. We argued that users might more easily choose a large listing when they compare both types. Though when deciding on using the directory or not, users might prefer smaller listings because these listings use less space for the information provided. Therefore, it makes the directory easier to handle and provides a better overview. We assumed that the number of looks is directly related to size. One can argue that the valuations r1 and r2 are also related to the size of a listing. After all, it is exactly the size of the listings that often blurs the informational value of a listing. Therefore, we can change usage equation 3.2 in order to include this nuisance eect of size. qR = v(q1 + q2 ) n(s1 q1 + s2 q2 ) (C1)
The rst term, with parameter v, captures the taste for variety, i.e. the pure network externality of adding another advertiser to the platform. The second term, with parameter n, captures the nuisance of the thickness of the directory. To simplify the analysis, we normalize n to 1. Note that we can rewrite equation C1 as qR = (v s1 )q1 + (v s2 )q2 , i.e. r1 = v s1 and r2 = v s2 . The ratio r is equal to vss22 . Since s1 < s2 , r1 > r2 vs always holds. Therefore, we can use the right panel of gure C1 to analyse the evolution of the equilibrium r-ratio.
106
Directories
Figure C2: Dierent solutions in the v-s-space
Proposition C.2 are no users.
If v s1 , there will be no platform since there
In all other cases, the platform will oer small listings. If s1 < v 4s1 3 , then the platform oers only small listings. 4s2 s22 s2 (15s+4s2 )s22 , then the platform implements the If v > 1+5s 4s2 s22 s2 (15s+4s2 )s22 4s1 comprehensive solution. If 3 < v , then 1+5s the platform is dierentiated. In gure C2, the outcomes are shown in function of v and s. The horizontal line is the size of the large listing s2 . Three out of the four solutions of the model in section 4.3 persist in this adapted model. In addition, if the taste for variety is not suciently large, then there will be no platform on the market. The claim that the absolute size of the listings does not matter in the choice of the quantities does not hold anymore. Proposition C.3 If size acts as a nuisance factor to the informational value of the platform, then an increase in s2 will lead to an increase in small listings and a decrease in large listings. This proposition holds for solutions 3 and 4 (comprehensive and differentiated platform). In the case where only one listing format is oered,
107
Figure C3: Relative prices (p = r=1000
p1 p2 ).
Left panel: r=0.5, right panel:
none of the parameters aect the solution. Note that an increase in s2 can also induce a shift between the solutions and makes it likelier that there will be no platform in the market. If size is not seen as a nuisance to the composition, but as a value added, the results change drastically. We can write the usage function as qR = v(q1 + q2 ) + n(s1 q1 + s2 q2 ). With n normalized to 1, the r of section 4.3 now reads as r = v+s1 or r = v+ss22 . v+s2 v+s Proposition C.4 If size generates a positive eect on usage, then the platform always oers both types. If the size dierence is large enough, then small listings are oered for free. (If v = 0, then only large listings will be oered.) Since r = v+ss22 and 0 s < 1, for every combination between v 0 v+s and s2 0, it holds that s r < 1. If we have a look at gure 3.1 again, then we see that the only possible solutions are the comprehensive and the dierentiated solution.
C3
Price discrimination
If a listing of size one costs one euro, then we would expect that a listing of twice the size would cost less than two euro. If this does not hold, every sender can buy two listings and obtain the same eect. This explains why the bundle costs less than the sum of the individual parts. Busse & Rysman (2005) nd the existence of this second degree price discrimination in Yellow Pages and link it to competition. If there are more platforms
108
Directories
in the market, then the price discrimination is larger, i.e. large ads are relatively cheaper vis-`-vis small ads. a Using the model, we can have a look at the price ratio in function of the size ratio. We note that the absolute size (s2 ) has no inuence on prices. Prices are only determined by the relative size (s) and the relative valuation (r). Proposition C.5 If both types are oered, then large listings always pay more per view than small listings. There is a cross-subsidization from large senders to small senders. The reason behind this cross-subsidization is, in the case two formats are oered, that the relative contribution of small listings to usage is larger than their visibility (r > s). Figure C3 shows the curvature, under r < 1 (left panel) and r > 1 (right panel). On the horizontal axes is the ratio of sizes (s). On the vertical axes, you nd p = p1 , the ratio of prices. If there is second degree p2 price discrimination, we would expect a price ratio above the 45 degree line. In both situations, we observe the reverse: if the size dierence is maximal (s=small), then p1 = 0 and therefore p = 0. The ratio gradually increases towards 1 when both sizes are the same. One of the reasons why we dont nd price discrimination is that we impose that senders buy at most one listing. Hence it is not possible to substitute a large ad by several small ads. Multiple listings are seldom observed in Yellow Pages, though this does not tell anything about the possibility. In sum, our model abstracts from the second degree price discrimination observed in Yellow Pages. On the contrary, the eects we investigate predict relative higher prices for the large ads.
C4
Results for a Ramsey Planner
We contrast the optimal quantities of the private monopoly with a public monopoly platform that maximizes welfare under a break-even constraint. Therefore, it takes into account the welfare of receivers and senders. The welfare of the receivers is equal to the sum of the net utilities (see equation 3.1). The welfare of the senders is also the sum of the net utilities and consists of two parts: senders buying a small listing and senders buying a large listing (see equation 3.4). Parameters and present the importance of sender welfare versus receiver welfare. The objective function of
109
Figure C4: Optimal welfare solutions in the r-s-space. Between brackets, the optimal values for q1 and q2 are shown.
the Ramsey planner reads: W = WA + WR (C2)
We impose the same restrictions as in the private monopoly case: q1 , q2 0 and q1 + q2 1. Proposition C.6 A welfare maximizing monopolist will always oer its listings for free and fully cover the market, though it oers only one type. s The platform will oer only small listings if r > ss22+R ; only large +R listings if r . Note that if the Ramsey planner cares only about receivers ( = 0), then = 1. If he cares only about senders ( = 0), then = 1 . If r < 1 or s r > 1 , then there is no conicting interest between receivers and senders: s the optimal solution is the same. But if 1 < r < 1 , then the Ramsey s planner has to disappoint either the receivers or the senders. This can be seen in gure C4. If a Ramsey planner can choose the size endogenously, then he chooses s s = 0 if r > ss22+R and s = 1 otherwise. In other words, s = 0 is +R
110 implemented if q2 = 1, s = 1 if q1 = 1.
Directories
C5
Competition
We assumed the market structure is a monopoly, i.e. there is a single platform. We can examine what happens if two or more platforms enter the market within the context of the framework introduced. Suppose that receivers singlehome, i.e. they use at most one platform to perform their searches. Competition will maximize the utility of the receivers. Therefore, platforms will oer only one listing format and give it away for free. If there exist an innitesimally small entry cost, then the incumbent monopoly stays a monopoly but it makes no longer prots26 . Proposition C.7 If users singlehome and an innitesimally small entry cost exist, then the incumbent monopolist becomes a contested monopolist. He will only oer the listing format that receivers like most and will provide it for free. This proposition is partly driven by the fact that receivers are homogenous (except for their opportunity cost). Therefore, they all choose the platform with the largest gross utility. This result can change if heterogeneous receivers are introduced. Note that competition causes the contested monopolist to implement solutions that are similar to the welfare solutions. Corollary C.8 If the monopoly is contested, then it implements the optimal quantities that prevail under welfare optimizations. This does not s hold if 1 < r < ss22+R . In this case, it oers the optimal welfare solution +R for readers only, but not for total welfare. This corollary can easily be explained by observing gure C4. Since we assume that readers singlehome, competition is focused on this side of the market. Since readers pay no price, the only option is to maximize welfare for them, leading the monopolist to implement always the reader welfare solution. The only conict zone of reader welfare and total welfare s is when 1 < r < ss22+R . Again, this result can be altered too, e.g. when +R singlehoming on the advertisement side is introduced.
there is no entry cost, results depend on how receivers decide using a certain platform or not when utility is exactly the same. If the market is split evenly among the platforms, there will be several platforms that all oer exactly the same directory. The argument is similar to the classical analysis of Bertrand competition.
26 If
111
Other Applications An interesting avenue for further research is to apply the framework to other markets. These markets should have the following in common with our telephone directories model. There should be two distinctly dierent demand sides that value each others presence. One side can access the platform under dierent forms and it is important that they can self-select their type. The other side values one type over another in the direct comparison, but potentially values them dierently on their contribution to the platform. One example is a shopping street. Consider a shopping street as a platform that connects buyers with shopkeepers. Assume now that shops are dierentiated by their display windows. Shops with large windows are more interesting to shopkeepers because they have a larger chance to be visited. It is a priori unclear whether buyers prefer small or large windows. Buyers can have a double preference for the size of the windows. On the one hand, a shop with a large window is more attractive than a shop with a small window. On the other hand, if a buyer considers the entirety of shops, i.e. the shopping street, the large windows might be relatively annoying because they reduce the overall visibility and make the shopping street considerably longer. Similar to the telephone directories example, the equilibrium number of shops with large and small windows can be described by parameters r and s, at least if there is one platform that controls all the shops in a certain area27 . Another example might be a dating event. Suppose that men and women are looking for a partner. Women are not charged. Men are charged but they can choose out of two types: they can appear at the event in suit or casual wear. If they compare them directly, women prefer men in suits over casually dressed men. But when they consider to go to the dating event or not, they might prefer an event with more casual men, because it means that they dont have to dress formally themselves. Again, the potential dierence between direct comparison and contribution to the entirety of the platform might drive the decision of the platform to dierentiate or not. With some adaptations, our framework can also be applied to search engines. Many search engines, such as Google and Bing, allow companies to buy the highest ranking in the search results. As such, companies have
27 While this is not so likely in an older city, it might be true in a shopping mall or newly developed area. For the implications of monopoly versus dispersed ownership of shops, see Nocke, Peitz & Stahl (2007).
112
Directories
the choice to appear as a regular entry or a sponsored entry. For the surfers, the number of regular and sponsored search results determine the use of a search engine. When they use the engine, it is likely that a surfer clicks more on sponsored links (at least on average). Therefore, our results might also hold for search engines. Note that the model can also be applied to online media. One of the big questions is whether online media can successfully introduce dierent prices, i.e. by charging at least one group of readers for preferential access. This so-called freemium model - a combination of free and premium - has been proposed as one of the solutions for online media. To analyze this model within our theoretical model, replace advertisers by readers and readers by (voluntary, non-paid) bloggers. QR becomes then the number of blog posts, s1 and s2 can then be interpreted as the amount of blog articles that can be read by both reader types respectively. Bloggers value the amount of exposure. Whether it will be optimal to charge dierent prices depends on the parameters of the model. Further adaptations can be made to upgrade the model to a real news platform with paid journalists. In that case, journalists will enter as a cost factor or they will become part of the platform management. In both cases, the prot function of the platform will change and therefore the result will be dierent from those presented in this article.
Proofs of Extensions
Proof of Proposition C.1 The rationale behind this proposition is similar to proposition 3.3.1. It can be seen as an extension or rectication of this proposition. As in proposition 3.3.1, solution 4 is the unconstrained optimum of the problem. The same constraints apply: q1 + q2 1 and q1 , q2 0. In the case where r2 > 0 and r1 < 0, it suces to investigate q1 0 since the platform will never set q2 0 or q1 + q2 > 1, even if that would be possible. The constraint q1 0 is violated if r s, also for values r < 0. Therefore, the analysis is not altered. Things are somewhat more complicated in the case where r1 > 0 and r2 < 0. The constraint q1 0 will never be violated since this would induce a no-receivers platform. To check whether q1 + q2 1 is not violated in solution 4, we have to investigate in which range the Lagrange operator is positive. We nd two r-regions where the assumptions are
D Proofs of Extensions
113
Figure D1: Dierent Solutions in the r s-space. Left panel: if r < 0, then r1 < 0, r2 > 0, right panel: if r < 0, then r1 > 0, r2 < 0.
violated: r
3
s4s2 2
s2 (15s+4s2 ) 14s
if r > 0 and r
if r < 0. We also have to check whether q2 =
3 2
+ zr(5s 3s 2s+ 2s6(sr)(1s)s(r2 +s2rs) z) 0 (with z = (r s) (s 1)s(3r + 6rs + s s2 4s)). This is violated in two r-regions: r > 4s3 in both cases (r < 0 3 and r > 0) and for all r-values if r < 0 and 4 < s < 1.
s2 (15s+4s2 ) 14s 3r 3 (s1)8r 2 (s1)s+2s2 6(sr)(1s)s(r 2 +s2rs) 2 2
s4s2 +2
Proof of Proposition C.2 We solve the same model as in section 4.3, though we alter the receiver demand. Since the new receiver equation can be written in a similar way as the old usage equation, we obtain qualitatively the same results. Though, solution 1, where (q1 , q2 ) = (0, 2 ) 3 is not an optimal solution anymore. This solution can be replaced by solution 5, (q1 , q2 ) = (0, 0), which prevents that qR would drop below zero. Rewriting the old parameters r1 and r2 as v s1 and v s2 , allows interpreting the solutions in the r s-space. r is equal to vss22 , i.e. r vs is determined by s (in the graph) and v and s2 (which are xed in the graph). There are two cases that should be investigated. First, if v > s2 , then both quantities are valued positively in the usage equation. Second, if v < s2 , then at least one quantity is valued negatively. If s1 < v < s2 , then small listings generate a positive eect on readership, large listings generate a negative eect. If v < s1 , then both listings generate a negative eect and it is optimal for the platform to oer no listings at all. In gure D1, the optimal solution choice is given for v > s2 (left panel) and v < s2 (right panel). The r(s) corresponding with the s-values is given
114
Directories
by the straight green line. In both cases, the relative value of v versus s2 determines the slope. If v > s2 , the larger v, the atter this line. If v < s2 , the larger v, the steeper this line. As can be seen in the left panel, the r determination crosses solution 3 and 4 (for v set twice as large as s2 ), but if v was smaller, then it would have crossed solution 2 as well. If we know the value s, then we can immediately derive from the graph which solution will prevail. The same holds for the case v < s2 . Note that if r(s) > 0, this indicates that both r1 and r2 are smaller than zero and that no listings are oered in optimum. If r(s) < 0, then the line always crosses the solution 2, 3 and 4. On the right panel, r(s) is drawn for v half as large as s2 . Proof of Proposition C.3 In solutions 3 and 4, the optimal quantities read: Solution 3:
q1 q2 (2+s)s2 +v (1+2s)s2 v+
(1s+s2 )s2 (1+s)s2 v+v2 2

3(1+s)s2 3(1+s)s2
(1s+s2 )s2 (1+s)s2 v+v2 2
Solution 4:
q1 q2 s(s2 v)v+ sv 2 (3v 2 +s(4s2 8s2 v+v 2 )) 2 sv 2 (3v 2 +s(4s2 8s2 v+v 2 )) 2 6s(ss2 (s2 2v)+v 2 ) ss2 (2s2 5v)v+3v 3 s2 6v (ss2 (s2 2v)+v 2 )
dq If we derive these quantities to s2 , then ds1 > 0 and 2 in the regions where these solutions prevail.
dq2 ds2
< 0, at least
Proof of Proposition C.4 We can draw the same graph as for proposition C.2. The r(s) corresponding with the s-values is given by the straight v green line. This line always ends in point (1,1). The starting point (0, v+s2 ) depends on the values of v and s2 and should lie between zero and one. If v = 0, then r = s and we obtain solution 1, i.e. only large listings are oered (see the proof of proposition 3.3.1. In graph D2, the r(s)-line is shown for v = 2 and s2 = 1. Proof of Proposition C.5 To check whether the price per view is 2 1 higher for large listings, one simply has to check whether p2 > p1 , for s s p1 2 every solution in the relevant parameter space. The ratios s1 and p2 s
D Proofs of Extensions
115
Figure D2: Dierent solutions in the r-s-space.
are plotted in gure D3. On the left panel, r = 0.5, on the right panel r = 1000. The s-values on the left hand side of the dotted line are the relevant ones, because on the right hand side of this line only one type is oered in the market. Proof of Proposition C.6 We solve equation C2 with the same KarushKuhn-Tucker conditions as under the private monopolist, i.e. q1 , q2 0 and q1 + q2 1. We obtain three groups of solutions: an unconstraint optimum, two solutions where either q1 or q2 is zero and two solutions where q1 + q2 = 1. We nd further that only the Lagrange operator related to this last solution is always binding in the region 0 < s < 1. Therefore, (1,0) and (0,1) are the only relevant solutions to our problem. s We obtain the critical r = ss22+R by comparing the two welfare +R levels, under the rst and second solution. Equivalently we can say that a Ramsey planner implements solution 1 if s < s2 +RrR and solution 2 rs2 otherwise. The critical values for the monopolist maximizing the welfare of one side of the market only, can be found easily by setting or equal to zero. Proof of Proposition C.7 Suppose that a platform would oer a quantity combination that does not maximize receiver welfare. Then another
116
Directories
Figure D3: Price ratios. Left panel: r=0.5, right panel: r=1000
platform can oer a combination that slightly improves receiver welfare and the rst platform will have no users and hence no senders. Since there are no costs in our model, it is always feasible to oer all listings for free. If we introduce costs to our model, then the platform can never oer the solution where there is only one type and that type is free of charge.
Chapter 4
Metaplatforms. The Eect of Aggregators on Content and Prots
118
Aggregators
4.1
Introduction
In this article, we investigate the role of aggregators or metaplatforms. Platforms, in the context of two-sided markets, bring together two distinctly dierent demands, e.g. buyers and sellers, readers and advertisers or card holders and merchants. Metaplatforms, as we dene them, are vehicles that aggregate one side of the market over all platforms, to serve the other side of the market. To clarify the meaning of metaplatforms, consider the market for online real estate classied ads. A traditional real estate platform brings potential buyers and sellers together, by setting up a website where sellers pay to announce their property for sale. Potential buyers might be charged to see these announcements. Metaplatforms collect these seller announcements from the dierent platforms and present them together on their own website. Potential buyers can access all announcements and are redirected to the original platform. A metaplatform creates both opportunities and threats for the traditional platform. In our theoretical model, we show the two adversary eects of a metaplatform. On the one hand, a metaplatform increases the number of buyers in the web ecosystem, but on the other hand some of the revenues related to the buyers are absorbed by the metaplatform, at the expense of the traditional platforms. Platforms can decide whether to allow or disallow the metaplatform. Our results show that despite this choice, platforms do not always take the most protable decision. Though disallowance of the platform would often be more protable, both platforms will allow the metaplatform, because they dont want to end up as the only platform not allowing (while the other does). Our results are particularly relevant for the current debates on aggregators, e.g. on news aggregators such as Google News and real estate aggregators such as Funda in the Netherlands and Google Maps in Australia. In these debates, believers and non-believers are diametrically opposed to each other. Believers are proponents of the so-called link economy, and therefore also of aggregators. As Jarvis (2008) argues: Links are the new currency of media. Links can be exploited and monetized; get links and you can grab audience and show ads and make money. But non-believers rmly think that aggregators steal more attention than they redistribute. Mishkin (2009) provides evidence that news aggregators retain the majority of the readers. Readers scan the headlines, but rarely click further.
4.1 Introduction
119
The same observation is made by research rm Outsell, as documented in Wauters (2010). It is clear that proponents of aggregators stress the rst eect we described, i.e. the market expansion, while the opponents stress the second eect, i.e. demand shifting in favor of the aggregator. News Corporation, owned by Rupert Murdoch, and the press agencies such as Associated Press, are well known as strong opponents of aggregators. Belgian newspapers also tried to restrict the use of links, without much success (Soureau, 2010a, 2010b). Whether traditional media gain or lose from aggregators, remains an open question. It is for sure, though, that aggregators play an important role in the content distribution and therefore also indirectly in the content production (see also Pew (2011)). Similar discussions are also held in court. In 2006, the Dutch society of real estate agencies (NVM) took two aggregators to court to stop the aggregating on the base of copyright law. In a preliminary statement, the court argued that copying the address, price and 1 to 1.5 rule description are not considered as a copyright infringement, but as a quote. In this article, we focus on the economic eects of metaplatforms. In Appendix B, we provide a discussion on the legal arguments1 . Metaplatforms and aggregators are by no means a brand new concept. For newspapers, aggregators such as Dialog and Lexis Nexis were erected in 1972 and 1973. The main dierence, though, is that they were no real competitors for newspapers, but served more as an archive than as a direct news medium. But also other real world examples, such as book stores can be interpreted as metaplatforms. Book stores aggregate books from dierent publishers, who are platforms themselves since they bring together authors and readers. We like to stress here the importance of IT for the concept of metaplatforms. At least two features give metaplatforms a lot more relevance
1 Note that the discussion on the economics of copyright infringements by aggregators is related to the discussion on digital music and movies a couple of years ago. Similar to the spidered websites, it has been shown that a potential copyright infringement for digital music should not necessary harm music distributors, due to the sampling eect (Peitz & Waelbroeck 2006). A similar discussion is whether the copying of pages from books by the Google Books project can be considered as fair use. Travis (2006) argues that like the samples on iTunes and similar digital music services, the primary utility of Google Book Search will be to enable internet users to preview works about which they lack the information to make a purchasing decision. [...] Google is salvaging entire libraries full of dusty, crumbling books while creating a highly ecient marketing platform for authors.
120
Aggregators
in the internet age: digitalization and hyperlinks. Digitalization makes it much more ecient to store and reproduce content. While the metaplatform described for online real estate classied ads would also be feasible for oine classied ads, it would cost a lot more eort to retrieve and reproduce the information. Due to a far-reaching digitalization, the use of so-called bots, crawlers, harvesters, scrapers or spiders makes it easy to ll databases with information. A second important, and possibly determining feature of the internet is the existence of hyperlinks. These hyperlinks make it possible to switch from one website to another; or from platform to platform. The possibility of websites to redirect consumers easily to other platforms might be perceived as a threat, because it is easier to lose consumers. But linking also increases the value of a website signicantly, without having to invest in own content or storage space. The remainder of this chapter is organized as follows. The next section provides an overview of the literature on aggregators and the economic concepts that are related to metaplatforms. Section 4.3 describes the model and the equilibrium results on the eects of a metaplatform. We also discuss some extensions to our model. Section 4.4 concludes and sheds light on the insight of our model in the context of the current debates on aggregators.
4.2
Related Literature
In this section, we discuss the two streams of literature that are related to our model: network economics and two-sided markets. As was discussed in the introduction, one of the dening features of the internet is the use of hyperlinks. Through these links, a network is formed. If links are costly to form and maintain, then individuals or rms will trade o the cost against the expected benets. Bala & Goyal (2000) investigate how equilibria are formed in social networks. Jackson & Wolinsky (1996) study the stability and eciency of these networks, also in a non-cooperative game where individuals can form or sever links. Jackson (2003, 2008) provides a survey of the literature. The article closest to ours is Dellarocas, Katona & Rand (2010). They apply these link formation theories to websites, and question whether sites will link to each other. They show that links can be benecial by reducing competition and duplication eort. Links do not always form, even though they would lead to more content and higher aggregate prots. Ag-
4.2 Related Literature
121
gregators prove to be protable if they oset the loss in ad revenues by attracting more trac. Their ndings also show that aggregators increase competition. When sites cannot link to each other, they react by overinvesting in content. One way out of this overinvesting is linking to another website. As the aggregator shrinks prots, it becomes more interesting to link to another website (and stop investing in own content). In an extension, they also make the decision to allow the aggregator endogenously. In a market where there is no outside option, sites are always better o if they coordinate to refuse links. From Dellarocas, Katona & Rand (2010), we borrow the search process of buyers. The set-up of their model deviates in three important points. First, they apply a two-sided market model with readers and sellers (whereas we apply a three-sided model with buyers, sellers and advertisers). The authors treat content as a cost factor (e.g. newspapers), while we perceive content as a revenue side (e.g. real estate platforms). Second, aggregators can at most copy one website (or link to it in their terminology) whereas in our model the aggregator can copy information of any website. Strictly speaking, their type of aggregator does not aggregate information, but acts as a quality lter for the reader. Third, platforms can also copy (or link) each others information. We abstract from the latter possibility.2 Despite the dierence in assumptions and subsequently the modeling, their results are qualitatively similar to ours. We also nd that seller quantity is increased if the metaplatform is allowed. We also show that there is a coordination failure, which is not limited to the case where there is no market expansion. This shows that the results of Dellarocas, Katona & Rand (2010) also extend to a framework with a real aggregator and a multi-sided market with externalities. Further economic literature on aggregators is scarce. George & Hogendorn (2011) also stress the market expansion eect of an aggregator, through a reduction of the transaction costs of viewers. Chiou & Tucker (2011) exploit a natural experiment to test the impact of Google News
2 Another article that allows platforms to link to each other, is Mayzlin & Yoganarasimhan (2010). They investigate blogs which can establish links between each other. Linking increases the value for readers, but might also lead to a loss in readership, on the short term directly because the reader follows the link and on the long term because credibility and promotion is given to a rival. Whether links are formed, depends on the magnitude of both eects. Unfortunately, they dont discuss the possibility of an aggregator.
122
Aggregators
on news websites. For a short period of time, AP was no longer available on Google News. Chiou & Tucker (2011) nd that Google News users were less likely to visit the news websites (compared to Yahoo News where AP was still allowed). Their results suggest that aggregators might prove useful to attract additional visitors. This is in line with our theoretical ndings, though we should add that increased (indirect) visitors do not automatically lead to higher revenues for the platforms, due to the potential loss in ad revenue. The two-sided market literature, started with Caillaud & Jullien (2003), Rochet & Tirole (2003) and Armstrong (2006), focuses on how a platform gets both sides on board. Many two-sided markets were inspired by internet platforms, since the internet made dierent business models plausible, such as pay per transaction (e.g. pay per click advertisements) (Caillaud & Jullien 2001). Internet industries that received substantial attention are B2B, B2C and C2C platforms (Yoo, Choudhary & Mukhopadhyay 2007), online publishing (McCabe & Snyder 2010), online advertising (Evans 2008), (open source) software (Economides & Katsamakas (2006), Gallaugher & Wang (2002)) and net neutrality (Economides & T (2008), Hahn & Wallsten (2006)). Not so much has been done ag yet on real estate platforms, the example extensively used in this chapter. Hendel, Nevo & Ortalo-Magn (2009) apply the matching model of e Coles & Muthoo (1998) to real estate marketing platforms. Starting from a stock-and-ow model, they test platform choice empirically by investigating a brokers platform and a for sale by owner platform in Madison, Winconsin, (United States). They nd that a listing on the more expensive platform does not lead to higher prices, but on average houses are sold quicker. Unfortunately, there is no aggregator active in this market. For an overview of the two-sided markets literature and the strategies applied, see Rochet & Tirole (2006), Rysman (2009) and Eisenmann, Parker & Van Alstyne (2006). Our contribution to the literature is that we complement the literature of two-sided markets with the introduction of a metaplatform, which remained undiscussed in the recent literature. In the link economy literature, we are one of the rst to analyse the eect of aggregators on platforms and its market sides. Besides, our model is related to other economic concepts. Our metaplatform can be interpreted as an adapter in the spirit of Katz & Shapiro (1985). Metaplatforms make platforms compatible, since a buyer can
4.2 Related Literature
123
access the platform content through the metaplatform. Pollock (2007) provides an interesting analysis on porting, the conversion of software or services developed for one platform to run on another. In a sense, this is also what a metaplatform does. Pollock (2007) gives the platforms the possibility to make porting more costly for other platforms. A monopolist platform decides on both prices and porting cost. The author shows that, under certain conditions, this decision on porting has far more impact on consumer welfare than the pricing decision. To some extent, this is also what we nd: for buyers, the decision whether to allow the metaplatform matters much more than the dierence in seller quantity set due to this decision. Katz & Shapiro (1985) investigate whether the construction of an adapter is welfare enhancing. Their results show that unequally sized rms dont have enough incentives to create an adapter while unequally sized rms might have socially excessive incentives. The focus of this strand of literature is more on the welfare aspects, while we focus on the impact for the platforms (and less on the demand sides). Another related concept is vertical separation. This means that the manufacturer does not sell its products directly to the consumers, but sells to a retailer who sells in turn to consumers3 . In two-sided markets, a platform normally connects one side (e.g. buyers) with the other side (e.g. sellers). Vertical separation can then be seen as two connected platforms, where one platform focuses on one side, whereas the other focuses on the other side. Consider again the market for real estate. A traditional platform aims at connecting both buyers and sellers. If a metaplatform comes in the market, it can be seen as vertical separation, since the metaplatform particularly aims at reaching buyers, while the traditional platform is now focused more on the seller side. Most scientic articles on aggregators are written either in the eld of information technology or law. The former mostly stresses the technical details to build an aggregator, or the impact of spiders on website performance (e.g. Boswell (2003)). Legal analysis focuses on the arguments used by courts (e.g. Jennings & Yates (2009), Fischer (2001)). Further details on this legal analysis can be found in Appendix B.
3 See (Bonanno & Vickers 1988) for the concept of vertical separation. Their analysis, though, is focused on the price eect of vertical separation and therefore the results are less related to ours.
124
Aggregators
Figure 4.1: Set-Up of the Game
4.3
4.3.1
Theoretical Model
Set-Up of the Model
Our metaplatform model consists essentially of ve groups of players: the platforms, the respective demand sides of these platforms and metaplatforms. In the context of real estate classied ads, these demand sides are potential buyers and sellers of houses and advertisers who try to reach the buyers (e.g. banks who are selling loans). The platform is a website where sellers can announce their property for sale. The website attracts buyers by collecting and representing this information on sellers. The metaplatform is also a website that is read by buyers; sellers cannot announce directly on the metaplatform, but their content appears on it if the metaplatform is allowed to copy the platform. Advertisers can place their ads both on platforms and metaplatforms. Our set-up is summarized in gure 4.1 and discussed in detail below. We maintain the real estate classied ads example throughout the chapter since it claries our model. It might be clear that our model is not limited to this particular example. It might be applied to other twosided markets, such as auction platforms, matching platforms for dating or employment or media platforms in general.
4.3 Theoretical Model 4.3.1.1 Framework without metaplatform
125
Sellers Sellers are house owners who try to sell their house. Sellers value the amount of buyers positively and the membership price negatively. Furthermore, we assume that sellers are multihoming (i.e. joining one or more platforms); buyers are singlehoming (i.e. joining at most one platform). This means that buyers visit only one platform; sellers that want to reach these potential buyers can only reach them by announcing on the platform chosen by these buyers. As such, the platform is a bottleneck for sellers: and platforms can act as monopolists on the potential buyers. This is a common assumption in two-sided markets (see Armstrong (2006)) and in line with the evidence presented by Evans (2008) on search engines. The utility a seller j derives from platform i is equal to:
S S Uij = ij QB pS i i
(4.1)
with QB the number of potential buyers on platform i, pS the membership i i S sellers pay on platform i and ij the value seller j derives from being seen S by a potential buyer on platform i. This ij is seller and platform specic. Sellers dier regarding to the chances they have to sell a house, or by the prots they can make to sell their house. The valuation of a buyer might also be dierent across platforms. Implicitly, we assume here that the type of information presented on both platforms diers. One platform might put more emphasis on the location or construction type, while the other platform presents more information on the look (e.g. with pictures). Some sellers value one type of information more than the other, because it aects their chances to sell the house. We assume that the valuation for S buyers ij is uniformly distributed over 0 and 1. Therefore, the demand on platform i reads: pS (4.2) QS = 1 iB i Qi Buyers Buyers visit only one platform and dont pay. To model how the buyers choose which platform to visit, we rely on the anchor site trac specication of Dellarocas, Katona & Rand (2010). They assume that the direct viewers are proportional to the utility UiB a website generates. QB = i
n k=1
UiB B Uk +
(4.3)
where 0 1 represents the outside option, e.g. searching for houses on the street, instead of using the platform. As argued by Dellarocas, Katona
126
Aggregators
& Rand (2010), this consumer demand can be the result of a dynamic search model. Suppose that new buyers are unaware of the quality of each platform and therefore choose randomly. Over time, these buyers can adopt various strategies to nd out the value of the platforms, such as a multi-armed bandit exploration-exploitation process. The static formula above can be considered as the steady state property of the dynamic game, where a number of new consumers are added each period. One can think of new generations that buy a property. This resembles a logit structure, a discrete choice model where the probability of choosing product A depends on the relative utility of A vis-`-vis the sum of the utility of all the products4 . a We assume that the utility of buyers is equal to: UiB = B QS i B is not buyer or platform specic and is normalized to 15 . Advertisers Advertisers are, similar to sellers, also looking for the attention of the buyers, but for other reasons. We assume that advertisers sell other products than houses. This might in principle be any product (think of advertisers in general newspapers). Logical advertisers in this context are companies that sell goods complementary to houses, such as banks oering mortgage loans, moving services, notaries or do-it-yourself shops. We dont model these advertisers explicitly; though it would not increase the complexity of the model, it would not add much insight. One can see the optimal price pA as the result of a maximization process on the advertiser market. An increase in pA might be the result of a positive external shock in the size of the advertiser market, or an increase in the willingness to pay of advertisers. We assume further that buyers and sellers are indierent about the amount of advertisers on the platform. As such, this third demand side does not generate externalities on the other two demand sides. Note that we chose for a dierent modeling approach for sellers, buyers and advertisers. Most two- or multi-sided markets articles treat all sides symmetric. While a symmetric approach is certainly most apt for general
4 One might even extend this framework to a nested logit, with platforms and metaplatforms in dierent nests. We abstract from this possibility. 5 This normalization has no qualitative impact on the results, but simplies the equilibrium analysis
(4.4)
127
models, we prefer a more ne-tuned approach for an applied model. Sellers have complete information about the game, and are allowed to multihome. Buyers, on the contrary, have incomplete information and can only singlehome. This structure, imposing that one side singlehomes whereas the other multihomes, is not unusual in two-sided markets. Indeed, if both would multihome, it would be possible for one group to stop multihoming without losing much (since they are still connected to all multihomers at the other side). By assuming that buyers do not all choose the largest platform, we avoid an outcome where all buyers would be attracted to the metaplatform, leaving the platform without direct buyers. Modeling the advertisers explicitly would not change our results, and would not add to a better understanding; therefore, we assume that pA is the result of a maximization process on the advertiser side.
Platforms The platforms are websites that connect buyers and sellers. Platforms generate revenues on both sides. We assume that platforms are quantity setters on the seller side. Therefore, equation 4.2 should be interpreted implicitly: pS (QS ) = QB (1 QS ). On the buyer side, the i i i platform does not gain direct revenues from the buyers since they are not charged. Platform can also gain revenues by charging advertisers. These advertisers pay for attention time, which is a function of the total time buyers spend on the website. We assume that the time a buyer spends on a website is proportional to the number of sellers. Therefore, the total amount of buyer attention tA is equal to the number of buyers that visit the platform times the time they spend on the website: tA = QB tQS , i i i where t is the average time a buyers looks at a seller announcement. For simplicity, we assume t is equal to one. pA is the price advertisers pay for every unit of attention time (tA ). As a result, the prot function of the platform reads: i = pS QS + pA QB QS i i i i (4.5)
Note that we have normalized costs to zero. The only decision variable in this prot equation is QS . To simplify the analysis, we assume that i there are two platforms active in the market, indexed 1 and 2. We relax this assumption in section 4.3.3.
128 4.3.1.2 Framework with metaplatform
Aggregators
The metaplatform is a website that automatically collects the information of other platforms. Therefore, it might also be labeled as an aggregator or crawler or spider. It copies the seller information to its own database and can represent this information on its website. This might provoke a legal discussion, because the copying of information might be considered as a copyright infringement. Other arguments against the automatic copying of information are the hot news doctrine and trespass to chattels (property infringement). We discuss these arguments in Appendix B. In the model, we assume that automatic copying is not allowed, and the platform should give its approval to the metaplatform to copy the information. This variable ci takes the value 1 if copying is allowed by platform i, and 0 otherwise. The metaplatform competes with normal platforms in the buyer market and advertiser market, but not in the seller market. Since some buyers will visit the metaplatform (in stead of the platform), the metaplatform can also attract advertisers and generate revenues. These potential buyers visiting the metaplatform are still interesting for platforms. Although they dont generate advertiser attention time, they watch the announcements of sellers, increasing the willingness to pay of the sellers (and therefore the price). The metaplatform has no own sellers; it relies only on the content provided by other platforms therefore it doesnt make money on the seller side. It is important to stress the dierence between direct and indirect buyers. Direct buyers (QB,d ) visit the platform directly, while indirect i buyers (QB,i ) are redirected by the metaplatform. The number of indirect i buyers depends on the number of buyers that visit the metaplatform, and the ratio of these buyers who watch the seller announcements of platform i. We assume that this ratio is proportionate to the ratio of sellers of both platforms on the metaplatform. Therefore, this ratio (ri ) is equal to ri = 0
ci QS i c1 QS +c2 QS 1 2
if c1 , c2 = 0 otherwise
An important question is how buyers value sellers that multihome. Sellers that join the two platforms will appear twice on the metaplatform. One can argue that buyers value these sellers only once, since the underlying property is the same. In the basic model, we assume that buyers
129
also value the second mentioning of the same seller. Since we implicitly assumed that both platforms are dierentiated in the type of information they collect on sellers, the second seller announcement will contain dierent information, and will therefore also be valuable for buyers. We discuss this issue further in section 4.3.4. The number of direct and indirect buyers are equal to: QB,d i QB,i i = = QS i (1 + c1 )QS + (1 + c2 )QS + 1 2 ri (c1 QS + c2 QS ) 1 2 (1 + c1 )QS + (1 + c2 )QS + 1 2 (4.6) (4.7)
The total amount of buyers QB is i QB = QB,d + QB,i i i i (4.8)
This formula replaces the denition of QB in equation 4.3. If c1 and c2 i are equal to zero, both denitions coincide. Sellers value both direct and indirect buyers and therefore equation 4.2 remains unchanged. Since the platform only gains advertisement revenue from direct buyers, we replace the prot function of the platform (equation 4.5) by6 i = pS QS + pA QB,d QS i i i i (4.9)
The metaplatform (indexed with m) only gains prot from advertisers, and therefore their prot function is equal to m = pA QB QS m m
c QS +c QS
(4.10)
1 1 2 where QB = (1+c1 )QS +(1+c22)QS + and QS = c1 QS + c2 QS . As can be m m 2 1 1 2 seen, the metaplatform cannot inuence its own prots (since it doesnt decide on the amount of sellers on its platform) and therefore it is not treated as a strategic player. An extension of the model might give the metaplatform tools to behave strategically, such as attracting own sellers, or paying side payments7 . In the basic model, we assume there is only one metaplatform. This assumption is relaxed in 4.3.3.
6 After
simplication,
(pA +(1+ci )(1QS ))(QS )2 i i (1+c1 )QS +(1+c2 )QS + 1 2 7 We discuss these extensions
the full form of the prot function reads i in sections 4.3.3 and 4.3.4.
130 4.3.1.3 Timing of the Game
Aggregators
Platforms maximize their prots with respect to the quantity of sellers QS i and the decision to allow copying by the metaplatform ci . Platforms rst decide simultaneously whether to allow the copying and second decide on the quantity of sellers. The timing of the game is: 1. Allowance metaplatform (c1 , c2 ) 2. Quantity sellers (QS , QS ) 1 2 In their extension where sites can veto links, Dellarocas, Katona & Rand (2010) apply a reverse timing. Sites rst decide how much they will invest in content and then decide whether to allow the link source our not. Both approaches are defendable, but it seems more natural to regard the allowance as a slow variable, since this probably implies the implementation of protection software by the website, whereas seller quantity (or content in their case) is easily adjustable.
4.3.2
Basic Results
In this section, we examine the optimal seller quantities for a market situation with two platforms and one metaplatform. We solve this game by backward induction, i.e. we rst solve for the equilibrium quantities (QS , QS ) for given decisions on the allowance (c1 , c2 ). 2 1 4.3.2.1 Stage 2: Optimal Seller Quantities
There are four possibilities concerning allowance: both platforms disallow the metaplatform (c1 , c2 = 0), both platforms allow the metaplatform (c1 , c2 = 1) or only one of the platforms allows (c1 = 1, c2 = 0 or c1 = 0, c2 = 1). We compute the equilibrium quantities in each case. Both platforms disallow the metaplatform The optimal seller quantities when both platforms disallow the metaplatform read: Lemma 4.3.1 1 10 (3(1 + pA ) + 2+ if pA 3+2 QS = i 1 2+ if pA > 3+2
9((1 + pA )2 + 2 ) + 22(1 + pA ))
131
These solutions are symmetric and unique; both platforms set a positive amount of seller quantity. If the price for advertisement attention is relatively high, then platforms attract all sellers. This means that the implied seller price goes to zero, i.e. the platform only generates revenues from advertisements and not from sellers anymore. Both platforms allow the metaplatform The optimal seller quantities when both platforms allow the metaplatform read: Lemma 4.3.2 1 20 (3(2 + pA ) + if pA 4+ 3+ QS = i 1 if pA > 4+ 3+
9((2 + pA )2 + 2 ) + 22(2 + pA ))
These solutions are quite similar in structure as when both disallow the metaplatform. One player allows, the other disallows the metaplatform Needless to say, case 3 and 4 are symmetric and therefore treated as one case. We relegate the equilibrium quantities to Appendix A because they are too lengthy to discuss here. The graphical presentation of the solutions in gure 4.2 shows that the equilibrium quantity of platform i, who asymmetrically allows the metaplatform (ci = 1, cj = 0), is close but below the equilibrium quantity when both allow the metaplatform. The quantity of the platform i, who asymmetrically disallows the metaplatform (ci = 0, cj = 1) is close but above the quantity when both disallow. The intuition behind this result is that less can be gained from advertisement revenues if the metaplatform is allowed, and therefore the quantity of ads is lowered to increase the implicit price of sellers. This increases the revenues for sellers, since platforms are over-providing seller quantity (compared to the case where a platform would only make prots from sellers) because more sellers also means more advertisement revenues (through the advertisement attention time). In the asymmetric case, the platform disallowing the metaplatform will set a seller quantity even higher than under the symmetric disallowance case, because it faces a stronger competition on the market for buyers and attempts to attract more buyers by increasing the number of sellers.
132
Aggregators
Figure 4.2: Equilibrium Seller Quantities for All Cases, with pA = 0.25
These results also allow for a comparative statics analysis of quantities and prots. Proposition 4.3.3 A higher and lower pA reduce the prots of the platforms. Increases in and pA lead to an increase in the optimal amount of sellers. This proposition holds for all four cases. A higher reduces the market for buyers and therefore puts pressure on prots. Moreover, less buyers are attracted for the same amount of sellers, and therefore the number of sellers is decreased. An increased advertiser demand increases prots and induces platforms to increase the number of sellers this lowers the implicit seller price but is oset by the higher advertiser price. 4.3.2.2 Stage 1: Optimal Allowance Decision
The strategic decision on allowance of copying of seller information can be represented by a two by two game-theoretic table (two players, two actions):
133
Player 1
c1 = 0 c1 = 1
Player 2 c2 = 0 c2 = 1 1 |(c1 , c2 = 0), 2 |(c1 , c2 = 0) 1 |(c1 = 0, c2 = 1), 2 |(c1 = 0, c2 = 1) 1 |(c1 = 1, c2 = 0), 2 |(c1 = 1, c2 = 0) 1 |(c1 , c2 = 1), 2 |(c1 , c2 = 1)
We search for equilibria in pure strategies in this game. We summarize the equilibrium conditions in the following lemma:
Lemma 4.3.4 c1 , c2 = 0 is an equilibrium if c1 , c2 = 1 is an equilibrium if ci = 1, cj = 0 is an equilibrium if i |(c1 , c2 = 0) > i |(c1 = 1, c2 = 0) i |(c1 , c2 = 1) > i |(c1 = 0, c2 = 1) i |(ci = 1, cj = 0) > i |(ci , cj = 0) and j |(ci = 1, cj = 0) > j |(ci , cj = 1)
Based on the equilibria in lemmas 4.3.1 and 4.3.2, we can state the following proposition. Proposition 4.3.5 For all values of pA and , there is always at least one symmetric equilibrium in pure strategies. If pA is relatively low, both platforms allow the metaplatform; if pA is relatively high, both platforms disallow the metaplatform. For moderate values of pA , both equilibria are supported. A higher increases the likelihood that the metaplatform will be accepted. The intuition behind this proposition is straightforward. A high means that a substantial fraction of the buyers is not in the ecosystem, i.e. there is a large potential for the metaplatform to attract new buyers. This makes the allowance of a metaplatform more protable. A high pA means that a platform has a lot to lose by allowing the metaplatform. The higher pA , the less likely that the loss in advertisement revenue is compensated by an increase in seller revenue (through the increased amount of buyers). Computing the sum of the prots of both player learns that they do not always take the prot maximizing allowance decision. Proposition 4.3.6 For some values of and pA , there is a coordination problem. Both platforms allow the metaplatform although a joint disallowance would be more protable. The proposition stated above is quite similar to the classical prisoners dilemma. If both platforms would be able to coordinate their actions, they would disallow the metaplatform whenever (disallow,disallow) this is
134
Aggregators
Figure 4.3: Equilibrium Zones for EQ=equilibrium P M =prot maximizing
Allowance
Decisions,
with
more protable. Though, even if (disallow, disallow) is the most protable for both, it might be more protable for one player to deviate to (allow, disallow). The reason why they both choose to allow the metaplatform is that they dont want to end up as the only platform disallowing. In gure 4.3, the four equilibrium zones are shown. In the upper and lower part of the graph, the platformss equilibrium actions are in line with the outcome which maximizes prot. In these cases, there is no coordination failure. In the lower intermediate zone, both players choose to allow the metaplatform, while joint disallowance would be more protable. This is the case described in proposition 4.3.6. In the upper intermediate zone, disallowance is the most protable, but both equilibria are supported. There might be a coordination problem, but it is also possible that the equilibrium actions are optimal. The results in lemmata 4.3.1 and 4.3.2 also allow to do a welfare analysis of the equilibrium outcome. Therefore, we compute the buyer and seller surplus, and the prots of platforms and metaplatforms. Total welfare is
4.3 Theoretical Model the sum of these demand surpluses and the (meta)platforms prot.
135
Lemma 4.3.7 Buyers are always better o if the metaplatform is allowed. Sellers favor the metaplatform if is relatively high and pA is relatively low. Total welfare is always higher if the metaplatform is allowed. The allowance of the metaplatform provokes the platforms to set a lower seller quantity, though for buyers, this is always compensated by the creation of an additional outlet to look for properties (i.e. the metaplatform). The result on the seller surplus is similar to the prot of platforms (which is also higher under the metaplatform if is high and pA low). In terms of gure 4.3, the seller surplus equivalence line (where surplus with or without the metaplatform is equal), is always below the platforms prot equivalence line. Based on lemma 4.3.7, we can review the impact of the coordination failure on all market players. Proposition 4.3.8 The coordination failure to disallow the metaplatform provokes a welfare transfer from sellers to buyers and from platforms to metaplatforms. Indeed, since the seller surplus equivalence line is always below the platforms prot equivalence line, sellers would even disallow when allowing is more protable and an equilibrium for the platforms. This guarantees that sellers always lose if the metaplatform is allowed due to the coordination failure.
4.3.3
Extensions
In this section, we alter the basic framework to investigate interesting extensions to the model. First, we investigate what happens if both platforms are jointly owned. Second, we extend our framework to a market situation with n (symmetric) platforms and m metaplatforms. We conclude the section with an analysis of side payments. Jointly Owned Platforms In the previous section, we analyzed the seller quantity decision of two competing platforms. Suppose now that these actions are taken by a joint entity, then we can state the following: Under joint ownership, the optimal quantities for both platforms read:
136
Aggregators
Lemma 4.3.9 1 8 (2 + 2pA 3 + (2 + 2pA + 3)2 + 8(1 + a)) 2+ if c1 , c2 = 0 and pA 2+2 1 2+ if c1 , c2 = 0 and pA > 2+2 S Qi = 1 A 16 (4 + 2p 3 + (4 + 2pA + 3)2 + 8(2 + a)) if c1 , c2 = 1 and pA 4+ 2+ 1 4+ if c1 , c2 = 1 and pA > 2+ As can be seen, these results are similar in structure as the duopoly case; the comparative statics for the duopoly case hold here as well. Comparing the quantities and prots with those from separately owned platforms, we can state: Proposition 4.3.10 If both platforms are jointly owned by one monopolist, the monopolist lowers the amount of sellers on the platform. The prot of the jointly owned platforms is higher than the total prot of the separately owned platforms. These results are in line with the standard monopoly models, where quantity is reduced, but prots increased. Note that the largest gain from the joint decision is made when there is a coordination failure. Indeed, the result of a merged decision entity for both platforms is double: rst, there is a monopoly prot gain (by setting the monopoly quantity) and second, there is a gain from implementing the prot maximizing solution (in terms of the allowance decision)8 . Generalization to n platforms We assumed that there are only two platforms active in the market, though reality learns that in many industries multiple platforms are operating. Besides, there might be more than one metaplatform available in the market. We generalize our model to m metaplatforms and n platforms. To keep the model tractable, we assume that rms are symmetric and that platforms decide to either allow all metaplatforms, or allow none. In this setting, the implicit price for sellers and prot function of the platforms remain the same, but the buyers quantity changes to
8 The prot equivalence line (where |(c , c = 0) = |(c , c = 1)) is slightly lower 1 2 1 2 than in the case where both platforms decide independently
137
QB,d n,i QB,i n,i with rn,i =
= =
QS i n (1 + mck )QS + k k
n k (1
(4.11) (4.12)
rn,i mci QS i + mck )QS + k if ci : ci = 0
0
ci mQS i n S k ck mQk
otherwise
Assuming that all platforms are symmetric, the equilibrium quantities for all rms are equal to: Lemma 4.3.11
1 ((2n 1)(1 + pA ) 3 + 4(2 + 2pA )(3n 1) + (1 + pA 2n 2npA + 3)2 ) 2(3n1) if (1) 1 if (2) (2n1)(1+m+pA )3+ 8(3n1)(1+pA +m)+(1+pA 2n2pA n+3) 2(1+m)(3n1) 1 if (3) if (4)
S Qi =
with (1): c1 , c2 = 0 and pA (3): c1 , c2 = 1 and pA

(1+m)(n+mn+) 2(n+mn+)(1+m) .
n+ 2n+21 , (2): (1+m)(n+mn+) 2(n+mn+)(1+m)
c1 , c2 = 0 and pA >
and (4): c1 , c2 = 1 and p >
n+ 2n+21 , A
Note that these general formulas boil down to the quantities of lemmas 4.3.1 and 4.3.2 for m = 1 and n = 2. Computing the comparative statics of these quantities and prots, we can state the following proposition. Proposition 4.3.12 The more platforms in the market, the lower the prots of these platforms. The eect of m on prots is ambiguous. If pA is low and/or is high, more metaplatforms increase the platformss prots. An increase in m decreases the amount of seller quantity, whereas an increase in n increases the seller quantity. It is not surprising that increased competition (n ) decreases prot. n increases competition on the buyer side. As a result, platforms increase the amount of sellers and prots go down. The eect of m is more nuanced. A higher m increases the impact of the allowance of metaplatforms, and therefore reduces prots of advertisements, but increases prots from sellers (through an increased number of buyers in the ecosystem). Therefore, this revenue shift is the most protable when the potential of metaplatforms to increase the amount of buyers in the ecosystem is high (i.e.
138
Aggregators
is high) and when revenues from advertisements are relatively low. Not surprising, platforms decrease the number of sellers as the relative overinvesting in seller quantity does pay o less because this is less compensated by advertisement revenues. There is no equilibrium in the deviating case (i.e. one platform disallows when all other platforms allow, or vice versa), therefore it is unclear what the optimal ci s are. It is clear, however, that an increase in m shifts the prot equivalence line upwards, whereas n decreases this line. Side Payments and Bargaining Until now, we assumed that there was no possibility for the metaplatforms to buy o the copying of their information on the metaplatform. From the previous analysis, we know that the existence of metaplatforms can be unprotable, therefore platforms might be tempted to obstruct the metaplatform. The possibility of side payments might convince these platforms to allow the metaplatform. Whether side payments can compensate the losses of the platforms, depends on the prot of the metaplatform (see equation 4.10). Proposition 4.3.13 The total prots in the ecosystem are always higher if a metaplatform can operate. Therefore, if side payments are feasible, then the metaplatform will always be allowed. Since total prots are higher under a metaplatform, it is always possible to compensate the potential losses of platforms occurred due to the existence of this metaplatform. This is in line with the analysis of Katz & Shapiro (1985). They also rst analyse compatibility as an exogenous characteristic of the market. Afterwards, they investigate the private (and social) incentives for compatibility. Since total prot is strictly higher if > 0, this additional prot might be divided by a bargaining process between the platforms and the metaplatform. It is interesting to see that Google Fast Flip uses such a compensation scheme. This website shows snap shots from news websites such as New York Times, Washington Post and Newsweek. Google Fast Flip shows advertisements on its website and shares the revenues with the news websites (Bharat 2009). Platforms can also ask indirectly for side payments. Legal actions might be inspired by the aim for a settlement on side payments. In the Belgian case Copiepresse vs Google9 , Sullivan (2007) argued that it was perfectly possible to stop Google accessing the news pages of
9A
discussion of this case can be found in Van Asbroeck & Cock (2007).
139
newspapers by simply putting a robots.txt le on their servers. It is likely that Google would have complied voluntarily with this request. Sullivan (2007) poses: This case was never about getting content out. It was about trying to blackmail Google into including content. Now the newspapers may get a large ne coming their way, but whether Google will feel it still wants to cut a deal with them remains unseen. Note also that it should not always be the metaplatform who compensates the platform for potential losses. It can also be the other way around. If a metaplatform is benecial to the platform, then the platform can pay the aggregator to be in the market. This is not the case in our model, but it might be if a xed cost is introduced. If the advertisement market is small, then the platform prots will be higher if a metaplatform exists, but the metaplatform will only make moderate prots. But again, to make compensation possible, total prot with the existence of a metaplatform should be higher, otherwise there is not enough to redistribute. Another way to settle side payments is a vertical integration between the metaplatform and one or more platforms. The additional prots are directly internalized, but side payments might still be needed for those platforms that do not share the same owner as the metaplatform. Some platforms have set up their own aggregator, such as Dutch real estate platform Funda, in reaction to other aggregators. Content providers of news, such as NewsCorp, were also examining the launch of an aggregator but the project was shelved in October 2010 (Levy 2010).
4.3.4
Future research
Even with our simple model, we were able to gain interesting insights on the eect of metaplatforms or aggregators. Here we present some extensions that might be interesting as well, though we only provide the intuition. These extensions concern multihoming, strategic metaplatforms, congestion, pricing and bias. Multihoming and doublecounting The specication in the previous sections did count a house that was multiple listed twice. This is reasonable, if both platforms contain other information about the properties, e.g. dierent photos, or present the information in a dierent way. Though,
140
Aggregators
one can also argue that a house that is multiple listed should be counted only once. A reasonable specication assumes that 1j and 2j are independently drawn from a uniform distribution between 0 and 1. This leads to a multihoming (or double-counting or overlap) of c1 c2 QS QS . Of 1 2 course, when at least one platform disallows the metaplatform, the overlap is zero. The problem though, is that adding this factor to the denominator of equation 4.7 makes the model untractable. Another set-up is to assume that the overlap is a constant fraction of the total amount of quantity on the metaplatform. This solution is unsatisfactory, but provides insight in the direction of the multihoming correction. Consider a market with n platforms and 1 metaplatform. Now we introduce (n), the overlap factor, which is equal to (n) = 0
n k
ck 1 n k ck
if ci : ci = 0 otherwise
The quantity of indirect buyers is corrected with this overlap factor: QB,i = n,i
n k (1
rn,i ci (1 (n))QS i + ck (1 (n)))QS + k
(4.13)
For n = 2, the equilibrium quantities in the case where both allow the metaplatform read Lemma 4.3.14
A 1 20(4) (12(2 + p ) +3( + 2) 9(4pA + ( 2)( 4))2 + 80( 4)( 4 2pA )) (4)(4+) if pA = 4(3+) 1 (4)(4+) A if p > 4(3+)
Qi
The equilibrium quantities for the other cases remain unchanged; in all dQS |(c1 ,c =1) other cases (n) = 0. The comparative statics show that i 2 > 0, i.e. the platform increases its quantity if there is more overlap. As ex1 pected, more overlap reduces the prots ( di |(c,c2 =1) < 0). The introduction of also moves down the prot equivalence line. This is logical, since decreases the prots when both allow, but doesnt aect the prots when both disallow the metaplatform. Metaplatforms as strategic players In our model, the metaplatform had no instruments to inuence its own prot. This might not correspond
141
to reality. Real estate aggregators, such as zoekallehuizen.nl and jaap.nl in the Netherlands, also oer the possibility to add a house to their database, without having to buy an announcement on a traditional platform. Moreover, metaplatforms might also have the possibility to refuse to copy content, or copy only a certain amount of the content of traditional platforms. Therefore, future research might consider a specication where the metaplatform can also decide to place its own seller quantity (QS ) and m choose whether to act as an aggregator: cm is 1 if it copies content from other platforms, cm is 0 otherwise10 . One possibility is to replace the non-strategic prot function of the metaplatform (equation 4.10) by m = pS QS + pA QB,a QS m m m m with pS m QB,a m = = QB,d (1 QS ) m m c1 cm QS + c2 cm QS + QS m 1 2 (1 + c1 cm )QS + (1 + c2 cm )QS + QS + m 1 2 (4.14)
In the demand function of the buyers, ci is then replaced by ci cm and QS is added to the nominator. m The timing becomes particularly relevant here. A logical timing would be that the metaplatform decides rst whether it is willing to copy content (cm ). Second, the platforms decide whether to allow the copying by the metaplatform and nally platforms and metaplatform set the content. Unlike in the case without metaplatform, the game would change substantially if the seller quantities are determined before the allowance decisions are made, for two reasons. First, the metaplatform might strategically set its quantity to inuence the allowance decision, and second, the platforms might strategically set quantity to inuence the willingness to copy of the metaplatform. This blur in the line between metaplatforms and platforms might also be particulary relevant to expand our model to other media markets, such as news websites. An interesting trend is that the most visited news aggregator, Yahoo!News, has expanded its sta with journalists in 2010. While the number of own stories remained low in 2010 (4 %), it is a relative huge increase with regards to 2009 where 1% of the stories came from the Yahoo sta (Pew 2011).
10 This copying might also be continuous, e.g. c m = 0.5 means that the metaplatform copies 50% of the seller announcements of traditional platforms.
142
Aggregators
Congestion In our model, the utility of buyers (sellers) does not depend on the number of buyers (sellers). This means that there is no congestion or push away eect from other advertisers. It is unlikely that this holds in reality. We would expect that the chance of a match or sale is reduced by an increase in seller announcements. This congestion eect is widely documented in the advertisment literature. As Comanor & Wilson (1974) put it: To the extent that the advertising of others creates noise in the market, one must shout louder to be heard, so that the eectiveness of each advertising message declines as the aggregate volume of industry advertising increases. Most two-sided markets articles, such as the seminal contribution of Rochet & Tirole (2003) ignore congestion or own-side eects. They assume that both sides of the market always interact once, irrespective the number of agents on each side. This means that a potential congestion eect is absent. Other authors put congestion at the heart of their exposition. Belleamme & Toulemonde (2009) investigate how negative intragroup externalities inuence the potential of a for-prot platform to enter successfully in a market where the two sides are already trading on a non-prot platform. The for-prot platform enters by subsidizing one side and charging the other11 . They show that entry is only possible if this externality is intermediate. Church & Gandal (1992) argue that software providers weigh the network eect (cross-group externality) with the competition eect of more agents on their side of the market, generating a lower chance of being seen or chosen. The same question pops up in the empirical article of Tucker & Zhang (2008): should platforms announce the number of sellers or not? On the one hand, a high number of sellers signals that there are also a high number of buyers (cross-group externality). On the other hand, it signals harsher competition for these buyers (congestion or own-side eect). Theoretically such a congestion eect can easily be operationalized in our model by adding a negative own side quantity term to our utility S S function, e.g. Uij = ij QB QS pS where measures the disutility of i i i congestion. The problem though, is that the inclusion makes our model intractable.
most articles do not allow for negative prices, subsidizing one side and charging the other side is a strategy very well suited for two-sided platforms. Caillaud & Jullien (2003) discuss these divide & conquer strategies and the eciency of the resulting equilibria.
11 While
143
Eects on Pricing and Competition within the Demand Sides Related to the previous extension, it might be interesting to investigate how a metaplatform is changing the competition on the demand sides and the subsequent prices and prots on this side. Rosenfeld (2002) provides a rst attempt to describe these eects. He argues that the likely eect of an aggregator is that prices drop, as was also suggested by the empirical work of Brown & Goolsbee (2002). Nevertheless, in his article Rosenfeld (2002) mainly warns for the potential negative eects of aggregators or spiders on competition. His argument is threefold. First, collusion might prevent the expected price drops. Second, spiders might trigger price wars. Third, other factors might still keep prices above marginal cost. We think that his claim that reduced search costs of the internet create as many dangers as they oer remedies is exaggerated12 . Further work needs to be done to disentangle the eects of a metaplatform on prices and competition within the demand sides. Search Results Bias In our analysis we implicitly assumed that every seller had equal chances of being reached by a buyer. In reality, though, the position in search results matters a lot (Battelle 2007). A platform, but also a metaplatform, might use his power to charge a menu of prices, or even auction its top positions, as is done by search engines (Varian 2007). Similarly to advertisement bias, buyers are probably worse o because of the reduced matching quality, but (meta)platforms can extract more revenues13 . In practice, this type of search results bias often occurs in
12 The arguments provided by Rosenfeld (2002) are not fully satisfactory. First, it is true that lower prices increase the gains from collusion, though it is not clear why spiders would make collusion likelier. It is more likely that other players enter with lower prices, since the spider eliminates many of the advertisement-related xed costs. Therefore, collusion will be harder to sustain. For the second argument, on price wars, they cite the article of Greenwald & Kephart (1999). In this article, Greenwald & Kephart (1999) argue that price wars might result in a framework where sellers have incomplete information or follow naive strategies. Though, when sellers are rational and behave game-theoretically, this does not lead to price wars; the arrival of the spider results in price drops. Third, Rosenfeld (2002) argues that the elimination of search costs is not sucient to reduce prices to marginal cost. Product heterogeneity, consumer heterogeneity, branding and loyalty programs might keep prices high. This is true, though the question is whether an aggregator makes these factors worse. There seems to be no reason to assume this is the case. 13 Note that many matching models would argue that introducing menu pricing increases the matching quality, e.g. see Damiano & Li (2007). Dierent prices allow players to signal a high type, whereas uniform pricing doesnt. If sellers prot is
144
Aggregators
real estate markets. The Dutch platform Funda.nl always puts the houses of its own brokers on top. The content they spider from other brokers is at the end of the search results. The Belgian metaplatform Zimmo.be applies a similar hierarchical system. On top appear the properties from brokers which website is developed by MaxImmo, a company aliated with Zimmo. Then are the brokers listed that are aliated, followed by the unaliated brokers. At the bottom of the list appear the properties that are privately sold (without a broker).
4.4
Conclusion
With the risk of stating the obvious, the internet has changed and transformed many businesses; especially those where information plays a big role. In this article, we shed light on what we named metaplatforms: platforms that aggregate the content of other platforms. We argued that the introduction of such an aggregator has two important opposite eects on a normal platforms prot. First, a metaplatform attracts and redirects customers that would not have visited the platform otherwise. Second, the metaplatform scores o some of the prots generated by these customers, e.g. by absorbing the advertising revenues. We show that a metaplatform is protable for normal platforms if the rst eect dominates the second. This will be the case if two conditions are fullled. First, the outside option for customers should be relatively important, i.e. the entrance of a metaplatform can attract many new buyers in the ecosystem. Second, the importance of the advertising market is relatively low, i.e. the platform doesnt lose a lot when he has to trade in direct customers for redirected customers. Even if platforms can decide to allow or disallow the metaplatform, they do not always choose the most protable option. For intermediate advertisement prices, it would be optimal for both platforms to disallow the metaplatform, but they decide both to allow the metaplatform. This is, similar to the prisoners dilemma, the result of a coordination failure. Both have an incentive to deviate from the disallow-disallow strategy, and no platform wants to end up as the only platform not allowing the metaplatform. If platforms and the metaplatform can bargain over the initiation and allowance of links, then the metaplatform will never be obstructed by the
uncorrelated with house quality, then this signalling is worthless for buyers.
4.4 Conclusion
145
platforms since total prots of all players is higher under the existence of a metaplatform. In other words, the existence of side payments overcomes the negative eect that a metaplatform might generate. An increase in the number of platforms and metaplatforms reduces the prot for each platform. An increase in platforms increases the amount of seller quantity set by the platform, the reverse occurs with an increase in metaplatforms. Adding more (meta)platforms does not change the basic insights on the market expansion and demand shifting eect of a metaplatform. With this article on metaplatforms, we hope we opened a new line of research on the functioning and eects of metaplatforms or aggregators. Further research might add more sophistication to our model. An interesting avenue might be to allow that the metaplatform can also attract sellers, and therefore behave strategically. Second, adding congestion to the model might make the matching role of the platform more realistically. The chance of a match, e.g. in the housing market, does not only depend on the number of actors on the other side of the market, but also on the own side of the market. Finally, an interesting application of this research might be the empirical testing of the hypotheses formed in this article. It would particularly interesting to see whether auctions, matching and media platforms t in the theoretical framework proposed here.
146
Aggregators
Appendix A Equilibrium quantities
The equilibrium quantities for the asymmetric case can be found on http://www.driesdesmet.be/appendix_a.pdf.
A Legal Discussion on Metaplatforms
Our simple theoretical model allowed to gain insight on the threats and opportunities of metaplatforms. The legal discussion on metaplatforms is often focused on the question who should be granted the right for links: the platform or the metaplatform. Our framework provides a partial answer to this question, because it sheds light on who can gain and who is harmed in dierent circumstances. But apart from these calculations, some general arguments in favor of platforms and metaplatforms have been developped. On the one hand, one can argue that the platform should decide whether to allow the metaplatform copying its content or not. The content owner can assert its intellectual property rights or argue that the consumption of its bandwidth capacity is a trespass to chattels. On the other hand, one can bring forward that the metaplatform should be able to use the information of the platform. From a welfare point, the metaplatform might enhance competition by reducing search costs. Denying access can be interpreted as foreclosure. Besides, one can argue that copying information, with deeplinking, is fair use of information. We provide these legal arguments below, because they are key to understand the past, current and future legal battles on metaplatforms. They complement the insights we derived from the model in the main part of this article.
B1
Arguments in favor of the Platform
Intellectual Property Rights Websites contain a lot of elements protected by copyright, such as texts and coding (literary works), photographs and animations (artistic works), music and audio content (musical works) and lay-out (Jennings & Yates 2009). As such, the content of a real estate platform is protected by these rights.
B A Legal Discussion on Metaplatforms
147
The economic rationale for intellectual property rights is to allow that those who invest in products or services can take advantage of it. Since most intellectual products can be reproduced at a small marginal cost, companies would have no incentives to invest in future innovations without intellectual property rights. This would reduce growth, i.e. reducing welfare for society. If real estate platforms do not invest anymore in brand familiarity, website quality and the collection of additional information, then the likely result is a less informed market, leading to an increase in search costs. This leads to a dead weight loss in society (Stahl 1989). Similar to the argument of copyright, some countries also have a database regulation in order to protect the content of database holders. The British regulation for example stipulates that a person infringes the database right if, without the consent of the owner of the right, he extracts or re-utilizes all, or a substantial part, of the contents of the database (Jennings & Yates 2009). This argument was used in the German case Stepstone vs Or and the French case Cadremploi.fr vs Keljob. Hot News Doctrine But even if content is not protected by copyright, it can still be prevented from distribution by the so-called Hot News Doctrine. News gets a specic treatment in US courts since the Supreme Courts decision on International News Service vs. Associated Press in 191814 . The court argued that news or facts are not copyrightable, because these facts are not the creation of a writer but is a report of matters that ordinarily are publici juris; it is the history of the day. But the court argued in favor of Associated Press, whose news was copied and rewritten by INS, saying that this missappropriation of information resulted in unfair competition. This protection is limited in time. The competitor might not engage in distribution of reproduction till the original distributor can valorize the news facts he has done research on. The digitalization of news makes it much easier to free-ride on timesensitive news. As argued by Freedman & Pozza (2010), the Hot News Doctrine has taken on new found signicance. Recent court decisions are mixed, e.g. in NBA vs. Motorola15 the court rejected the claim of NBA while in the case Barclays Capital vs. Theyonthewall.com16 they ruled in favor of the content provider. The debates between proponents (such
News Service vs. Associated Press, 248 US 215 (1918) Basketball Assoc. v. Motorola, Inc., 105 F.3d 841 16 Barclays Capital Inc. v. Theyonthewall.com, Inc., No. 06 Civ. 4908(DLC), 2010 WL 1005160
15 National 14 International
148
Aggregators
as AP, New York Times and Time) and opponents (such as Google and Twitter) of the Doctrine are ongoing. Trespass to Chattels Spidering can be considered a trespass to chattels, i.e. the unauthorized use, dispossession or interference with the tangible property of another17 . While this early common law required a physical touching of anothers chattel, modern interpretation stipulates that indirect touching is sucient. In recent history, trespass to chattels has been used in many technological cases, such as telephone lines, radio, television broadcastings and e-mail spam (Fischer 2001). This claim was used in the case eBay vs Bidders Edge and Oyster Software vs Forms Processing18 . Bidders Edge collected data on auctions of eBay and other platforms. While eBay rst granted the right to spider to Bidders Edge, after a while eBay asked to cease collecting information. Bidders Edge complied, but resumed later. Technical protection measures of eBay failed, therefore the case ended up in court (see Fischer (2001) for a discussion of the case). The focus of the case was based on the illegal access of Bidders Edge to eBays computer system. The court argued that eBays servers are private property, conditional access to which eBay grants the public. Therefore, Bidders Edge has to comply with the terms of use of eBay. While other interests might have played in the eBay vs Bidders Edge case19 , it is a legitimate question of platforms to ask protection of their bandwidth. If spiders overuse the platforms servers, then this reduces the website speed, reducing the surng experience of potential buyers.
B2
Arguments in favor of the Metaplatform
Fair use A counterargument to the intellectual property right argument, is the fair use doctrine. This doctrine, only applicable in the U.S., says that the fair use of a copyrighted work for purposes such as criticism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, or research, is not an infringement of copyright20 . This is similar to the French droit de citation and British right
17 Restatement 18 Oyster
(Second) of Torts 217. Software vs Forms Processing, 2001 WL 1736382 (N.D.Cal. 2001), United
States. 19 See Fischer (2001) and Cisneros (2007). 20 US Copyright Act 1976 (USC), 17, section 107
B A Legal Discussion on Metaplatforms
149
of parody (Depoorter & Parisi 2002). Gordon (1982) argues that use of a copyrighted work should be allowed without remuneration, if there is a market failure. This means that the hassle cost of selling or transfering money would be so hard that usage would be discouraged. In a broader argumentation, one can see fair use as a correction to the copyright law. As the court argued in the case Bill Graham Archives vs Dorling Kindersley, the ultimate test of fair use should be whether the copyright laws goal of promoting the Progress of Science and useful Arts would be better served by allowing the use than by preventing it (quoted in Travis (2006)). The interpretation of real estate content as Science or Art is a bit far stretched, but the citation right was used in the case NVM vs Zoekallehuizen.nl21 . The Dutch court argued in a preliminary statement that copying the address, price and 1 to 1.5 rule description are not considered as a copyright infringement, but as a quote. The same holds for pictures, given that their size is suciently reduced22 . Foreclosure Foreclosure means that a dominant rm denies access to an essential good it produces, with the intent of extending monopoly power from that segment to an adjacent segment (Rey & Tirole 2007). Translated to our theoretical model, one can argue that the platform forecloses the direct seller market, in order to extend its monopoly to the broker market. If the metaplatform cannot survive without the content of the platform, then the metaplatform exits (or doesnt enter) the market. As such, the platform doesnt face competition, neither in the brokers market nor in the buyers market. Such a monopoly power might reduce welfare and therefore a competition authority will act against foreclosure. Note that metaplatforms have to prove that the content of the traditional platform is essential, i.e. it cannot easily or cheaply be reproduced. In the metaplatform case, one can argue that it is not so hard for the metaplatform to set up a traditional platform as well to attract direct sellers. But in the case of the Dutch case of Funda.nl (the platform of NVM brokers), it can be argued that it is hard to reproduce the supply of NVM brokers if Funda has an exclusivity contract with NVM brokers.
nr AV5236, Court Arnhem, 16-03-2006. Available on www.rechtspraak.nl. the Kelly vs. Ariba Soft Corporation, the court also decided that using compressed pictures by the image search engine was considered as fair use and therefore allowed (California, US, Case SA CV 99-560).
22 In 21 Arrest
150
Aggregators
Chapter 5
General Conclusion
152
General Conclusion
In this dissertation, I investigated three related two-sided markets: newspapers, telephone directories and internet platforms. Each of these markets shared the same characteristics. One or more platforms serve two distinctly dierent groups, readers and advertisers. Since these demand sides have dierent utility and prot functions, their interests do not always align. As a result, the platform will trade-o to what extent it will serve (and therefore attract) both sides. This trade-o is made by optimizing its own prots. When the newspaper implements advertiser bias, it serves advertisers but it chases away readers. But the existence of the other side of the market is not necessarily bad for the demand side. In the case of the real estate platforms in chapter 4, the price is lower for sellers because of the feedback eect they generate on buyers. More sellers means more buyers and therefore the platform lowers the price compared to a market without these feedback eects. While the framework used shares many similarities, in each chapter I tried to solve dierent questions. The answers provided in each chapter are summarized below. Summary Chapter 2 In chapter 2, I investigated the existence of a wall between the advertisement department and the editorial sta of a newspaper. Most scholars agree that the advertising department should not exert pressure on the editorial sta to shape articles. And if they do the professional ethics of journalists should prevent any direct or indirect inuence of advertisers on the selection of news. The analogy between a church-state separation of journalism and commerce is never far away. Newspaper owners and editors might nevertheless have strong incentives to tear down the wall between the advertisement department and the editorial sta. My theoretical model shows that it is protable for both monopoly and duopoly newspapers to oer a bundle of bias and advertisement if the eectiveness of bias is relatively high for advertisers and if the annoyance of bias is relatively low for readers. If duopoly newspapers are closer substitutes for readers, then newspapers set less bias; while the eect on advertisements is ambiguous. But, as Davies (2008) argues, to the outsiders eye, [advertising bias] is very tempting. Advertisers have money, the media outlets need money, so they must be vulnerable to some kind of pressure from the advertisers. Its a ne theory, but its truth is very limited. Whether bias will prevail is indeed ultimately an empirical question. I have tried to answer this empirical question for the Belgian Dutch-
153 language newspapers. I investigated to which extent advertisements aect the coverage of companies in these newspapers in the period 2001-2005. Depending on the specication, I nd that spending 54000 to 212000 euro in a newspaper in one month generates on average one additional article in this newspaper in the same month. My results appear robust to alternative specications. My results are unsurprising and surprising at the same time. On the one hand, newspapers are prot maximizing business. And as my theoretical model predicts, pleasing advertisers might generate higher prots. Therefore, it is not unlogical that newspapers insert bias. But on the other hand, newspapers are no normal business. Correctness of information generates a lot of externalities on society. Many decisions, be it voting, consumption or investment, are based on information retrieved from newspapers. Therefore, newspaper owners and editors should not allow that advertisers or the advertisement department attempt to inuence editorial content.
Summary Chapter 3 In chapter 3, I examined the main conditions of the universal service obligation for telephone directories. These conditions are comprehensiveness and availability. Telephone directories should contain all telephone subscribers and should be available to all end-users. I investigate whether one of these conditions is redundant, i.e. if imposing one condition also implies that one automatically gets the other. I nd that imposing a reader price equal to zero (and hence availability to all end-users) results in the publishing of a comprehensive directory. Conversely, it is not guaranteed that imposing a comprehensive directory results in availability to all end-users. An important caveat to my model is that costs are ignored and that an exogenously imposed minimum size could soften my redundancy result. Beyond this specic result on universal service regulation, I also showed that my model can be extended to cover other interesting topics that might be playing in the market for telephone directories. In Appendix C, I proposed extensions to the basic model to gain additional insights on the functioning of the telephone directory market. I relaxed some assumptions (e.g. on the sign of the valuation of advertisements), endogenized parameters, and looked at issues such as price discrimination and welfare implications.
154
General Conclusion
Summary Chapter 4 With the risk of stating the obvious, the internet has changed and transformed many businesses; especially those where information plays a big role. In this article, I shed light on what I named metaplatforms: platforms that aggregate the content of other platforms. I argued that the introduction of such an aggregator has two important opposite eects on a normal platforms prot. First, a metaplatform attracts and redirects customers that would not have visited the platform otherwise. Second, the metaplatform scores o some of the prots generated by these customers, e.g. by absorbing the advertising revenues. I show that a metaplatform is protable for normal platforms if the rst eect dominates the second. This will be the case if two conditions are fullled. First, the outside option for customers should be relatively important, i.e. the entrance of a metaplatform can attract many new buyers in the ecosystem. Second, the importance of the advertising market is relatively low, i.e. the platform doesnt lose a lot when he has to trade in direct customers for redirected customers. Even if platforms can decide to allow or disallow the metaplatform, they do not always choose the most protable option. For intermediate advertisement prices, it would be optimal for both platforms to disallow the metaplatform, but they decide both to allow the metaplatform. This is, similar to the prisoners dilemma, the result of a coordination failure. Both have an incentive to deviate from the disallow-disallow strategy, and no platform wants to end up as the only platform not allowing the metaplatform. If platforms and the metaplatform can bargain over the initiation and allowance of links, then the metaplatform will never be obstructed by the platforms since total prots of all players is higher under the existence of a metaplatform. In other words, the existence of side payments overcomes the negative eect that a metaplatform might generate. An increase in the number of platforms and metaplatforms reduces the prot for each platform. An increase in platforms increases the amount of seller quantity set by the platform, the reverse occurs with an increase in metaplatforms. Adding more (meta)platforms does not change the basic insights on the market expansion and demand shifting eect of a metaplatform. With this article on metaplatforms, I hope I opened a new line of research on the functioning and eects of metaplatforms or aggregators. Further research might add more sophistication to my model. An interesting avenue might be to allow that the metaplatform can also attract
155 sellers, and therefore behave strategically. Second, adding congestion to the model might make the matching role of the platform more realistically. The chance of a match, e.g. in the housing market, does not only depend on the number of actors on the other side of the market, but also on the own side of the market. Finally, an interesting application of this research might be the empirical testing of the hypothesis formed in this article. It would particularly interesting to see whether auctions, matching and media platforms t in the theoretical framework proposed here.
156
General Conclusion
Acknowledgements
I would like to thank the many people who supported me through the ve years I worked on my PhD. In the rst place, I thank my advisor, Patrick Van Cayseele, for the excellent guidance in the last ve years. I especially liked the inspirational talks and the abundance of good research ideas. It is clear that without Patricks insights and expertise, I wouldnt have been able to write this PhD. Much appreciation goes to the members of committee, Simon Anderson, Fabrizo Germano, Jo Swinnen and Frank Verboven, for the advice during my PhD and on the pre-defence. As one of the members noticed correctly, the pre-defence often resembled a brainstorming session and provoked many valuable suggestions to improve my work. I am also indebted to Stijn Vanormelingen, my co-author of the second chapter. Despite the travel distance between Barcelona and Mannheim, we managed to work together in an ecient way. For this chapter, I also like to thank the Vereniging voor Vlaamse Journalisten (VVJ) and Mediargus. Laurens De Koster provided much appreciated research assistance, which was partly published as De Koster (2010). I would like to show my gratitude to the Center of Economic Studies (KU Leuven), Licos (KU Leuven) and ZEW (Mannheim) who provided nancial support to start and nish this dissertation. I also take the opportunity to thank the colleagues in Leuven and Mannheim of which many became good friends. With the lunch-partners in crime of the rst hour Thomas, Kristof, Peter and Eline and those who joined sooner or later Laura, Annelore, Ellen, Gerd, Thomas, Joris and many others I held amusing, but neverending discussions often followed by mails in the afternoon which proved that Peter was right, again. With my room mates in Leuven Thomas,
158
Acknowledgements
Alex, Laura and Maryam and Mannheim Mario, Matthias and Uli I did not only share the oce, but also many thoughts, questions and ideas. My fellow MASE students who also joined the PhD programm in Leuven Thijs and Jo remained much appreciated sounding boards throughout the PhD. Besides, I would also thank all my other colleagues in Leuven and Mannheim. Thanks to An, Andr, Anna, Anna, Annie, Aysegul, Bert, e Bram, Catherine, Catherine, Christophe, Damiaan, Dirk, Dirk, Eline, Els, Francine, Frank, Giulia, Heidi, Isabelle, Jacqueline, Jan, Javier, Jeroen, Jeroen, Julien, Karen, Karla, Kobe, Koen, Kris, Kristine, Kristof, Laura, Leonardo, Liesbeth, Lotte, Lydia, Maarten, Marijke, Mark, Mathias, Nicolas, Pablo, Pieter, Priscilla, Rien, Romain, Sandra, Saskia, Sebastian, Simon, Sophie, Stef, Tom, Toon, Wim, Wouter, Xavier and Yuemei; and to Georg, Isabel, Kai, Martin, Nina, Tobias, Ulrich and Vigen. I am also indebted to those who brought the necessary diversion throughout my PhD: the leaders of the youth movements KSA Ter Straeten and Pirlewiet, my squash partner Geert and all the members of the Begijnhof committee. I also thank numerous friends for the support, even those who kept on asking How is your PhD going? and When do you get a real job?. I also express my gratitude for those who have helped me to learn how business and journalism work in practice. Many thanks to Ari, Hilde, Paul and Peter of Antwerp by Bike and Bram, Georges, Jan, Tim and Tom of Apache. I am most grateful to my parents, grandparents, brothers, sister, sistersin-law and parents-in-law who supported me in any respect during the completion of this project. Lastly, and most importantly, I wish to thank my wife Nele. Many thanks for the never-ending support and critical reection during my PhD, and thanks to be an inexhaustible source of warmth and joy in the past nine years. I dedicate this PhD to her.
About the Author

Dries De Smet studied Economic Sciences at the KU Leuven. Afterwards, he completed a one-year master in Journalism at the Erasmushogeschool in Brussels and wrote for a short time for the Belgian newspaper De Standaard. In 2006, he enrolled for the PhD program and entered the research program Master in Science of Economics. His PhD, titled Essays on the economics of media platforms investigates media such as newspapers, telephone directories and websites as two-sided platforms attracting both advertisers and readers. Dries works since September 2010 at ZEW (Mannheim, Germany) and is also involved in the online news magazine Apache. His main research interests are media economics and competition policy.
160
About the Author
Bibliography
An, S. & Bergen, L. (2007), Advertiser pressure on daily newspapers. a survey of advertising sales executives, Journal of Advertising 36(2), 111121. Anderson, S. & Gabszewicz, J. (2006), Handbook of the Economics of Art and Culture, Elsevier Science, Amsterdam, chapter The media and advertising: a tale of twosided markets. Anderson, S. P. & Coate, S. (2005), Market provision of broadcasting: A welfare analysis, Review of Economic Studies 72(4), 947972. Anderson, S. P. & Gans, J. (2010), Platform siphoning: Ad-avoidance and media content, CEPR Discussion Papers 7729, C.E.P.R. Discussion Papers. Argentesi, E. & Filistrucchi, L. (2007), Estimating market power in a two-sided market: The case of newspapers, Journal of Applied Econometrics 22(7), 12471266. Armstrong, M. (2006), Competition in two-sided markets, RAND Journal of Economics 37(3), 66891. Armstrong, M. (2008), Access pricing, bypass and universal service in post, Review of Network Economics 7(2), 172187. Armstrong, M. & Weeds, H. (2005), Public service broadcasting in the digital world, Industrial Organization 0507010, EconWPA. ASNE (1999), Examining our credibility. Perspectives of the public and the press., Technical report, American Society of Newspaper Editors. Bagdikian, B. H. (2004), The New Media Monopoly, Beacon Press. Baker, C. E. (1994), Advertising and a Democratic Press, Princeton University Press. Bala, V. & Goyal, S. (2000), A noncooperative model of network formation, Econometrica 68(5), 11811230. Baldasty, G. J. (1992), The commercialization of news in the nineteenth century, The University of Wisconsin Press. Baron, D. P. (2006), Persistent media bias, Journal of Public Economics 90(1-2), 1 36. Battelle, J. (2007), The Search: how Google and Its Rivals Rewrote the Rules of Business and Transformed Our Culture, Portfolio. Belleamme, P. & Toulemonde, E. (2009), Negative intra-group externalities in twosided markets, International Economic Review 50(1), 245272.
162
BIBLIOGRAPHY
Besanko, D., Donnenfeld, S. & White, L. J. (1988), The multiproduct rm, quality choice, and regulation, Journal of Industrial Economics 36(4), 41129. Bharat, K. (2009), Read news fast with Google fast ip, The Ocial Google Blog (14-09-2009). http://googleblog.blogspot.com/2009/09/read-newsfast-with-google-fast-flip.html. Bignon, V. & Miscio, A. (2009), Media bias in nancial newspapers: Evidence from early 20th century france, Technical report. Blasco, A., Pin, P. & Sobbrio, F. (2010), Paying positive to go negative: Advertisers competition and media reports, Working paper, Universita di Bologna. Bogart, L. (2000), Commercial culture: the media system and the public interest, Transaction Publishers. Bonanno, G. & Vickers, J. (1988), Vertical separation, Journal of Industrial Economics 36(3), 25765. Boswell, D. (2003), Distributed high-performance web crawlers: A survey of the state of the art. Brown, J. R. & Goolsbee, A. (2002), Does the internet make markets more competitive? Evidence from the life insurance industry, Journal of Political Economy 110, 481507. Busse, M. & Rysman, M. (2005), Competition and price discrimination in yellow pages advertising, RAND Journal of Economics 36(2), 378390. Caillaud, B. & Jullien, B. (2001), Competing cybermediaries, European Economic Review 45(4-6), 797808. Caillaud, B. & Jullien, B. (2003), Chicken and egg: Competition among intermediation service providers, RAND Journal of Economics 34(2), 30928. Chiou, L. & Tucker, C. (2011), News and Online Aggregators, SSRN eLibrary . Church, J. & Gandal, N. (1992), Network eects, software provision, and standardization, Journal of Industrial Economics 40(1), 85103. Cisneros, O. (2007), Ebay ghts spiders on the web, Wired Magazine (31-07-2000). http://www.wired.com/politics/law/news/2000/07/37 643. Claussen, D. S. (2002), American journalism: History, principles, practices, McFarland, chapter Economics, business, and nancial motivations, pp. 106115. Cochez, T. (2010), De post laat niet in zijn kaarten kijken over perssteun, Apache (10-08-2010). http://www.apache.be/2010/08/de-post. Coles, M. G. & Muthoo, A. (1998), Strategic bargaining and competitive bidding in a dynamic market equilibrium, Review of Economic Studies 65(2), 23560. Comanor, W. S. & Wilson, T. A. (1974), Advertising and Market Power, Harvard University Press. Cremer, H., Gasmi, F., Grimaud, A. & Laont, J.-J. (1998), Universal services: An economic perspective, Annals of Public and Cooperative Economics 72(1), 543. Crew, M. & Kleindorfer, P. (2002), Postal and Delivery Services: Pricing, Productivity, Regulation and Strategy, Kluwer.
BIBLIOGRAPHY
163
Crmer, J. (2000), Network externalities and universal service obligation in the intere net, European Economic Review 44(4-6), 10211031. Cunningham, A. & Haley, E. (2000), A look inside the world of advertising-free publishing: A case study of ms. magazine, Journal of Current Issues & Research in Advertising 22(2), 1730. Damiano, E. & Li, H. (2007), Price discrimination and ecient matching, Economic Theory 30(2), 243263. Damiano, E. & Li, H. (2008), Competing matchmaking, Journal of the European Economic Association 6(4), 789818. Davies, N. (2008), Flat Earth News. An Award-winning Reporter Exposes Falsehood, Distortion and Propaganda in the Global Media, Chatto & Windus. De Donder, P. (2006), Access pricing in the postal sector: theory and simulations, Review of industrial organization 28(3), 307326. De Koster, L. (2010), An empirical study of commercial media bias in Belgian newspapers, Masters thesis, K.U.Leuven. De Smet, D. & Van Cayseele, P. (2010), Product dierentiation on a platform: the informative and persuasive role of advertising, Technical Report 10-03, CES KULeuven. De Smet, D. & Van Cayseele, P. (2011), The universal service obligation for telephone directories: Regulating the redundant, Technical Report 2011-01, Amsterdam Center for Law & Economics. Dellarocas, C., Katona, Z. & Rand, W. (2010), Media, aggregators and the link economy: Strategic hyperlink formation in content networks, Working Papers 10-13, NET Institute. DellaVigna, S. & Gentzkow, M. (2009), Persuasion: Empirical evidence, NBER Working Papers 15298, National Bureau of Economic Research, Inc. DeLorme, D. & Fedler, F. (2005), An historical analysis of journalists attitudes toward advertisers and advertisings inuence, American Journalism 22(2), 740. Depoorter, B. & Parisi, F. (2002), Fair use and copyright protection: a price theory explanation, International Review of Law and Economics 21(4), 453473. Di Tella, R. & Franceschelli, I. (2009), Government advertising and media coverage of corruption scandals, NBER Working Papers 15402, National Bureau of Economic Research, Inc. Doms, M. & Morin, N. (2004), Consumer sentiment, the economy, and the news media, Technical report. Downes, T. & Greenstein, S. (2007), Understanding why universal service obligations may be unnecessary: The private development of local internet access markets, Journal of Urban Economics 62(1), 226. Durante, R. & Knight, B. (2009), Partisan control, media bias, and viewer responses: Evidence from Berlusconis Italy, NBER Working Papers 14762, National Bureau of Economic Research, Inc.
164
BIBLIOGRAPHY
Economides, N. & Katsamakas, E. (2006), Two-sided competition of proprietary vs. open source technology platforms and the implications for the software industry, Management Science 52(7), 10571071. Economides, N. & T J. (2008), Net neutrality on the internet: A two-sided market ag, analysis, Working Paper Series 727, Research Institute of Industrial Economics. Eisenmann, T., Parker, G. & Van Alstyne, M. W. (2006), Strategies for two-sided markets, Harvard Business Review p. 12. Ellman, M. & Germano, F. (2009), What do the papers sell? a model of advertising and media bias, Economic Journal 119(537), 680704. European Commission (2010), Antitrust: Commission probes allegations of antitrust violations by Google, http://europa.eu/rapid/pressReleasesAction.do?ref erence=IP/10/1624. Evans, D. S. (2004), The antitrust economics of two-sided markets, SSRN eLibrary . Evans, D. S. (2008), The economics of the online advertising industry, Review of Network Economics 7(3). Evans, D. & Schmalensee, R. (2008), Markets with two-sided platforms, Issues in Competition Law and Policy 1(28), 667693. Fico, F., Lacy, S. & Rie, D. (2008), A content analysis guide for media economics scholars, Journal of Media Economics 21, 114130. Fischer, S. (2001), When animals attack: Spiders and internet trespass, Minnesota Intellectual Property Review 2(2), 139182. Fisman, R. (2010), Industrial espionage. how the CIA got the world to buy American during the Cold War., Slate (24-05-2010). Fleetwood, B. (1999), The broken wall: How newspapers are selling their credibility to advertisers., Washington Monthly 1(1). Freedman, J. O. & Pozza, D. C. (2010), Renewed interest in hot news misappropriation claims against online aggregators of news and information, Intellectual Property & Technology Law Journal 22(10), 16. Gallaugher, J. & Wang, Y.-M. (2002), Understanding network eects in software markets: Evidence from web server pricing, MIS Quarterly 26(4), 303327. Gambaro, M. & Puglisi, R. (2009), What do ads buy? Daily coverage of listed companies on the italian press, Departemental Working Papers 2009-36, Department of Economics University of Milan Italy. Garbacz, C. & Thompson, H. G. (2005), Universal telecommunication service: A world perspective, Information Economics and Policy 17(4), 495512. Gentzkow, M. & Shapiro, J. M. (2006), Media bias and reputation, Journal of Political Economy 114(2), 280316. Gentzkow, M. & Shapiro, J. M. (2010), What drives media slant? Evidence from U.S. daily newspapers, Econometrica 78(1), 3571. George, L. & Hogendorn, C. (2011), Aggregators, search and the economics of new media institutions, in 9th ZEW Conference on the Economics of Information and Communication, ZEW.
BIBLIOGRAPHY
165
Germano, F. & Meier, M. (2010), Concentration and self-censorship in commercial media, Technical Report 1256, Universitat Pompeu Fabra. Gomes, R. (2010), Mechanism design in two-sided markets: Auctioning users. mimeo. Goodman, E. (2006), Stealth marketing and editorial integrity, Texas Law Review 85(1), 83152. Gordon, W. (1982), Fair use as market failure: A structural and economic analysis of the betamax case and its predecessors, Columbia Law Review 82, 1600. Graham, S., Cornford, J. & Marvin, S. (1996), The socio-economic benets of a universal telephone network: A demand-side view of universal service, Telecommunications Policy 20(1), 310. Greenwald, A. & Kephart, J. O. (1999), Shopbots and pricebots. Groseclose, T. & Milyo, J. (2005), A measure of media bias, The Quarterly Journal of Economics 120(4), 11911237. Gurun, U. G. & Butler, A. W. (2010), Dont believe the hype: Local media slant, local advertising, and rm value, Technical report, AFA 2010 Atlanta Meetings. Haan, M. A. & Moraga-Gonzalez, J. L. (2009), Advertising for attention in a consumer search model, IESE Research Papers D/794, IESE Business School. Hagiu, A. & Lee, R. S. (2008), Exclusivity and control, SSRN eLibrary . Hahn, R. W. & Wallsten, S. (2006), The economics of net neutrality, The Economists Voice 3(6). Hamilton, J. M. & Krimsky, G. A. (1996), Hold the Press: The Inside Story on Newspapers, Lousiana State University Press. Hamilton, J. T. (2004), All The News Thats Fit To Sell, Princeton, New Jersey. Hart, P. (2009), Fear & favor 2008: Financial woes accelerate corporate pressure in the newsroom, Extra! . Hart, P. (2010), Fear & favor 10th annual report: Hidden interference in the newsroom, Extra! . Hart, P. & Hollar, J. (2004), Fear & favor 2003 the fourth annual report: More examples of medias vulnerability to power, Extra! . Hart, P. & Hollar, J. (2005), Fear & favor 2004 the fth annual report: How power shapes the news, Extra! . Hayakawa, S. (1940), Language in thought and action, Harcourt, Brace & Company. Hays, R. G. & Reisner, A. E. (1990), Feeling the heat from advertisers: Farm magazine writers and ethical pressures, Journalism Quarterly 67(4), 936942. Hendel, I., Nevo, A. & Ortalo-Magn, F. (2009), The relative performance of real e estate marketing platforms: Mls versus fsbomadison.com, American Economic Review 99(5), 187898. Hollar, J., Jackson, J. & Goldstein, H. (2006), Fear & favor 2005 the sixth annual report: Outside (and inside) inuence on the news, Extra! . Hollings, J., Lealand, G., Samson, A. & Tilley, E. (2007), The big NZ journalism survey: underpaid, under-trained, under-resourced, unsure about the future but still idealistic, Pacic Journalism Review 13(2), 175197.
166
BIBLIOGRAPHY
Jaag, C., Koller, M. & Trinkner, U. (2008), Calculating the cost of the universal service obligation: The need for a global approach, Working Papers 0010, Swiss Economics. Jackson, J. (2007), Fear & favor 2006 the seventh annual report: Encroachment without apology, Extra! . Jackson, J. & Hart, P. (2001), Fear & favor 2000 the rst annual report: How power shapes the news, Extra! . Jackson, J. & Hart, P. (2002), Fear & favor 2001 the second annual report: How power shapes the news, Extra! . Jackson, J., Hart, P. & Coen, R. (2003), Fear & favor 2002 the third annual report: How power shapes the news, Extra! . Jackson, M. (2008), Social and Economic Networks, Princeton University Press. Jackson, M. O. (2003), A survey of models of network formation: Stability and eciency, Game Theory and Information 0303011, EconWPA. Jackson, M. O. & Wolinsky, A. (1996), A strategic model of social and economic networks, Journal of Economic Theory 71(1), 4474. Jarvis, J. (2008), The link economy v. the content economy, BuzzMachine (18-062008). http://www.buzzmachine.com/2008/06/18/the-link-economy-v-the-cont ent-economy/. Jennings, F. & Yates, J. (2009), Scrapping over data: are the data scrapers days numbered?, Journal of Intellectual Property Law & Practice 4, 120129. Kaiser, U. & Song, M. (2009), Do media consumers really dislike advertising? An empirical assessment of the role of advertising in print media markets, International Journal of Industrial Organization 27(2), 292301. Kaiser, U. & Wright, J. (2006), Price structure in two-sided markets: Evidence from the magazine industry, International Journal of Industrial Organization 24(1), 1 28. Katz, M. L. & Shapiro, C. (1985), Network externalities, competition, and compatibility, American Economic Review 75(3), 42440. Kelsey (2009), Global yellow pages 2009-2010, Technical report, Kelsey Group. Levy, K. (2010), News corp dumps news aggregator plan, Campaign 42, 11. Lott Jr., J. R. & Hassett, K. A. (2004), Is newspaper coverage of economic events politically biased?, http://www.aei.org/docLib/20040913_588453[1].pdf. Maskin, E. & Riley, J. (1984), Monopoly with incomplete information, RAND Journal of Economics 15(2), 171196. Mayer, M. (2010), Do not neutralise the webs endless search, Financial Times (1407-2010). Mayzlin, D. & Yoganarasimhan, H. (2010), Link to success: How blogs build an audience by promoting rivals, Technical report, Yale School of Management. McCabe, M. & Snyder, C. (2010), The economics of open-access journals, Working Papers 914525, SSRN.
BIBLIOGRAPHY
167
McClung Lee, A. (1937), The Daily Newspaper in America, Macmillan, New York. McCluskey, J. J. & Swinnen, J. F. M. (2010), The Oxford Handbook of Business and Government, Oxford Handbooks, chapter Media Economics and the Political Economy of Information. Milgrom, P. & Roberts, J. (1982), Limit pricing and entry under incomplete information: An equilibrium analysis, Econometrica 50(2), 443459. Mirabel, F. & Poudou, J.-C. (2004), Mechanisms of funding for universal service obligations: the electricity case, Energy Economics 26(5), 801823. Mishkin, A. (2009), The fallacy of the link economy, paidContent.org (13-08-2009). http://paidcontent.org/article/419-the-fallacy-of -the-link-economy/. Mullainathan, S. & Shleifer, A. (2005), The market for news, American Economic Review 95(4), 10311053. Mussa, M. & Rosen, S. (1978), Monopoly and product quality, Journal of Economic Theory 18, 30117. New York Times (2010), The Google algorithm, The New York Times (14-07-2010). http://www.nytimes.com/2010/07/15/opinion/15thu3.html. Nocke, V., Peitz, M. & Stahl, K. (2007), Platform ownership, Journal of the European Economic Association 5(6), 11301160. Parker, G. & Van Alstyne, M. (2005), Two-sided network eects: A theory of information product design, Management Science 51(10). Peitz, M. & Valletti, T. M. (2008), Content and advertising in the media: Pay-tv versus free-to-air, International Journal of Industrial Organization 26(4), 949965. Peitz, M. & Waelbroeck, P. (2006), Why the music industry may gain from free downloading the role of sampling, International Journal of Industrial Organization 24(5), 907913. Petrova, M. (2009), Newspapers and parties: How advertising revenues created an independent press, Working Papers w0131, Center for Economic and Financial Research (CEFIR). Pew Pew (2000), Self censorship: http://people-press.org/report/39/. How often coverage, and say why,
(2004), Bottom-line pressures now hurting http://people-press.org/report/214/.
journalists,
Pew (2008), Financial woes now overshadow all other concerns for journalists, http://people-press.org/report/403/. Pew (2010), State of the media 2010: News leaders and the future, http://www.stateofthemedia.org/2010/specialreport s_survey_executives.php.
Pew (2011), State of the news media 2011: An annual report on American journalism, http://stateofthemedia.org/2011/. Pollock, R. (2007), The control of porting in two-sided markets, Cambridge Working Papers in Economics 0754, Faculty of Economics, University of Cambridge.
168
BIBLIOGRAPHY
Ppress (2010), Onderzoek naar de opleidingsbehoeften van magazine- en vrije pers journalisten, mimeo. Price, C. J. (2003), Interfering owners or meddling advertisers: How network television news correspondents feel about ownership and advertisers inuence on news stories, The Journal of Media Economics 16(3), 175188. Quetelet, A. (1846), Letters addressed to H.R.H. the Grand Duke of Saxe Coburg and Gotha, on the Theory of Probabilities, Charles & Edwin Layton. Raad voor de Journalistiek (2010), Code van de Raad voor de Journalistiek, http://www.rvdj.be/sites/default/files/pdf/code20 10.pdf. Reuter, J. (2009), Does advertising bias product reviews? An analysis of wine ratings, Journal of Wine Economics 4(3), 125151. Reuter, J. & Zitzewitz, E. (2006), Do ads inuence editors? Advertising and bias in the nancial media, The Quarterly Journal of Economics 121(1), 197227. Rey, P. & Tirole, J. (2007), A Primer on Foreclosure, Vol. 3 of Handbook of Industrial Organization, Elsevier, chapter 33, pp. 21452220. Rinallo, D. & Basuroy, S. (2009), Does advertising spending inuence media coverage of the advertiser?, Journal of Marketing 73(6), 3346. Riordan, M. (2001), Handbook of Telecommunications Economics, Elsevier Science, chapter Universal Residential Telephone Service. Rochet, J.-C. & Tirole, J. (2003), Platform competition in two-sided markets, Journal of the European Economic Association 1(4), 9901029. Rochet, J.-C. & Tirole, J. (2006), Two-sided markets: A progress report, RAND Journal of Economics 37(3), 645667. Rodriguez, F. & Storer, D. (2000), Alternative approaches to estimating the cost of the uso in posts, Information Economics and Policy 12(3), 285 299. Rosenfeld, J. (2002), Spiders and crawlers and bots, oh my: The economic eciency and public policy of online contracts that restrict data collection, Stanford Technology Law Review 3. Roy, A. & Chattopadhyay, S. (2010), Stealth marketing as a strategy, Business Horizons 53(1), 6979. Rysman, M. (2004), Competition between networks: A study of the market for yellow pages, Review of Economic Studies 71(2), 483512. Rysman, M. (2009), The economics of two-sided markets, Journal of Economic Perspectives 23(3), 12543. Seldes, G. (1943), The Facts Are... A Guide to Falshood and Propaganda in the Press and Radio, 5 edn, In Fact. Singhal, A. (2011), Supporting choice, ensuring economic opportunity, http://googleblog.blogspot.com/2011/06/supporting-choiceensuring-economic.html. Sobbrio, F. (2009), A citizens-editors model of news media, MPRA Paper 18213, University Library of Munich, Germany.
BIBLIOGRAPHY
169
Soley, L. C. & Craig, R. L. (1992), Advertising pressures on newspapers: A survey, Journal of Advertising 21(4), 110. Soureau, B. (2010a), Krantenuitgevers snappen het internet niet, Apache.be (12-072010). http://www.apache.be/2010/07/krantenuitgevers-snappen-het-interne t-niet/. Soureau, B. (2010b), Reprocopy past richtlijn dieplinken aan, Apache.be (23-092010). http://www.apache.be/2010/09/reprocopy-past-richtlijn-dieplinken -aan/. Spence, M. (1976), Product dierentiation and welfare, American Economic Review 66(2), 40714. Stahl, D. (1989), Oligopolistic pricing with sequential consumer search, American Economic Review 79(4), 700712. Suen, W. (2004), The self-perpetuation of biased beliefs, Economic Journal 114(495), 377396. Sullivan, D. (2007), Google loses in Belgium newspaper case, SearchEngineLand.com (13-02-2007). http://searchengineland.com/google-loses-in-belgi um-newspaper-case-10500. Travis, H. (2006), Google book search and fair use: iTunes for authors, or Napster for books?, University of Miami Law Review 61, 601681. Tsetsura, K. (2005), Bribery for news coverage: Research in Poland., Technical report, Institute for Public Relations Online: International Research. Tucker, C. & Zhang, J. (2008), Decomposing the congestion eect and the crossplatform eect in two-sided networks: A eld experiment, Technical Report 8-12, NET Institute. Upshaw, J., Chernov, G. & Koranda, D. (2007), Telling more than news: Commercial inuence in local television stations, Electronic News 1(2), 6787. Valletti, T. M., Hoernig, S. & Barros, P. P. (2002), Universal service and entry: The role of uniform pricing and coverage constraints, Journal of Regulatory Economics 21(2), 16990. Van Asbroeck, B. & Cock, M. (2007), Belgian newspapers v Google news: 2-0, Journal of Intellectual Property Law & Practice 2(7), 463466. van Caspel, M., Moerman, B., Vermeer, L. & Hermans, E. (2002), Eindrapport. universele telecomdiensten. marktconsultatie vraagzijde en aanbodzijde, Technical report, Ministerie van Economische Zaken. Van Cayseele, P. & Vanormelingen, S. (2009), Prices and network eects in two-sided markets: the Belgian newspaper industry, Technical Report 1404392, SSRN. Varian, H. R. (2007), Position auctions, International Journal of Industrial Organization 25(6), 11631178. Verlinden, A., Billiet, J., Smedts, D., Pyck, H., Page, H. & Van de Velde, M. C. (2005), Doctoreren in vlaanderen. verslag van de survey aan de universiteit gent en de katholieke universiteit leuven, http://www.ugent.be/nl/onderzoek/doctoreren/visie/rapportdoctoreren/sur vey2006.pdf.
170
BIBLIOGRAPHY
Viecens, M. F. (2006), Two-sided platforms with endogenous quality dierentiation, Economics Working Papers we061204, Universidad Carlos III, Departamento de Econom a. Warsh, D. (2007), Everything you wanted to know (but were afraid to ask) about two-sided markets, EconomicPrincipals.com. Wauters, R. (2010), Report: 44 percent of Google news visitors scan headlines, dont click through, TechCrunch (19-01-2010). Weller, D. (1999), Auctions for universal service obligations, Telecommunications Policy 23(9), 645674. Weyl, E. G. (2010), A price theory of multi-sided platforms, American Economic Review 100(4), 164272. Wilbur, K. C. (2008a), How the digital video recorder changes traditional television advertising, Journal of Advertising 37(1), 143149. Wilbur, K. C. (2008b), A two-sided, empirical model of television advertising and viewing markets, Marketing Science 27(3), 356378. Williams, S. (2009), Measuring company A. A case study and critique of a news media content analysis program, Technical report, Institute for Public Relations. Williams, W. S. (1992), For sale! real estate advertising & editorial decisions about real estate news, Newspaper Research Journal 13, 160169. Wood, M. L. M., Nelson, M. R., Cho, J. & Yaros, R. A. (2004), Tonights top story: Commercial content in television news, Journalism & Mass Communication Quarterly 81(4), 807823. Xavier, P. (2003), Should broadband be part of universal service obligations?, info 5(1), 825. Yoo, B., Choudhary, V. & Mukhopadhyay, T. (2007), Electronic b2b marketplaces with dierent ownership structures, Management Science 53(6), 952961. Zellner, A. (1962), An ecient method of estimating seemingly unrelated regression equations and tests for aggregation bias, Journal of the American Statistical Association 57, 348368. Zitzewitz, E. (2009), Forensic economics, http://www.dartmouth.edu/~ericz/forensi c.pdf.
Doctoral dissertations from the Faculty of Business and Economics

(from August 1, 1971 till July 4, 2011) For a more recent list, see http://www.kuleuven.be/doctoraatsverdediging/. 1. GEPTS, Stefaan Stability and efficiency of resource allocation processes in discrete commodity spaces, 1971. 2. PEETERS, Theo Determinanten van de internationale handel in fabrikaten, 1971. 3. VAN LOOY, Wim Personeelsopleiding: een onderzoek naar investeringsaspekten van opleiding, 1971. 4. THARAKAN, Mathew Indian exports to the European community: problems and prospects, 1972. 5. HERROELEN, Willy Heuristische programmatie: methodologische benadering en praktische toepassing op complexe combinatorische problemen, 1972. 6. VANDENBULCKE, Jacques De studie en de evaluatie van data-organisatiemethodes en data-zoekmethodes, 1973. 7. PENNYCUICK, Roy A. The economics of the ecological syndrome, 1973. 8. KAWATA, T. Bualum Formation du capital dorigine belge, dette publique et stratgie du dveloppement e e au Zaire, 1973. 9. DONCKELS, Rik Doelmatige orintering van de sectorale subsidiepolitiek in Belgi: een theoretisch e e onderzoek met empirische toetsing, 1974. 10. VERHELST, Maurice Contribution to the analysis of organizational information systems and their financial benefits, 1974. 11. CLEMEUR, Hugo Enkele verzekeringstechnische vraagstukken in het licht van de nutstheorie, 1974. 12. HEYVAERT, Edward De ontwikkeling van de moderne bank- en krediettechniek tijdens de zestiende en zeventiende eeuw in Europa en te Amsterdam in het bijzonder, 1975. 13. VERTONGHEN, Robert Investeringscriteria voor publieke investeringen: het uitwerken van een operationele theorie met een toepassing op de verkeersinfrastructuur, 1975. 14. Niet toegekend. 15. VANOVERBEKE, Lieven Microeconomisch onderzoek van de sectorile arbeidsmobiliteit, 1975. e 16. DAEMS, Herman The holding company: essays on financial intermediation, concentration and capital market imperfections in the Belgian economy, 1975. 17. VAN ROMPUY, Eric Groot-Brittanni en de Europese monetaire integratie: een onderzoek naar de e gevolgen van de Britse toetreding op de geplande Europese monetaire unie. Leuven, Acco, 1975. XIII, 222 pp. 18. MOESEN, Wim Het beheer van de staatsschuld en de termijnstructuur van de intrestvoeten met een toepassing voor Belgi, 1975. e 19. LAMBRECHT, Marc Capacity constrained multi-facility dynamic lot-size problem, 1976. 20. RAYMAECKERS, Erik De mens in de onderneming en de theorie van het producenten-gedrag: een bijdrage tot transdisciplinaire analyse, 1976. 21. TEJANO, Albert Econometric and input-output models in development planning: the case of the Philippines, 1976. 22. MARTENS, Bernard Prijsbeleid en inflatie met een toepassing op Belgi, 1977. e 23. VERHEIRSTRAETEN, Albert Geld, krediet en intrest in de Belgische financile sector, 1977. e 24. GHEYSSENS, Lieven International diversification through the government bond market: a risk-return analysis, 1977. 25. LEFEBVRE, Chris Boekhoudkundige verwerking en financile verslaggeving van huurkooptransacties e en verkopen op afbetaling bij ondernemingen die aan consumenten verkopen, 1977. 26. KESENNE, Stefan Tijdsallocatie en vrijetijdsbesteding: een econometrisch onderzoek, 1978.
172
Dissertations
27. VAN HERCK, Gustaaf Aspecten van optimaal bedrijfsbeleid volgens het marktwaardecriterium: een risico-rendementsanalyse, 1978. 28. VAN POECK, Andre World price trends and price and wage development in Belgium: an investigation into the relevance of the Scandinavian model of inflation for Belgium, 1979. 29. VOS, Herman De industrile technologieverwerving in Brazili: een analyse, 1978. e e 30. DOMBRECHT, Michel Financial markets, employment and prices in open economies, 1979. 31. DE PRIL, Nelson Bijdrage tot de actuarile studie van het bonus-malussysteem, 1979. e 32. CARRIN, Guy Economic aspects of social security: a public economics approach, 1979. 33. REGIDOR, Baldomero An empirical investigation of the distribution of stock-market prices and weakform efficiency of the Brussels stock exchange, 1979. 34. DE GROOT, Roger Ongelijkheden voor stop loss premies gebaseerd op E.T. systemen in het kader van de veralgemeende convexe analyse, 1979. 35. CEYSSENS, Martin On the peak load problem in the presence of rationizing by waiting, 1979. 36. ABDUL RAZK ABDUL Mixed enterprise in Malaysia: the case study of joint venture between Malysian public corporations and foreign enterprises, 1979. 37. DE BRUYNE, Guido Coordination of economic policy: a game-theoretic approach, 1980. 38. KELLES, Gerard Demand, supply, price change and trading volume on financial markets of the matching-order type. = Vraag, aanbod, koersontwikkeling en omzet op financile markten van het Eue ropese type, 1980. 39. VAN EECKHOUDT, Marc De invloed van de looptijd, de coupon en de verwachte inflatie op het opbrengstverloop van vastrentende financile activa, 1980. e 40. SERCU, Piet Mean-variance asset pricing with deviations from purchasing power parity, 1981. 41. DEQUAE, Marie-Gemma Inflatie, belastingsysteem en waarde van de onderneming, 1981. 42. BRENNAN, John An empirical investigation of Belgian price regulation by prior notification: 1975 1979 - 1982, 1982. 43. COLLA, Annie Een econometrische analyse van ziekenhuiszorgen, 1982. 44. Niet toegekend. 45. SCHOKKAERT, Eric Modelling consumer preference formation, 1982. 46. DEGADT, Jan Specificatie van een econometrisch model voor vervuilingsproblemen met proeven van toepassing op de waterverontreiniging in Belgi, 1982. e 47. LANJONG, Mohammad Nasir A study of market efficiency and risk-return relationships in the Malaysian capital market, 1983. 48. PROOST, Stef De allocatie van lokale publieke goederen in een economie met een centrale overheid en lokale overheden, 1983. 49. VAN HULLE, Cynthia ( /08/83) Shareholders unanimity and optimal corporate decision making in imperfect capital markets, 1983. 50. VAN WOUWE, Martine (2/12/83) Ordening van risicos met toepassing op de berekening van ultieme ru nekansen, 1983. 51. DALCANTARA, Gonzague (15/12/83) SERENA: a macroeconomic sectoral regional and national account econometric model for the Belgian economy, 1983. 52. DHAVE, Piet (24/02/84) De vraag naar geld in Belgi, 1984. e 53. MAES, Ivo (16/03/84) The contribution of J.R. Hicks to macro-economic and monetary theory, 1984. 54. SUBIANTO, Bambang (13/09/84) A study of the effects of specific taxes and subsidies on a firms R & D investment plan, 1984. 55. SLEUWAEGEN, Leo (26/10/84) Location and investment decisions by multinational enterprises in Belgium and Europe, 1984. 56. GEYSKENS, Erik (27/03/85) Produktietheorie en dualiteit, 1985. 57. COLE, Frank (26/06/85) Some algorithms for geometric programming, 1985. 58. STANDAERT, Stan (26/09/86) A study in the economics of repressed consumption, 1986. 59. DELBEKE, Jos (03/11/86) Trendperioden in de geldhoeveelheid van Belgi 1877-1983: een theoretische e en empirische analyse van de Banking school hypothese, 1986. 60. VANTHIENEN, Jan (08/12/86) Automatiseringsaspecten van de specificatie, constructie en manipulatie van beslissingstabellen, 1986. 61. LUYTEN, Robert (30/04/87) A systems-based approach for multi-echelon production/inventory systems, 1987. 62. MERCKEN, Roger (27/04/87) De invloed van de data base benadering op de interne controle, 1987. 63. VAN CAYSEELE, Patrick (20/05/87) Regulation and international innovative activities in the pharmaceutical industry, 1987. 64. FRANCOIS, Pierre (21/09/87) De empirische relevantie van de independence from irrelevant alternatives. Assumptie indiscrete keuzemodellen, 1987. 65. DECOSTER, Andr (23/09/88) Family size, welfare and public policy, 1988. e 66. HEIJNEN, Bart (09/09/88) Risicowijziging onder invloed van vrijstellingen en herverzekeringen: een theoretische analyse van optimaliteit en premiebepaling, 1988. 67. GEEROMS, Hans (14/10/88) Belastingvermijding. Theoretische analyse van de determinanten van de belastingontduiking en de belastingontwijking met empirische verificaties, 1988. 68. PUT, Ferdi (19/12/88) Introducing dynamic and temporal aspects in a conceptual (database) schema, 1988. 69. VAN ROMPUY, Guido (13/01/89) A supply-side approach to tax reform programs. Theory and empirical evidence for Belgium, 1989. 70. PEETERS, Ludo (19/06/89) Een ruimtelijk evenwichtsmodel van de graanmarkten in de E.G.: empirische specificatie en beleidstoepassingen, 1989. 71. PACOLET, Jozef (10/11/89) Marktstructuur en operationele efficintie in de Belgische financile sector, e e 1989. 72. VANDEBROEK, Martina (13/12/89) Optimalisatie van verzekeringscontracten en premieberekeningsprincipes, 1989.
173
73. WILLEKENS, Francois () Determinance of government growth in industrialized countries with applications to Belgium, 1990. 74. VEUGELERS, Reinhilde (02/04/90) Scope decisions of multinational enterprises, 1990. 75. KESTELOOT, Katrien (18/06/90) Essays on performance diagnosis and tacit cooperation in international oligopolies, 1990. 76. WU, Changqi (23/10/90) Strategic aspects of oligopolistic vertical integration, 1990. 77. ZHANG, Zhaoyong (08/07/91) A disequilibrium model of Chinas foreign trade behaviour, 1991. 78. DHAENE, Jan (25/11/91) Verdelingsfuncties, benaderingen en foutengrenzen van stochastische grootheden geassocieerd aan verzekeringspolissen en -portefeuilles, 1991. 79. BAUWELINCKX, Thierry (07/01/92) Hierarchical credibility techniques, 1992. 80. DEMEULEMEESTER, Erik (23/3/92) Optimal algorithms for various classes of multiple resourceconstrained project scheduling problems, 1992. 81. STEENACKERS, Anna (1/10/92) Risk analysis with the classical actuarial risk model: theoretical extensions and applications to Reinsurance, 1992. 82. COCKX, Bart (24/09/92) The minimum income guarantee. Some views from a dynamic perspective, 1992. 83. MEYERMANS, Eric (06/11/92) Econometric allocation systems for the foreign exchange market: Specification, estimation and testing of transmission mechanisms under currency substitution, 1992. 84. CHEN, Guoqing (04/12/92) Design of fuzzy relational databases based on fuzzy functional dependency, 1992. 85. CLAEYS, Christel (18/02/93) Vertical and horizontal category structures in consumer decision making: The nature of product hierarchies and the effect of brand typicality, 1993. 86. CHEN, Shaoxiang (25/03/93) The optimal monitoring policies for some stochastic and dynamic production processes, 1993. 87. OVERWEG, Dirk (23/04/93) Approximate parametric analysis and study of cost capacity management of computer configurations, 1993. 88. DEWACHTER, Hans (22/06/93) Nonlinearities in speculative prices: The existence and persistence of nonlinearity in foreign exchange rates, 1993. 89. LIN, Liangqi (05/07/93) Economic determinants of voluntary accounting choices for R & D expenditures in Belgium, 1993. 90. DHAENE, Geert (09/07/93) Encompassing: formulation, properties and testing, 1993. 91. LAGAE, Wim (20/09/93) Marktconforme verlichting van soevereine buitenlandse schuld door private crediteuren: een neo-institutionele analyse, 1993. 92. VAN DE GAER, Dirk (27/09/93) Equality of opportunity and investment in human capital, 1993. 93. SCHROYEN, Alfred (28/02/94) Essays on redistributive taxation when monitoring is costly, 1994. 94. STEURS, Geert (15/07/94) Spillovers and cooperation in research and development, 1994. 95. BARAS, Johan (15/09/94) The small sample distribution of the Wald, Lagrange multiplier and likelihood ratio tests for homogeneity and symmetry in demand analysis: a Monte Carlo study, 1994. 96. GAEREMYNCK, Ann (08/09/94) The use of depreciation in accounting as a signalling device, 1994. 97. BETTENDORF, Leon (22/09/94) A dynamic applied general equilibrium model for a small open economy, 1994. 98. TEUNEN, Marleen (10/11/94) Evaluation of interest randomness in actuarial quantities, 1994. 99. VAN OOTEGEM, Luc (17/01/95) An economic theory of private donations, 1995. 100. DE SCHEPPER, Ann (20/03/95) Stochastic interest rates and the probabilistic behaviour of actuarial functions, 1995. 101. LAUWERS, Luc (13/06/95) Social choice with infinite populations, 1995. 102. WU, Guang (27/06/95) A systematic approach to object-oriented business modeling, 1995. 103. WU, Xueping (21/08/95) Term structures in the Belgian market: model estimation and pricing error analysis, 1995. 104. PEPERMANS, Guido (30/08/95) Four essays on retirement from the labor force, 1995. 105. ALGOED, Koen (11/09/95) Essays on insurance: a view from a dynamic perspective, 1995. 106. DEGRYSE, Hans (10/10/95) Essays on financial intermediation, product differentiation, and market structure, 1995. 107. MEIR, Jos (05/12/95) Het strategisch groepsconcept toegepast op de Belgische financile sector, 1995. e 108. WIJAYA, Miryam Lilian (08/01/96) Voluntary reciprocity as an informal social insurance mechanism: a game theoretic approach, 1996. 109. VANDAELE, Nico (12/02/96) The impact of lot sizing on queueing delays: multi product, multi machine models, 1996. 110. GIELENS, Geert (27/02/96) Some essays on discrete time target zones and their tails, 1996. 111. GUILLAUME, Dominique (20/03/96) Chaos, randomness and order in the foreign exchange markets. Essays on the modelling of the markets, 1996. 112. DEWIT, Gerda (03/06/96) Essays on export insurance subsidization, 1996. 113. VAN DEN ACKER, Carine (08/07/96) Belief-function theory and its application to the modeling of uncertainty in financial statement auditing, 1996. 114. IMAM, Mahmood Osman (31/07/96) Choice of IPO Flotation Methods in Belgium in an Asymmetric Information Framework and Pricing of IPOs in the Long-Run, 1996. 115. NICAISE, Ides (06/09/96) Poverty and Human Capital, 1996. 116. EYCKMANS, Johan (18/09/97) On the Incentives of Nations to Join International Environmental Agreements, 1997. 117. CRISOLOGO-MENDOZA, Lorelei (16/10/97) Essays on Decision Making in Rural Households: a study of three villages in the Cordillera Region of the Philippines, 1997. 118. DE REYCK, Bert (26/01/98) Scheduling Projects with Generalized Precedence Relations: Exact and Heuristic Procedures, 1998. 119. VANDEMAELE Sigrid (30/04/98) Determinants of Issue Procedure Choice within the Context of the French IPO Market: Analysis within an Asymmetric Information Framework, 1998.
174
Dissertations
120. VERGAUWEN Filip (30/04/98) Firm Efficiency and Compensation Schemes for the Management of Innovative Activities and Knowledge Transfers, 1998. 121. LEEMANS Herlinde (29/05/98) The Two-Class Two-Server Queueing Model with Nonpreemptive Heterogeneous Priority Structures, 1998. 122. GEYSKENS Inge (4/09/98) Trust, Satisfaction, and Equity in Marketing Channel Relationships, 1998. 123. SWEENEY John (19/10/98) Why Hold a Job ? The Labour Market Choice of the Low-Skilled, 1998. 124. GOEDHUYS Micheline (17/03/99) Industrial Organisation in Developing Countries, Evidence from Cte dIvoire, 1999. o 125. POELS Geert (16/04/99) On the Formal Aspects of the Measurement of Object-Oriented Software Specifications, 1999. 126. MAYERES Inge (25/05/99) The Control of Transport Externalities: A General Equilibrium Analysis, 1999. 127. LEMAHIEU Wilfried (5/07/99) Improved Navigation and Maintenance through an Object-Oriented Approach to Hypermedia Modelling, 1999. 128. VAN PUYENBROECK Tom (8/07/99) Informational Aspects of Fiscal Federalism, 1999. 129. VAN DEN POEL Dirk (5/08/99) Response Modeling for Database Marketing Using Binary Classification, 1999. 130. GIELENS Katrijn (27/08/99) International Entry Decisions in the Retailing Industry: Antecedents and Performance Consequences, 1999. 131. PEETERS Anneleen (16/12/99) Labour Turnover Costs, Employment and Temporary Work, 1999. 132. VANHOENACKER Jurgen (17/12/99) Formalizing a Knowledge Management Architecture MetaModel for Integrated Business Process Management, 1999. 133. NUNES Paulo (20/03/2000) Contingent Valuation of the Benefits of Natural Areas and its Warmglow Component, 2000. 134. VAN DEN CRUYCE Bart (7/04/2000) Statistische discriminatie van allochtonen op jobmarkten met rigide lonen, 2000. 135. REPKINE Alexandre (15/03/2000) Industrial restructuring in countries of Central and Eastern Europe: Combining branch-, firm- and product-level data for a better understanding of Enterprises behaviour during transition towards market economy, 2000. 136. AKSOY, Yunus (21/06/2000) Essays on international price rigidities and exchange rates, 2000. 137. RIYANTO, Yohanes Eko (22/06/2000) Essays on the internal and external delegation of authority in firms, 2000. 138. HUYGHEBAERT, Nancy (20/12/2000) The Capital Structure of Business Start-ups, 2000. 139. FRANCKX Laurent (22/01/2001) Ambient Inspections and Commitment in Environmental Enforcement, 2001 140. VANDILLE Guy (16/02/2001) Essays on the Impact of Income Redistribution on Trade, 2001 141. MARQUERING Wessel (27/04/2001) Modeling and Forecasting Stock Market Returns and Volatility, 2001. 142. FAGGIO Giulia (07/05/2001) Labor Market Adjustment and Enterprise Behavior in Transition, 2001. 143. GOOS Peter (30/05/2001) The Optimal Design of Blocked and Split-plot experiments, 2001. 144. LABRO Eva (01/06/2001) Total Cost of Ownership Supplier Selection based on Activity Based Costing and Mathematical Programming, 2001. 145. VANHOUCKE Mario (07/06/2001) Exact Algorithms for various Types of Project Scheduling Problems. Nonregular Objectives and time/cost Trade-offs, 2001. 146. BILSEN Valentijn (28/08/2001) Entrepreneurship and Private Sector Development in Central European Transition Countries, 2001. 147. NIJS Vincent (10/08/2001) Essays on the dynamic Category-level Impact of Price promotions, 2001. 148. CHERCHYE Laurens (24/09/2001) Topics in Non-parametric Production and Efficiency Analysis, 2001. 149. VAN DENDER Kurt (15/10/2001) Aspects of Congestion Pricing for Urban Transport, 2001. 150. CAPEAU Bart (26/10/2001) In defence of the excess demand approach to poor peasants economic behaviour. Theory and Empirics of non-recursive agricultural household modelling, 2001. 151. CALTHROP Edward (09/11/2001) Essays in urban transport economics, 2001. 152. VANDER BAUWHEDE Heidi (03/12/2001) Earnings management in an Non-Anglo-Saxon environment, 2001. 153. DE BACKER Koenraad (22/01/2002) Multinational firms and industry dynamics in host countries : the case of Belgium, 2002. 154. BOUWEN Jan (08/02/2002) Transactive memory in operational workgroups. Concept elaboration and case study, 2002. 155. VAN DEN BRANDE Inge (13/03/2002) The psychological contract between employer and employee : a survey among Flemish employees, 2002. 156. VEESTRAETEN Dirk (19/04/2002) Asset Price Dynamics under Announced Policy Switching, 2002. 157. PEETERS Marc (16/05/2002) One Dimensional Cutting and Packing : New Problems and Algorithms, 2002. 158. SKUDELNY Frauke (21/05/2002) Essays on The Economic Consequences of the European Monetary Union, 2002. 159. DE WEERDT Joachim (07/06/2002) Social Networks, Transfers and Insurance in Developing countries, 2002. 160. TACK Lieven (25/06/2002) Optimal Run Orders in Design of Experiments, 2002. 161. POELMANS Stephan (10/07/2002) Making Workflow Systems work. An investigation into the Importance of Task-appropriation fit, End-user Support and other Technological Characteristics, 2002. 237 pp. 162. JANS Raf (26/09/2002) Capacitated Lot Sizing Problems : New Applications, Formulations and Algorithms, 2002. 163. VIAENE Stijn (25/10/2002) Learning to Detect Fraud from enriched Insurance Claims Data (Context, Theory and Applications), 2002.
175
164. AYALEW Tekabe (08/11/2002) Inequality and Capital Investment in a Subsistence Economy, 2002. 165. MUES Christophe (12/11/2002) On the Use of Decision Tables and Diagrams in Knowledge Modeling and Verification, 2002. 166. BROCK Ellen (13/03/2003) The Impact of International Trade on European Labour Markets, 2002. 167. VERMEULEN Frederic (29/11/2002) Essays on the collective Approach to Household Labour Supply, 2002. 168. CLUDTS Stephan (11/12/2002) Combining participation in decision-making with financial participation : theoretical and empirical perspectives, 2002. 169. WARZYNSKI Frederic (09/01/2003) The dynamic effect of competition on price cost margins and innovation, 2003. 170. VERWIMP Philip (14/01/2003) Development and genocide in Rwanda ; a political economy analysis of peasants and power under the Habyarimana regime, 2003. 171. BIGANO Andrea (25/02/2003) Environmental regulation of the electricity sector in a European Market Framework, 2003. 172. MAES Konstantijn (24/03/2003) Modeling the Term Structure of Interest Rates Across Countries, 2003. 173. VINAIMONT Tom (26/02/2003) The performance of One- versus Two-Factor Models of the Term Structure of Interest Rates, 2003. 174. OOGHE Erwin (15/04/2003) Essays in multi-dimensional social choice, 2003. 175. FORRIER Anneleen (25/04/2003) Temporary employment, employability and training, 2003. 176. CARDINAELS Eddy (28/04/2003) The role of cost system accuracy in managerial decision making, 2003. 177. DE GOEIJ Peter (02/07/2003) Modeling Time-Varying Volatility and Interest Rates, 2003. 178. LEUS Roel (19/09/2003) The generation of stable project plans. Complexity and exact algorithms, 2003. 179.MARINHEIRO Carlos (23/09/2003) EMU and fiscal stabilisation policy : the case of small countries, 2003. 180. BAESENS Bart (24/09/2003) Developing intelligent systems for credit scoring using machine learning techniques, 2003. 181. KOCZY Laszlo (18/09/2003) Solution concepts and outsider behaviour in coalition formation games, 2003. 182. ALTOMONTE Carlo (25/09/2003) Essays on Foreign Direct Investment in transition countries : learning from the evidence, 2003. 183. DRIES Liesbeth (10/11/2003) Transition, Globalisation and Sectoral Restructuring: Theory and Evidence from the Polish Agri-Food Sector, 2003. 184. DEVOOGHT Kurt (18/11/2003) Essays On Responsibility-Sensitive Egalitarianism and the Measurement of Income Inequality, 2003. 185. DELEERSNYDER Barbara (28/11/2003) Marketing in Turbulent Times, 2003. 186. ALI Daniel (19/12/2003) Essays on Household Consumption and Production Decisions under Uncertainty in Rural Ethiopia, 2003. 187. WILLEMS Bert (14/01/2004) Electricity networks and generation market power, 2004. 188. JANSSENS Gust (30/01/2004) Advanced Modelling of Conditional Volatility and Correlation in Financial Markets, 2004. 189. THOEN Vincent (19/01/2004) On the valuation and disclosure practices implemented by venture capital providers, 2004. 190. MARTENS Jurgen (16/02/2004) A fuzzy set and stochastic system theoretic technique to validate simulation models, 2004. 191. ALTAVILLA Carlo (21/05/2004) Monetary policy implementation and transmission mechanisms in the Euro area., 2004. 192. DE BRUYNE Karolien (07/06/2004) Essays in the location of economic activity, 2004. 193. ADEM Jan (25/06/2004) Mathematical programming approaches for the supervised classification problem., 2004. 194. LEROUGE Davy (08/07/2004) Predicting Product Preferences : the effect of internal and external cues., 2004. 195. VANDENBROECK Katleen (16/07/2004) Essays on output growth, social learning and land allocation in agriculture : micro-evidence from Ethiopia and Tanzania, 2004. 196. GRIMALDI Maria (03/09/004) The exchange rate, heterogeneity of agents and bounded rationality. , 2004. 197. SMEDTS Kristien (26/10/2004) Financial integration in EMU in the framework of the no-arbitrage theory, 2004. 198. KOEVOETS Wim (12/11/2004) Essays on Unions, Wages and Employment, 2004. 199. CALLENS Marc (22/11/2004) Essays on multilevel logistic Regression, 2004. 200. RUGGOO Arvind (13/12/2004) Two stage designs robust to model uncertainty, 2004. 201. HOORELBEKE Dirk (28/01/2005) Bootstrap and Pivoting Techniques for Testing Multiple Hypotheses, 2005. 202. ROUSSEAU Sandra (17/02/2005) Selecting Environmental Policy Instruments in the Presence of Incomplete Compiance, 2005. 203. VAN DER MEULEN Sofie (17/02/2005) Quality of Financial Statements : Impact of the external auditor and applied accounting standards, 2005. 204. DIMOVA Ralitza (21/02/2005) Winners and Losers during Structural Reform and Crisis : the Bulgarian Labour Market Perspective, 2005. 205. DARKIEWICZ Grzegorz (28/02/2005) Value-at-risk in Insurance and Finance : the Comonotonicity Approach, 2005. 206. DE MOOR Lieven (20/05/2005) The Structure of International Stock Returns : Size, Country and Sector Effects in Capital Asset Pricing, 2005. 207. EVERAERT Greetje (27/06/2005) Soft Budget Constraints and Trade Policies : The Role of Institutional and External Constraints, 2005.
176
Dissertations
208. SIMON Steven (06/07/2005) The Modeling and Valuation of complex Derivatives : the Impact of the Choice of the term structure model, 2005. 209. MOONEN Linda (23/09/2005) Algorithms for some graph-theoretical optimization problems, 2005. 210. COUCKE Kristien (21/09/2005) Firm and industry adjustment under de-industrialisation and globalization of the Belgian economy, 2005. 211. DECAMPS MARC (21/10/2005) Some actuarial and financial applications of generalized diffusion processes, 2005. 212. KIM HELENA (29/11/2005) Escalation games: an instrument to analyze conflicts. The strategic approach to the bargaining problem, 2005. 213. GERMENJI ETLEVA (06/01/2006) Essays on the economics of emigration from Albania, 2006. 214. BELIEN JEROEN (18/01/2006) Exact and heuristic methodologies for scheduling in hospitals: problems, formulations and algorithms, 2006. 215. JOOSSENS KRISTEL (10/02/2006) Robust discriminant analysis, 2006. 216. VRANKEN LIESBET (13/02/2006) Land markets and production efficiency in transition economies, 2006. 217. VANSTEENKISTE ISABEL (22/02/2006) Essays on non-linear modelling in international macroeconomics, 2006. 218. WUYTS Gunther (31/03/2006) Essays on the liquidity of financial markets, 2006. 219. DE BLANDER Rembert (28/04/2006) Essays on endogeneity and parameter heterogeneity in crosssection and panel data, 2006. 220. DE LOECKER Jan (12/05/2006) Industry dynamics and productivity, 2006. 221. LEMMENS Aurlie (12/05/2006) Advanced classification and time-series methods in marketing, e 2006. 222. VERPOORTEN Marijke (22/05/2006) Conflict and survival: an analysis of shocks, coping strategies and economic mobility in Rwanda, 1990-2002, 2006. 223. BOSMANS Kristof (26/05/2006) Measuring economic inequality and inequality aversion, 2006. 224. BRENKERS Randy (29/05/2006) Policy reform in a market with differentiated products: applications from the car market, 2006. 225. BRUYNEEL Sabrina (02/06/2006) Self-econtrol depletion: Mechanisms and its effects on consumer behavior, 2006. 226. FAEMS Dries (09/06/2006) Collaboration for innovation: Processes of governance and learning, 2006. 227. BRIERS Barbara (28/06/2006) Countering the scrooge in each of us: on the marketing of cooperative behavior, 2006. 228. ZANONI Patrizia (04/07/2006) Beyond demography: Essays on diversity in organizations, 2006. 229. VAN DEN ABBEELE Alexandra (11/09/2006) Management control of interfirm relations: the role of information, 2006. 230. DEWAELHEYNS Nico (18/09/2006) Essays on internal capital markets, bankruptcy and bankruptcy reform, 2006. 231. RINALDI Laura (19/09/2006) Essays on card payments and household debt, 2006. 232. DUTORDOIR Marie (22/09/2006) Determinants and stock price effects of Western European convertible debt offerings: an empirical analysis, 2006. 233. LYKOGIANNI Elissavet (20/09/2006) Essays on strategic decisions of multinational enterprises: R & D decentralization, technology transfers and modes of foreign entry, 2006. 234. ZOU Jianglei (03/10/2006) Inter-firm ties, plant networks, and multinational firms: essays on FDI and trade by Japanse firms., 2006. 235. GEYSKENS Kelly (12/10/2006) The ironic effects of food temptations on self-control performance, 2006. 236. BRUYNSEELS Liesbeth (17/10/2006) Client strategic actions, going-concern audit opinions and audit reporting errors, 2006. 237. KESSELS Roselinde (23/10/2006) Optimal designs for the measurement of consumer preferences, 2006. 238. HUTCHINSON John (25/10/2006) The size distribution and growth of firms in transition countries, 2006. 239. RENDERS Annelies (26/10/2006) Corporate governance in Europe: The relation with accounting standards choice, performance and benefits of control, 2006. 240. DE WINNE Sophie (30/10/2006) Exploring terra incognita: human resource management and firm performance in small and medium-sized businesses, 2006. 241. KADITI Eleni (10/11/2006) Foreign direct investments in transition economies, 2006. 242. ANDRIES Petra (17/11/2006) Technology-based ventures in emerging industries: the quest for a viable business model, 2006. 243. BOUTE Robert (04/12/2006) The impact of replenishment rules with endogenous lead times on supply chain performance, 2006. 244. MAES Johan (20/12/2006) Corporate entrepreneurship: an integrative analysis of a resource-based model. Evidence from Flemish enterprises, 2006. 245. GOOSSENS Dries (20/12/2006) Exact methods for combinatorial auctions, 2006. 246. GOETHALS Frank (22/12/2006) Classifying and assessing extended enterprise integration approaches, 2006. 247. VAN DE VONDER Stijn (22/12/2006) Proactive-reactive procedures for robust project scheduling, 2006. 248. SAVEYN Bert (27/02/2007) Environmental policy in a federal state, 2007. 249. CLEEREN Kathleen (13/03/2007) Essays on competitive structure and product-harm crises, 2007. 250. THUYSBAERT Bram (27/04/2007) Econometric essays on the measurement of poverty, 2007. 251. DE BACKER Manu (07/05/2007) The use of Petri net theory for business process verification, 2007. 252. MILLET Kobe (15/05/2007) Prenatal testosterone, personality, and economic behavior, 2007.
177
253. HUYSMANS Johan (13/06/2007) Comprehensible predictive models: New methods and insights, 2007. 254. FRANCKEN Nathalie (26/06/2007) Mass Media, Government Policies and Economic Development: Evidence from Madagascar, 2007. 255. SCHOUBBEN Frederiek (02/07/2007) The impact of a stock listing on the determinants of firm performance and investment policy, 2007. 256. DELHAYE Eef (04/07/2007) Economic analysis of traffic safety, 2007. 257. VAN ACHTER Mark (06/07/2007) Essays on the market microstructure of financial markets, 2007. 258. GOUKENS Caroline (20/08/2007) Desire for variety: understanding consumers preferences for variety seeking, 2007. 259. KELCHTERMANS Stijn (12/09/2007) In pursuit of excellence: essays on the organization of higher education and research, 2007. 260. HUSSINGER Katrin (14/09/2007) Essays on internationalization, innovation and firm performance, 2007. 261. CUMPS Bjorn (04/10/2007) Business-ICT alignment and determinants, 2007. 262. LYRIO Marco (02/11/2007) Modeling the yield curve with macro factors, 2007. 263. VANPEE Rosanne (16/11/2007) Home bias and the implicit costs of investing abroad, 2007. 264. LAMBRECHTS Olivier (27/11/2007) Robust project scheduling subject to resource breakdowns, 2007. 265. DE ROCK Bram (03/12/2007) Collective choice behaviour: non parametric characterization, 2007. 266. MARTENS David (08/01/2008) Building acceptable classification models for financial engineering applications, 2008. 267. VAN KERCKHOVEN Johan (17/01/2008) Predictive modelling: variable selection and classification efficiencies, 2008. 268. CIAIAN Pavel (12/02/2008) Land, EU accession and market imperfections, 2008. 269. TRUYTS Tom (27/02/2008) Diamonds are a girls best friend: five essays on the economics of social status, 2008. 270. LEWIS Vivien (17/03/2008) Applications in dynamic general equilibrium macroeconomics, 2008. 271. CAPPELLEN Tineke (04/04/2008) Worldwide coordination in a transnational environment: An inquiry into the work and careers of global managers, 2008. 272. RODRIGUEZ Victor (18/04/2008) Material transfer agreements: research agenda choice, co-publication activity and visibility in biotechnology, 2008. 273. QUAN Qi (14/04/2008) Privatization in China: Examining the endogeneity of the process and its implications for the performance of newly privatized firms, 2008. 274. DELMOTTE Jeroen (30/04/2008) Evaluating the HR function: Empirical studies on HRM architecture and HRM system strength, 2008. 275. ORSINI Kristian (05/05/2008) Making work pay: Insights from microsimulation and random utility models, 2008. 276. HOUSSA Romain (13/05/2008) Macroeconomic fluctuations in developing countries, 2008. 277. SCHAUMANS Catherine (20/05/2008) Entry, regulation and economic efficiency: essays on health professionals, 2008. 278. CRABBE Karen (21/05/2008) Essays on corporate tax competition in Europe, 2008 279. GELPER Sarah (30/05/2008) Economic time series analysis: Granger causality and robustness, 2008 280. VAN HOVE Jan (20/06/2008) The impact of technological innovation and spillovers on the pattern and direction of international trade, 2008 281. DE VILLE DE GOYET Cdric (04/07/2008) Hedging with futures in agricultural commodity mare kets, 2008 282. FRANCK Tom (15/07/2008) Capital structure and product market interactions: evidence from business start-ups and private firms, 2008 283. ILBAS Pelin (15/09/2008) Optimal monetary policy design in dynamic macroeconomics, 2008 284. GOEDERTIER Stijn (16/09/2008) Declarative techniques for modeling and mining business processes, 2008 285. LAMEY Lien (22/09/2008) The private-label nightmare: can national brands ever wake up?, 2008 286. VANDEKERCKHOVE Jan (23/09/2008) Essays on research and development with spillovers, 2008 287. PERNOT Eli (25/09/2008) Management control system design for supplier relationships in manufacturing: Case study evidence from the automotive industry, 2008 288. AERTS Kris (16/10/2008) Essays on the economics of evaluation: public policy and corporate strategies in innovation, 2008 289. BOUDT Kris (27/11/2008) Estimation of financial risk under non-normal distributions, 2008 290. MARKIEWICZ Agnieszka (01.12.2008) Essays in exchange rate economics, 2008 291. GLADY Nicolas (08.12.2008) Customer profitability modeling, 2008 292. MIERZEJEWSKI Fernando (11.12.2008) Essays on liquidity-preference in markets with borrowing restrictions, 2008 293. VAN BEVEREN Ilke (15.12.2008) Globalization and firm dynamics, 2008 294. VERBRUGGEN Marijke (16.12.2008) The role of career counseling in the new career era, 2008 295. CORNIL Anneleen (15.12.2008) Essays on financial statement quality: The influence and the effects of accounting restatements, 2008 296. SOENEN Helke (16.12.2008) Between vulnerability and creativity: an anthropological economic analysis of undocumented migrants in Brussels, 2008 297. VERGOTE Olivier (17.12.2008) Financial transaction data and volatility, 2008 298. COLANTONE Italo (18.12.2008) Essays on entrepreneurship and industrial dynamics in an open economic context, 2008 299. LETEN Bart (19.12.2008) Technology scope decisions in large firms, 2008 300. HERBOTS Jade (19.12.2008) Dynamic project portfolio management: Selection and capacity-allocation decisions, 2008
178
Dissertations
301. JURZYK Emilia (22.12.2008) Essays on the effects of foreign bank entry into Central and Eastern European countries, 2008 302. CARDOEN Brecht (23.01.2009) Operating room planning and scheduling: solving a surgical case sequencing problem, 2009 303. VERMEULEN Bart (27.01.2009) Design issues in conjoint analysis for market and non-market valuation, 2009 304. JIE YU (09.02.2009) Optimal design methodology for choice experiments in the presence of model uncertainty and consumer heterogeneity, 2009 305. DE RORE Lotte (03.03.2009) Measuring productivity and improving efficiency in software development environments, 2009 306. LIU Fang (20.03.2009) The prediction power of interest rates with respect to exchange rates and consumption, 2009 307. SLOOTMAEKERS Veerle (20.03.2009) Institutions and market structure in transition countries, 2009 308. DE WITTE Kristof (31.03.2009) On analyzing drinking water monopolies by robust non-parametric efficiency estimations, 2009 309. CONSENTINO Fabrizio (27.04.2009) Modelling the missingness: estimation, testing and model selection, 2009 310. COENE Sofie (26.05.2009) Routing problems with profits and periodicity, 2009 311. CREEMERS Stefan (15.06.2009) Appointment-driven queueing systems, 2009 312. CHENG Xiaoqiang (19.06.2009) Legal system, financial development and economic performance: evidence from China, 2009 313. VERRIEST Arnt (22.06.2009) Institutional factors, firm profitability and financial reporting quality, 2009 314. DECANCQ Koen (15.06.2009) Essays on the measurement of multidimensional inequality, 2009 315. VAN DEN BERGH Bram (02.07.2009) Basis instinct: the fire of desire in economic decisions, 2009 316. GIJSENBERG Maarten (06.07.2009) Timing is money. In the search of the role of timing in marketing decisions and effectiveness, 2009 317. CALLAERT Julie (16.09.2009) Enterprising academics: (when) does the combination of scientific and patenting activities make sense?, 2009 318. JOCHMANS Koen (18.09.2009) Essays on inference in econometric models with heterogeneity and endogeneity, 2009 319. FERRARI Stijn (23.09.2009) Empirical models of restricted entry: efficiency in ATM deployment and retail magazine distribution, 2009 320. HAESEN Raf (21.10.2009) Designing information system services in information-intensive organizations, 2009 321. VAN CAMPENHOUT Bjrn (29.10.2009) Sample splitting and threshold estimation techniques. Apo plications in development studies, 2009 322. DUTILLIEUX Wouter (04.11.2009) Audit research opportunities in European institutional settings, 2009 323. LUYPAERT Mathieu (17.12.2009) Mergers and acquisitions: when do they really create value?, 2009 324. CHEN Xinliang (17.12.2009) Super-replicating exotic options, allocating capital and approximating aggregate distributions using the comonotonic approach, 2009 325. GRINE Slimane (18.01.2010) Multi-layer model of correlated instruments. Application to energy markets, 2010 326. DUNKERLEY Fay (25.02.2010) Essays in transport economics, 2010 327. VERGHOTE Patrick (24.02.2010) Essays on international account management, 2010 328. BUI THI NGOC Tuan (25.02.2010) Spot and forward prices in the Brussels stock exchange: Differential costs, informativeness and price discovery, 2010 329. ZARNIC Ziga (19.03.2010) Trade, institutional measures and competitiveness, 2010 330. VANORMELINGEN Stijn (31.03.2010) Essays on empirical industrial organization, 2010 331. STOUTHUYSEN Kristof (17.05.2010) Formal control in interfirm service exchanges, 2010 332. BINZ Hanna (25.05.2010) Who does (not) want to pay for innovation - the financing gap story revisited, 2010 333. PERSYN Damiaan (28.05.2010) Essays on unionised labour markets in a globalising world, 2010 334. MASSCHELEIN Stijn (31.05.2010) Information and control mechanisms in collaborations: three experimental studies, 2010 335. NGUYEN THI TUONG Van (31.08.2010) Essays on commodity prices, 2010 336. ZULUAGA DIAZ Blanca (03.09.2010) Essays on schooling investment decisions and poverty, 2010 337. FATICA Serena (03.09.2010) Three essays on firms location choices and public policy, 2010 338. XIANG Tao (10.09.2010) Food standards and development: partial and general equilibrium analysis with applications to China, 2010 339. DEBLAERE Filip (17.09.2010) Resource-constrained project scheduling under uncertainty, 2010 340. TALLA NOBIBON Fabrice (24.09.2010) Algorithms for selection and graph-coloring problems with applications in marketing and micro-economics, 2010 341. VAN DE VOORDE CARINE (24.09.2010) Essays on risk adjustment in health insurance, 2010 342. MAHIEU Koen (08.10.2010) Robut smoothing and forecasting of multivariate time series, 2010 343. ZHAONING Shang (08.10.2010) A recursive approach to general diffusion processes: with applications to financial and insurance pricing, 2010 344. CARBONEZ Katelijne (11.10.2010) On the behavior of commodity prices: statistics and economics, 2010 345. KUPPER Gerd (27.10.2010) Contributions to European energy policy, 2010 346. VAN DER LOO Saskia (29.10.2010) Transport infrastructure pricing and investment issues, 2010 347. LECOCQ Cathy (26.10.2010) Technological performance of regions (and firms). The case of biotechnology, 2010 348. VAN LANDEGHEM Bert (22.11.2010) Essays on subjective well-being, 2010
179
349. POELMANS Jonas (13.12.2010) Essays on using formal concept analysis in information engineering, 2010 350. SABBE Jeroen (14.12.2010) Revealed preference analysis of consumption and production behavior: applications of the collective model, 2010 351. MONSIEUR Geert (17.12.2010) Pattern-based coordination in process-based service compositions, 2010 352. VISNJIC Ivanka (22.12.2010) Conceiving and implementing service business models within manufacturing firms, 2010 353. CLAERHOUT Diederik (17.01.2011) The supply chain operations planning: the lead time setting problem, 2011 354. VAN LAERE Elisabeth (26.01.2011) Capital regulation of financial institutions, the role of ratings and the tension field between regulation and economic reality, 2011 355. JI YUEMEI (01.03.2011) Three essays on the economics of education: a behavioral approach, 2011 356. NEYENS Inge (03.03.2011) Composing and managing technological alliance portfolios, 2011 357. ROVIRA KALTWASSER Pablo (02.03.2011) Heterogeneous expectations in the foreign exchange market, 2011 358. THORWARTH Susanne (18.03.2011) Productivity effects of research, development and design activities, 2011 359. DE JAEGER Simon (13.05.2011) Assessing incentive-based environmental policies for reducing municipal solid waste disposal, 2011 360. SLAETS Leen (26.05.2011) Analyzing phase and amplitude variation of functional data, 2011 361. VANDEMOORTELE Thijs (27.05.2011) Political and economic theory of standards, 2011 362. VAN WEERT Koen (10.06.2011) Optimal portfolio selection: the comonotonic approach, 2011 363. ROGGE Nicky (14.06.2011) Evaluating teaching and research performances: a non-parametric benefitof-the-doubt approach, 2011 364. VANDEPLAS Anneleen (16.06.2011) The efficiency and equity effects of emerging agricultural value chains: four essays on institutions and development, 2011 365. WANG Lihong (28.06.2011) Value creation through financing and governance: a comparison of listed SOEs and private companies in China, 2011 366. THARMARATNAM KUKATHARMINI (04.07.2011) Robust estimation and model selection in semiparametric regression models, 2011

Essays On The Economics of Media Platforms

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Essays On The Economics of Media Platforms

Загружено:

Авторское право:

Доступные форматы

FACULTEIT ECONOMIE EN BEDRIJFSWETENSCHAPPEN

KATHOLIEKE UNIVERSITEIT LEUVEN

ESSAYS ON THE ECONOMICS OF MEDIA PLATFORMS

FACULTEIT ECONOMIE EN BEDRIJFSWETENSCHAPPEN

KATHOLIEKE UNIVERSITEIT LEUVEN

ESSAYS ON THE ECONOMICS OF MEDIA PLATFORMS

Dries De Smet Mannheim, July 2011

5 General Conclusion Acknowledgements About the Author Bibliography Dissertations

The Advertiser is Mentioned Twice. Media Bias in Belgian Newspapers

2.2 Theoretical Model

The Newspapers Incentives for Advertising Bias

2.2 Theoretical Model utility of reader k reads uRk = 0 n 1 b 2 qA wk

Newspapers Therefore, the persuasion per reader is = 1 b + 2 qA (2.3)

2.2 Theoretical Model

Newspapers We summarize these results in the following proposition.

2.2 Theoretical Model

2 (2+2 4+2 ) 1 < 1 2 2 22 1 (2+2 4+2 ) 1 22 2 1 1 > 1+2 2

2.2 Theoretical Model

2.2 Theoretical Model

Proposition 2.2.7 The quantities that maximize welfare are b= 0

2.3 Empirical Evidence

Conclusion of the Model

Empirical Evidence of Advertising Bias in Belgian Newspapers

2.3 Empirical Evidence

Table 2.1: Summary Statistics of Advertising and Coverage

2.3 Empirical Evidence

The results of estimating equation 2.13 are displayed in table 2.2.

Table 2.2: Coverage explained by advertising and dummies

0.00657*** [0.00242] 0.00185*** [0.000702]

Adv X Newsp Newsp X Month Adv X Month 24624 0.628

2.3 Empirical Evidence

2.3 Empirical Evidence

Table 2.3: Robustness Checks

Adv X Newsp Newsp X Month 20976 0.406

Adv X Newsp Newsp X Month 21546 0.368

Adv X Newsp Newsp X Month 24094 0.458

2.3 Empirical Evidence

Table 2.4: Granger Causality Test

Advertisingt4 N R2 Joint signicance (F) Lagged Ads (Prob > F)

Table 2.5: Coverage explained by advertising of last 12 months

Total Ads (lag year)

2.4 Survey of Belgian Journalists

Survey of Belgian Journalists

2.4 Survey of Belgian Journalists

Age 20-30 31-40 41-50 50+

Table 2.6: Summaries Survey Number 91 21 19 23 28 90 26 14 50 63 26 21 16 89 23 11 25 30

2.4 Survey of Belgian Journalists

2.4 Survey of Belgian Journalists

Figure 2.2: Suggestion to Write about Advertiser

2.4 Survey of Belgian Journalists

Table 2.7: Survey Results per Group

Media Group (N=63) Persgroep (N=26) Corelio (N=21) Concentra (N=16)

2.5 Advertising Bias Channels

Advertising Bias Channels

Supply and demand channels

Appendix A Monopoly prots

(2 1)2 21 1 > 1+2 2 2

Figures & Tables

Figure B1: Total Advertising Expenditure over Time

Figure B2: Total Coverage over Time

Table B1: Coverage explained by logged advertising and dummies

B Figures & Tables

Table B2: Coverage explained by lagged advertising

0.00657* [0.00242] 0.00185* [0.000702]