The Economics of Superstars and Celebrities

Stephan Nüesch The Economics of Superstars and Celebrities

GABLER EDITION WISSENSCHAFT Markt- und Unternehmensentwicklung Herausgegeben von Professor Dr. Dres. h.c. Arnold Picot, Professor Dr. Professor h.c. Dr. h.c. Ralf Reichwald, Professor Dr. Egon Franck und Professorin Dr. Kathrin Möslein

Der Wandel von Institutionen, Technologie und Wettbewerb prägt in vielfältiger Weise Entwicklungen im Spannungsfeld von Markt und Unternehmung. Die Schriftenreihe greift diese Fragen auf und stellt neue Erkenntnisse aus Theorie und Praxis sowie anwendungsorientierte Konzepte und Modelle zur Diskussion.

Stephan Nüesch

The Economics of Superstars and Celebrities

Deutscher Universitäts-Verlag

Bibliografische Information Der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über abrufbar.

Dissertation Universität Zürich, 2007 Die Wirtschaftswissenschaftliche Fakultät der Universität Zürich gestattet hierdurch die Drucklegung der vorliegenden Dissertation, ohne damit zu den darin ausgesprochenen Anschauungen Stellung zu nehmen. Zürich, 7. Februar 2007 Der Dekan: Prof. Dr. H. P. Wehrli

1. Auflage August 2007 Alle Rechte vorbehalten © Deutscher Universitäts-Verlag | GWV Fachverlage GmbH, Wiesbaden 2007 Lektorat: Frauke Schindler / Sabine Schöller Der Deutsche Universitäts-Verlag ist ein Unternehmen von Springer Science+Business Media. www.duv.de Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlags unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Die Wiedergabe von Gebrauchsnamen, Handelsnamen, Warenbezeichnungen usw. in diesem Werk berechtigt auch ohne besondere Kennzeichnung nicht zu der Annahme, dass solche Namen im Sinne der Warenzeichen- und Markenschutz-Gesetzgebung als frei zu betrachten wären und daher von jedermann benutzt werden dürften. Umschlaggestaltung: Regine Zimmer, Dipl.-Designerin, Frankfurt/Main Gedruckt auf säurefreiem und chlorfrei gebleichtem Papier Printed in Germany ISBN 978-3-8350-0849-6

Foreword

V

Foreword In 1981 University of Chicago economist Sherwin Rosen wrote his famous paper “The Economics of Superstars” addressing the question why a small number of performers are paid so much more than the average person in their field – far more than their close competitors. Rosen’s paper provided a careful explanation of the superstar phenomenon focusing on two distinct economic mechanisms: imperfect substitution among different sellers of the service and the substantial scale economies of joint consumption technologies. In 1985 Moshe Adler added an important insight to the superstar discussion. Whereas Rosen’s imperfect substitution among sellers was based on objective talent differentials, Adler pointed to the fact that perceived talent differentials may be a result of past consumption and network externalities among consumers with similar consumption capital. As a consequence, superstars could emerge among performers of equal talent in a bandwagon process triggered by pure luck or media strategies. The thesis of Stephan Nüesch picks up these classical superstar explanations and tries to expand their scope by introducing an additional interaction benefit in the consumption of a star’s performance. Debriefing the performance of a performer in the family, with friends, workmates or acquaintances may generate additional value. In contrast to the consumption benefit, the interaction benefit is not at all dependent on the performer’s talent. A mere “gossip externality” suffices to give rise to the phenomenon of celebrities who are basically known for their well-knowness. Casting shows, docusoaps or reality-based game shows are the common platforms of celebrity creation used by the media industry.

The thesis contains both theoretical and empirical chapters covering the emergence and creation of superstars and celebrities. Stephan Nüesch addresses questions like the following: Which mechanisms drive the emergence of superstars and celebrities? What marginal revenue are superstars able to generate? Why may “manufactured” celebrities be more lucrative for the media than “self-made” superstars?

VI

Foreword

The soccer industry serves as a labor market laboratory in this thesis. Stephan Nüesch confronts the different theories of superstar formation with detailed soccer data.

Without going further into the details here, the book is recommendable both because of its creativity and because it addresses a highly relevant real-world phenomenon. Scholars interested in the application areas and techniques explored in the essays as well as everybody else simply wondering about the superstar and celebrity phenomena in everyday-life will find rich food for thought.

Prof. Dr. Egon Franck

Preface

VII

Preface This thesis combines two years of exciting and enriching research. I have had the privilege to pursue my studies at a very good and stimulating institution, namely the Institution for Strategy and Business Economics, respectively the Chair of Business Management and Policy at the University of Zürich. I have experienced an intensive and very fruitful collaboration with my supervisor, Prof. Dr. Egon Franck. He sharpened my economic thinking and inducted me into the hard work of scientific research. I highly profited from his profound knowledge in the broader field of institutional economics and in superstar economics in particular. In addition, he enabled me to participate in international conferences where I could present my results and exchange ideas with other researchers from all over the world. I would also like to thank Prof. Dr. Helmut Dietl not only for being my co-advisor, but for the good cooperation and the valuable suggestions regarding the models in this thesis as well. Additionally, I want to thank my colleague, Leif Brandes, for the friendly and supportive atmosphere, for the helpful discussions concerning econometrical issues and for the good collaboration referring to the “star attraction”-paper. For the valuable spelling checks, I am grateful to Annemarie Vit-Meister.

Finally, I want to thank my parents who financially supported me during my studies. Special thanks go, of course, to my wife, Hanna, for her deep love and encouragement.

Zürich, January 2007

Stephan Nüesch

Contents

IX

Contents Contents..................................................................................................... IX Tables......................................................................................................XIII Figures ....................................................................................................XIII 1

Introduction ........................................................................................ 1 1.1

2

The Economics of Superstars .........................................................................3

1.2

The Economics of Celebrities ........................................................................9

1.3

Overview of the Thesis.................................................................................13

Talent, Past Consumption and/or Popularity – Are Outstanding German Soccer Players Rosen or Adler Stars?............................. 17 2.1

Introduction ..................................................................................................17

2.2 Theories of Superstar Formation ..................................................................18 2.3

Related Literature .........................................................................................21

2.4

Hypothesis ....................................................................................................23

2.5

Data and Stylized Facts on German Soccer..................................................24

2.6

Empirical Framework ...................................................................................26

2.7

3

2.6.1

Dependent Variables.........................................................................26

2.6.2

Independent Variables ......................................................................27

2.6.3

Results ..............................................................................................33

Conclusion ....................................................................................................38

Local Heroes and Superstars – An Empirical Analysis of Star Attraction in German Soccer .......................................................... 41 3.1

Introduction ..................................................................................................41

3.2

Related Literature .........................................................................................42

3.3

Stylized Facts on German Soccer.................................................................44

3.4

Star Attraction ..............................................................................................47

3.5

3.4.1

Star Performance ..............................................................................47

3.4.2

Star Popularity ..................................................................................48

Econometric Framework ..............................................................................49 3.5.1

Data and Dependent Variable...........................................................49

X

Contents

3.5.2

3.6

3.7

4

5

3.5.3

Estimation Approach ........................................................................52

3.5.4

Results ..............................................................................................54

Robustness Analysis .....................................................................................57 3.6.1

Increasing the Number of Superstars and Local Heroes ..................57

3.6.2

Alternative Measures for Star Performance .....................................61

Conclusion ....................................................................................................63

Superstar Earnings in Soccer – Are Voluntary Salary Cap Agreements Self-Enforcing?............................................................ 65 4.1

Introduction ..................................................................................................65

4.2

Related Literature on Salary Cap Agreements .............................................68

4.3

The Model ....................................................................................................70

4.4

Discussion.....................................................................................................75

4.5

Conclusion ....................................................................................................77

4.6

Appendix 1: Nash Equilibrium in a One-Shot Interaction ...........................78

4.7

Appendix 2: Asymmetric Clubs ...................................................................79

Superstars versus Celebrities – Big Man or Big Name?............... 83 5.1

Introduction ..................................................................................................83

5.2

Superstar Emergence ....................................................................................84

5.3

5.2.1

Economic Superstar Theories...........................................................84

5.2.2

A Simple Model of Superstars .........................................................88

Celebrity Emergence ....................................................................................90 5.3.1

5.4

6

Controls ............................................................................................50

“Gossip Consumption” .....................................................................91

5.3.2

The Role of Media in Celebrity Emergence.....................................92

5.3.3

A Simple Model of Celebrities.........................................................94

Conclusion ....................................................................................................95

Different Star Strategies in the Media – Why “Manufactured” Celebrities are More Lucrative than “Self-Made” Superstars .... 97 6.1

Introduction ..................................................................................................97

6.2

Pop Idol – An Example of “Manufacturing” Celebrities ...........................100

6.3

A Strategy Framework of Star Attraction in the Media .............................107 6.3.1 The Rosen Explanation for the Viewer Drawing Capability of Superstars .......................................................................................108

Contents

6.4

6.5

XI

6.3.2

The Adler Explanation for the Viewer Drawing Capability of Superstars .......................................................................................110

6.3.3

Bargaining Power of Superstars .....................................................111

„Manufactured“ Celebrities........................................................................112 6.4.1

Viewer Drawing Capability of “Manufactured” Celebrities ..........114

6.4.2

Bargaining Power of “Manufactured” Celebrities .........................115

6.4.3

Market Segmentation......................................................................116

Conclusion ..................................................................................................116

7

Summary and Outlook................................................................... 119

8

References........................................................................................ 125

9

Index ................................................................................................ 139

Tables and Figures

XIII

Tables Table 1:

Estimates of earning differences due to publicity differences...................... 11

Table 2:

Variables and descriptive statistics............................................................... 32

Table 3:

Estimates of the logarithm of the players’ market values ............................ 35

Table 4:

Pearson correlation coefficients.................................................................... 37

Table 5:

Comparison of Bayern Munich and Hansa Rostock in the 2003/04 season. 45

Table 6:

Variables and descriptive statistics............................................................... 52

Table 7:

Estimates of a team’s star attraction (2% superstar definition) .................... 56

Table 8:

Estimates of a team’s star attraction (5% superstar definition) .................... 58

Table 9:

Estimates of a team’s star attraction (8% superstar definition) .................... 59

Table 10: Joint significance tests .................................................................................. 60 Table 11: Estimates of a team’s star attraction using BEST11..................................... 62

Figures Figure 1: Theories of the demand for superstar services ............................................. 20 Figure 2: Density allocation of the logarithm of market values .................................. 25 Figure 3: Profits with and without honoring the salary cap agreement if D ! 0.75 .... 74 Figure 4: Articles mentioning the American Idol finalists of season 1...................... 105 Figure 5: Articles mentioning the American Idol finalists of season 2...................... 106 Figure 6: Articles mentioning the American Idol finalists of season 3...................... 107

Introduction

1

1 Introduction

Recently, the income inequality of most western societies is increasing.1 “The phenomenon of Superstars, wherein relatively small numbers of people earn enormous amounts of money and dominate the activities in which they engage, seems to be increasingly important in the modern world” (Rosen, 1981, p. 845).2 The starting sentence of Sherwin Rosen’s seminal paper “The Economics of Superstars” applies more than ever. Technological change has enlarged the scope and the intensity of so-called winner-takes-all markets3 during the last decades.4 Rosen (1981) explicitly mentions the worlds of sports, arts and letters or the show business5 as examples of markets dominated by a few superstars. Radio, television, reproduction equipment, and advancements in communication technology have increased the scope of a performer’s audience and expanded their markets. Nowadays superstar performances may be broadcasted and consumed world-wide. Media 1

The top 1% income share in the US rose from 7.9% in 1976 to 14.6% in 1998. The income share of the top 0.01% richest people quintupled from 0.56% in 1976 to 2.57% in 1998 (Piketty & Saez, 2003). Growing inequality is not confined to the US. For example, the richest 20% in the United Kingdom earned seven times as much as the poorest 20% in 1991, compared with only four times as much in 1977 (Frank & Cook, 1995). In China the top 10% income share increased from 18.7% to 27.2% and the top 1% income share more than doubled from 2.8% in 1986 to 6.0% in 2003 (Piketty & Quian, 2006).

2

The accentuation is introduced by the author.

3

Frank and Cook (1995) defined winner-takes-all markets as markets in which a few suppliers capture the bulk of revenues.

4

The recent and widespread rise in income inequality is mostly explained by technological improvement and/or globalization effects (for a literature review see e.g. Burtless (1995), Brauer and Hickok (1995), or Johnson (1997)). In this thesis, technological change is considered as the dominant driving force of rising inequality, which also influences the globalization.

5

Alain Kruger’s analysis of superstar effects in the market for rock concerts indicate that “the top 5% of revenue generators took in 62% of concert revenue in 1982 and 84% in 2003” (Krueger, 2005, p. 15).

2

Introduction

technology makes it possible for large parts of the world to be served by one person, who is paid accordingly (Borghans & Groot, 1998). Why listen to the second-best pop star or a mediocre local singer, if the best is available? As the demand for home entertainment is increasing, it is likely that superstar earnings from records, television videos and film continue to rise, which will cause even greater skewness in the future earnings distribution (Towse, 1997).

The economic literature basically agrees that superstars serve a large market share because they provide services of perceived superior quality (Rosen, 1981; Adler, 1985; MacDonald, 1988; Borghans & Groot, 1998). This thesis tries to expand the rather narrow, talent and quality driven consideration of superstars by introducing an interaction benefit in consuming a star performance. Thus, the benefit from watching a superstar performance is only one part of the total benefit associated with consumption. Debriefing the performance in the family, with friends, workmates or acquaintances generates additional value for those involved. Unlike the direct consumption benefit, the interaction benefit is no longer dependent on the star’s talent. This fact paves the way for celebrities who are defined by their fame and publicity rather than by any inherent talent. Celebrities are well-suited for gossip because they are known by nearly everyone, and rumors and stories about celebrities are easy to find and share. The interaction benefit of gossip offers a plausible explanation why we have experienced a boom of very famous but rather trivial celebrities through reality television; confessional talk formats, casting contests, docu-soaps or reality-based game shows.

This thesis deals with questions like: What characteristics drive the emergence of superstars and celebrities? What marginal revenue product are superstars able to generate? Are voluntary salary cap agreements an adequate policy to curb inefficiently high superstar earnings? What are the differences between superstars and celebrities? Why are “manufactured” celebrities more lucrative for the media than “self-made” superstars? The present thesis contains both empirical and theoretical chapters. In the sections 2 and 3, I empirically confront two different superstar theories in the field of sports. The sports business is an ideal labor market laboratory,

Introduction

3

since it offers a unique opportunity for labor market research in general and for empirical testing of specific theories in particular: “There is no research setting other than sports where we know the name, face, and life history of every production worker and supervisor in the industry. Total compensation packages and performance statistics for each individual are widely available, (…)” (Kahn, 2000, p. 75). In chapter 5, I introduce the concept of celebrities and explain its differences compared to superstars. Both concepts are applied to the media industry in section 6. In the following, I will give a short introduction to the economics of superstars and celebrities and a brief overview of this thesis.

1.1 The Economics of Superstars Alfred Marshall (1947) had already pointed out that innovations in technology and mass production would lower the per unit price of quality goods and ultimately allow higher quality goods to obtain a greater market share. In 1981 Sherwin Rosen named this effect the “superstar phenomenon”. Extraordinary salaries earned by superstars are driven by a market equilibrium which rewards talented people with increasing returns to ability. The key to the high earnings of superstars lies in the vast extent of the audience they are able to reach, afforded by the existence of scale economies. Superstars arise in markets in which the production technology allows for jointconsumption. For example, if one person watches a tennis game on television, it does not diminish someone else’s opportunity of watching it as well. In superstar industries production costs do not rise in proportion to the size of the seller’s market. Since most of the costs are up-front, average costs decrease with consumed output. Such economies of scale enable a few or just one supplier to serve the whole market. And since non-paying customers may still be excluded, there are no issues of free riding. Scale economies explain why a sports or music superstar earns a multiple of a school teacher, even if he or she is the best teacher in town. The teacher’s income is restricted by the small number of students who can be taught in a classroom. However, Rosen and Sanderson (2001) suggest that it is all in the technology. If teachers use the Internet to personally teach millions of students at one time, star teachers are possibly able to earn at least as much as star athletes. But large

Introduction

4

economies of scale do not guarantee high salaries for a small number of stars unless the market demand becomes highly concentrated on their services. In the superstar literature the demand for superstar services is basically driven by two distinct but not mutually exclusive factors: superior talent and network externalities.

Firstly, the market demand may be concentrated on superstars because they just have superior talent. Rosen (1981) argues that people prefer consuming fewer high-quality services rather than more of the same service at moderate quality levels: “(…) hearing a succession of mediocre singers does not add up to a single outstanding performance” (Rosen, 1981, p. 846). Thus, poorer quality is only an imperfect substitute for higher quality. If a surgeon is 10% more successful in saving lives than her peers, most people would be willing to pay more than a 10% premium for her services. This imperfect substitutability applies in particular to status goods or gifts: To celebrate a special occasion, people do not search for an average restaurant meal or bottle of wine but for the best. We give two ounces of Russian caviar, not forty pounds of frozen whitefish costing the same (Frank & Cook, 1995).6

Network externalities offer a second explanation why the demand may be highly focused on the services of a few superstars. Network externalities induce feedback effects – processes in which success breeds success.7 In contrast to the typical standardization literature (e.g. Katz & Shapiro, 1985 or Farrell & Saloner, 1985) network externalities of superstars are by no means confined to issues of technological compatibility or a larger variability of complements. Moshe Adler (1985) rather suggests a cognitive and social form of network externalities. According to Adler (1985) the phenomenon of superstars is the consequence of a learning process. He believes that the marginal utility from consuming a superstar service

6

In a comment on Rosen’s superstar article Bowbrick (1983) criticizes that a star’s talent is not always unambiguously assessable. Therefore, evaluation processes must be taken into account. Rosen (1983) replied that different tastes diminish the extent of the seller’s market. However, the basic mechanisms for superstar formation remain unchanged according to Sherwin Rosen.

7

The sociologist Robert Merton (1968) calls this phenomenon the “Matthew effect”, after the scripture in the book of Matthew: “For unto everyone that hath shall be given, and he shall have abundantly, but from him that hath not shall be taken away even that which he hath.”

Introduction

5

increases with the ability to appreciate it, which depends not only on the star’s talent but also on the amount of star specific knowledge the consumer owns. This specific knowledge – called consumption capital – is accumulated by past consumption activities or by discussing the star’s performance with other likewise knowledgeable individuals. The latter effect gives rise to positive network externalities. The more popular the artist is, the easier it gets to find other fans. Searching cost economies imply that consumers are better off patronizing the most popular star as long as others are not perceived as clearly superior. “It is plausible to assume that the cost of searching for knowledgeable discussants is minimized if one chooses the most popular artist. (…) To reemphasize, the star need not possess greater talent. Stardom is a market device to economize on learning costs in activities where “the more you know the more you enjoy.” Thus stardom may be independent of the existence of a hierarchy of talent” (Adler, 1985, p. 208-209). According to Adler (1985), luck (by luck, he means factors other than talent) determines who amongst equally talented artist will snowball into a star. Superstars may emerge because initially (slightly) more people happen to know one artist than any other artists of possibly equal talent.

Even though the economic superstar literature is currently more than 20 years old, the basic principles formulated by Rosen (1981) and Adler (1985) have remained omnipresent. Nevertheless, other researchers have still made notable contributions to the superstar literature, which will be summarized below.

In 1982 Sherwin Rosen expanded his basic superstar theory to managerial reward distributions across ranks in and among hierarchical firms. Starting point is the fact that earnings of top executive officers of large firms are enormous in magnitude and positively correlated with firm size. Rosen (1982) argues that assigning persons of superior talent to top positions increases productivity by more than the increments of their abilities because greater talent filters through the entire corporation by a recursive chain of command technology. Chief executives exercise a great deal of economic power. A small difference in the quality of a decision of a CEO can

Introduction

6

translate into an enormous value difference. Decisions in the top position of a large hierarchy influence the productivity of a great number of subordinates. “The most capable foot soldier is not very effective if he is fighting the wrong war” (p. 312). By subordinating authority through many levels in a deeper organization, the diseconomy of direct supervision is relaxed. “Subordination through a chain of command economizes on the limited time of more talented individuals, who imperfectly “clone” themselves by transferring part of their talent to their immediate subordinates, who, in turn, transfer it to their subordinates and so on down the line” (p. 313). It is clear that under these circumstances it pays to assign the most talented persons to positions of greatest power and influence. According to Rosen (1982) multiplicative productivity interactions in firms cause the rewards to be more skewed than the underlying distribution of talent.

In 1988 Glenn MacDonald provided a dynamic version of Rosen’s superstar model. MacDonald (1988) describes how young artists, whose uncertainty of talent is high, firstly perform to small audiences and earn net returns below what they could earn outside the industry. The audience feedback they receive provides useful information about their probable future performance and likelihood of becoming a superstar. Since the quality of their performances is serially correlated, the knowledge of firstperiod reviews reduces the uncertainty of quality. These reviews have predictive power for the second period’s performance. Those performers who have been recipients of good reviews stay in the industry, earn larger incomes and play to bigger crowds than before. The audience pays a high price for the assurance that they are less likely to be dissatisfied. The less fortunate young performers leave the industry. MacDonald (1988) postulates, therefore, earnings to be an increasing convex function of a performer’s talent. However, in contrast to Rosen (1981) he considers this function to have rather stochastic than deterministic properties.

Michael Kremer’s O-ring theory offers an additional explanation why wages and output may steeply rise with the skill of a worker. The O-ring theory applies to

Introduction

7

production technologies with a lot of complementary tasks, in which all tasks must be successfully completed for the product to have full value. Prime examples of such a multiplicative production function are space flights. In 1986 the space shuttle Challenger exploded due to malfunctions of just one of many thousands components, the O-rings. For example, firms may fail as a result of bad marketing, even if the product design, manufacturing, and accounting are excellent. As a result of the multiplicative production function, it is not possible to substitute several low-talented workers for one high-skill worker.8 Since workers of similar skill will be matched together in equilibrium, small differences in worker skill lead to large differences in output and wages. Firms are willing to expend resources screening a number of applicants for a single position in order to find an employee of the right skill level rather than picking just someone and paying the estimated marginal revenue product. Thus, wage and productivity differentials between work groups, firms, regions or countries with different skill levels tend to be enormous (Kremer, 1993).

The book “The Winner-take-all Society” of Robert Frank and Philip Cook in 1995 made a rather popular contribution to the analysis of markets in which rewards are highly concentrated among a few individuals. They call them “winner-takes-all markets”. According to Frank and Cook (1995) winner-takes-all payoff structures do not only apply to well-known examples of professional sports, pop culture or arts but affect a wide variety of economic activities like for example education. Frank and Cook (1995) name several driving forces of winner-takes-all markets: production cloning, network economies and other self-reinforcing processes, decision leverage, habit formation or acquired taste, psychological rationales of gifts and avoidance of regret, purely positional concerns or concentrated purchasing power; however, unfortunately without giving serious theoretical or empirical evidence. Frank and Cook (1995) argue that entry and efforts spent competing for the large rewards in winner-takes-all markets are socially inefficient: Since people tend to be overconfident about their own abilities and fail to consider the negative externality of reducing the rivals’ chances of winning, too many people enter the winner-takes-all

8

Skill refers to the probability that a worker will successfully complete a task.

Introduction

8

markets and invest too much for winning the tournament. Too many non-productive resources are devoted to the contest itself, for example, by rent-seeking behavior. Therefore, the authors advocate highly progressive general income taxation or active collective restrictions to curb wasteful competition. Frank and Cook (1995) believe that there is no equity-efficiency trade-off, because reducing the rewards of winnertakes-all markets decrease the number of contestants, which may only cause a small decline in the quality of the superstar service, but increases the society’s total income.9

Borghans & Groot (1998) argue that superstars emerge in the combination of two conditions: Firstly, superstars need to be more talented than others. Secondly, there must be a certain degree of monopolistic power which emerges due to the numberone position of superstars. In contrast to the neoclassical model, Borghans and Groot (1998) state that superstars earn more than their marginal product based on talent differences alone. It is rather monopolistic power as a result of being the best that explains the enormous salaries of superstars. “The extra reward is not due to the way in which the superstar performs, but to the fact that people are not satisfied to watch other players once the superstar has been identified” (Borghans & Groot, 1998, p. 555). Superstars receive a de facto “property right” on the number-one position that is scarce and hence constitutes market power. However, there is a competition for the lucrative number-one position beforehand, leading to inefficiently high investments levels. High superstar incomes can survive only if some people have higher chances of winning the competition. If everybody had the same talent and, therefore, the same initial probability of winning, all rents connected to the superstar position would be absorbed by investment costs to reach superstar position. Borghans and Groot (1998, p. 570) conclude: “Thus the phenomenon of superstar incomes can be explained by the temporary monopolistic power of the best in the activities in which they excel.”

9

Even though Rosen (1986) acknowledges that high top prices for the winners lead to socially useless signalling actions, he disagrees with Frank and Cook (1995) about the equity-efficiency trade-off. Investigating the incentive properties of prizes in sequential elimination events, Rosen (1986) argues that an extra reward for the overall winner is necessary to maintain performance incentives throughout the game. “Extra weight on top-ranking prizes is required to induce competitors to aspire to higher goals independent of past achievements” (Rosen, 1986, p. 713).

Introduction

9

Egon Franck (2001) attributes the enormous earnings of superstars not only to scale economies or network externalities but also to the suitability of superstars as quality monitors. Referring to markets of experience goods like the movie sector, he argues that star actors have both the power and the incentives to assure excellence by monitoring the film production. Star actors mostly play a leading part in the plot and may have voice concerning the script and the allocation of the roles. And since their (future) reputation is strongly influenced by the success or failure of the film, star actors have strong incentives to pick only promising scenarios and to make an effort in guaranteeing high quality throughout the recording of the film. Therefore, star actors serve as a credible signal of high quality to the consumers who cannot ex ante identify the quality of the film. The studios pay the star actors accordingly.

The existence of superstars who are defined by their enormous income is based on the provision of services with perceived superior quality. Even though the precise talent or quality is often difficult to identify and to measure, all star theories basically agree that superstars have a somehow exceptional talent.

1.2 The Economics of Celebrities The most widely quoted definition of celebrity was given by Daniel Boorstin (1961, p. 57): “The celebrity is a person who is known for his well-knowness.” According to Boorstin (1961) appearances of celebrities are pseudo-events; they seem to be meaningful but are in fact insubstantial. The assumption that “(…) no celebrity possesses any meaning of consequence” (Marshall, 1997, p. 11) is a heroic simplification, of course. In reality, the boundary between talented superstars and trivial celebrities is more blurred. Most celebrities may also have a moderate level of talent and superstars have often profited from publicity platforms too. But the fact that the well-knowness of celebrities has become a viable commodity all by itself is intrinsic to their meaning. Fame may stand independent of accomplishment, heroics, or talent (Rein, Kotler, Hamlin, & Stoller, 2006). Celebrities may enjoy enormous

Introduction

10

fame and popularity without necessarily having any special talent or exhibit any particular achievement other than the attraction of public interest.10

Celebrity is a subject mostly addressed in marketing or consumer research (see e.g. McCracken, 1989; Agrawal & Kamakura, 1995; Mathur, Mathur & Rangan, 1997; Erdogan, 1999). In cultural, social, psychological or media studies celebrity is a widely discussed issue too (see e.g. Boorstin, 1961; Gamson, 1994; Marshall, 1997; Holmes, 2004a&b; Turner, 2004; Turner, 2006; Young & Pinsky, 2006). However, I did not find any clear economic theory of celebrity formation. This thesis tries to fill the gap.

The value created by celebrities primarily consists of the interaction benefit rather than the consumption benefit. Celebrities generate gossip opportunities. The pleasure of gossip lies in the exchange of interpretations, evaluations, and judgments, or in the development of new story lines. Gossip is a way of sharing social judgments and of processing social behavior. Unlike acquaintance gossip, there is no danger of repercussions or accountability in celebrity gossip (Gamson, 1994). The interest in the details of celebrities’ lives is based, among other things, on bringing them down to the level of ordinary human beings and on the imagination of them as part of the extended family (Turner, 2004). Gossip neither relies on any extraordinary talent nor on any special achievement of the celebrity; pure fame suffices. The collective aspect of gossip gives rise to snowball effects. The more popular a celebrity gets, the easier gossip circulation becomes. This creates a self-energizing bandwagon effect11: a famous celebrity becomes more and more valuable to each viewer as he or she attracts ever more viewers. Once a rising celebrity shows plausible signs of ascent, a lot of fans will jump on the bandwagon and the performer might achieve stardom

10

Pop diva Kylie Minogue once said: “Fame used to be a by-product. Now it’s like “What do you want to be when you grow up?” “Famous.” “What for?” “It doesn’t matter.”” (cited from Turner, 2004, p. 52).

11

Leibenstein (1950) named the observation that people often follow the crowd as „bandwagon effect“. The bandwagon effect emerges if people’s valuations of a commodity increase when they observe others consuming the same commodity.

Introduction

11

virtually overnight. However, fame bubbles can burst as quickly as they formed (Cowen, 2000). The popularity and fame of celebrities is often fleeting.

I clearly distinguish between superstars and celebrities for two reasons: Firstly, celebrities enjoy enormous stardom without automatically earning enormous salaries. Secondly, celebrity status is artificially producible by media publicity, whereas superstars are “self-made” as their status is based upon an inherent exceptional characteristic of the person.

My first assumption is the result of a small study in which I analyzed the influence of publicity and fame on earnings concerning the Forbes top 100 celebrities. The webpage www.forbes.com annually lists the top 100 celebrities and delivers not only earnings estimates (y) but also web mentions on Google (WEB), press clips compiled by LexisNexis (PRESS), TV/radio mentions compiled by Factiva (TV), or appearances on the cover of any 17 major consumer magazines (COVER). Since unobservable individual effects like e.g. the celebrity’s charisma may influence both earnings and publicity measures, estimates possibly suffer under omitted variable bias. Therefore, I used a two period12 first-differencing equation. In doing so the unobserved individual effect is incorporated in the fixed effect that drops out in the following first-differencing equation: 'yi

E 'WEB  E 'PRESS  E 'TV  E 'COVER  H i

1

2

i

i

3

i

4

i

The estimates of the differences in earnings are listed in Table 1: Variable

ȕ-coef.

Std.Error

WEB PRESS TV COVER

-0.59 -362.72 16'965.73 32'494.86

3.64 580.88 17'811.89 944'107.60

Adjusted R

2

0.11

Overall significance Number of observations

Table 1: 12

F-value: 0.53

(p-value: 0.72)

59

Estimates of earning differences due to publicity differences

The collected data corresponds to the years 2004 and 2005.

(1)

Introduction

12

Using a first-differencing equation, none of the four publicity variables have a significant influence on earnings. The independent variables are even jointly insignificant. Since we fail to reject the null hypothesis, there is no evidence that any of the publicity variables help to explain the earnings of celebrities. In other words: An increase or decline of earnings is not systematically linked to high profile.13 Therefore, celebrities defined as very famous people are not automatically rich superstars too. Of course, most of the celebrities are able to translate their popularity into financial prosperity. Other celebrities, however, may be extremely popular without making big money.

Secondly, celebrity status is artificially producible by media publicity, whereas superstars are always “self-made” to some extent. In this thesis celebrities are considered as media commodities which are consciously created and marketed.14 Celebrity status is seen as a direct consequence of media coverage which provides large public attention. Probably the best examples of such “manufactured” celebrities are participants of reality television shows. The winners of television casting contests like Big Brother or Pop Idol enjoy enormous fame and publicity for a short time, but mostly fade into obscurity once the program went off the air.

Since celebrities have high viewer drawing potential, the “manufacturing” and the “trading” of celebrities has become a commercial strategy for all kind of media corporations. The importance and relevance of celebrities in everyday life has grown as the mass media, in particular the visual media, disseminated. Recent information technology has even further enhanced the capacity to deliver images of celebrities in real time around the globe. Satellite and digital television, computer technology and 13

Unfortunately, we do not have a random sample; instead journalists of the Forbes webpage consciously select the top 100 celebrities. Whenever random sampling is absent, the issue of selection bias may cause biased estimators. But by choosing first-differencing, the issue of selection bias is moderated since fixed-effects analysis allows for the reason of selection to be correlated with the unobserved effect (Wooldridge, 2002, p. 469).

14

Of course, this is not the only approach to the phenomenon of celebrities. According to Rojek (2001) celebrity status is not only attributed by the media, but also ascribed through blood relations (e.g. royals) or achieved in open competitions (e.g. sport stars). Turner (2004) argues that celebrities generate para-social interactions which operate as a means of compensating for changes in the social construction of the communities we live in.

Introduction

13

the Internet extended the opportunities to create, transmit and market celebrity status (Smart, 2005).

1.3 Overview of the Thesis The thesis deals with both empirical and theoretical issues of superstars and celebrities. Chapter 2 and 3 empirically test two competing superstar theories in German soccer. Section 4 provides a theoretical model about the usefulness of voluntary salary cap agreements to reduce potential Pareto inefficiencies in superstar markets. In chapter 5 the theory of superstars is confronted with the less established concept of celebrities. Section 6 applies both concepts to the media industry, and chapter 7 summarizes the main findings of the thesis and gives an outlook. Chapters 2 to 6 are all self-contained, which means that each section starts with a new introduction, discussion and conclusion and does not require any previous knowledge. The different chapters are not directly interconnected. This modular approach allows the reader to just pick out some sections and to understand them without having to read the whole thesis. However, if you read the whole thesis through, do not be surprised at recurrences that may occur.

Chapter 2 empirically analyzes the competing theories of superstar formation proposed by Rosen (1981) and Adler (1985) in German football. Using data on market values and individual player performance and publicity data, I differentiate between Rosen and Alder stars and examine if firsthand observable talent or other factors like the past consumption of consumers or the player’s popularity influence the emergence of superstars. Running quantile regression I find that Adler’s superstar theory better explains the emergence of superstars in European football. Thus, not only investments in physical talent but also the cultivation of popularity may be an adequate strategy in order to become a superstar.

Chapter 3 deals with the marginal revenue product of superstars. How do superstars influence their teams’ match attendances? Since match attendance is largely driven by geographical proximity, I distinguish between superstars and local heroes. Whereas

14

Introduction

superstars belong to the 2% most valuable players of the league, a local hero is just the most expensive player within a team that has no superstars. While Rosen (1981) suggests that star players attract audiences by an extraordinary individual performance which translates into a high winning percentage of his team, Adler (1985) postulates that the stars may draw attention due to their popularity and charisma beyond actual playing ability. Chapter 3 again empirically reviews the competing superstar theories of Rosen (1981) and Adler (1985), however, not by analyzing individual market values but rather by examining star attraction on consumers. I find that superstars facilitate fan support mainly because of their outstanding field performances, whereas local heroes enhance home game attendance by mere popularity.

Chapter 4 deals with the issue that positional externalities may lead to inefficiently high superstar earnings in sports. Since the financial rewards of soccer teams for example are largely determined by relative league rank, clubs try to increase winning probability by hiring the most talented superstars. However, any action that increases one contestant’s chances of winning must necessarily reduce the chances of others. Clubs tend to overinvest in playing talent, because they fail to consider the negative externality of reducing its rivals’ revenues by winning. In almost any European football league annual growth rates of player salaries exceeded annual growth rates of revenues during the last decade. In 2002 the leading European football clubs reacted to the increasing player salaries by signing a voluntary agreement to limit player salaries to 70% of revenues. I analyze under which condition such a voluntary salary cap is self-enforcing. Using a simple model with two identical profit-maximizing clubs I derive that the voluntary salary cap is very unlikely to be self-enforcing in European football.

The subsequent section 5 is a theoretical chapter about the differences between superstars and celebrities. While the theory of superstars is widely discussed in the economic literature since 1981, the theory of celebrities is rather blurred and shared among many disciplines namely sociology and psychology. Chapter 5 attempts to give a rather systematic economic theory of media celebrities who are just people

Introduction

15

known for being well known. I argue that the demand for celebrities is based on the human desire to gossip; namely the desire to discuss, to share interpretations or judgments. The more popular a celebrity is, the easier gossip circulation becomes, which then fuels further popularity and creates a self-energizing bandwagon effect. Media plays a crucial role in selecting for whom it triggers this bandwagon effect.

The sixth chapter applies the concept of superstars and celebrities to the media sector. Media corporations generally enjoy increasing returns-to-scale if more customers watch a program. In this context media firms use stars as attention-getting individuals in order to increase audiences. Media corporations may choose between the audience attracting capability of highly talented – and therefore “self-made” – superstars and “manufactured” and thus rather trivial celebrities. Illustrating the casting show Pop Idol and comparing the abilities of superstars and celebrities to generate and to capture value, I show that “manufacturing” celebrities is a very lucrative business. “Self-made” superstars do not only attract large audiences based on the perceived excellence of their services, they also capture the bulk of the profits as a result of their bargaining power. Celebrities, however, are interchangeable and have thus low market power to obtain value. But the market potential of “manufactured” celebrities is limited because they typically prevail only in “talent free” entertainment.

Finally, chapter 7 summarizes the results of the thesis and mentions present limitations and future research topics regarding the economics of superstars and celebrities.

Are Outstanding German Soccer Players Rosen or Adler Stars?

17

2 Talent, Past Consumption and/or Popularity – Are Outstanding German Soccer Players Rosen or Adler Stars?

2.1 Introduction While clubs overbid each other and pay enormous transfer fees and salaries for socalled superstars, other players receive comparably low remuneration. But what makes a soccer player a superstar? In the literature there are basically two competing theories of superstar formation proposed by Rosen (1981) and Adler (1985).1 Whereas Rosen (1981) stresses clearly observable talent superiority in order to explain the emergence of superstars, Adler (1985) maintains that besides talent, also past consumption and popularity influence stardom. The question to be addressed in this chapter is: Are outstanding soccer players Rosen or Adler stars? Using data on individual market values and a set of personal characteristics of all soccer players appearing in the first German league in the 2004/05 season for more than half an hour, I differentiate between Rosen’s and Adler’s theory of superstar formation. Running quantile regressions I find empirical evidence that variables associated to Adler’s theory contribute to the explanation of market value differentials in German soccer. Thus, not only investments in physical talent but also the cultivation of popularity is an adequate strategy for becoming a superstar.

1

MacDonald’s theory of superstar formation is not treated separately in this section, since he basically presents a dynamic version of Rosen’s superstar model (MacDonald, 1988).

18

Are Outstanding German Soccer Players Rosen or Adler Stars?

The remainder of the chapter is organized as follows. Section 2.2 illustrates the two alternative theories of superstar formation. Section 2.3 presents the related literature. In section 2.4 the hypothesis is motivated. Subsequently, I explain the main features of the data and give some stylized facts on German soccer. The variables and the method used as well as the results are presented in section 2.6. Section 2.7 concludes.

2.2 Theories of Superstar Formation Theories of superstar formation agree that superstars emerge in the provision of certain services where large economies of scale on the supply side are combined with high appreciation on the demand side.

The technology of soccer games facilitates the reproduction of the service at low cost. The cost of production is largely independent of the size of the audience (Lucifora & Simmons, 2003). Since most of the costs are up-front, average costs decrease with consumed output. Large soccer stadiums and various media allow many paying spectators to observe a soccer game simultaneously, while at the same time enabling teams to exclude non-paying customers. Thus, there are no issues of free riding due to non-exclusion. The World Cup, the European Championship or even just a game of the German Bundesliga can attract a remarkably large audience all over the world by television broadcast. As a result of these large economies of scale, only few sellers are needed to serve the whole market. However, large economies of scale do not guarantee high salaries for a restricted number of players. In addition, these players have to be perceived as very scarce so that demand becomes highly concentrated on their services (Rosen & Sanderson, 2001).

While on the supply side both Rosen (1981) and Adler (1985) agree on the necessity of large economies of scale, their explanation of the demand for superstar services is different. Rosen (1981) considers a performer’s talent as costlessly observable to all potential consumers. Since lower talent is an imperfect substitute for higher talent, the artist or sportsman who has slightly higher talent than his competitors may attract the whole market demand under ceteris paribus conditions.

Are Outstanding German Soccer Players Rosen or Adler Stars?

19

Adler (1985) explains the phenomenon of superstars as a learning process that occurs if consumption requires knowledge. A performer’s talent is rather considered as a hidden characteristic than as a clear feature. Based on the notion of “consumption capital” introduced by Stigler and Becker (1977), Adler (1985) argues that appreciation increases with knowledge: “… the more you know the more you enjoy” (Adler, 1985, p. 208-209). Stigler and Becker (1977) use good music as an example of how past consumption activities lead to beneficial addiction through an accumulation of consumption capital. By having exposed themselves to good music in the past, consumers have built up consumption capital that enables them to derive more pleasure from listening to good music in the present. Adler (1985) extends this well-known Stigler/Becker-framework by adding the element of discussing consumption with likewise knowledgeable individuals. A person interested in soccer may increase player specific knowledge by both watching games (Stigler/Beckereffect) and discussing it with other people who know about it (Adler-effect). The more popular the sportsman in question is, the lower the searching costs to find competent discussants will consequently be. These positive network externalities explain why stars may even emerge among equally talented performers. Searching cost economies imply that one is always better off patronizing a well-known star as long as other sportsmen are not perceived as superior by an order of magnitude. Given that consumers face certain budget constraints, the more popularity a specific player already enjoys, the more player specific consumption capital will be accumulated. In Adler’s theory the demand for superstar services depends both on hidden talent characteristics and on consumption capital which itself is affected by both past consumption (Stigler/Becker-effect) and the player’s popularity (Adlereffect). Hence, a potential advantage in knowledge about the talent of a non-star would have to be balanced against the higher searching costs for discussants if one were to abandon the already popular star.

20

Are Outstanding German Soccer Players Rosen or Adler Stars?

Talent costlessly observable talent

Roseneffect

Past consumption

Popularity

hidden characteristics

+ +

Stigler/ Beckereffect

+

+

Adlereffect

Consumption capital

+ Demand for superstar services

Figure 1:

Theories of the demand for superstar services

According to Adler (1985), luck (by luck, he means factors other than talent) determines who amongst equally talented people will snowball into a star. Stars may be born because initially (slightly) more people happen to know one player than any other players of possibly equal talent. However, more than twenty years later, Adler (2006) dismisses the idea of luck as the only possible mechanism driving the initial selection among equally talented people. Just as the suppliers in other businesses prone to superstar effects, sportsmen too do not usually entrust this choice to pure chance. Instead, they consciously use publicity, such as appearances on talk shows and coverage in tabloids, magazines and the Internet to strengthen their popularity. Adler (2006) emphasizes that the acquisition of consumption capital occurs not only by exposure to the activity itself, or through discussing it with friends or acquaintances, but also by reading about it in newspapers, magazines and the Internet.

Are Outstanding German Soccer Players Rosen or Adler Stars?

21

2.3 Related Literature The theories of superstar formation have their origin in the field of arts2, which was also the subject of various empirical investigations of superstar effects (e.g. Hamlen, 1991; Hamlen, 1994; Chung & Cox, 1994). Schulze (2003), however, mentions that in sports the empirical analysis of the superstar phenomenon is even more promising, because in most sports talent is easier to measure than in art or entertainment activities.3 Hausman and Leonard (1997) were the first to empirically analyze superstar effects in professional sports.4 They found out that the mere presence of stars had a substantial positive impact on club revenues, even after controlling for team quality measured by the number of All-Star players in a team. By analyzing all NBA local and national television ratings as well as match attendances, Hausman and Leonard (1997) singled out that – back in 1993 – the estimated value of Michael Jordan for the National Basketball Association (NBA) was $ 53 million. Berri, Schmidt, and Brook (2004) or Berri and Schmidt (2006) extended the work of Hausman and Leonard (1997) by investigating the two-sided relationship between match attendance and both team performance and the team’s employment of star players in the NBA. Their results suggest that it is performance on the court, not star power, which attracts the fans. However, all three studies only cover superstar effects on a team level and not on an individual basis. The question why superstars arise is faded out.

Using longitudinal individual data from two North American team sports leagues – the National Hockey League (NHL) and the National Basketball Association (NBA) – Frick (2001) analyzed the salary differentials between superstars – defined as

2

Rosen (1981) uses examples of full-time comedians or classical music, and Adler (1985) mentions singing and painting as artistic activities generating superstars.

3

Seaman (2003) analyzed the similarities between the arts and the sports literature. He strongly suggests fruitful collaboration and extensive cross-referencing between these two areas of application.

4

Noll (1974), Scott, Long, and Scompii (1985), Brown, Spiro, and Keenan (1991), as well as Burdekin and Idson (1991) already controlled for the effect of star attraction in their analyses of match attendance prior to Hausman and Leonard (1997). However, they did not emphasize the superstar effects. Of these studies only Brown et al. (1991) were able to find a statistically significant relationship between a measure of consumer demand and a team’s star attraction.

22

Are Outstanding German Soccer Players Rosen or Adler Stars?

players who received all-star vocations – and “benchwarmers”. His results show that performance measures like the numbers of scores, rebounds, steals, assists or blocks are good predictors of the observed salary differentials. Frick (2001) found evidence for Rosen’s explanation of superstars. However, a final answer whether Rosen’s or Adler’s theory of star formation applies is still open. His empirical investigation does not differentiate between these two standpoints, since variables measuring the Adlereffect are missing in the set of independent variables.

Lucifora und Simmons (2003) investigated wage determination looking for superstar effects among professional soccer players appearing in the Italian league. The authors used rare data on individual salaries as dependent variable and individual performance indicators, experience, reputation and team quality as regressors. They found empirical evidence for Rosen’s theory. Talent – measured by goals and assists – exercises significant influence on the skewness of the salary distribution of Italian forwards and midfield players. Lucifora and Simmons (2003) do not control for popularity.

Lehmann and Schulze (2005) tested the competing predictions of existing superstar theories in German soccer. Using various measures for individual player’s performance and an indicator for media presence they find that neither performance nor publicity can explain the salaries of superstars. This study extends the paper of Lehmann and Schulze (2005) in several ways: Firstly, I divide a player’s performance into firsthand observable talent measures which are identifiable without costs and indirect quality measures capturing hidden talent characteristics. Secondly, indicators for past consumption and three different popularity measures that specify media presence in more than 20 German newspapers and magazines as well as publicity in the Internet are included. Thirdly, I use market values as endogenous variable since they are a proxy for the total value generated by a player. In this sense they can be interpreted as incorporating salaries, signing fees, bonuses, potential transfer fees and a remaining producer surplus. And last but not least, the analysis of a unique data set delivers new results. I am able to find empirical evidence for Adler’s superstar theory in professional team sports.

Are Outstanding German Soccer Players Rosen or Adler Stars?

23

2.4 Hypothesis Both Rosen (1981) and Adler (1985) believe that talent is an important determinant of stardom. However, while in Rosen’s sense superstars necessarily have superior talent, Adler (1985) delivers an explanation which allows for the superstar phenomenon to arise even among equally talented people. Rosen (1981) treats talent as observable without cost by all economic agents, while Adler (1985) makes clear that superstars only exist if the consumption of their services requires knowledge. According to Adler (1985), a player’s talent is a hidden characteristic that has to be discovered through personal and interpersonal learning processes. The appreciation of a particular player grows with the knowledge consumers have acquired about him. The assumption of observable talent marks a key difference between Rosen’s and Adler’s theory. The appropriateness of a certain theory, therefore, largely depends on the relevance of knowledge for consumption.

In individual sports talent is generally more observable than in team sports. In an Olympic 100 meter sprint finale for example, there is less uncertainty about its participants’ talent than in a soccer game. Talent is clearly measured by milli-seconds which tip the scales between success and loss. Consumers do not need specialized knowledge to single out the best sprinter. In line with Rosen (1981) even small differences in talent are leveraged into disproportionate differences in earnings.

In team sports like soccer, however, every game is a team product. Team production is characterized by the fact that it is difficult to determine each individual’s contribution to the output of the cooperating inputs (Alchian & Demsetz, 1972). Soccer is a highly interactive game based on the combination of complementary player skills. Together with relatively low scores and limited ‘set’ plays, the interactivity does not facilitate decomposition, record and measurement (Carmichael, Thomas, & Ward, 2000; Carmichael, Thomas, & Ward, 2001). A playing team consists of one goalkeeper plus ten outfield players who can generally be categorized as defenders, midfielders and attackers. A player’s performance always depends on complementary skills of other team-mates. Even the best goalkeeper hardly manages

24

Are Outstanding German Soccer Players Rosen or Adler Stars?

to impede opposition’s goal scoring, if the defense is virtually nonexistent. Or even outstanding attackers become lame ducks if they are not supported by offensive passes of midfielders or defenders. In soccer, all outfield players are involved in all aspects of the game to varying degrees. A player’s talent involves many hardly measurable capabilities like passing the ball to free-standing team-mates, retaining possession of the ball, running or dribbling with the ball, creating goal-scoring chances, tackling opponents, blocking or intercepting opposition’s passes and shots, or clearing the ball from pressure situations (Carmichael et al., 2001). The exact talent of a soccer player is fuzzy and requires much player specific knowledge to be properly discovered and assessed. I therefore expect German soccer players to be Adler stars whose market values depend on hidden talent characteristics, past consumption of the consumers and the player’s popularity.

2.5 Data and Stylized Facts on German Soccer In contrast to US leagues, which are generally ‘hermetic’, the composition of European soccer leagues changes annually through promotion and relegation. The best teams from a lower league are promoted to the next higher league, while the weakest in the latter are demoted to the next lower league. Due to the profile of the first Bundesliga as the highest German soccer league, I rule out superstar status to players appearing in lower leagues. While the first Bundesliga had an average match attendance of 35’183 in the 2004/05 season, the next lower league only attracted 12’074 fans on average.5 For the empirical analysis I concentrate, therefore, solely on players of the first Bundesliga. The sample contains all players who played for at least half an hour during the 2004/05 season6 – in total 427 players. These players or rather their teams generated an estimated turnover of € 1.1 billion in the 2004/05 season.7 The first German league is the third largest European soccer league after the English Premier League and the Italian Serie A (Jones, 2005). I chose the German 5

Average match attendance was calculated by the Kicker soccer magazine.

6

Unfortunately, I am not able to include further seasons because popularity data on previous seasons was partially not available.

7

If lower leagues are included as well, the revenue even topped € 1.5 Mrd. in the 2004/05 season.

Are Outstanding German Soccer Players Rosen or Adler Stars?

25

league because of its well documented games in the specialized press and two independent institutions that assess the market values of all players appearing in the first German league. Data on a set of personal player characteristics (e.g. goals, assists, appearances, tactical position, team, age, or race) is available from two special editions of the Kicker soccer magazine covering the 2004/05 season.

The analysis of the market values of 427 players appearing in the first German league reveals a highly unequal distribution with a substantial skewness. The Ginicoefficient is 0.56, which indicates high inequality. Figure 2 illustrates the density allocation of the logarithm of the market values in the first German league during the 2004/05 season. The distribution of the logarithm of market values is skewed to the right. The fatter (upper) right tail indicates the presence of a restricted number of star players with very high market values. While the median player is valued € 1.25 million, star players at the 95% quantile are exchanged for € 9 million.

Density .4

.3

.2

.1

0 10

12

14

16

18

Logarithm of market values

Figure 2:

Density allocation of the logarithm of market values

26

Are Outstanding German Soccer Players Rosen or Adler Stars?

The market value of Michael Ballack, who was the winner of the “Player of the Year”-award8 in the 2004/05 season, amounts to € 30 million. This corresponds to 600 times the lowest market value in the sample equaling € 0.05 million. The skewness of the distribution is lower than in many individual sports like for example in tennis,9 but higher than in other team sports like in American football, baseball, hockey or basketball.10 The earnings distribution in individual sports is expected to be more skewed than in team sports, because in individual sports no prize money awaits the bottom finisher, but at least a minimum salary is available to rookies in team sports (Scully, 1995). The fact that the distribution of market values in European soccer is more skewed than the distribution of salaries in typical US team sports at least partly depends on the different institutional restrictions in the US leagues (e.g. salary caps).

2.6 Empirical Framework 2.6.1 Dependent Variables The dependent variable in our study is the logarithm of a player’s market value at the end of the 2004/05 season. The used market values are estimated by industry experts of a team independent institution that runs the webpage www.transfermarkt.de.11 The used market values do not only incorporate salaries but also signing fees, bonuses, transfer fees and a remaining producer surplus. They reflect the total value generated by a particular player for his team and equal, therefore, the team’s maximal willingness to pay. The player himself appropriates a part of this value through salary

8

“Player of the Year” is an award assigned by sports journalists to the best player in the German league or the best German player in any other league.

9

In 1997 Kubat (1998) calculated a Gini-coefficient of 0.73 for the distribution of prize money to tennis players.

10

Scully (1995, p. 74) provides an extensive analysis of the distribution of player earnings in the US Major Leagues: The listed Gini-coefficients for the US Major Leagues vary between 0.22 (Hockey, 1978) and 0.51 (Baseball, 1990).

11

The same data source was also used by Eschweiler and Vieth (2004) or Torgler, Schmidt and Frey (2006).

Are Outstanding German Soccer Players Rosen or Adler Stars?

27

payments, bonuses and signing fees,12 whereas the selling club receives potential transfer fees. The buying club retains a possible producer surplus. However, the market values do not include individual endorsement fees. In order to explore the reliability of our market value data, we compared it with the market values provided by a second independent source, namely the Kicker soccer magazine. The two estimations are strongly correlated (correlation is 0.89), which indicates high reliability.13 Both data sources have been widely used for empirical research studies in the past (see for instance Lehmann & Weigand, 1999; Hübl & Swieter, 2002; Eschweiler & Vieth, 2004; Lehmann & Schulze, 2005; or Torgler, Schmidt & Frey, 2006).

2.6.2 Independent Variables I distinguish between four groups of independent variables: Talent variables, variables of past consumption, popularity variables and control variables. While the first three groups of variables are employed to differentiate between the Rosen-, Stigler/Becker- and the Adler-effect, the control variables are used to eliminate alternative explanations such as age, contractual status, race, and club or position characteristics.

In soccer, one performance characteristic that is clearly identifiable and measurable is goal scoring. The number of goals scored by each team of a particular fixture including those unintentionally scored by the opponent team determines the result of a game. Goal scoring and preventing the opposition to score are the critical success factors in a soccer game. Even though there are many constructive elements in a game which enable the teams to score goals, the public’s attention is largely concentrated on the players who finally score. There is no need for specialized 12

Unfortunately data about salaries, signing fees, bonuses, or transfer fees is not available in grand scale. In the 1999/00 season, salary data of players appearing in the German league was collected and published in a special edition of the magazine Sportbild (Lehmann, 2000; Schulze & Lehmann, 2005) and in the newspaper Welt am Sonntag (Kern & Süssmuth, 2005). However, these salaries do not include any bonuses, signing fees, or transfer fees.

13

In addition, the Pearson ǒ2 independence test rejects independency at the 0% level of significance. Thus, the two institutions seem to deliver corresponding data.

28

Are Outstanding German Soccer Players Rosen or Adler Stars?

consumer knowledge in order to ascertain the goal scorer. Since the sequence of goal scoring is replaid and analyzed several times in the live broadcast of a game, in the television newscast or on large screens in modern stadiums, not only the goal scorer but also the player making the final pass (called assist) prior to a goal being scored is easily identified. Thus I label GOALS and ASSISTS as firsthand observable talent measures, because they are clearly identifiable and measurable by the spectators without requiring significant specialized knowledge. They fit into Rosen’s conception of talent that is based on factors observable without cost. In contrast to the study of Lucifora and Simmons (2003) my firsthand observable talent measures GOALS and ASSISTS are not constructed as per game ratios, because the mere fact of low appearances should not have a positive impact on these performance measures. According to the law of large numbers, starters would have a lower chance to randomly achieve high scores than newcomers if the firsthand observable talent measures were per game ratios. As firsthand observable talent measure for goalkeepers I used OPPGOAL counting the number of opponent’s goals per game of a particular goalkeeper. Here I employed OPPGOAL as per game ratio to control for the effect that the number of opponent’s goals increases on average with the number of appearances, even though the latter is generally a sign of high talent.

A completely different set of talent variables is needed to control for the possibility that outstanding soccer players are Adler stars. In an Adler conception of the star phenomenon, talent is not easily identifiable because it is rather hidden than observable. Talent depends on many hardly measurable factors like e.g. physical characteristics, fitness, form, technical and social abilities and motivation. Thus, assessing a player’s true talent may imply a learning process that requires a lot of observations, reading and discussions with other competent individuals. In order to handle this complexity, consumers often rely on indirect talent indicators like expert opinions. Reinstein and Snyder (2005) show that expert opinions are an important source of “product” information especially for goods with high quality uncertainty. In European soccer, expert opinions often appear as comments by professional critics or journalists. They deliver valuable information that help consumers to indirectly assess a player’s talent.

Are Outstanding German Soccer Players Rosen or Adler Stars?

29

Three different expert appraisals are used in the study: Average match evaluation published by the Kicker soccer magazine (GRADE), votes for the “Player of the Year”-election among sports journalists (PLAYOTY), and membership of the national team (NAT).

In German soccer every match performance of a player who plays more than half an hour is individually evaluated by sports experts. The grades, which are published in the Kicker soccer magazine, vary between 1.0 (excellent) and 6.0 (very bad). But since I use the average grade of all evaluated match performances in the study, the variable GRADE spreads only from 2.5 to 5.

The Kicker soccer magazine also organizes an annual voting for the “Player of the Year”. At the end of the 2004/05 season approximately 3400 sports journalists were asked to vote for any player in the German league or any German player in any other league. PLAYOTY measures how many votes a player received. In total 995 valid votes entered the investigation. Compared to the variable GRADE the measure PLAYOTY considers more general overview impressions of players than precise match analyses.

A further indicator of exceptional talent is the membership of the national team. The national coach and his assistants screen potential players and select the most talented ones to form an excellent team for international team competitions like the European Championship or the World Cup. The membership of the national team is thus a sign of a remarkably high talent.14

The variables measuring appearances in the first German league during the 2004/05 season (APP) and prior to that season (PRIAPP) are used as proxies for past consumption. According to Byers, Peel, and Thomas (2001) spectators range in type from the committed regulars, who make up the “core” of attendance, to the “floaters” 14

I also tried to weight the national membership dummy with the FIFA-Ranking of the particular team in order to consider quality differences between national teams. However, this did not change the results in any significant way. Due to the ease of interpretation I use the unweighted dummies.

30

Are Outstanding German Soccer Players Rosen or Adler Stars?

whose attendance is determined by the attractiveness of a particular fixture. Since the percentage of attendance having a season ticket varies between 10% and 40% (Roy, 2004), I assume that the “core” of a club support attending match after match regardless of the team’s current form or star attraction is small. Most of the fans are “floaters”, however, within the same league. Potential accumulated knowledge, therefore, depends on the number of appearances in the first German league.15 The more often a particular player appeared on field, the higher is the expected consumption capital a fan may have accumulated. Not only the current productivity attracts fans but also memories of past performances (Rosen & Sanderson, 2001). In order to specifically analyze the consumption capital of the “core” of a team support, I also experimented with the separate effect of appearances for the present team only. However, I dropped this (insignificant) variable from the model because long contract duration of a player does not only facilitate the accumulation of player specific consumption capital, it also seems to be a sign of low talent.16 The work of Carmichael, Forrest, & Simmons (1999) shows that favorable performance measures increase the probability of being transferred. Unfortunately, more detailed variables measuring past consumption are not available. It is impossible to quantify the amount of time effectively used by all potential spectators in watching a particular player. As a result of missing alternatives I use APP and PRIAPP although potential distortion could result from a direct talent enhancing effect due to greater field experience.17

The Internet offers new and promising indicators of the popularity of a player. I collected data whether a player has a personal homepage (HOMEP) which provides the opportunity to directly address large groups18 with personal statements, personal characteristics or club information. In summer 2005, 23% of the players already ran a personal homepage and several planned to start one. I held an extensive interview

15

I assume that past consumption of player performances in foreign leagues is negligible.

16

Lehmann (2000), who analyzed wage determination in the first German league, found no significant influence of the appearances for the present team on salaries.

17

Lucifora and Simmons (2003) used the number of appearances as a variable measuring the experience of a player.

18

The homepages of well-known players are visited more than 100’000 times a month.

Are Outstanding German Soccer Players Rosen or Adler Stars?

31

with the head of a company that operates every fourth homepage in the sample.19 He told me that the main reason why players instruct him to design and operate a personal Internet platform is to have a channel of information which is self-controlled and suits, therefore, best to put oneself in the right light. Nowadays, personal Internet platforms seem to be indispensable in comprehensive public relation activities in order to increase one’s own popularity.

General publicity in the Internet was measured by the logarithm of results given by the Google search engine (LNGOOGR) searching for “name” and “Bundesliga”.20 If there were multiple players having the same name, I included the first name in the search job too. Thus, I minimized potential distortion to an acceptable level.

In addition, I analyzed the media presence in the German press. The variable PRESS indicates how often players are quoted with name and first name21 in over 20 German newspapers and magazines between the first July 2004 and 30th June 2005.22 In Table 2 the whole set of variables as well as the descriptive statistics are listed.

I use several control variables to eliminate alternative explanations, such as age, contractual status, race, team effects or position effects. I control for age (AGE) because several studies show that a player’s age has a positive but diminishing impact on salaries (Lehmann & Weigand, 1999; Frick, 2001; Lucifora & Simmons, 2003). To capture this nonlinearity I also control for age square (AGESQ). Even though empirical studies of North American Major Leagues typically do not include both appearances and age, in European soccer it is appropriate to utilize age and

19

The interview was held on 18th August 2005.

20

Both data on homepages and the results of the Google search were collected between 25th and 30th August 2005.

21

This way I minimize the distortions coming from the short match reviews in which players are quoted only by name. I excluded citations by name alone in order to prevent issues concerning multicollinearity with appearances and scores.

22

The used database contains quality nationwide newspapers (including Frankfurter Allgemeine Zeitung, Süddeutsche Zeitung, Stuttgarter Zeitung, Hamburger Abendblatt, Die Welt, taz, Berliner Morgenpost, Financial Times Deutschland) and weekly magazines (including Der Spiegel, Stern, Bunte).

32

Are Outstanding German Soccer Players Rosen or Adler Stars?

appearances separately, because players are not drafted and can, therefore, enter the industry at many different ages. Using age and experience together does not generate perfect multicollinearity (Lucifora & Simmons, 2003).

Variable

Description

Mean

SD

14.13

1.09

Firsthand observable talent measures: GOALS Goals ASSISTS Assists OPPGOAL Opponent's goals per game of a goalkeeper

2.07 1.75 0.12

3.46 2.50 0.42

Indirect talent measures: GRADE Average match grade by the Kicker sports magazine PLAYOTY Votes for the "Player of the Year"-election for the 2004-2005 season NAT Membership of the national team (dummy)

3.77 2.09 0.32

0.46 25.69

19.33 67.23

9.94 76.91

Personal homepage (dummy) Logarithm of results of the google search Citations in over 20 German newspapers and weekly magazines

0.23 9.32 155.83

1.10 266.35

Player's age Squared term of AGE Contract ends in summer 2005 (dummy) Contract ends in summer 2006 (dummy) Foreign player from a European country (dummy) Foreign player from a non-European country (dummy) attacker (dummy) defender (dummy)

27.31 763.39 0.34 0.27 0.41 0.12 0.24 0.33

Dependent variables LNVALUE

Logarithm of a player's market value

Independent variables Talent variables

Proxies for past consumption APP PRIAPP

Appearances in the 2004/05 season Accumulated appearances prior to the 2004/05 season

Popularity variables HOMEP LNGOOGR PRESS Control variables AGE AGESQ LASTY LASTBOY FOREU FORNONEU ATTACKER DEFENDER

4.18 232.12

Note: The model also includes 17 team dummies that are not reported.

Table 2:

Variables and descriptive statistics

In addition, I control for the contractual status of a player using two dummy variables. The first dummy variable (LASTY) indicates if the contract ends in summer 2005 (coded 1) and the second (LASTBOY) if the player contract ends in summer 2006 (coded 1). The impact of contract duration on market values is controversial: some scholars (e.g. Lehn, 1982; Scoggings, 1993) say that guaranteed multi-year contracts

Are Outstanding German Soccer Players Rosen or Adler Stars?

33

reduce player effort due to a moral-hazard effect while others (e.g. Kahn, 1993; Maxcy, 2004) argue that only the better players receive comparably long contracts (self-selection effect).

Two dummy variables concerning a player’s race are included: FOREU coded 1 for European players that are not German and FORNONEU coded 1 for non-European players. Since cost considerations (screening costs, mobility costs etc.) speak for hiring the German player among two equally talented players, I predict that nonGerman players who actually got engaged in the German league have superior talent and thereby higher market values. In addition, the variable FORNONEU also controls for the effect that German teams are still not allowed to select more than three nonEuropeans to simultaneously play in a game. By restricting the number of nonEuropean players, this regulation has the effect that only the very best from the talent distribution of non-Europeans will be employed at all.

I take account of team-specific effects by using team fixed effects estimations assigning unobserved team effects to team dummies. Team effects are supposed to have significant influence on player market values (Idson & Kahane, 2000). Somebody who is in the squad of the team winning the championship race enjoys much greater publicity and finances than someone in the team being relegated to the next lower league.

Position dummies are used to control for specific effects resulting from the tactical position of a player. Lehmann and Weigand (1999) for instance find evidence that in the German league midfielders earn significantly more money than other players.

2.6.3 Results A standard approach is to specify the unknown parameters of a linear regression using the method of ordinary least squares (OLS) or least absolute deviation (LAD). Both methods lead to an approximation to the mean (OLS) or median (LAD) and represent the “averaging” behavior or the “central” tendency of a conditional

34

Are Outstanding German Soccer Players Rosen or Adler Stars?

distribution. However, they tell little about the tail behavior (Kuan, 2004). The ordinary least squares (OLS) procedure that tests on the mean value will, therefore, not be able to capture the superstar phenomenon correctly (Lehmann & Schulze, 2005). The quantile regression approach, originally developed by Koenker and Bassett (1978), allows characterizing a particular point of the conditional (asymmetric) distribution. It minimizes an asymmetrically weighted sum of absolute errors, where the weights are functions of the quantile of interest. The standard errors are estimated using the bootstrap procedure.23

Sherwin Rosen defined superstars as “the relatively small numbers of people who earn enormous amounts of money and dominate the activities in which they engage” (Rosen, 1981, p. 845). Obviously Rosen (1981) bases his definition of superstardom on the distribution of earnings among the suppliers of a certain good or service. However, Rosen does not propose a clear percentage number as “boundary” between “normal” suppliers and superstars. I, therefore, decided to analyze different quantiles in order to examine the robustness of the results. In addition, I also present the OLS estimates with White-robust standard errors as comparison in Table 3.

Table 3 shows that the coefficients generally loose statistical significance by moving from OLS to quantile regression (see also Schulze & Lehmann, 2005). Whereas the number of goals scored and the opponent’s goals per game of a goalkeeper significantly influence the market values on the mean, the same does not hold for the top 10%, 5% or 2% quantiles. The firsthand observable talent measures (goals, assists and opponent’s goals per game of a goalkeeper) do not significantly affect the market values of stars. This result does not change if interaction terms between GOALS and ASSISTS with tactical position dummies are included. Thus, not even the market values of forwards are driven by firsthand observable scores but rather by indirect talent measures like e.g. the Kicker grade. If the average grade of the match evaluations given by Kicker sports journalists is one score better, this increases the

23

I ran 1000 replications so that the estimates of standard errors are rather stable (see Koenker & Hallock, 2000).

Are Outstanding German Soccer Players Rosen or Adler Stars?

35

star’s value more than 30%.24 The coefficient of the variables PLAYOTY and NAT have the expected positive sign; however, both variables are statistically insignificant regarding the quantile regressions. It seems that the accurate match evaluations better detect and represent the talent of a star player than rather global judgments of sports journalists or of the coaches of national teams.

OLS Variable

ȕ-coef.

GOALS ASSISTS OPPGOAL

0.0179 0.0125 -0.1748 *

GRADE PLAYOTY NAT APP PRIAPP

98% quantile

Std. Error

ȕ-coef.

Std. Error

ȕ-coef.

Std. Error

0.0120 0.0152 0.0822

0.0227 0.0088 -0.0815

0.0211 0.0236 0.1449

0.0219 0.0022 -0.1476

0.0237 0.0274 0.1492

0.0221 0.0135 -0.1371

0.0227 0.0282 0.1617

-0.4263 ** 0.0931 -0.0001 0.0006 0.1338 * 0.0680

-0.3397 * -0.0021 0.0239

0.1458 0.0071 0.1494

-0.3393 * 0.0005 0.0523

0.1538 0.1605 0.1605

-0.3029 * 0.0003 0.1002

0.1589 0.0071 0.1612

0.0294 ** 0.0039 -0.0009 * 0.0005

0.0176 ** 0.0067 + -0.0017 0.0010

0.0156 * + -0.0015

0.0077 0.0011

0.0136 * -0.0012

0.0078 0.0011

0.1609 * 0.0740 0.0944 * 0.0447 0.0007 ** 0.0002

0.2332 * 0.1146 + 0.0988 0.0628 0.0012 ** 0.0004

0.2068 0.1291 0.1413 * 0.0636 0.0010 ** 0.0004

0.1159 0.1492 * 0.0009 *

0.1419 0.0708 0.0004

** 0.0806 ** 0.0014 0.0761 0.0776 ** 0.0850 ** 0.1010 0.0948 0.0722

0.6139 ** 0.1395 -0.0114 ** 0.0025 0.0052 0.1176 -0.0627 0.1178 0.3953 ** 0.1331 0.2618 0.1889 0.1404 0.1660 + 0.2412 0.1266

0.4982 ** 0.1623 -0.0093 ** 0.0029 0.0365 0.1302 -0.0720 0.1216 0.3706 * 0.1483 0.2853 0.2103 0.2406 0.1770 + 0.2476 0.1395

0.4494 ** 0.1705 -0.0085 ** 0.0031 0.0310 0.1326 -0.0776 0.1366 0.3609 * 0.1495 0.3155 0.2150 0.2360 0.1820 + 0.2596 0.1474

5.8471 ** 1.0685

0.6430 ** 2.0741

7.7799 ** 2.3423

9.3597 ** 2.5219

HOMEP LNGOOGR PRESS

0.6225 -0.0115 -0.0928 -0.0567 0.3130 0.4042 -0.1016 -0.0398

Constant Team fixed effects Pseudo R

95% quantile

ȕ-coef.

+

AGE AGESQ LASTY LASTBOY FOREU FORNONEU ATTACKER DEFENDER

90% quantile

Std. Error

yes **

2

Number of observations

yes *

+

yes

+

yes **

0.72

0.56

0.57

0.61

427

427

427

427

+

Note: Significance levels: 10%, * 5%; ** 1%; Significance tests are one-tailed for directional independent variables and two-tailed for control variables.

Table 3:

Estimates of the logarithm of the players’ market values

Analyzing the variables of past consumption, we see that the number of appearances in the 2004/05 season (APP) correlates with the dependent variable at least at a 5% significance level. The stronger fans specialize on a particular star player, the higher the appreciation of this player gets. The coefficient of the variable PRIAPP measuring 24

However, we have to be cautious with the generalization of the interpretation, since it implies that a person who happens to be in a specific quantile of one conditional distribution will also find himself in the same quantile had his independent variables changed (Buchinsky, 1998).

36

Are Outstanding German Soccer Players Rosen or Adler Stars?

prior appearances is negative (at the 10% level of significance or lower). It seems that only recent experience displays positive influence on market values.25

A special focus of my study lies on the popularity variables. Table 3 shows that all popularity measures used in my study have the expected positive impact on a star’s market value. For star players in the 95% quantile, the existence of a personal homepage (HOMEP) increases market values by 20.7%. One percent more hits in the Google search enhance the demand by 0.14%, and every press citation leads to an increase of 0.1%. This means that the media presence of star players significantly enhance their market values even when all the talent and performance variables are held constant. Therefore, we find evidence that Adler’s theory of superstar emergence is supported for German soccer stars. It seems that the hardly measurable task of soccer players requires player specific knowledge in order to be properly evaluated and appreciated. Hence, the demand for a star player is not only determined by his talent, but also by his popularity and the past consumption opportunities for the fans.

The significant influence of the control variables AGE and its square confirms what a general human capital earnings function would predict: The market value of star players rises with age but at a decreasing rate. The turning point for star players slightly increases if the star definition is modified to encompass 5% or even 10% of the players. A superstar at the 98% quantile reaches his personal peak already at the age of 26.4, while a star at the 90% quantile arrives at the maximum strength at the age of 26.9. Beyond that age, higher consumption capital is in general offset by worsening talent concerning physical performance, reduced speed and fitness. The analysis of the control variables FOREU and FORNONEU confirms the prediction that overall non-German players have higher market values than German players. The premium for non-German European star players is even higher than for nonEuropeans. The latter coefficient though is not significant regarding the chosen quantile regressions.

25

Appearances can also be interpreted as an indicator of the star’s (not the consumers’) experience that might be subject to diminishing (or even negative) returns.

Are Outstanding German Soccer Players Rosen or Adler Stars?

37

Since we have both firsthand observable and hidden talent variables, I have to respond to the issue of multicollinearity. Table 4 thus provides a correlation matrix which shows that a correlation above 0.7 is found only for AGE and AGESQ. Since the variance inflation factors (VIFs) for all regressors except AGE and AGESQ are well below 10, I am not concerned with multicollinearity due to different operationalizations of a player’s talent in the model.26

Variable 1 LNVALUE 2 GOALS 3 ASSISTS 4 OPPGOAL 5 GRADE 6 PLAYOTY 7 NAT 8 APP 9 PRIAPP 10 HOMEP 11 LNGOOGR 12 PRESS 13 AGE 14 AGESQ 15 LASTY 16 LASTBOY 17 FOREU 18 FORNONEU 19 ATTACKER 20 DEFENDER

1 1.000 0.469 0.461 -0.129 -0.369 -0.196 0.440 0.537 0.206 0.319 0.536 0.558 0.011 -0.017 0.089 0.193 -0.119 -0.027 0.048 -0.083

Variable 11 LNGOOGR 12 PRESS 13 AGE 14 AGESQ 15 LASTY 16 LASTBOY 17 FOREU 18 FORNONEU 19 ATTACKER 20 DEFENDER

11 1.000 0.497 0.110 0.103 -0.033 -0.017 -0.054 0.081 0.040 -0.114

Table 4:

2

3

4

5

1.000 0.624 -0.161 -0.201 0.345 0.280 0.441 0.079 0.172 0.382 0.477 0.023 0.009 -0.043 -0.032 0.042 0.136 0.380 -0.276

1.000 -0.190 -0.251 0.293 0.270 0.494 0.175 0.278 0.392 0.339 0.050 0.039 -0.075 0.036 0.028 -0.075 0.181 -0.287

1.000 -0.349 0.080 -0.142 -0.054 0.039 0.057 -0.040 -0.003 0.164 0.176 0.026 -0.040 -0.153 -0.106 -0.156 -0.193

1.000 -0.357 -0.147 -0.312 -0.202 -0.172 -0.258 -0.303 -0.114 -0.125 0.085 0.069 0.147 0.001 0.210 0.044

12

13

14

15

6

7

8

9

10

1.000 0.145 1.000 0.287 0.259 1.000 0.113 0.095 0.298 1.000 0.294 0.199 0.251 0.266 1.000 0.343 0.279 0.467 0.368 0.340 0.521 0.301 0.359 0.321 0.345 0.075 0.010 0.200 0.573 0.040 0.076 -0.005 0.185 0.581 0.039 -0.033 0.009 -0.072 -0.009 -0.061 -0.001 -0.054 -0.018 0.091 0.030 -0.097 0.250 -0.020 -0.104 -0.137 0.008 0.108 0.041 -0.046 -0.008 0.025 0.079 0.028 -0.090 0.047 -0.073 -0.035 -0.102 -0.065 -0.105 16

17

1.000 0.093 1.000 0.090 0.996 1.000 -0.068 0.198 0.193 1.000 -0.001 -0.060 -0.058 -0.443 1.000 -0.132 0.152 0.141 0.104 -0.005 1.000 0.131 -0.052 -0.063 -0.041 -0.080 -0.335 0.083 -0.126 -0.134 -0.049 0.071 0.102 -0.111 0.003 0.003 0.001 -0.105 -0.010

18

19

1.000 0.110 1.000 0.014 -0.394

20

1.000

Pearson correlation coefficients

Despite the high correlation between AGE and AGESQ I do not drop AGESQ from the model for two reasons: Firstly, even high multicollinearity (as long as it is not

26

A commonly given rule of thumb says that only VIFs above a value of 10 may be a reason of concern (see e.g. Williams, 2006).

38

Are Outstanding German Soccer Players Rosen or Adler Stars?

perfect) does not violate unbiased estimates. Actually, worrying about high degrees of correlations between the independent variables is the same as worrying about a small sample size: both work to increase the variance of the coefficient estimates and might lead to statistical insignificance (Wooldridge, 2003). However, in my model both AGE and AGESQ display high significance. Secondly, a general Mincer-type human capital formulation expects that player salaries increase with age at a decreasing rate and that salaries would fall with age as players experience declining speed and athleticism (Lucifora & Simmons, 2003). Thus, dropping the variable AGESQ would lead to biased estimates, because it really belongs to the model.

Whenever correlational designs are used, concerns about internal validity such as possible reverse causality may be raised as well. However, since most of the independent variables concern the whole 2004/05 season, while the market values were estimated at the end of the 2004/05 season, the issue of reverse causality is appeased by this lag structure.

2.7 Conclusion Rosen’s theory of superstar formation stressing the importance of firsthand observable talent is not supported for German soccer stars. Easy measurable and identifiable talent indicators like goals and assists have no significant impact on their market values. The specific contribution to a soccer game and hence the exact talent of a star player is indeed difficult to determine. A soccer match is a typical team product. It seems that the assessment of soccer players requires specific consumption capital as stipulated by Adler’s theory of superstar emergence. The market values of German soccer stars are better predicted by expert evaluations revealing hidden talent characteristics than by firsthand observable talent measures. We also find clear empirical evidence that both past consumption of the spectators (Stigler/Beckereffect) and the player’s popularity (Adler-effect) are significant predictors of the stars’ market values. Publicity in the press and in the Internet increases demand. I believe that the predictive power of the popularity measures is even underestimated in

Are Outstanding German Soccer Players Rosen or Adler Stars?

39

my study, because the used market values did not include any individual endorsement fees which are highly contingent on a player’s popularity.

According to Adler’s superstar theory, two different strategies for becoming a superstar arise: players can either intensify their investments in physical talent in order to receive better expert appraisals or they can make higher popularity investments. The best example for a player following the second strategy is David Beckham, who was seen as the world’s most famous soccer player in 2004. He was never winner of the “FIFA World Player of the Year”-award27 and hardly any soccer expert considers him as the world’s most talented player. Nevertheless, David Beckham led the Forbes list 2004 of the highest-paid soccer player.28 In summer 2004 he switched teams from Manchester United to the Spanish club Real Madrid in a $41 million transfer. Real Madrid signed the soccer star not so much for his on-the-field prowess, but for his ability to attract fans due to his popularity, in particular in Asia. In Japan, Beckham has already achieved name recognition of over 90% (James, 2003). Beckham consciously presents himself not only as a soccer star but as a pop icon as well. His face has launched a thousand of tabloids and his marriage with the famous pop singer Victoria Adams definitively made him a celebrity gold. Real Madrid’s coach, however, was not as taken with Beckham and in autumn 2006 relegated him to the bench for much of the season. Beckham had to recognize that his abilities as a player were on the wane. They did no longer meet the high requirements of one of the strongest team in Europe. Thus, in January 2007 Beckham agreed to a five-year deal with the Los Angeles Galaxy in American Major League Soccer, which has failed to gain the popularity soccer receives in most other countries so far. Beckham’s signing represents a confluence of marketing and pop culture with a little bit of soccer in the perfect locale – southern California. It is obvious that moving in Los Angeles’ celebrity circles will better satisfy his continuing enthusiasm for life in the limelight than being a benchwarmer in Madrid. The profile of David Beckham is 27

“FIFA World Player of the Year” is a soccer award annually given to the male and female player who is thought to be the best in the world based on votes by coaches and captains of international teams.

28

See for this information www.forbes.com/athletes2004 visited on second October 2005. Forbes earning estimates include salaries, bonuses, prize money, endorsements and appearance fees.

40

Are Outstanding German Soccer Players Rosen or Adler Stars?

expected to rise in Hollywood land, not least because of his nearby celebrity friends Tom Cruise and Katie Holmes. Los Angeles is the ideal place to cash in on his celebrity appeal. The press reports that the five-year deal with the Los Angeles Galaxy includes salary and commercial opportunities worth $250 million (Bell, 2007). David Beckham is a typical Adler-star – talented, of course, but above all famous and glamorous.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

41

3 Local Heroes and Superstars – An Empirical Analysis of Star Attraction in German Soccer1

3.1 Introduction Team composition plays a fundamental role in facilitating fan support: 69% of the European soccer fans say that their identification with and affiliation to a team is largely determined by the particular players the team engages (Sportfive, 2004). Recent studies in the widely and fast growing literature on the demand for sports2 clearly indicate that outstanding players – so-called stars – play an important role in attracting fans (see i.e. Hausman & Leonard, 1997; Mullin & Dunn, 2002; Berri, Schmidt, & Brook, 2004; Berri & Schmidt, 2006). Since soccer fans tend to form attachments to particular teams mostly on the basis of geographic proximity (Szymanski, 2003b), this chapter argues that not only well-known superstars but also “local heroes” may play an important role in enhancing fan interest. Defining superstars as players whose market values are in the top 2% quantile of the league’s distribution of market values3 and a “local hero” as the most valued player of a particular team that has no superstars, I want to shed more light

1

This chapter was published in the Journal of Sports Economics (see Brandes, Franck, & Nüesch, 2007). Even though I have written this section together with Leif Brandes and Egon Franck, I still use the singular due to consistency reasons. The same applies for the chapters four and six which were written with the co-authors Helmut Dietl (chapter four) and Egon Franck (chapters four and six).

2

See Szymanski (2003a) or Borland and MacDonald (2003) for a review.

3

The used market values were collected from special editions of the Kicker soccer magazine at the beginning of each season. For the season 1997/98 the market values were published in the weekly Kicker edition No. 61 in 1997.

42

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

on the still quite obscure relationship between star players and match attendance. In the theoretical star-literature it is controversial whether stars drive demand by their talent superiority (see Rosen, 1981; MacDonald, 1988) or simply by their comparably higher popularity (see Adler, 1985). Analyzing longitudinal match attendance data of all clubs in the first German soccer league during the seasons 1995/96 – 2003/04, this chapter explores star attraction by both a star’s field performance and his popularity. Furthermore, I distinguish between locally dominating stars and national superstars and I investigate their star attraction both in home games and on the road. The data shows that local heroes attract fans only in home games, namely due to their popularity. Superstars, however, facilitate fan support both at home and on the road – not because of mere celebrity status but rather because of their outstanding talent. However, robustness checks reveal that a star attraction analysis requires a precise definition of superstardom.

3.2 Related Literature Roger Noll (1974) was the first to analyze star attraction by introducing a superstar variable in his match attendance study. This superstar variable captured the effect of stars on attendance beyond their contribution to team victories. However, it was not significant. Gerald Scully (1974) stated that players can influence club revenues in Major League Baseball in a twofold way: “Ability contributions to team performance and victories raise gate receipts. (…) Additionally, it is possible that some players may attract fans over and above their individual contribution through the team” (p. 916). Unfortunately Scully (1974) did not include the latter effect in his econometric framework. Using a two-equation model, he only related player specific performance statistics to team success and, in a second step, team revenue to the team’s win-loss record and other market characteristics. Scully (1974) did not consider star attraction by sheer popularity in his econometric framework.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

43

Hausman and Leonard (1997) empirically analyzed superstar effects on team revenues in professional basketball.4 They found that the mere presence of stars had a substantial positive impact on club revenues even after controlling for team quality measured by the number of All-Star players in a team. By analyzing all NBA local and national television ratings as well as match attendances, Hausman and Leonard (1997) singled out that – back in 1993 – the estimated value of Michael Jordan for the National Basketball Association (NBA) was $ 53 million. The study of Hausman and Leonard (1997), however, does not analyze whether the star’s performance and/or his popularity increases team revenues.

Mullin and Dunn (2002) define “star quality” in the Major League Baseball as the residual in a fit of a player’s card prices to performance statistics. They acknowledge that star quality brings fans to the stadium and impacts team revenues in a significant way beyond pure field productivity. Mullin and Dunn (2002) determine a player’s marginal revenue product running a three-step process involving the sequential determination of (1) the effect of an individual’s performance on team performance; (2) the effect of team performance on winning percentage; and (3) the impact of winning percentage and a player’s star quality on attendance and hence on revenues. They found clear evidence that stars may influence gate revenues both by their talent which is translated into field success and by their popularity.5

Berri et al. (2004) investigated the two-sided relationship between match attendance and both team performance and the team’s mere employment of star players in the NBA. By choosing a multiplicative model, they regressed a team’s home gate revenue on team performance, star popularity measured with received All-Star votes, franchise and market characteristics. Their results suggest that it is performance on the court, not star popularity, which attracts fans.

4

Scott, Long, and Scompii (1985), Brown, Spiro, and Keenan (1991), as well as Burdekin and Idson (1991) already controlled for the effect of a team’s star attraction in their analyses of match attendance in the NBA prior to Hausman and Leonard (1997). However, the existence of a potential superstar effect was not their main focus. Of these studies only Brown et al. (1991) were able to find a statistically significant relationship between match attendance and the number of stars in a team.

5

Note that the term “star quality” from above only reflects the popularity aspect of stars.

44

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

Berri and Schmidt (2006) extended the study of Hausman and Leonard (1997) via an examination of road attendance in the NBA. They found evidence of a superstar externality. Whereas an additional All-Star vote increases aggregate road attendance by only 0.005 fans, each team win leads to an estimated 1’011 increase in attendance on the road. According to Berri and Schmidt (2006) Michael Jordan’s productivity for example was worth approximately $ 2.2 million while his star appeal only generated $ 156’123. Thus, they suggest that showmanship cannot replace actual court performance. However, the studies of Berri et al. (2004) as well as Berri and Schmidt (2006) treat team wins as exogenously given by the stars’ talent. They do not analyze how stars exactly influence team performance.

3.3 Stylized Facts on German Soccer German soccer enjoys high popularity. According to a representative survey of the Sportfive-company, 77% of the German population are interested in soccer. 39% of them quote that they cannot even imagine a life without soccer (Sportfive, 2004). This high enthusiasm is reflected in hard facts: The financial turnover of the German soccer leagues topped € 1.5 billion in the 2004/05 season (Bundesliga, 2006). At the same time, average match attendance in the first Bundesliga increased to 36’900. No other soccer league in Europe attracts more fans at the gate than the first Bundesliga (Jones & Boon, 2005).

Most soccer supporters express allegiance to a particular club. Their attendance is largely an expression of support for that club. Spectators who attend out of purely neutral interest tend to represent a minority at soccer matches (Simmons, 1996). Supporters are often organized into supporter clubs, which raise the social component of a sports event. The geographical distribution of fan bases varies largely between different teams of the league. While some are more locally rooted, others have supporter clubs all over Germany (Czarnitzki & Stadtmann, 2002). Bayern Munich for example appeals rather nationally. Only 29% of all Bayern fans actually live in Munich. Hansa Rostock, on the other hand, has strong local roots. 68% of their fan basis live in Rostock. Even though Bayern Munich had an average home match attendance of

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

45

54’882 in the 2003/04 season, this only represents 9.1% of Munich’s male population. In the case of Hansa Rostock, however, match attendance corresponds to 22.9% of the male population in the home town (see Table 5). Average match attendance Portion of local attendance Male population Match attendance in percentage of male population Number of superstars

Bayern Munich 54'882 29% 602'708

Hansa Rostock 22'323 68% 97'567

9.1%

22.9%

6

0

Number of players in national teams

14

4

Number of players nominated for the "Player of the Year"-election

6

0

Table 5:

Comparison of Bayern Munich and Hansa Rostock in the 2003/04 season (Source: Sportfive, 2004, own calculations)

While Bayern Munich had six superstars with a market value in the top 2% quantile of the league and six players were nominated for the “Player of the Year”-election6 in the season 2003/04, Hansa Rostock had none of these superstars. However, supporters of Hansa Rostock are very unlikely to regard a Bayern Munich match as a perfect substitute for watching “their” team. Explanations for this imperfection may be found either in economic reasons like travel costs or in the intangible allegiance or loyalty to a particular team. Therefore, the market for admission to Hansa Rostock home games bears features of a local monopoly. Of course, Hansa competes for spectators with other clubs (including those in other leagues) and with other leisure attractions. No club has a monopoly in an absolute sense (Forrest, Simmons, & Feehan, 2002). However, the high affiliation of local fans leads to a situation, in which Hansa Rostock has discretion over a level of admission prices. And, therefore, outstanding players of small teams, such as Hansa Rostock, may attract fans without having a nationwide appeal. I call them “local heroes”. A local hero is defined as the most expensive player in a team, given that his market value does not belong to the highest two percent of the league. Therefore, the definition of a local hero does not necessarily imply that the respective player is a

6

The Kicker soccer magazine organizes an annual voting for the “Player of the Year”. At the end of the 2003/04 season approximately 3400 sports journalists were asked to vote for any player in the German league or any German player in any other league.

46

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

home-grown young player who has just made the starting team. Local heroes may not achieve superstardom league-wide. However, they take the number-one-position within their teams.

The possibility that both superstars and local heroes exist in the same team is ruled out because of the following reasons. Firstly, as already mentioned, fans largely focus on the team they support. Within a particular team the star attraction of a superstar is expected to dominate over the potential star attraction of a local hero. Secondly, talented players consciously select a team in order to maximize their individual utility. In doing so, not only the salary or the absolute quality of a team, but also the relative position and rank in the team enter their utility function. People in general constantly compare themselves to others and enjoy a sense of well-being when they out-perform their peers (see e.g. Clark & Oswald, 1996; or Ferrer-i-Carbonell, 2005). Therefore, given a certain team quality, I assume a local hero to prefer being the best within a team instead of being just an interchangeable player among stars.

In order to check the robustness of the results, I also run the regressions using broader star definitions. One alternative model defines the highest 5% quantile of the league’s market value distribution as superstars and the two most valuable players of a team which has no superstars as local heroes. A second alternative model accounts even the 8% players with the highest market values as superstars and the three most valuable players in teams without superstars as local heroes. This sensitivity analysis is necessary since it is not a priori clear how many players within a team may exhibit a particular star attraction. Or taken the other way round: On how many star players are viewers able to focus simultaneously? On theoretical grounds, I clearly prefer the 2% superstar definition because the superstar literature argues that one single actor – the best – dominates the whole market (Schulze, 2003). Based on the fact that the market for gate attendance may be considered as a local monopoly, it makes sense to assume a strong concentration of viewer interest on a very restricted number of players per team.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

47

3.4 Star Attraction The existing theoretical literature on superstars (Rosen, 1981; Adler, 1985; MacDonald, 1988) suggests two main ways how stars attract fans: by outstanding talent and exceptional performance and/or by remarkable popularity.

3.4.1 Star Performance Sherwin Rosen, who wrote a seminal paper on “The Economics of Superstars” in 1981, derives the existence of superstars from the premise that consumers consider lower quality as an imperfect substitute for higher quality. According to Rosen, spectators want to see the best players under the ceteris paribus assumption. Watching for example a succession of mediocre dribblings does not add up to a single outstanding dribbling performance. Therefore, small differences in talent translate into large differences in fan support. In line with Rosen (1981), I postulate that stars attract fans and generate disproportional high match attendance by their outstanding field performance.

Soccer is a highly interactive game based on the combination of complementary player skills. Together with relatively low scores and limited ‘set’ plays, the interactivity does not facilitate decomposition, record and measurement (Carmichael et al., 2001). Hence, in soccer we do not have the depth of player performance indicators available for more individualistic North American team sports such as baseball and basketball (Lucifora & Simmons, 2003). However, one performance characteristic that is clearly identifiable and measurable is goal scoring. Since winning depends on a positive goal difference, goal scoring and preventing the opposition to score are the critical success factors of a game. In my empirical study I, therefore, measure field performance by counting the goals and the assists defined as final pass before a goal is scored. Since forwards and midfielders are more likely to score than defenders, the performance of each star is divided by the league average of goals and assists of players in the same tactical position in the corresponding season. The sum of weighted goals and assists – namely the weighted scoring points – of a local hero (SCORELHWP) or of a superstar

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

48

(SCORESSWP) in a particular team serve as Rosen talent variables. In addition, a dummy is incorporated if a team has a local hero or a superstar as goalkeeper.

3.4.2 Star Popularity In contrast to Sherwin Rosen, Moshe Adler stated that stars do not necessarily need to have superior talent. They may just be more popular and attract fans by their high profile and celebrity status. In Adler’s logic the appreciation of a star’s performance increases with the knowledge the consumer has about the star. The more popular a soccer player is, the easier it gets to accumulate this so-called “consumption capital”7. According to Adler (1985) there is more than mere quality that attracts fans. Mullin and Dunn (2002) describe the star’s popularity of a baseball player as an intangible characteristic that attracts fans who pay to see these stars even when their playing performance is not more than mediocre: “Star quality thus consists of both reputation based on past performance and charisma above and beyond actual playing ability” (p. 621). Stars may have a “personal appeal” that activates fan interest even after controlling for their team’s (increased) quality (Hausman & Leonard, 1997).

To identify the Adler-star effect, I measure a player’s popularity by counting how often star players are quoted with name and first name in more than 20 German newspapers and magazines (MEDIALHP and MEDIASSP).8 Of course, press citation rather reflects publicity and is only a proxy of a player’s popularity. However, publicity such as coverage in tabloids, magazines or newspapers is strongly related to popularity (Adler, 2006).

7

The notion „consumption capital“ was introduced by Stigler and Becker (1977).

8

The used database contains quality nationwide newspapers (including Frankfurter Allgemeine Zeitung, Süddeutsche Zeitung, Stuttgarter Zeitung, Hamburger Abendblatt, Die Welt, taz, Berliner Morgenpost, Financial Times Deutschland) and weekly magazines (including Der Spiegel, Stern, Bunte).

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

49

3.5 Econometric Framework 3.5.1 Data and Dependent Variable The analyzed sample contains data on all 18 clubs in the first German league over nine seasons – beginning with the 1995/96 season and concluding with the 2003/04 season. Due to the high profile of the first Bundesliga as the highest German soccer league, substantial star attraction for players appearing in lower leagues is ruled out.9 The composition of European soccer leagues changes annually through promotion and relegation. The three best teams from the second Bundesliga are promoted to the first league in the following year, while the weakest three clubs of the first Bundesliga are relegated. The sample consists of 28 clubs in total. Some of them played only one season in the highest soccer league (Uerdingen, SSV Ulm) while others like Bayern Munich, Hamburg or Leverkusen were never relegated.

Studies about the star attraction in sports either concentrate on home games (Noll, 1974; Scully, 1974; Mulin & Dunn, 2002; Berri et al., 2004) or away games (Hausman & Leonard, 1997; Berri & Schmidt, 2006). This section analyzes star attraction of both home and away attendance in two separate models, because I assume the star effect to be different based on where the game is played. The dependent variable for home games is the logarithm of the aggregate seasonal match attendance. The logarithm of the sum of attendance of all away games of a particular team denotes the dependent variable analyzing star attraction on the road.

To identify the relationship between a team’s star performances and a team’s star popularity with match attendance, a set of control variables is needed to eliminate alternative explanations such as team or market characteristics.

9

The average match attendance in the second Bundesliga is approximately one third of the match attendance in the first Bundesliga.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

50

3.5.2 Controls Besides a simple time trend, I also control for club idiosyncratic factors like a team’s reputation (REP20) or the stadium capacity (CAPACITY) and market characteristics. Czarnitzki and Stadtmann (2002), who analyzed the determinants of match attendance in the first German soccer league for the seasons 1995/96 and 1996/97, identified a strong relationship between reputation, measured by past field success, and match attendance. Teams that enjoyed success in the past are expected to have stronger fan support than other teams which had less success. The measure REP20 takes into account the performance of a particular team over the last twenty years according to the following formula: 20

REP 20

18

¦x t 1

t

t

(1)

xt is the team’s final rank in the championship t years ago. In case that the team did not play in the first German league in season t , the corresponding summand is set equal to zero. By weighting the rankings with the square root of the number of years past, the index is constructed to reflect the depreciating effect of time (Czarnitzki & Stadtmann, 2002).

The aggregate seasonal stadium capacity (CAPACITY) is expected to have a positive impact on a team’s gate attendance. The variable CAPACITY for home games is calculated using the weighted average of all sold-out games of a particular club in a given season, which is then multiplied with the number of home games. In doing so, capacity changes within a season for example due to stadium reconstruction are incorporated. Concerning attendance on the road, the variable CAPACITY is the sum of stadiums' capacities of all other clubs in a given season. In general, I expect that the higher a stadium capacity, the more people may attend a game without increasing ticket prices. Berri et al. (2004) found a significant positive relationship of stadium capacity and gate revenues in the NBA.

In addition to the mentioned team characteristics, I also use three variables controlling for specific market characteristics like the male population (MEN) and the

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

51

unemployment rate (UNEMP) in the home town and the competitive balance of the league (CB).

European soccer fans typically tend to form attachments to particular teams on the basis of geographic closeness (Szymanski, 2003b). Thus, the size of the population in the potential market for a particular team is expected to positively relate to gate attendance (Borland & MacDonald, 2003; Falter & Pérignon, 2000). Schmid and Berri (2001) suggest that the size of the metropolitan statistical area is a common proxy for the size of a team’s market.10 Since soccer is rather a men’s game11, I only count the number of males in the home town.12 Borland and MacDonald (2003) claim that attendance at sporting events may constitute a social outlet for unemployed persons, so that (other things equal) attendance is higher as the rate of unemployment increases. On the other hand average income, which is positively associated to match attendance, decreases. Therefore, the forecasted effect of the unemployment rate on match attendance is not clear.

In addition, I also control for seasonal competitive balance using the Herfindahl-Index which measures the concentration of points among the participating teams. The higher the Herfindahl-Index, the lower the competitive balance. According to the uncertainty of outcome-hypothesis (Rottenberg, 1956), higher competitive balance increases fan interest.

The control variable BUTT denotes a dummy variable for the goalkeeper Hans-Jörg Butt. Butt is a peculiarity in German soccer because he scored 23 goals in the

10

However, Buraimo, Forrest and Simmons (2006) state that the metropolitan statistical area may be a flawed measure for market size. For example, if a club is located in a town with twice the population of another, it cannot be considered as having double market size. The bigger the town, the higher the mean travel costs for residents to reach the stadium, implying that the ticket demand will not linearly increase with the size of the home town. Therefore, Forrest et al. (2002) or Buraimo et al. (2006) employed modern GIS software to measure population within a certain distance from the stadium and also included a measure of competition from neighboring clubs. Unfortunately, I could not obtain the corresponding data for Germany.

11

Stollenwerk (1996) shows that the share of women among spectators in Bundesliga matches usually varies between 3% and 18%.

12

This further enables me to control for differences in relative shares of men in the population.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

52

considered time period – all through penalties. In Table 6 the set of variables as well as descriptive statistics are listed.

Table 6 indicates that teams with superstars have 50.7% more home attendance and 13.9% more attendance on the road than teams with a local hero. While a local hero accounts for almost three times as many goals and assists, a superstar accounts for even more than 6 times as many goals and assists as the league average of players in the same position. Superstars enjoy high popularity. They have 292 citations in the German press whereas a local hero is mentioned 155 times on average.

Variable

Description

Teams with Local Heroes

Teams with Superstars

Mean

SD

Mean

SD

13.03 13.18

0.35 0.10

13.44 13.31

0.39 0.12

Dependent variables LNATTHOME LNATTAWAY

Logarithm of match attendance at home Logarithm of match attendance on the road

Independent variables Star-Performance SCRORELHWP SCORESSWP GKLH GKSS

Weighted goals and assists of a local hero within a team Weighted goals and assists of a superstar within a team Dummy = 1 if goalkeeper is a local hero Dummy = 1 if goalkeeper is a superstar

Star-Popularity MEDIALHP Average citations of a local hero in the German press (in 100) MEDIASSP Average citations of a superstar in the German press (in 100) Control variables REP20 Reputation: weighted average of final rankings in the past 20 years CAPACITY Aggregate seasonal stadium capacity (in 10'000) MEN Male population in the hometown (in 10'000) UNEMP Unemployment rate (in %) CB Competitive balance (Herfindahl-Index in %) BUTT Dummy = 1 if goalkeeper is Hansjörg Butt

2.88 0.04

6.15

3.21

0.13

0.34

0.20 ___

1.55

___

2.75 ___

14.49 68.53 31.30 12.30 5.93 0.00

___

2.04 ___

12.33 27.65 33.22 3.79 0.10 0.00

2.92

2.25

44.10 84.57 36.83 12.08 5.92 0.01

26.25 31.55 40.16 3.85 0.11 0.09

Note: The model also includes a time trend which is not reported.

Table 6:

Variables and descriptive statistics

3.5.3 Estimation Approach Recall from above that the dataset contains all teams which played in the first Bundesliga during the period 1995/96 – 2003/04. It is well known that panel data

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

53

structures like ours require special econometric modeling, namely fixed-effects or random-effects. I choose the fixed-effects models as estimation approach. Let me first quickly restate the underlying assumptions of these models in order to show why I believe this choice to be appropriate for my analysis. The fixed-effects model assumes the following specification: yit

D i  x'it E  H it ,

(2)

where xit is a K-dimensional vector of explanatory variables. Unlike the fixed-effects model, the random-effects model does not allow the fixed effects ( D i ) and the regressors to be correlated, i.e. cov( D i , xit )

0; t = 0,1,.., T . However, in my empirical

setting this assumption does not seem reasonable. The fact that Bayern Munich may always have higher attendance than Hansa Rostock can be expected to be correlated with some regressors. For example, higher attendance might come from a higher degree of continuity in different fan generations. This might well be correlated with the team’s reputation. Thus, I expect the fixed-effects model to provide superior performance. This reasoning is supported by the empirical results of the Hausman specification test.13

Besides this specification test, I also test for strict exogeneity of the regressors where I take the results from the Hausman test into account. Following Wooldridge (2002), I specify the following regression equation: yi

D i  x'it E  w'it 1 G  H it ,

(3)

where w'it 1 denotes a subset of x'it for club i in the subsequent year t  1 . A test of the null hypothesis of strict exogeneity is equivalent to testing H 0 : G

0 . First I have to

choose the relevant elements of w'it 1 . Here, it is crucial to analyze for which regressors future values might be correlated with H it . Therefore, I decided to include all of the

13

Performing the Hausman specification test (Hausman, 1978) which compares the fixed-effects model with the random-effects model, we can reject the null hypothesis for home and away games on a 1% level of significance. The Hausman specification test, therefore, confirms that team-level effects are inadequately modeled by a random-effects model since they are correlated with the explanatory variables.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

54

regressors except for MEN and UNEMP.14 For the latter two, it seems highly implausible that a shock in current match attendance should be correlated with future values for inhabitants or the unemployment rate. Based on the specification from equation (3), I am not able to reject the null hypothesis on the 10% level of significance for both attendances at home and on the road.15 Thus, the regressors may be considered adequate.

A final aspect lies in the nature of my data set: Due to promotion and relegation we have an unbalanced panel, because some teams do not always play in the first Bundesliga. Since the reason why a team gets promoted or relegated (called attrition) is not random and therefore expected to be correlated with the idiosyncratic error – those unobserved factors that change over time and affect match attendance – resulting sample selection possibly causes biased estimators. However, through my choice of a fixed-effects model this problem is already moderated as fixed-effects analysis allows for the attrition to be correlated with the unobserved effect (Wooldridge, 2003). Therefore, I explore star attraction only within one team. While model 1 estimates the influence of star attraction on home game attendance, model 2 measures star attraction of local heroes and superstars on the road.

3.5.4 Results Table 7 shows all the ȕ-coefficients, the estimated White-robust standard errors as well as the levels of significance of both home and away games.16

14

For example, future values of media coverage might be correlated with current shocks in match attendance for the following reason: A positive shock in match attendance might lead the media to increase their coverage of a team’s players as they could expect a higher interest from consumers. Similar relationships could be derived for other regressors.

15

If a broader definition of superstars is chosen (see section about the robustness checks), the null hypothesis of strict exogeneity is rejected at the 5% significance level. I take this as an additional confirmation of a rather narrow definition of stars.

16

Whenever correlational designs are used, concerns about internal validity such as possible reverse causality may be raised. However, the issue of reverse causality (impact of revenues on the number of stars a team engages) is appeased by the lag structure of our model. I identify the stars in the beginning of a season whereas the performance, popularity, and attendance data is collected during the season.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

55

Table 7 shows that superstar performance significantly increases both home game and away game attendance. If a superstar scores one more goals than the average of players in the same position, match attendance at home increases by 1.4% and on the road by 0.7%.17 A local hero draws viewers into the home stadium by his popularity, but he does not have a nationwide appeal. MEDIALH does not significantly increase attendance on the road. An evaluation of the marginal effects E SCORESSWP and E MEDIALHP at the corresponding mean values for SCORESSWP and MEDIALHP shows that based on pure star attraction, superstars have a greater impact on home attendance than local heroes.18

17

The fact that the scores of superstars even increase attendance on the road may be puzzling at first glance. On the one hand, match attendance in German soccer is generally dominated by home team supporters who want “their” team to win (see e.g. Borland and MacDonald (2003) for the overwhelming evidence that attendance is positively related to home-team winning percentage). The more goals the opposite stars score, the lower the winning probability of the home team becomes, which then decreases demand. On the other hand, a greater value of SCORESSWP should be related with a higher expected match quality, which would increase match attendance. Based on the results in Table 7, it seems as if the latter effect dominates. I view this result to be in line with previous results that show the dominating influence of team quality variables for match attendance (see Garcia and Rodriguez (2002) for evidence from Spanish soccer). In the NBA, Berri and Schmidt (2006) found a positive impact of the winning percentage of the visiting team on attendance on the road.

18

The average marginal effect of SCORESSWP is 8.4% (0.0137*6.15) whereas the popularity of the local hero increases home attendance only by 1.1% (0.007*1.55).

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

56

Match Attendance at Home

Match Attendance on the Road

Variable

ȕ-coef.

SCORELHWP SCORESSWP GKLH GKSS

-0.0046 0.0137 ** -0.0549 -0.0606

0.0071 0.0043 0.0939 0.5220

0.0023 0.0066 ** -0.0137 -0.4785

0.0028 0.0028 0.0386 0.0357

MEDIALHP MEDIASSP

0.0070 * -0.0127

0.0041 0.0091

0.0032 -0.0036

0.0043 0.0049

REP20 CAPACITY MEN UNEMP CB YEAR BUTT

-0.0048 0.0070 ** 0.0493 0.0113 0.0960 0.0114 * 0.2757 *

0.0025 0.0008 0.0294 0.0091 0.1000 0.0056 0.1295

0.0013 0.0133 ** -0.0130 0.0109 * 0.0797 0.0151 ** -0.0254

0.0014 0.0022 0.0087 0.0050 0.0476 0.0027 0.0431

2

R within F statistic Number of observations

0.39 12.63 162

Std.Error

ȕ-coef.

Std.Error

0.62 24.68 162

Note: Significance levels: * 5% ; ** 1%; Significance tests are one-tailed for directional independent variables and two-tailed for control variables. Standard errors are Whiteheteroscedasticity robust standard errors.

Table 7:

Estimates of a team’s star attraction (2% superstar definition)

Concerning the control variables we see that German soccer enjoys increasing fan interest. The data delivers a significant positive time trend. The aggregate seasonal capacity strongly influences match attendance at home and on the road. The unemployment rate in the home town is positively related to attendance on the road. It seems that the lower opportunity costs of unemployed persons dominate over the negative income effect. But this only applies to away games where travel time is considerably higher. The greatest, statistically significant impact is derived for HansJörg Butt concerning home gate attendance. It seems that the peculiarity of a goalkeeper shooting penalties is an exciting and thus viewer drawing spectacular.

The pooling of both superstar and local hero teams allow a direct comparison of the superstar versus local hero coefficients. The null hypothesis of an equivalent star attraction of a superstar and of a local hero is rejected both for the performance and the

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

57

popularity variables. This indicates that these two star groups significantly differ in the way they activate fan interest.19

3.6 Robustness Analysis Within this section I perform several robustness tests. Robustness is analyzed with respect to two different aspects: (1) the number of stars per team and (2) different measures for star performance.

3.6.1 Increasing the Number of Superstars and Local Heroes A natural starting point for robustness checks is to ask whether the estimation results would be affected by a change in the number of superstars (or local heroes) employed by the teams. In the literature there exist a lot of different star definitions: In studies of the NBA, superstars are often defined as players who have made the All-Pro team or the All-Star Game for certain times (Scott et al., 1985; Brown et al., 1991; Burdekin & Idson, 1991; Berri et al., 2004, Berri & Schmidt, 2006). Hausman and Leonard (1997) defined two players only, Michael Jordan and Shaquille O’Neal, as superstars for whom they assumed a positive externality on the attendance of other clubs. In soccer, Lucifora and Simmons (2003) defined a superstar as a player who scored more than 0.25 goals per game. Sherwin Rosen, who is seen as the founder of the “economics of superstars”, however, defined superstars as “the relatively small numbers of people who earn enormous amounts of money and dominate the activities in which they engage” (Rosen, 1981, p. 845). Obviously Rosen (1981) bases his definition of superstardom on the distribution of earnings among the suppliers of a certain good or service. However, Rosen does not propose a clear percentage number as “boundary” between “normal” suppliers and superstars. Therefore, I decided to shift the “boundary” in order to increase the number of both superstars and local heroes for a sensitivity analysis. In the first alternative model, superstars are defined as the 5% most valuable players in the

19

Slope equality of SCORELHWP and SCORESSWP (MEDIALHP and MEDIASSP) is rejected on the 5% level of significance.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

58

league. The two most expensive players in teams without superstars are denoted as local heroes.

Match Attendance at Home Variable

ȕ-coef.

Std.Error

SCORELHWP SCORESSWP GKLH GKSS

-0.0022 0.0042 0.0392 0.0387

0.0062 0.0052 0.0626 0.0871

MEDIALHP MEDIASSP

0.0262 * 0.0091

0.0129 0.0113

REP20 CAPACITY MEN UNEMP CB YEAR BUTT

-0.0054 * 0.0058 ** 0.0456 0.0125 0.0618 0.0044 0.1815

0.0023 0.0011 0.0259 0.0093 0.1043 0.0061 0.1079

2

R within F statistic Number of observations

0.34 6.83 162

Match Attendance on the Road ȕ-coef.

Std.Error

-0.0012 0.0005 -0.0095 -0.0461 * 0.0168 ** 0.0026 0.0012 0.0131 -0.0101 0.0113 0.0861 0.0124 -0.0417

** * * **

0.0028 0.0022 0.0193 0.0217 0.0047 0.0056 0.0013 0.0023 0.0080 0.0056 0.0432 0.0028 0.0218

0.60 31.32 162

Note: Significance levels: * 5% ; ** 1%; Significance tests are one-tailed for directional independent variables and two-tailed for control variables. Standard errors are Whiteheteroscedasticity robust standard errors.

Table 8:

Estimates of a team’s star attraction (5% superstar definition)

Finally, I defined the 8% most expensive players in the league as superstars. For this broader definition of stardom, the number of local heroes per team without superstars corresponds to three. The estimations from this specification are displayed in Table 9.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

Match Attendance at Home

Match Attendance on the Road

Variable

ȕ-coef.

Std.Error

ȕ-coef.

Std.Error

SCORELHWP SCORESSWP GKLH GKSS

-0.0069 0.0105 * 0.0227 -0.0246

0.0107 0.0061 0.0385 0.0351

0.0006 0.0030 -0.0195 0.0008

0.0042 0.0025 0.0141 0.0177

0.0675 0.0065

0.0505 0.0110

0.0352 * 0.0067

0.0174 0.0050

0.0021 0.0011 0.0245 0.0087 0.1043 0.0059 0.0868

0.0007 0.0138 ** -0.0120 0.0098 0.0587 0.0115 ** -0.0426

0.0012 0.0023 0.0076 0.0056 0.0436 0.0028 0.0246

MEDIALHP MEDIASSP

-0.0052 * 0.0062 ** 0.0483 0.0120 0.0556 0.0056 0.2599 **

REP20 CAPACITY MEN UNEMP CB YEAR BUTT 2

R within F statistic Number of observations

0.37 7.31 162

59

0.60 31.00 162

Note: Significance levels: * 5% ; ** 1%; Significance tests are one-tailed for directional independent variables and two-tailed for control variables. Standard errors are Whiteheteroscedasticity robust standard errors.

Table 9:

Estimates of a team’s star attraction (8% superstar definition)

If the superstar definition is modified to encompass 5% or even 8%, the field performance of superstars does no longer significantly increase attendance (except for home games in the 8% superstar definition20). The more balanced talent distribution in top teams could offer an explanation for this finding: Bayern Munich for example engaged 4 to 12 (5% model) or even 7 to 14 (8% model) superstars. It is easier for the 2% most expensive players to stick out than for a larger group of superstars who sometimes even have to compete to be in the starting squad. By enlarging the number of players covered by the superstar definition, the (average) talent of superstars looses its distinctive feature to differentiate them from the rest of the team.

20

However, based on test results on joint significance, see Table 10, I do not put too much emphasis on this finding. The same applies to the negative coefficient of GKSS in Table 8.

60

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

The popularity variable of local heroes is not very robust. In the original model MEDIALHP has a significant positive impact on home game attendance. At the 5% superstar level it increases attendance both for home and away games. Concerning the 8% model only attendance on the road is significantly influenced by the popularity of local heroes. However, the fact that local heroes draw viewers by popularity and not by outstanding talent is unambiguous.

High robustness is seen for the control variable YEAR referring to attendance on the road and for CAPACITY. These variables have positive coefficients at the 1% significance level regardless of the exact star definitions.

A test of joint significance of performance variables, i.e. SCORESSWP and GKSS, confirms that superstar performances do not affect match attendance any more in the 5% and 8% superstar models. As can be seen from Table 10, this result is independent of the location of the match. This finding may be driven by the fact that a larger basis of “superstars” results in the inclusion of non-superstars, which leads to the insignificance of the performance measures.

Match Attendance at Home F-statistic

P-value

Match Attendance on the Road F-statistic

P-value

Main model (2% superstars) Joint signficance of a local hero's performance Joint significance of superstar's performance

0.35 7.05 **

0.7060 0.0010

0.63 8.40 **

0.5366 0.0004

0.27 0.38

0.7609 0.6818

0.17 2.57

0.8417 0.0810

0.45 1.68

0.6387 0.1909

1.04 0.72

Alternative models 5% superstar definition/two local heroes Joint signficance of a local hero's performance Joint significance of superstar's performance

+

8% superstar definition/three local heroes Joint signficance of a local hero's performance Joint significance of superstar's performance

0.3564 0.4872

+

Note: Significance levels: 10%, * 5% ; ** 1%.

Table 10:

Joint significance tests

However, the fact that a broader range for the superstar definition should result in flawed estimations of star attraction is also mirrored in the specification tests about

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

61

strict exogeneity. Whereas we could not reject the assumption of strict exogeneity in the 2% definition, the same does not hold for the broader star definitions. In both cases, strict exogeneity is rejected concerning match attendance on the road. Thus, I am extremely careful about implications from these estimates. In particular, it seems as if serial correlation is introduced to the away attendance model by a move towards a broader star definition. I consider this as further support for my choice of a narrow star definition.

Having discussed the consequences of relying on varying specifications, we now turn to a robustness check of the applied measures themselves.

3.6.2 Alternative Measures for Star Performance The choice of the position-weighted scores (measured by the sum of goals and assists) was motivated by the ease of evaluating goals and assists. As already said, the talent of soccer players is rather blurred, because the game specific characteristics complicate the measurement of individual performance. Therefore, I decided to apply another performance measure in order to test the robustness of the performance variables. On every match day the Kicker soccer magazine publishes the “team of the day” comprehending the 11 players who played best. I estimated a model which includes the number of appearances of a superstar or local hero in the BEST11 team. An important advantage of this measure is that players of each tactical position have similar chances to be elected for the BEST11 team. However, a disadvantage of the BEST11 variable is the fact that it is published by the same source that estimates the players’ market values.

Table 11 shows the estimates of a team’s star attraction using the BEST11 variable instead of scores. As in the original model, superstars are defined as the 2% quantile in the league’s market value distribution whereas the most valuable player in teams without superstars counts as local hero.

62

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

Match Attendance at Home

Match Attendance on the Road

Variable

ȕ-coef.

BEST11LH BEST11SS GKLH GKSS

-0.0038 0.0212 ** -0.0392 -0.0747

0.0069 0.0068 0.0971 0.0533

-0.0030 0.0153 ** -0.0179 -0.0311

0.0040 0.0037 0.0277 0.0276

MEDIALHP MEDIASSP

0.0065 -0.0092

0.0046 0.0087

0.0037 -0.0080 *

0.0032 0.0042

REP20 CAPACITY MEN UNEMP CB YEAR BUTT

-0.0059 * 0.0067 ** 0.0532 0.0129 0.0731 0.0104 0.2557

0.0027 0.0008 0.0280 0.0093 0.1014 0.0055 0.1312

0.0002 0.0132 ** -0.0092 0.0113 0.0738 0.0152 ** -0.0336

0.0013 0.0020 0.0071 0.0049 0.0418 0.0024 0.0312

2

R within F statistic Number of observations

0.37 13.13 162

Std.Error

ȕ-coef.

Std.Error

0.65 26.35 162

Note: Significance levels: * 5% ; ** 1%; Significance tests are one-tailed for directional independent variables and two-tailed for control variables. Standard errors are Whiteheteroscedasticity robust standard errors.

Table 11:

Estimates of a team’s star attraction using BEST11 as performance indicator

Table 11 illustrates that the average number of nominations of superstars for the “team of the day” significantly influences their team’s gate attendance both at home and on the road. This confirms the results from Table 7. The use of another performance measure does not change the finding that only superstars are able to draw spectators by their outstanding field performance. However, analyzing the popularity variables we see that the results change. MEDIALHP no longer has statistical significance and the press citations of superstars decrease attendance on the road. The latter effect may be explained from the ceteris paribus interpretation of the coefficient estimates: The negative coefficient on MEDIASSP refers to a change of media citations while holding the number of nominations for the “team of the day” constant. The results in Table 11 indicate that non-performance related media citations of superstars lower attendance on

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

63

the road. It seems that publicity might not be attendance-improving per se.21 Alternatively, we could say that a superstar substituting media presence for performance will lower his positive externality for matches on the road.22

3.7 Conclusion Analyzing seasonal match attendance data I find evidence for star attraction in the first German soccer league. However, the exact channel of generating this attraction (by field performance or popularity) largely differs depending on firstly whether a player is a nationwide superstar or a local hero and secondly whether attendance at home or on the road is investigated. While superstars enhance attendance both at home and on the road, the star attraction of local heroes is limited to home games. Superstars attract fans by outstanding field performances, whereas local heroes facilitate fan support by mere popularity. Robustness tests reveal that the estimations of star attraction are influenced by the chosen star definition. If the superstar and local hero categories are extended to encompass larger number of players, i.e. the superstar category to account for the 5% or 8% (instead of 2%) most expensive players of the league, specific performance-related star attraction is no longer observed in the data. Superstardom is a “small number” phenomenon, just as postulated in the economic star literature (see Rosen, 1981).

The results indicate that superstars produce a positive externality for home teams when playing on the road. This is an important finding with respect to the question of who should bear the costs of paying these superstars (see also the recent paper by Berri and 21

The consumer response possibly differs between positive and negative publicity. For the latter see e.g. Dean (2004).

22

The reader might wonder why a similar reasoning would not apply to the specification including SCORESSWP. However, recall that the BEST11 measure accounts for a potential bias of SCORESSWP towards midfielders and strikers in spite of the standardization with respect to the player’s position. Thus, it is possible that an exceptional performance of a defender might be the reason for the increased number of media citations which would not automatically be reflected in an increase in SCORESSWP. As argued earlier, such a performance of a defender would most likely result in a nomination for the “team of the day”. In other words, for midfielders and strikers, SCORESSWP should be a much better predictor for a nomination for the “team of the day” than for defenders and goalkeepers. As a result, the impact of MEDIASSP differs for the BEST11 and SCORESSWP specifications, as, under the ceteris paribus condition, an increase of MEDIASSP might well refer to different types of media coverage in the two models.

Local Heroes and Superstars – An Empirical Analysis of Star Attraction

64

Schmidt, 2006). More precisely, this externality gives rise to a de-facto system of revenue sharing in the German Bundesliga between teams employing superstars and those who do not. Based on my results, the value of the average superstar externality was about € 430.00023 per season. Furthermore, the results show that there is a trade-off for a league between an increase of its level of competitive balance and allowing for dominant teams.

Regarding future research, I believe two aspects to be especially worthwhile being followed. Firstly, it would be interesting to see whether the positive externality imposed by visiting superstars varies with the number of superstars in the home team, which would allow for a more precise quantification of the value of this externality. A second aspect relates more directly to the superstars themselves: Although this study provides new evidence for different types of star attraction in German soccer, it does not explicitly address transitions between local heroes and superstars. I am not able to link individual career paths with the team’s financial or field success. Hence, it still remains to be examined how player-specific star attraction changes as rising stars climb the career ladder.

23

Superstars increase match attendance on the road by 4%, which results in 24’484 tickets sold additionally. Given an average admission prize of € 17.50, this sums up to € 428’470.

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

65

4 Superstar Earnings in Soccer – Are Voluntary Salary Cap Agreements Self-Enforcing?1

4.1 Introduction The 20 teams of the English Premier League paid € 1.093 billion in total for wages and salaries in the 2002/03 season (Jones & Boon, 2004). This enormous number might pose the question: How can it be justified that sport stars achieve earnings so obviously disproportionate to those of other occupational groups, including doctors, nurses, teachers or police officers? At a first view the market for soccer players may be considered in the same way as any other labor market – by analyzing the interaction between supply and demand. In the neoclassical model, market mechanisms allocate players to clubs and determine wages by matching bids and offers for the players’ services. In doing so, the highest wage a profit-maximizing club is willing to offer equals the marginal revenue product of this player, namely the amount he would add to the club’s revenue if he was signed (Dobson & Goddard, 2001). And since soccer provides a technology with large economies of scales, potential star attraction of outstanding players is of high value. However, according to textbook economic theory a market solution is only Pareto efficient2 if externalities, monopolistic power or public goods are inexistent. In the following I will argue that the players’ earnings in professional team sports are inefficiently high due to positional externalities inherent in the league tournament. 1

A shorter version of this chapter was published in the European Sport Management Quarterly (see Dietl, Franck, & Nüesch, 2006).

2

Market equilibrium is Pareto efficient if it is impossible to make one person better off without making at least some others worse off (Frank, 2003).

66

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

In association football rewards are largely determined by relative performance. Successful clubs do not only receive a higher share of the broadcasting revenue3, they are also more likely to qualify for the very lucrative UEFA Champions League that provides significant extra-money. In the season 2004/05 the qualified clubs received in total € 414.1 million broadcasting income and generated substantial extra matchday turnover (Jones & Boon, 2005). Clubs increase their probability of winning by investing in playing talent.4 However, any action – for example hiring a star player – that increases one contestant’s chances of winning must necessarily reduce the chances of others. Positional externalities lead to “rat races” (Akerlof, 1976) or “positional arms races” (Frank, 2003) in which the economic agents exhibit a tendency to overinvest. Clubs compete for the “cheese” by poaching superstars and paying comparably higher salaries. However, not considering the externality of reducing its rivals’ revenues, clubs tend to overinvest and give rise to inefficiently high salaries.5

In almost any European football league annual growth rates of player salaries exceeded annual growth rates of revenues during the last decade. In the English Premier League the wage/turnover ratio increased from 50% in 1995 to 62% in 2001. During the same period it rose from 57% to 90% in Italy’s Serie A (Jones & Boon, 2003).6 In order to keep up with this general trend, a lot of clubs were forced to increase their liabilities. The accumulated debt of all Italian Serie A clubs, for example, increased to approximately € 2.5 billion. This compromised profitability. Italian teams generated a cumulated loss of € 1.2 billion between 1995/96 and 2002/03. French teams lost € 0.3 billion in the same period (Jones & Boon, 2004). Several clubs, for example Fiorentina, Servette Geneva, SW Bregenz, Lausanne Sports, and FC Lugano went bankrupt and

3

25% of the broadcasting income of the English Premier League and 50% of the broadcasting revenue in the German Bundesliga is distributed on the basis of league rank (Szymanski & Késenne, 2004; Feess & Muehlheusser, 2003). In Italy and Spain clubs individually negotiate broadcast agreements for their home matches, which leads to an even stronger concentration of rewards among the most successful clubs.

4

Michie and Oughton (2004) calculated a strong positive influence of relative wage expenditures on the championship performance of clubs in the English Premier League.

5

Strictly speaking, this negative externality of winning due to the talent superiority of superstars must be set against a potential positive externality of star attraction on the road (see chapter 3).

6

Of course, not only superstars but regular players as well load the wage bill.

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

67

numerous other clubs, such as Leeds United, FC Sevilla, Borussia Dortmund, AS Rome, AC Parma, and Lazio Rome are or were close to bankruptcy. Bohemians Prague could only be bailed out because fans donated more than € 100’000.

Whitney (1993) suggests that the labor market for professional sports athletes is subject to “destructive competition” which drives some participants out of the market even though this is inefficient for the league as a whole. Overinvestment does not imply that clubs will necessarily go bankrupt. It is well-known that some European clubs like Manchester United or Bayern Munich are rather profitable. Theoretically, overinvestment only means that clubs will deviate from joint profit-maximization in the Nash equilibrium7 emerging in unrestricted contests (Dietl, Franck, & Roy, 2003; Dietl, Franck, & Lang, 2005). In this sense, overinvestment may contribute to the financial troubles of many clubs. They would be better off financially if they invested less than in the Nash equilibrium.

In November 2002 the leading European football clubs, organized as the so-called G-14, reacted to the increasing player salaries by signing a voluntary agreement to limit (annual) salaries to 70% of (annual) revenues (Késenne, 2003). Salary caps are a novelty in Europe. In the US, where the first salary cap was introduced in 1984 by the National Basketball Association, clubs and leagues have more experience with salary caps. In the US case the existing salary cap agreements are backed up by some form of centralized enforcement mechanism. There is evidence that these enforcement mechanisms are not perfectly effective in practice because clubs still succeed in circumventing the imposed limitations for example through deferred or unreported payments (Fort & Quirk, 1995; Staudohar, 1998). Salary caps may be considered as incomplete contracts. A breach of an incomplete contract is observable by insiders but cannot be verified by a court of law (Hart, 1988).

In contrast to the US case, the G-14 did not foresee any kind of enforcement mechanism at all, but decided to entirely rely on self-enforcement instead. The G-14 salary cap

7

See Appendix 1 for the definition of the Nash equilibrium.

68

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

agreement raises different questions. Firstly, since it is not a collective bargaining agreement it may not hold under European competition law (Heermann, 2004). Secondly, the G-14 represents only a small fraction of all European football clubs. The effectiveness of the G-14 salary cap is in danger to become corrupted by other clubs which did not commit themselves to limit their salaries. Thirdly, since European football clubs compete at an international level, salary caps necessarily have to be Europe-wide. However, any Europe-wide system would face the obstacle of diversified living conditions, tax rates and administrative systems (Szymanski, 2003a). In addition to these and other practical obstacles the G-14 salary cap agreement faces a rather general challenge. Will self-enforcement be effective? This question would still remain, even if the other practical obstacles could be overcome. This chapter focuses entirely on the exploration of the preconditions for self-enforcing salary caps. If the probability of self-enforcement turns out to be very low given the specific situation of European football, it will not pay to invest anything in the solution of the more practical issues raised so far.

I develop a simple model of a professional sports league with two profit-maximizing clubs competing for an endogenously determined league prize. Based on the model I show that the self-enforcing character of salary caps increases with (1) the clubs’ valuation of future profits; and (2) the importance of competitive balance. The application of these findings to the situation encountered in European football leagues leads to the conclusion that the clubs are unlikely to honor a salary cap agreement.

The remainder of this chapter is organized as follows. Section 4.2 presents the existing literature on salary caps. Section 4.3 explains the model. Subsequently, the results are discussed in section 4.4. Section 4.5 concludes.

4.2 Related Literature on Salary Cap Agreements Staudohar (1998, 1999) explores the history of salary cap agreements in four US Major Leagues and points to numerous loopholes in the agreements which made them soft in reality. Fort and Quirk (1995), Vrooman (1995, 2000), and Késenne (2000) illustrate

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

69

different salary cap regulations in a Walrasian fixed-supply conjecture model (Szymanski, 2004). Looking at different cross-subsidization schemes, Fort and Quirk (1995) identify payroll caps in combination with a revenue sharing arrangement as an effective instrument for improving competitive balance in a league. However, since salary caps are inconsistent with league-wide revenue maximization, enforcement problems arise. In practice, creative accounting and deferred or unreported payments undermine salary cap agreements. Vrooman (1995, 2000) also uses a Walrasian model, however, with imperfect competition in the labor market. Therefore, salary caps may help the league to maximize the combined profits of all its teams at the expense of less competitive balance within the league. Késenne (2000) distinguishes between a payroll cap defined as a fixed percentage of the total league revenue divided by the number of teams and an individual cap. He shows that only the payroll cap increases competitive balance. In his model of asymmetric clubs, both salary cap regulations decrease total league revenue by impeding a market clearing remuneration and equal marginal player productivity. However, Késenne (2000) mentions that potential negative externalities would make salary caps necessary in order to maximize total league revenue.

Although the effects and the enforcement of salary caps have been recognized as important issues in the literature, they have never been analyzed in a Contest-Nash model. Since, in contrast to the Walrasian model, the supply of players is not fixed in the Contest-Nash model, it better fits the real situation in European football where several national leagues compete for the most talented players. While clubs are price takers in the Walrasian model (i.e. clubs cannot influence the salary level), in the Contest-Nash model they choose independently how many players to hire and how much to pay them (Szymanski, 2004). However, in the Nash equilibrium clubs do not maximize joint profits. The reason is that each team fails to internalize the externality of reducing its rival’s revenue by winning. If honored, voluntary salary caps would allow the clubs to correct these negative externalities. In this chapter I analyze under which conditions a voluntary salary cap agreement is self-enforcing. My theoretical approach is based on Telser (1980) and Bull (1987). They declare that non-contractual

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

70

agreements are enforceable in a repeated game-theoretic setting if the short-term gain from reneging8 does not exceed the associated long-term loss.

4.3 The Model The model consists of two identical profit-maximizing9 clubs, A and B . Each club decides how much to spend on player talent. In order to keep the model simple, I assume that each unit of money buys one unit of playing talent. Consequently, x A ( xB ) denotes the total amount of player salaries as well as the playing strength of club A ( B ). Both clubs decide simultaneously and non-cooperatively. The investment decisions of both clubs determine the league’s total revenue R according to the revenue function:

O x  x  O ( x  x ) , R( x , x ) t 0 2

R ( x A , xB )

1

A

B

2

A

B

A

B

(1)

O and O are parameters that indicate the relative importance of the league’s total 1

2

investments in player talent and of competitive balance, respectively. O1 and O2 have positive values. According to this revenue function total league revenue increases if (1) on an aggregate level teams spend more on player salaries and if (2) playing talent is more evenly distributed among teams.

The investment levels x A and xB do not only determine total league revenue. They also determine the probability that club A (or B ) will win the league’s championship. In particular, I assume that the winning probability of club A is given by the following logit contest function:

8

To renege means to fail to keep an agreement.

9

I assume profit-maximization as a best-case scenario for self-enforcing salary caps.

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

PA ( x A , xB )

xA x A  xB

71

(2)

The winning probability of the competitor B is:

PB ( x A , xB ) 1  PA ( x A , xB )

xB x A  xB

(3)

The profit function of club i  ^A, B` is given by:

3 i ( x A , xB ) D ˜ Pi ( x A , xB ) ˜ R( x A , xB )  (1  D ) ˜ (1  Pi ( x A , xB )) ˜ R( x A , xB )  xi

(4)

D  >0.5,1@ represents the portion of league revenue received by the winning club. While D 1 describes a “winner-takes-all” league, D

0.5 denotes perfect revenue sharing.

Firstly, I compute the league optimum (LO) which is the maximum of the aggregated profits 3 A and 3 B . The league optimum ( x ALO , xBLO ) is given by the following first-order condition: w3 i ( x ALO , xBLO ) wxi

O

1

2 2 xi

1 0

(5)

Note that the second-order condition is negative:

w 2 3 i ( x ALO , xBLO ) w 2 xi



2O1 8 xi

3

0

The optimal salary levels are therefore:

(6)

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

72

xiLO

O

2

(7)

1

8

In the league optimum each club generates an expected profit of:

3 iLO ( x ALO , xBLO )

2

O

(8)

1

8

Next, I compute the non-cooperative Nash solution in a one-shot interaction. Solving for the symmetric Nash equilibrium ( x ANE , xBNE ) I get:10

xiNE

O

2

1

32

(4D  1) 2

(9)

In the symmetric Nash equilibrium each club generates an expected profit of:

3 iNE ( x ANE , xBNE )

O

2

1

32

(16D 2  24D  5)

(10)

Comparing the Nash equilibrium with the league optimum, I derive the following results: In a non-cooperative setting, the league optimum will only be attained if

D

0.75 . If D ! 0.75 , the winning club receives more than three quarters of the league

revenue. This “winner-takes-most” regime creates strong incentives for both clubs to increase their winning probability by investing more in playing talent, i.e. attracting better players by paying higher salaries. As a result, both clubs will overinvest compared to the league optimum. If, on the other hand, D  0.75 , a large portion of the league revenue is shared regardless of field success. In this revenue sharing regime each club has strong 10

In the Appendix 1 the symmetric Nash equilibrium of equation 9 is explained in more detail.

73

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

incentives to free-ride on the other club’s talent investment. As a result, both clubs will underinvest compared to the league optimum.

If European football clubs face an overinvestment problem, the model needs to have an

D ! 0.75 . In this case, both clubs would benefit if they agreed to invest x

LO

i

instead of

xi ! xi . Such a salary cap agreement, however, is highly unstable. Each club has NE

LO

strong incentives to unilaterally break the agreement. In fact, in a one-shot interaction breaking the agreement strictly dominates honoring the agreement: regardless of what the other club does, breaking the agreement always results in higher expected profits than honoring the agreement. Formally, if club B honors the salary cap agreement, the optimal deviation investment level of club A ( x ADEV ) is determined by the following equation:

w3 A ( x ADEV , xBLO ) wx A

2

(2D  1)

>

O /8 O x  O / 8  O ( x  O / 8) ( x  O / 8) 1

2

A

2

2

A

1

1

2

A

2

1

2

@

1

Dx  (1  D )O / 8 ª O   2O ( x  O / 8) « x  O /8 «¬ 2 x  O / 8 2

A

2

1

1

2

A

1

2

A

1

2

A

1

2

º » 1 0 »¼

(11)

Given that D ! 0.75 , xiDEV exceeds xiLO and the deviation profit ( 3 iDEV ) exceeds 3 iLO . In a one-shot interaction both clubs will always cheat on the salary cap agreement and invest more than xiLO .

Can the league optimum be attained in a repeated setting? To answer this question I analyze whether the following trigger strategy constitutes a Nash equilibrium of the infinitely repeated game: honor the salary cap in the first period and continue to honor the agreement as long as the other player honored the agreement in the previous period. If the other player deviated in the previous period, never honor the agreement in any future period.

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

74

Suppose that club B follows this trigger strategy and honors the salary cap agreement by choosing xBLO in the first period. Club A then faces the following dilemma: It can maximize its first-period profit by deviating, i.e. by choosing x ADEV . If club A deviates in the first period, however, club B will never honor the agreement in any future period. The agreement will break down and both clubs cannot do any better than repeatedly play the non-cooperative one-shot game. Club A has to trade off the short-term gain  3 LO ) against the long-term loss of realizing 3 NE instead of 3 LO from deviating (3 DEV A A A A

in all future periods. Figure 3 highlights this trade off.

profit

3 DEV A

3 LO A 3 NE A

periods Figure 3:

Profits with and without honoring the salary cap agreement if D ! 0.75

Given that future profits are discounted by the factor G , salary cap agreements are selfenforcing if the following condition holds:

3 iLO

1 G t 3 iDEV  3 iNE 1 G 1 G

(12)

The left hand side of inequality (12) represents the present value of a club’s future profits given that both clubs honor the agreement. The right hand side is the sum of the singular profit from deviating and the present value of a club’s future profits given that both clubs will no longer honor the agreement once deviation has happened.

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

75

By rearranging (12) I conclude that in repeated interaction salary cap agreements are self-enforcing if the discount factor G is sufficiently high.

Gt

3 iDEV  3 iLO 3 iDEV  3 iNE

(13)

4.4 Discussion The discount factor G can be interpreted in three different ways: Firstly, the discount factor G

1 /(1  r ) is the present value of a money unit to be received one period later,

where r is the interest rate per period. Secondly, G can be used to reinterpret what is called an infinitely repeated game as a repeated game that ends after a random number of repetitions (Gibbons, 1992). The probability of future interactions, therefore, influences the discount factor. The likelihood of an interaction’s end may be affected by different factors like the financial health of the clubs, the strictness of the licensing procedure, or the openness of the league for example. Thirdly, G is influenced by the frequency of interaction. A more frequent interaction between A and B corresponds to an increase in G (Tirole, 1988). A low discount factor translates into a low likelihood of self-enforcing salary caps. In the following sections I will illustrate how the discount factor G may be influenced by two peculiarities of European football leagues.

In European football leagues the best teams from a lower league are promoted to the next higher league, while the weakest teams in the latter are demoted to the former. The composition of an open league changes annually through promotion and relegation. Promotion and relegation decrease the probability of future interaction resulting in a stronger devaluation of future profits (lower G ). Self-enforcing salary caps are ceteris paribus less likely in open leagues than in hermetic leagues.

76

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

A second peculiarity that influences the discount factor G is the restriction of transfer windows. The threat of punishment will only be effective if the punishment comes fairly soon after the deviant behavior. Theoretically, punishment might be delayed for two reasons: Firstly, clubs may notice the rival’s breach of agreement with a lag. Secondly, infrequent interaction delays the punishment and makes current deviation more attractive (Tirole, 1988). In professional team sports the second reason may be more important. The frequency of interaction is determined by the regulations governing transfer periods. European football leagues have introduced two transfer periods: one in the winter and one in the summer break. This implies that clubs can only change their rosters twice a year. If a club realizes that the competitor has broken the salary cap agreement in the winter break, it has to wait until the end of the championship race to increase its investments in player talent too. This infrequent interaction corresponds to a low G , which makes self-enforcing salary caps less likely. A third factor affecting the self-enforcing aspect of salary caps is the impact of competitive balance on league revenue. The uncertainty of outcome hypothesis (Rottenberg, 1956) – an increase of fan interest through even contests – is a well disputed issue (Szymanski, 2003a; Borland & MacDonald, 2003). My revenue function considers the controversy of the uncertainty of outcome hypothesis by introducing an undefined parameter O2 . If competitive balance strongly influences league revenue, O2 is high. A low O2 , however, represents a league whose fans do not care about even contests. The higher O2 is, the stronger the deviation profit 3 iDEV declines towards 3 iLO holding everything else equal. The fraction in inequality (13) decreases with an increase of O2 because 3 iNE and 3 iLO are not affected by the parameter O2 . Therefore, the stronger the influence of competitive balance on league revenue is, the lower the minimal discount factor for self-enforcing salary caps becomes. The self-enforcing character of salary caps increases with the importance of competitive balance. But how important is competitive balance for demand in European football?

All in all the empirical evidence in favor of the uncertainty of outcome hypothesis is far from convincing in European football as the cases treated in the review of Szymanski

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

77

(2003a) show. On purely theoretical grounds European football leagues should be able to deal with a greater imbalance of their teams than typical US Major Leagues without loosing fan interest. Due to the fact of promotion and relegation European leagues may capture fan interest by presenting two competitions simultaneously. Less endowed teams at the bottom of the league may activate fan interest by competing with each other against being relegated. At the same time the top teams compete to qualify for promotion to the next higher league or to international club competitions like the Champions League or the UEFA Cup. By providing several focal points for fan interest, European football leagues are less likely to become boring even if competitive imbalance is high. Since competitive balance is less important for European fans due to this peculiarity, O2 is lower and self-enforcing salary caps are less likely.

4.5 Conclusion Voluntary salary caps have to be self-enforcing in order to be effective in limiting player expenses. A repeated game-theoretic contest model with two identical profitmaximizing clubs indicates that the self-enforcing character of salary caps increases with the clubs’ discount factor and the importance of competitive balance.

A voluntary salary cap agreement is unlikely to be self-enforcing in European football for several reasons: Promotion and relegation decrease the probability of future interactions and, therefore, complicate self-enforcing contracts. Moreover, restricted transfer windows reduce the frequency of interaction, which translates into delayed punishments for cap breakers. Finally, simultaneous competitions in European football leagues provide entertainment even if imbalance within a league is high. Low importance of competitive balance for fan interest fosters deviant behavior.

Assuming identical profit-maximizing clubs, I built my model as a best-case scenario for self-enforcing salary caps. Since the result that European salary caps are unlikely to be self-enforcing is derived from a best-case model, the prediction seems even more plausible when additional difficulties for self-enforcement are considered. As is shown

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

78

in Appendix 2, differences in club productivity or cost functions make the establishment of self-enforcing salary caps even more difficult. Moreover, since European clubs are rather seen as winning-percentage-maximizers than profit-maximizers (see Sloane, 1971; Fort, 2000), they often do not mind paying high salaries and transfer fees as long as these expenditures promise field success.

4.6 Appendix 1: Nash Equilibrium in a One-Shot Interaction The Nash equilibrium is the combination of strategies in a game such that neither player has any incentive to change strategies given the strategy of his opponents. Each player’s choice is a best response to the strategies actually played by his rivals (Mas-Colell, Whinston, & Green, 1995).

Given the expenditure of club B ( xB ), the profit-maximizing salary level of club A ( x A ) is determined by:

NE

NE

w3 A ( xA , xB ) wxA

(2D  1) ˜

Dx  (1  D ) xB wR( xA , xB ) xB ˜ R ( x A , xB )  A ˜ 1 0 x A  xB ( x A  xB ) 2 wxA (A1)

The chosen salary level of club B is analogous:

NE

NE

w3 B ( xA , xB ) wxB

(2D  1) ˜

Dx  (1  D ) xA wR( xA , xB ) xA ˜ R ( x A , xB )  B ˜ 1 0 x A  xB ( xA  xB ) 2 wxB (A2)

Both equations (A1 & A2) can only be solved simultaneously if xA we have identical clubs, the Nash equilibrium is symmetric.

NE

NE

equals xB . Since

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

79

4.7 Appendix 2: Asymmetric Clubs Asymmetries among the economic agents complicate self-enforcing agreements in repeated interaction (Tirole, 1988). In the following I prove under additional simplifying assumptions that productivity differences hamper self-enforcing salary caps.11

Let two clubs ( A and B ) compete for an exogenous league revenue R . Since R does not depend on the league’s total investment in player talent, the league optimum is given by xi

LO

0 . In this simplified case every expenditure level above zero can be seen

as an overinvestment which reduces the club’s profits. Since the winning probability of club i  ^A, B` is not defined in the case of xi

0 , I assume that both clubs receive an

equal share of R in the league optimum and generate a profit of:

LO

LO

3 i ( x A , xB )

R 2

(A3)

To introduce club asymmetries, I formulate the winning probability of club A more generally:

PA ( xA , xB )

g ( xA ) g ( x A )  h ( xB )

(A4)

g ( x A ) and h( xB ) denote how efficiently club A and club B transform money into

playing talent. So far I have assumed the easiest case in which g ( xi )

h( xi )

analyze different transformation efficiencies. I define that g ( x A ) h ( xB )

11

xi . Now I

kxA and that

xB . The winning probability of club A is therefore:

The following model is an adapted version of the model presented by Dietl et al. (2003).

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

80

PA ( xA , xB )

kxA kxA  xB

(A5)

Analogous to the main model, the profit function of club i  ^A, B` is given by:

3 i ( xA , xB ) D ˜ Pi ( xA , xB ) ˜ R  (1  D ) ˜ (1  Pi ( xA , xB )) ˜ R  xi

(A6)

In a non-cooperative one-shot interaction the profit-maximizing expenditure level of club A ( x A ) is determined by the following first-order condition:

NE

NE

w3 A ( x A , xB ) wx A

>DR  (1  D ) R@ ˜

kxB 1 0 (kxA  xB ) 2

(A7)

The optimal salary level of club B ( xB ) is determined in the same way:

NE

NE

w3 B ( x A , xB ) wxB

>DR  (1  D ) R@ ˜

kxA 1 0 (kxA  xB ) 2

Both equations (A7) and (A8) can only be solved, if xA

(A8)

NE

NE

xB . The Nash equilibrium

is therefore:

xA

NE

xB

NE

>DR  (1  D ) R@ ˜

k (1  k ) 2

(A9)

Superstar Earnings – Are Voluntary Salary Cap Agreements Self-Enforcing?

81

The salary level in the Nash equilibrium is maximized if k equals 1. Therefore, the overinvestment is comparably lower and the club’s profit higher if club productivity is asymmetric (which means k z 1 ). Given that 3 i

asymmetries,

12

LO

and 3 i

DEV

are not influenced by club

I conclude that under the presented assumptions salary caps are less

likely to be self-enforcing if club productivities differ.

12

If a club honors the salary cap agreement and invests nothing, the optimal deviation investment level leads to winning probability of 1 regardless of the club’s productivity.

Superstars versus Celebrities – Big Man or Big Name?

83

5 Superstars versus Celebrities – Big Man or Big Name?

“The hero was a big man; the celebrity is a big name” (Boorstin, 1961, p. 61).

5.1 Introduction Information technology and mass media have opened the gates for a new type of stars: celebrities. Celebrities are individuals who are known for their well-knowness. Until recently, stars were considered as exceptionally gifted and highly talented individuals who earn enormous amounts of money. This paradigm has been challenged at the latest when several television casting shows like for example Big Brother experienced a boom. Through these pseudo-events anyone can become famous. It is no longer necessary to have demonstrated great talent, since fame itself has obtained tremendous commercial value. But why are millions of people spending a lot of time and money (for the voting procedure) to see ordinary1 people singing, dancing or just performing themselves?

Existing economic theories of superstar formation (Rosen, 1981; Adler, 1985, MacDonald, 1988) concentrate on the perceived quality of the star’s performance. While Rosen (1981) and MacDonald (1988) argue that superstars necessarily have superior talent, Adler (1985) states that popularity also enhances star emergence. According to Adler (1985) a star’s popularity facilitates the accumulation of

1

Evans and Wilson (1999, p. 1) speak about the “democratization“ of fame.

84

Superstars versus Celebrities – Big Man or Big Name?

consumption capital, which itself increases the valuation of a star’s performance. However, consumption capital has no value of its own and thus popularity cannot completely replace missing talent. Therefore, existing superstar theories fail to explain the occurrence of celebrities who enjoy enormous fame but may not have any special talent at all. In this section I suggest that social interaction does not only provide new consumption capital in the sense of Adler (1985), but that people directly benefit from interacting as well. Celebrities are well suited for “gossip consumption” (Gamson, 1994). In gossip, the pleasure comes from the exchange of stories, interpretation or judgments. Thanks to media promotion celebrities are known by nearly everyone and there is no danger of repercussion. The more popular a celebrity is, the easier gossip circulation becomes. Popularity breeds popularity resulting in “popularity-driven stars”. Such celebrities may be extremely famous independent of their (lack of) talent.

This chapter is organized as follows: First, economic theories of superstar formation and a simple model of superstars are presented. Subsequently I explain the emergence of celebrities using the concept of gossip. I illustrate the role of the media and provide a simple model of celebrities. Finally, I conclude.

5.2 Superstar Emergence 5.2.1 Economic Superstar Theories Sherwin Rosen (1981, p. 845) defines superstars as “relatively small numbers of people who earn enormous amounts of money and dominate the activities in which they engage.” Among superstar theorists it is indisputable that on the supply side the existence of a superstar is based on a technology that allows for joint-consumption. A superstar activity can easily be reproduced and the cost of production does not rise in proportion to the size of the seller’s market. Hence the marginal cost function

Superstars versus Celebrities – Big Man or Big Name?

85

necessarily decreases or at least remains constant with an increasing demand. Otherwise cost considerations would prevent any superstars to satisfy a high market share.2

However, it is controversial what determines the demand of superstar performances. Large economies of scale do not guarantee high salaries for a restricted number of stars unless the demand becomes highly concentrated on their services. In the following, I shortly introduce the three most prevalent theories of superstar demand by Sherwin Rosen, Moshe Adler, and Glenn MacDonald.

Rosen’s superstar theory is based on two basic premises: Firstly, lower quality is an imperfect substitute of higher quality. If a surgeon is 10 percent more successful in saving lives than his peers, most people would be willing to pay more than a 10 percent premium for his services. Secondly, talent or quality is costlessly identifiable and observable by all potential consumers. Therefore, given the large economies of scale on the supply side, small differences in talent are magnified into large differences in earnings. In Rosen’s model, a single superstar (or a single group of superstars) – the best –serves the whole market (Schulze, 2003).

The plausibility of Rosen’s assumptions largely depends on the sector or job in which a star is engaged. The performance of a 100 meter sprinter, for example, is clearly and unambiguously determined by the running time. The sprinter’s talent is easily identifiable and measurable. And in general people favor watching the finales in the Olympic Games rather than ten runs at mediocre levels. Concerning artistic activities, however, quality determination is a lot more difficult. Consumers have manifold tastes and their understanding of quality is highly diversified. While some people love the music of Madonna, others may hate it. Commonly accepted and clearly measurable talent indicators are often not available. Thus Rosen’s second assumption is less

2

The “personal scale of operations” explains why a soccer star for example earns a multiple of a school teacher, even if he or she is the best teacher in town. However, Rosen and Sanderson (2001) suggested that it is all in the technology. If a teacher used the Internet to personally teach millions of students at the same time, star teachers would earn at least as much as star athletes.

86

Superstars versus Celebrities – Big Man or Big Name?

plausible in arts. Hamlen (1991, 1994) or Salganik, Dodds, and Watts (2006) fail to find empirical evidence for Rosen’s superstar explanation in the popular music industry.

In contrast to Rosen (1981), Adler (1985) did no longer consider consumer preferences as time-invariant. In fact in Adler’s superstar theory, the appreciation of a star’s performance increases with the consumption capital of the consumer: “… the more you know the more you enjoy” (Adler, 1985, p. 208-209). The star specific consumption capital may be accumulated both by past consumption and by discussing the star’s performance with likewise knowledgeable individuals. Since the discussion is easier if all participants share common prior knowledge, a star’s popularity facilitates the accumulation of additional consumption capital. According to Adler (1985) popularity increases demand. The notion of “consumption capital” was introduced by Stigler and Becker (1977), who explained how past consumption activities may lead to beneficial addiction through an accumulation of specific knowledge. Stigler and Becker (1977) themselves referred to Marshall (1923) who had written: “… the more good music a man hears, the stronger is his taste for it likely to become.”3 When discussing the taste for “good” music, Alfred Marshall had probably some distinguished operas or classical music in mind. But is it also possible to accumulate consumption capital with respect to “bad” music? Is consumption capital concerning television reality show celebrities imaginable? How much does consumption capital depend on the quality of the star’s performance? These interesting questions are left unanswered.

Adler’s theory is based on the assumption that stars only exist where consumption requires knowledge. He drops Rosen’s second premise of perfectly observable talent. Otherwise knowledge would not be of any concern. Rosen’s first assumption, however, persists. In Adler’s star model talent or quality assessment still matters a lot. Popularity only indirectly feeds star attraction by simplifying the accumulation of consumption capital. But consumption capital has no value of its own; it only generates a benefit by increasing the valuation of the star’s performance. But the underlying quality of the

3

Original statement in Marshall (1923, p. 94) quoted in Stigler and Becker (1977, p. 78). The accentuation is introduced by the author.

Superstars versus Celebrities – Big Man or Big Name?

87

performance still depends on the star’s talent. Thus according to Adler (1985) even enormous popularity cannot completely replace missing talent.

MacDonald (1988) provides a dynamic version of Rosen’s superstar model by adding an information accumulating process about the performer’s talent. MacDonald (1988) describes how young artists, whose uncertainty of talent is high, perform to small audiences and earn net returns below what they could earn outside the industry. Since the quality of their performances is serially correlated, the knowledge of first-period reviews reduces the quality uncertainty. These reviews have predictive power for the second period’s performance. Those performers who have been recipients of good reviews stay in the industry, earn larger incomes and play to bigger crowds than before. The less fortunate young performers leave the industry. Overall, there are few superstars in the industry who serve a large fraction of the audience and obtain an even larger share of the returns. In line with Rosen’s model, MacDonald (1988) postulates earnings to be an increasing convex function of talent. However, he considers this function to have rather stochastic than deterministic properties. In the initial period a performer’s talent is characterized by high uncertainty. But through performing, useful information of the likely quality of a subsequent performance is obtained. Rosen’s second premise of costlessly observable talent is weakened. His first premise of a star’s talent superiority, however, is still a key element in MacDonald’s star model: Only bad luck may hinder the most talented performers to become superstars.

My short literature review shows that the existing theories of superstar formation concentrate on the perceived quality of the star’s performance. Therefore, they cannot explain the high attention and demand for celebrities who may not have any exceptional talent at all.

88

Superstars versus Celebrities – Big Man or Big Name?

5.2.2 A Simple Model of Superstars Assume consumers have identical preferences concerning the non-identical producers called artists.4 Following the existing superstar theories, the consumer’s utility function e

of an artist’s service shall be given by its consumption value v(t j , h( x j )) . The e

consumption value function v(t j , h( x j )) depends on the artist’s talent t j , which is perfectly observable, and on the consumer’s consumption capital ( h( x j ) )5 of a e

particular artist. It is taken to be twice continuously differentiable in both arguments. In line with Rosen (1981) I postulate that small differences in talent become magnified into large value differences near the top end of the scale. The consumption value function is thus convex in talent t j . As Adler (1985) postulated it is also positively e

influenced by the artist specific consumption capital h( x j ) . The higher the expected e

number of consumers ( x j ) of a particular performance, the easier it gets to accumulate additional consumption capital by discussing an artist’s performance with likewise e

e

knowledgeable individuals. Due to the positive network externalities, wh( x j ) / wx j is positive and – to keep it simple – assumed to be constant. The supply of an artist’s performance is characterized by large economies of scale. Since I model constant marginal costs g j and positive fix costs G j , average costs decrease with the number of consumers.

In the following I illustrate the value generated by an artist’s activity prone to superstar effects. Focusing on the value creation the price determination for the superstar activity looses importance. In my setting the price does not primarily influence the value creation but rather its redistribution, namely the division of total rent in a consumer and a producer surplus.6 Consumers generally choose the artist for which their consumer

4

The model applies to any other occupation prone to superstar effects like for instance to athletes, doctors, or managers.

5

Assuming homogenous individuals I neglect past consumption activities as additional source of consumption capital.

6

Of course, this requires an efficient price, which is disputable especially given the properties of a „natural monopoly“ (see e.g. Mas-Colell, Whinston, & Green, 1995).

89

Superstars versus Celebrities – Big Man or Big Name?

surplus is maximized. The artist who creates the highest value added is then also able to provide the highest consumer surplus and survives therefore in a competitive environment.

e

The value creation of an artist’s performance – denoted as 3 (t j , x j ) – is determined by:

e

e

3 (t j , x j )

e

v(t j , h( x j )) ˜ x j  G j  g j x j

e

(1)

Superstar services are provided as long as the aggregated consumption benefit exceeds total costs. To calculate the optimal market size of an artist, I differentiate the value e

e

creation function, 3 (t j , x j ) , with respect to the expected number of consumers ( x j ) and derive:

e

w3 (t j , x j ) e wx j

ª wv(t j , h( x j e )) wh( x j e ) º e e ˜ « » ˜ x j  v(t j , h( x j ))  g j e e w x w x j j ¬ ¼

(2)

As a logical consequence of positive network externalities and constant marginal costs, e

the marginal value creation function of an artist’s performance increases in x j :

e

w 2 3 (t j , x j ) e w2 xj

2

e e ª wv(t j , h( x j e )) wh( x j e ) wv(t j , h( x j e )) w 2 h( x j e ) º e wv(t j , h( x j )) wh( x j ) ˜ ˜  ˜ 2 e » ˜ xj  2 ˜ « e e e e e wx j wx j w x j »¼ wx j wx j wx j «¬ e e e ª wh( x j ) e º wv(t j , h( x j )) wh( x j ) ˜ x j  2» ˜ !0 ˜ « e e e w wx j x wx j j ¬ ¼

(3)

Hence, one artist should serve the whole market, because the existence of multiple artists competing with each other is inefficient. However, which artist will succeed in becoming a superstar?

90

Superstars versus Celebrities – Big Man or Big Name?

Let’s first assume a situation with two equally talented artists. Differences in the perceived quality of the performance only depend on differentials of the star specific e

consumption capital h( x j ) and, therefore, on the expected number of consumers of an artist’s service. Consumers simultaneously decide with respect to the expected “fan community” of an artist.7 The artist who has slightly higher expected popularity will snowball into a superstar.

The situation changes when star heterogeneity is introduced. If an artist’s talent is unambiguously distinguishable, consumers maximize their utility in adopting the most talented artist, since he or she is expected to have the highest “fan community”. Thus, higher talent t j comes along with higher popularity, which clearly leads to a superior consumption value. Consequently, the more talented artist will be leveraged to a superstar. I conclude that superstars emerge in the combination of exceptionally high talent and large economies of scale. Celebrities, however, somehow manage to catch high attention without outstanding talent. In the following section an explanation of this phenomenon is offered.

5.3 Celebrity Emergence Boorstin (1961, p. 57) defines a celebrity as “a person who is known for his wellknowness”. While a superstar is distinguished by some kind of special achievement; the celebrity is characterized by an image, fame or a trademark. Superstars all share admirable qualities – qualities that somehow set them apart form the rest of us – whereas celebrities need not do anything special (Monaco, 1978). Intrinsic to the meaning of celebrities is the fact that their well-knowness has become a viable commodity all by itself. It can stand by itself, independent of accomplishment, heroics, or talent (Rein, Kotler, Hamlin, & Stoller, 2006).

7

Thus expectation management becomes crucial. In general expectation management is critical whenever the services themselves are not clearly distinguishable. In a very real sense, a new artist who is expected to become a star will become a star. Self-fulfilling expectations are a typical manifestation of bandwagon effects (Shapiro & Varian, 1999).

Superstars versus Celebrities – Big Man or Big Name?

91

5.3.1 “Gossip Consumption” While in Adler’s sense discussion with knowledgeable friends increases the utility of consuming a star’s performance due to higher consumption capital, I propose that people directly benefit from interacting. This means that the pure consumption benefit is only part of the total benefit. Talking about a celebrity with friends, workmates, or acquaintances generates additional value for those involved.8

Gamson (1994) names this interaction benefit as “gossip consumption”. For gossip it does not matter how celebrities got there, or even how they manage to stay there. “In gossip pleasure comes from the activity of circulating information and forming evaluations” (Gamson, 1994, p. 175). The pleasure lies in the exchange, in the development of new story lines, in discussing, sharing interpretations or judgments. It is not necessary for gossip consumption that the information is demonstrably true; in fact, too much truth can stop the gossip game (Gamson, 1994). Celebrities are in many ways better objects for this game than other people like e.g. neighbors. Celebrities are like neighbors whom nearly everybody knows, in nearly every social setting, and news about them is easier to find and share than information about friends or colleagues. More important, celebrity gossip is a much freer realm than acquaintance gossip: there is no danger of repercussions and accountability (Gamson, 1994).

The interaction benefit increases with the number of people knowing the tidings of a particular celebrity. The activity of discussion, story telling, interpretation, or judgment is typically subject to network externalities. The more popular a celebrity is, the easier gossip circulation becomes. The interaction benefit is, therefore, an increasing function of the celebrity’s popularity. This creates a self-energizing virtuous cycle: a celebrity with a large and popular “fan community” becomes more and more valuable to each fan as he attracts ever more fans. Leibenstein (1950) named the observation that people often follow the crowd as “bandwagon effect”. The bandwagon effect emerges if

8

Frank and Cook (1995, p. 34) shortly addressed this point, writing that: „(…) one valuable part of the experience of reading a book is discussing it with a friend who has also read it. (…) Similar considerations apply to movies, plays, music, spectator sports, and a host of other interactive consumer activities.“

92

Superstars versus Celebrities – Big Man or Big Name?

people’s valuations of a commodity (and thus demand for that good) increase when they observe others consuming the same commodity. Banerjee (1992, p. 798) defines this herd behavior as “everyone doing what everyone else is doing.” Individuals decide whether or not to follow a rising celebrity depending on the number of people currently paying attention to this person. The more popular a new celebrity is expected to be, the more valuable she becomes for others, and this fuels further popularity in a virtuous cycle. Popular support for an individual artist, athlete or personality may thus suddenly gain momentum and escalate.

5.3.2 The Role of Media in Celebrity Emergence Since the prerequisite for fame no longer is high birth, or the gift of great talent, or some valiant deed, the first step of celebrity emergence consists of nothing but somehow finding one’s way into the media (Franck, 1998). Anyone can become a celebrity, if only he or she gets into the news. Before the “graphic revolution” it was generally necessary to have demonstrated great deed or action in order to become well-known. With the development of mass-media the production and dissemination of fame began to be manufactured by the media. Recent information technology has further enhanced the capacity to deliver vivid images of individuals in real-time around the globe. Satellite and digital television, computer technology and the Internet have considerably extended the capacity to make, transmit and disseminate images of celebrities (Smart, 2005). “We can fabricate fame, we can at will … make a man or woman well known; but we cannot make him great. We can make a celebrity, but we can never make a hero. … The hero created himself; the celebrity is created by the media” (Boorstin, 1961, p. 48+61). The power of the media lies in the decision for whom it triggers the bandwagon effect of popularity. Superstars manage to catch attention by their superior talent. The higher quality of a superstar’s performance suffices as selection criteria. Celebrities, however, may not have any qualities that set themselves apart. According to Boorstin (1961) they are superficial, trivial, bereft of distinction and, therefore, in a constant battle of attracting or maintaining the media’s attention. Typically celebrity status is fleeting and

Superstars versus Celebrities – Big Man or Big Name?

93

needs to be continually regenerated in order to remain in the public eye. Celebrities are destined to disappear and to be quickly replaced (Smart, 2005). “Whereas superstars emerge with the passage of time, through a process of gestation in which their feats have to withstand the test of time, celebrity is forever ‘now’, by definition contemporary. Celebrity is forged through media attention, through the cultivation and projection of image. Celebrity needs the oxygen of publicity” (Smart, 2005, p. 14). Appearances in talk shows or coverage in tabloids, magazines are, therefore, essential for keeping celebrity status. Sometimes even liaisons between Hollywood film partners or personal sex tapes on the Internet are arranged to catch and maintain the media’s attention.

Since well-knowness itself has obtained tremendous commercial value – wholly divorced from great deeds and accomplishments – there is a whole industry today that manages the business of transforming unknowns into celebrities. Pop artist Andy Warhol mentioned that in the future, everyone will be a celebrity for fifteen minutes – an allusion to the explosion of print and broadcast media, which must incessantly fill their space and time with people’s stories. The man who rescues a boy from drowning or the woman who wins the state lottery – everyone and every story is potential grist for the news mill (Rein et al., 2006). In contrast to superstars, celebrities can be entirely “fabricated” resulting in minor, short-lived, “flash in the pan” socialites (Rindova, Pollock, & Hayward, 2006). A New York publicist and a socialite decided to test the limits of celebrity making. The two ran a successful PR firm that launched a variety of products. They began outfitting a clerk – a twenty five year old woman who worked in a boutique – in designer clothes, deliver her to the most exclusive parties, limo her to movie and theatrical premieres. At each event, they made sure that she was photographed draped with other famous people, and in general, leak to the press her hotness and prominence. In a short period of time, she was arguably a well-known celebrity giving quotes to gossip columnists or being interviewed on her hairstyle by Vogue. (Rein et al., 2006).

94

Superstars versus Celebrities – Big Man or Big Name?

5.3.3 A Simple Model of Celebrities In contrast to superstars, celebrities do not need to have any special talent. I rather consider celebrities as a media process that turns ordinary people into well-known socialites. Therefore, the consumption benefit of celebrities is negligible. However, celebrities are well-suited for “gossip consumption”. The utility of gossip about a e

celebrity j depends on x j which denotes the expected combined number of media recipients and other people who indirectly hear about the celebrity’s rumors.

e

U (xj )

e

f (xj )

(4)

In line with the network goods literature (see e.g. Katz & Shapiro, 1985), we model an e

increasing but concave interaction benefit function f ( x j ) . Hence the pleasure of circulating information and forming evaluations and rumors increases with the people who are able to join in these gossip discussions, but at a diminishing marginal rate.

Media produces information goods: content, advertisements, or in our case celebrity news. The production of information goods are typically subject to high fix costs G j and no (or low) variable costs. Since the consumption of celebrity tidings is non-rival, the costs are independent of the market size. The total value generated in the provision of celebrity news is given by:

e

3( x j )

e

e

f (xj ) ˜ xj  Gj

(5)

As long as equation (5) is positive, it is efficient to provide celebrity information. And celebrity news becomes ever more valuable as an increasing number of individuals are e

paying attention. Due to this bandwagon effect the marginal value increases in x j :

e

w3 ( x j ) e wx j

e

wf ( x j ) e e ˜ xj  f (xj ) ! 0 e wx j

(6)

Superstars versus Celebrities – Big Man or Big Name?

95

Equation (6) indicates that the value generated by celebrities is maximized if in the extreme case one celebrity dominates the market for gossip news. In general, a specialization on a few media-chosen celebrities is definitely more efficient than a diversification on numerous socialites.

To sum up; media plays a crucial role in creating celebrity status. The media executes a coordinative function by orchestrating the public’s attention to a few celebrities for whom it triggers the self-energizing bandwagon effect of popularity. Consumers maximize utility by paying attention to the most popular celebrity, because he or she provides the highest interaction benefit.

5.4 Conclusion With the evolution of mass-media and information technology, talent superiority or any special achievement is no longer a precondition for high attention. A new type of stars has arisen: Celebrities – individuals who are just known for their well-knowness. While the demand for superstars is directly linked with the consumption benefit depending on perceived superior quality of the star’s performance, celebrities need not do anything special to attract demand. In this chapter I suggest that social interaction does not only provide new consumption capital as it is the case in Adler’s superstar model, but that people rather directly benefit from interacting. Discussing, telling rumors or the development of new story lines about the martial relations, sexual habits, dressing fashions, or appearance of socialites generate value of their own for those involved. Celebrities qualify well for gossip because they are well-known and tidings about them are most easily to find and share. The higher the popularity of the celebrity, the easier gossip circulation and the higher the interaction benefit become. This process fuels further popularity in a virtuous cycle.

Since celebrities are mostly not able to set themselves apart by any special achievements, they need the media to trigger a self-energizing bandwagon effect. The media plays a fundamental role in celebrity emergence. Popular support by the media and general publicity through television casting shows, talk shows or coverage in

96

Superstars versus Celebrities – Big Man or Big Name?

tabloids, magazines and the Internet may suddenly gain momentum and escalate – creating most famous celebrities.

Different Star Strategies in the Media

97

6 Different Star Strategies in the Media – Why “Manufactured” Celebrities are More Lucrative than “Self-Made” Superstars1

6.1 Introduction In the advertiser-supported media sector media content like television programs, newspaper copies, or magazine articles are sold to media consumers and the attracted audiences can be packaged, priced and sold to advertisers. Therefore, audiences are the main currency for media firms (Doyle, 2002). Broadcasters do not only broadcast programs; they are in particular in the business of producing audiences (Owen and Wildman, 1992). In general the profits of media companies increase with the number of viewers per program since mass media typically operate under increasing returns to scale. For example, once a television program has been produced, the extra cost of an additional viewer is very small. The audience attracting capability of stars is one of the traditional instruments employed to increase the number of viewers. The literature distinguishes between two different types of stars: highly talented and therefore “selfmade” superstars (see Rosen, 1981; Adler, 1985; MacDonald, 1988; Borghans & Groot, 1998), and “manufactured” and thus rather trivial celebrities (see e.g. Boorstin, 1961; Gamson, 1994; Marshall, 1997; Cowen, 2000; Turner, 2004). Both kinds of stars increase audience interest and draw attention. But whereas “self-made” superstars set themselves apart by superior talent, celebrities draw people by pure fame “fabricated” by media publicity. 1

This chapter was published in Kyklos (see Franck & Nüesch, 2007).

Different Star Strategies in the Media

98

“Self-made” superstars attract audiences based on their perceived excellence in the provision of a certain service. Placido Domingo has convinced the opera fans around the world of his exceptional voice just as Diego Maradona has persuaded the international football fans of his outstanding technical abilities on the pitch. Both became superstars because they were considered to be the (or among the) most talented performers in their field. Hausman and Leonard (1997) for example found out that the mere presence of superstars like Michael Jordan had a substantial positive impact on national television ratings of NBA matches. Several studies in the movie sector clearly indicate that superstars promote the success of the films in which they play (see e.g. Wallace, Seigerman & Holbrook, 1993; Prag & Cassavant, 1994; Albert, 1998; Franck & Opitz, 2003; Elberse, 2006). Media provide access to superstars in many different ways, for example by broadcasting a top sports competition, by airing an interview with an excellent singer or by inviting a successful actor to a talk-show.

Recently broadcasted reality television shows like e.g. Pop Idol are based on a different star concept: They “create” stars out of anonymous performers by providing them a media platform and allowing the viewers to pick a singer to be groomed as a star. The format Pop Idol which was first aired 2001 in England has had tremendous success. Meanwhile, the show is broadcasted in 110 countries. Pop Idol is just one format – but possibly the most successful – of hundreds of reality television shows recently flooding the television programs. The reality shows range from reality soaps like Expedition Robinson, Big Brother, or I’m a Celebrity, Get Me Out of Here!, other casting shows like America’s Next Top Model, to docu-soaps starring celebrities like The Anna Nicole Smith Show, The Simple Life (with Paris Hilton and Nicole Richie), or Newlyweds (with Jessica Simpson) just to name a few. Reality television has rapidly come to occupy a place at the forefront of contemporary television culture (Holmes, 2004a). Of course, the idea of manufacturing celebrities is not new: For example, this strategy was already used by television quiz shows during the 1950s or by major recording labels during the late 1960s to “create” celebrities like the Monkees as a televised alternative for the Beatles.2

2

I am grateful to Stephen Lacy for raising this issue.

Different Star Strategies in the Media

99

Before the “graphic revolution” it was generally necessary to have demonstrated great deed or action in order to become a well-known and attention-drawing star (Smart, 2005). Mass-media, however, allow stardom to be artificially created by media publicity and promotion. Recent information technology and the Internet have even extended the capacity to create, transmit and disseminate images of celebrities. Whereas there is still a loose connection between talent, winning the contest and the ensuing celebrity status in casting shows like Pop Idol, other reality formats like e.g. Big Brother do not even claim to select the winner according to any special achievements. Through such pseudoevents anyone may become famous. Boorstin (1961) speaks of people who are just known for being well-known. Apart from their fame, “manufactured” celebrities may be trivial and superficial. But the rating success of most of the reality television shows still proves that celebrities have high viewer drawing capability. Thus, the attraction of large audiences is not necessarily based on exceptional talent; pure fame suffices.

In this chapter, I compare the basic economic mechanisms which explain the emergence of traditional “self-made” stars and “manufactured” celebrities. Based on these mechanisms I show that “self-made” stars and celebrities are comparable in their potential to generate value in the media industry. However, whereas “self-made” stars become endowed with market power through the very mechanisms which create them, celebrities have inferior opportunities to capture the value created by their appearance. Therefore, the media companies are able to capture the bulk of the profits from “manufacturing” celebrities.

The remainder of this section is structured as follows: First, I describe the reality television show Pop Idol as a case study of “manufacturing” celebrities. In the second part of the chapter, I analyze the economic mechanisms explaining the emergence of “self-made” stars as well as of celebrities and compare their consequences with respect to viewer drawing capability and bargaining power. In a last section, the findings are summarized.

100

Different Star Strategies in the Media

6.2 Pop Idol – An Example of “Manufacturing” Celebrities Pop Idol originated 2001 in the UK as a public singing contest to determine the best undiscovered young singers in a country. The format starts with televised auditions where contestants are gradually selected by sarcastic judges. The final participants perform live on television each week and viewers then vote by phone or text message for their favorite. The singer with the least votes leaves each week until a winner is crowned. The winner then receives a management contract with 19 Entertainment and a recording contract with BMG (Mortimer, 2004). Pop Idol was an immediate rating success. The final episode attracted a viewing audience of 13.2 million, or a total audience share of 57%. The first final of American Idol – the US version of Pop Idol – generated record ratings of 23 million viewers, which was the biggest audience for a non-sports program at that time in over ten years. American Idol also generated an unprecedented 110 million telephone votes over the progress of the first final shows (Dann, 2004). And the success has continued: The average number of viewers of American Idol increased from 26.5 million in 2005 (Daly, 2006) to 30 million in 2006 (Zeitchik & LaPorte, 2006).3

Pop Idol was devised by the British artist manager Simon Fuller and a director of BMG, Simon Cowell. Simon Fuller began his career as a talent scout in the 1980s. In 1985 he launched the keyboard maestro Paul Hardcastle and guided his song, “19”, to the number one spot. In the same year he founded the company 19 Entertainment4 which has grown and diversified to become a group of numerous companies covering television, music management, music publishing, recording, artist and producer management, sponsoring and promotion (Sanghera, 2002). For example Fuller created and managed the Spice Girls and launched teen act S Club 7. 19 Television was formed in 1997 as a subsidiary company of 19 Entertainment to produce or co-produce television shows or films. In 2001 it started Pop Idol in the UK in cooperation with

3

The British Pop Idol show was replaced by the casting format The X Factor after the second series. Simon Fuller claimed that The X Factor was a copy of his own show and filed a lawsuit against the producers of The X Factor. In November 2005 an out-of-court settlement was reached.

4

In the beginning, Fuller’s company was named 19 Management. But he later changed the company’s name to 19 Entertainment.

Different Star Strategies in the Media

101

Thames TV which internationally trades as Fremantle Media. Since then, Pop Idol has spun off several successes such as American Idol, Canadian Idol, Australian Idol to name just a few. In March 2005 Simon Fuller sold 19 Entertainment to Robert Sillerman’s company CKX for $ 196 million (Serwer, 2005). Fuller, however, has remained the chief executive.

In just a few years the Pop Idol format has developed to a multi-million pound brand operating all over the world. In 2006, 34 local versions of Pop Idol air in 110 countries. The music expert Michael Learmonth (2006) names Pop Idol a “diamond-studded annuity that generates in excess of $ 1 billion a year worldwide through advertising, sponsorships, license fees, merchandising, telephone voting, record sales and touring”. The Los Angeles Times estimated that already the annual global advertising revenues of the Pop Idol format exceed $ 1 billion (Hardy, 2004). Exact data is not available. However, it is undoubted that the Pop Idol shows are a very lucrative business.

In the UK, Fuller’s 19 Television company and Thames Television have an equal share of ownership in the Pop Idol format. Internationally, the television rights of the Pop Idol format are held by 19 Television for two-thirds in conjunction with Fremantle Media which owns one-third. Fremantle Media is a television production subsidiary of Europe’s largest television and radio group RTL, itself 90% owned by German media conglomerate Bertelsmann. Over the period 2002-2003 19 Television and Fremantle Media received over $ 250 million in format fees (Hardy, 2004). However, broadcasters did not miss out. They obtained considerable Idol-related advertising revenues. All the global Pop Idol shows in the years 2002-2003 generated over $ 2 billion advertising revenues according to the Los Angeles Times. The US-based Fox network for example gained $ 200 million advertising income from the first two seasons of American Idol (Hardy, 2004). During the fourth American Idol series in 2005 Fox sold ads at an average price of about $ 600’000 per 30-second spot. According to Lieberman (2005) this summed up to at least $ 444 million advertising income.

Another significant revenue stream is derived from merchandising the Idol brand, which is split between 19 Entertainment and Fremantle Media. According to Fremantle

102

Different Star Strategies in the Media

Media, in the last three month of 2003 more than $ 45 million of merchandise associated with American Idol was sold in the US (Hardy, 2004). Consumers can drink from their idol’s mug, download the artists’ ring tones for their own phone or sing along to the branded karaoke system. Even video games, cell phone faceplates, or perfumes are sold as licensed products. In 2004 consumers spent about $ 215 million on Idollicensed products mostly for typical pop-culture products: toys, candies, trading cards, games, a magazine and books (Lieberman, 2005). The sales of albums, singles and music videos associated with the Pop Idol format in the US, UK and Germany totaled some additional $ 170 million from 2003 to 2004 (Hardy, 2004).

Even though the winners of Pop Idol enjoy enormous fame and publicity, financially they do not profit likewise. A very detailed contract between the participants and Fuller’s 19 Entertainment guarantees on a clear “take it or leave it” basis that the young performers are wrapped up for recording, management and merchandising under very restrictive terms for three years. Gary Fine, a music attorney, made this “particularly aggressive” contract public as he came into possession of it when a mother of a young man who was interested in being on the show brought it for his perusal (Olsen, 2002). The first clause, for example, says that the producers can record any and all behavior of the contestant “in and in connection with the series” and use the contestant’s likeness, voice and biographical material, whether true or false any way they want to. The producers own all this material forever and everywhere. The second clause says that all information regarding the show and this contract is strictly confidential and if a contestant breaches this confidentiality, it will cause damages assumed to be in excess of $ 5 million. A further clause requires each finalist to enter into agreements exclusively with 19 Recording for recording of solo albums;5 19 Merchandising for advertising, endorsement, merchandising and sponsorship; and 19 Management for the management of his or her career. All this was entirely at the option of Fuller’s 19 Entertainment, save for the winner, who was guaranteed this result. Another clause even states that the Idol winner has to appear at the later World Idol program6, for a total fee

5

The recording rights, however, are mostly licensed out to BMG.

6

In the World Idol program the winners of the various national Idol shows compete against each other.

Different Star Strategies in the Media

103

of just $ 1’400 (Olsen, 2002). 19 Entertainment virtually controls all aspects of the singers under contract. Some music experts say that the careers of Pop Idol finalists are literally not their own. An example: 19 Entertainment arranged for Kelly Clarkson, winner of the first American Idol show, to sing the American national anthem at the first September 11 anniversary in Washington D.C. Several newspapers and prominent persons criticized it, questioning if a day of national mourning shall be turned into a giant promotional opportunity. In response, Clarkson wanted to cancel her obligation. She was quoted in the New York Times, saying: “If anyone thinks I’m trying to market anything, well, that’s awful. I am not going to do it – I am not going to sing on September 11.”7 The next day 19 Management issued a press release and clarified that Clarkson will sing the national anthem on September 11 in Washington D. C. and that media reports to the contrary are incorrect (Dann, 2004). Kelly Clarkson had no other choice. The very restrictive contracts between the Pop Idol singers and 19 Entertainment restrain the singers’ careers financially as well. 19 Entertainment receives 10% of their recording revenue (Serwer, 2005). The management fee is estimated to be an additional 15-20% (James, 2002). And 19 Merchandising also generously partakes of the merchandising and touring8 revenue. In a rare interview with The Associated Press, Simon Fuller described his manager-client relationships with the Idol contestants as “partnerships”, in which he receives between 25% and 50% percent of their earnings. The industry standard, however, is a 20% management fee (Ehlers & Writer, 2004).

19 Entertainment is effectively structuring a global base for Pop Idol, which generates money not simply from the sale of the format and the exploitation of the promoted celebrities, but also from multimedia platforms, phone calls and the Internet (Holmes, 2004b). According to Learmonth (2006), 19 Entertainment and Fremantle Media together receive 50% of revenues from cellular phone calls and instant messaging. Given the tremendous numbers of votes, this is considerable money. In 2003, 7.5 million viewers of American Idol cast votes by text messages. One year later the 7

See Kuczynski (2002) quoted in Dann (2004, p. 17)

8

According to the Billboard magazine ticket buyers spent more than $ 28 million in 2004 to see the finalists of American Idol (Lieberman, 2005).

104

Different Star Strategies in the Media

number increased to 13.5 million and then another year later more than tripled to 41.5 million text messages (Zeitchik & Laporte, 2006). In 2003 there were 60 million phone calls for American Idol. The chief executive of Fremantle Media Licensing Worldwide estimates that more than a billion phone votes have been cast globally for Idol contestants in the years 2003 and 2004. In 2006 580 million votes were cast during the fifth season of American Idol. In the UK alone, each series was estimated to generate $ 9 million by phone and text message voting (Mortimer, 2004).

The business model of Pop Idol is rather simple: Take unknown but ambitious young individuals who are willing to sign recording, management or merchandising rights away, equip them with stardom and sell access to the “fabricated” celebrities. Every year ten thousands of individuals audition for various Idol contests. Few if any of those amateur or semi-professional singers have experience or track record in the music industry. Their bargaining power is very low. Most candidates will sign anything the show’s producer puts in front of them, because potential earnings of Idol finalists are still higher than alternative earnings as workaday Janes and Joes (Piccoli, 2006). The big profiteers, however, are both production companies – 19 Entertainment and Fremantle Media – and the television broadcasters. According to Learmonth (2006) the two production companies have retained an “unusually large stake in the myriad revenue streams” the show generates. Pop Idol turned Simon Fuller into the second richest “millionaire in film and television” with an estimated fortune of $ 540 million (Beresford & Boyd, 2006). The television broadcasters have profited as well. According to various network estimates, reality shows like Pop Idol cost about half as much to produce – about $ 600’000 per hour – as typical new dramas or sitcoms. Pop Idol does not require new sets, or stars who ask for $ 1 million-an-episode salaries, as the actors of the series Friends successfully did in 2003. And Pop Idol tends to be popular with the audience advertisers desire most: young women (Farhi, 2003).

Pop Idol explicitly distinguished itself from prior casting shows (e.g. Popstars) in its invocation of audience interactivity and popular taste. The creators promoted Pop Idol insisting “But this time, you choose!” (Holmes, 2004b). As a result of this procedure rather popularity-driven celebrities emerge, because the candidate with the largest fan-

Different Star Strategies in the Media

105

support wins. Becoming the winner of Pop Idol is less an issue of talent and more one of sympathy and compatibility to popular taste. Experience shows that the “manufactured” celebrities in general could not sustain audience interest in their work when they lost the publicity generated by appearing weekly on prime time television. While the initial success of some of the individual singles or albums of Pop Idol finalists has been extraordinary (the first UK, Australian or American singles, in particular, went to number one immediately), most of the singers have not been able to repeat their initial success nor to construct a continuing career (Turner, 2004). They have had only mixed achievements outside the safety of the created popularity bubble.

To examine general publicity of the American Idol finalists, I measured how often they were mentioned in the press by conducting a text analysis of articles in numerous quality and tabloid newspapers as well as weekly magazines.9 Figures 4 to 6 illustrate the number of monthly articles which mention the finalists of the first three American Idol series by name.

1200 1100 1000

01. Kelly Clarkson

900

02, Justin Guarini

800

03. Nikki McKibbin 04. Tamyra Gray

700

05. R.J. Helton

600

06. Christina Christian

500

07. Ryan Starr

400

08. A.J. Gil

300

09. Jim Verraros

200

10. EJay Day

100

Ju n Au -0 2 gO 02 ct De 02 cFe 0 2 bAp 03 rJ u 03 n Au -0 3 gO 03 ct De 03 cFe 0 3 bAp 04 rJ u 04 n Au -0 4 gO 04 ct De 04 cFe 0 4 bAp 05 rJ u 05 n Au -0 5 g O - 05 ct De 05 cFe 0 5 bAp 06 rJ u 06 n Au -0 6 g06

0

Figure 4:

9

Articles mentioning the American Idol finalists of season 1

The used database contains quality newspapers (including e.g. USA Today, The New York Times, International Herald Tribune, Los Angeles Times, Chicago Tribune, The Boston Globe, Chicago Daily Herald, The Denver Post, Detroit Free Press, Florida Today, The Kansas City Star, Miami Herald, The Washington Post), tabloid newspapers (e.g. The Edmonton Sun, The Boston Herald), press agencies like The Associated Press, weekly magazines (including Life, The Economist, Houston Press) or music magazines (including e.g. Billboard, BBC Music Magazine, or Variety).

106

Different Star Strategies in the Media

Figure 4 indicates that only the winner of the first American Idol season, Kelly Clarkson, has had lasting visibility in the press. After the television show in summer 2002, she attracted large media attention by releasing a successful solo album in April 2003, appearing on the World Idol contest in December 2003 and by launching a very successful second album in spring 2005.10 In February 2006 she won two Grammies and enjoyed more publicity than ever before. Kelly Clarkson was able to use the show’s publicity bubble as a stepping stone for a successful individual pop career. By receiving honors of the prestigious Recording Academy, Clarkson definitely achieved superstar status and emancipated herself from the televised contest that originally made her famous. However, all other finalists of the first series of American Idol have vanished into thin air.

1200 1100

01. Ruben Studdard

1000

02. Clay Aiken

900

03. Kimberley Locke

800

04. Joshua Gracin 05. Trenyce

700

06. Carmen Rasmusen

600

07. Kimberly Caldwell

500

08. Rickey Smith

400

09. Corey Clark

300

10. Julia DeMato

200

11. Charles Grigsby

100

12. Vanessa Olivarez

Ja n0 M 3 ar M 03 ay -0 Ju 3 lSe 03 p0 No 3 v0 Ja 3 n04 M ar M 04 ay -0 Ju 4 lSe 04 p0 No 4 v0 Ja 4 n0 M 5 ar M 05 ay -0 Ju 5 lSe 05 p0 No 5 v0 Ja 5 n0 M 6 ar M 06 ay -0 Ju 6 lSe 06 p06

0

Figure 5:

Articles mentioning the American Idol finalists of season 2

During the second American Idol contest both the winner and the runner-up snowballed into famous celebrities with monthly publicity scores of over 800 articles. Their media coverage after the show has been – although very fluctuating – generally decreasing. Besides Ruben Studdard and Clay Aiken, who were able to call attention by releasing rather successful personal singles and albums or by extensive touring, the other finalists’ celebrity status disappeared once the show went off the air. The publicity peak 10

The second album of Kelly Clarkson stayed in the Billboard 200 album charts for more than 100 weeks.

Different Star Strategies in the Media

107

of Corey Clark in May 2005 was an exception. It proved that pure rumors suffice to attract media attention. Although his recordings never reached the charts, he prominently placed himself in the media spotlight by claiming that he and one of the show’s judge Paula Abdul had an affair during the contest. He thereby profited from the enormous popularity of the fourth season of American Idol in which Paula Abdul was on the air at that time. However, the fame bubble of Corey Clark burst as quickly as it formed.

1200 1100

01. Fantasia Barrino

1000

02. Diana DeGarmo

900

03. Jasmine Trias

800

04. LaToya London 05. George Huff

700

06. John Stevens

600

07. Jennifer Hudson

500

08. Jon Peter Lewis

400

09. Camile Velasco

300

10. Amy Adams

200

11. Matthew Rogers

100

12. Leah LaBelle

Ja n04 M ar -0 4 M ay -0 4 Ju l-0 4 Se p04 No v04 Ja n05 M ar -0 5 M ay -0 5 Ju l-0 5 Se p05 No v05 Ja n06 M ar -0 6 M ay -0 6 Ju l-0 6 Se p06

0

Figure 6:

Articles mentioning the American Idol finalists of season 3

Figure 6 gives a typical illustration of the media coverage of “manufactured” and thus rather trivial celebrities: The enormous publicity created during the television show rapidly decreased after the show. No one of the finalists has ever reached comparable publicity scores since then. Fantasia Barrino, Diana DeGamo, and Jasmine Trias lost their fame as quickly as it came.

6.3 A Strategy Framework of Star Attraction in the Media From a simple strategy perspective the success of a company depends on two elements: value creation and value capture. A media enterprise has a competitive advantage if it is able to create and obtain more economic value than the marginal (breakeven) competitor in its product market (Peteraf & Barney, 2003). In order to prosper, a firm

Different Star Strategies in the Media

108

must not only be able to create value, but to capture the value it creates (Saloner, Shepard, & Podolny, 2001). Referring to stars this means that the profits of media firms are positively related to the viewer drawing capability of stars and negatively to their bargaining power as external resource suppliers.

The media provide a production technology suitable to create, promote or exploit stardom.11 The consumption of media content is generally non-rival. If a person watches a television broadcast, it does not diminish someone else’s opportunity of watching it as well. Media content does not get used up or destroyed in the act of consumption (Doyle, 2002). The production of media content is subject to large economies of scale, because the production costs are largely independent of the size of the audience. The same content may be marketed under a windowing process in which it is delivered to consumers via multiple distribution channels sequentially in different time periods. The expenses involved in generating the first copy tend to be considerable. However, once the first copy of the program has been created, it then costs little or nothing to reproduce it to extra customers. The value of the program to the viewer is unaffected by the number of viewers, but the value of the commercial to the advertiser is directly and positively linked to audience size. Thus increasing returns will be enjoyed as more and more customers watch a program. The larger the audience, the more profitable it will become for the producer (Doyle, 2002). Therefore, the viewer drawing capability is crucial. Superstars are providers of media content with high viewer drawing capability. Economic theory offers two distinct explanations why this might be the case.

6.3.1 The Rosen Explanation for the Viewer Drawing Capability of Superstars In his seminal paper on “The Economics of Superstars” Sherwin Rosen defines superstars as “relatively small numbers of people who earn enormous amounts of money and dominate the activities in which they engage” (Rosen, 1981, p. 845). Given

11

It is no surprise that the occupations of stars are generally closely related to the media: e.g. actors, musicians, or athletes (Borghans & Groot, 1998).

Different Star Strategies in the Media

109

a production technology that allows for joint consumption and scale economies,12 output may be concentrated among a few individuals who have the most talent. Rosen’s superstar theory is based on two basic premises: Firstly, lower quality is an imperfect substitute of higher quality. People prefer consuming fewer high-quality services rather than more of the same service at moderate quality levels: “(…) hearing a succession of mediocre singers does not add up to a single outstanding performance” (Rosen, 1981, p. 846). Most people tend not to be satisfied with the performance of a less talented but cheaper artist when they are able to enjoy the performance of a top artist even if the cost is somewhat higher (Frey, 1998). Secondly, Rosen (1981) assumes that talent or quality is costlessly observable by all potential consumers. Therefore, small differences in talent are magnified into large differences in earnings. In Rosen’s model, a single superstar (or a single group of superstars) – the best – serves the whole market (Schulze, 2003). Superstars attract audiences by providing performances of comparably higher quality.

The plausibility of Rosen’s assumptions largely depends on the sector or job in which a star is engaged. The performance of a 100 meter sprinter, for example, is clearly and unambiguously determined by the running time. The sprinter’s talent is easily identifiable and measurable. And in general, people favor watching the finales in the Olympic Games rather than ten runs at mediocre levels. Concerning artistic activities, however, quality determination is a lot more difficult. Consumers have manifold tastes and their understanding of quality is highly diversified. While some people love the songs of Madonna, others may hate them. Commonly accepted and clearly measurable talent indicators are often not available. Hence Rosen’s second assumption is less plausible in arts. Hamlen (1991, 1994) or Salganik, Dodds, and Watts (2006) fail to find empirical evidence for Rosen’s superstar explanation in the popular music industry.

12

Media typically provide such a technology. Of course, public performances of a classical concert for example may exhibit a unit cost decrease with rising audience size too. However, there will be congestion costs at some point as a classical live concert is more enjoyable in a medium-sized concert hall than in a large sport arena (Schulze, 2003). Media eliminate congestion, since the superstar activity can easily be replicated through CD productions, television performances, videos, movies or books. These “canned performances” display higher scale economies and a higher personal scale of operations (Schulze, 2003). Media technology makes it possible for large parts of the world market to be served by one person.

Different Star Strategies in the Media

110

6.3.2 The Adler Explanation for the Viewer Drawing Capability of Superstars Adler (1985) explains the phenomenon of superstars as a learning process that occurs if consumption requires knowledge. Based on the notion of “consumption capital” by Stigler and Becker (1977), Adler (1985) argues that appreciation of a star’s performance increases with knowledge: “… the more you know the more you enjoy” (Adler, 1985, p. 208-209). Stigler and Becker (1977) use music as an example of how past consumption activities lead to beneficial addiction through an accumulation of consumption capital. By having exposed themselves to music in the past, consumers have built up consumption capital that enables them to derive more pleasure from listening to the same music in the present. Stigler and Becker (1977) themselves referred to Marshall (1923) who had written: “(…) the more good music a man hears, the stronger is his taste for it likely to become.”13 When discussing the taste for good music, Alfred Marshall had probably some distinguished operas or classical music in mind. For example, Beethoven connoisseurs feel great pleasure in listening to symphonies, concertos, or operas of Beethoven, since specific consumption capital allows them to appreciate subtle details and delicacies of his compositions. This explains why consumers will not diversify indefinitely either across activities, or across individuals within a given activity; however, it does not explain why everybody would choose to have the same superstars. Adler (1985) supplemented the Stigler/Becker-framework by adding the element of discussing consumption with likewise knowledgeable individuals. Star specific consumption capital is not only accumulated by past consumption activities, but also by discussing the star’s performance with other people who know about it. The more popular the superstar in question is, the lower the searching costs to find competent discussants will consequently be. Searching cost economies imply that one is always better off patronizing a well-known star as long as others are not perceived as superior by an order of magnitude. These positive network externalities explain why superstars may even emerge among equally talented performers.

13

Original statement in Marshall (1923, p. 94) quoted in Stigler and Becker (1977, p. 78). The accentuation is introduced by the authors.

Different Star Strategies in the Media

111

6.3.3 Bargaining Power of Superstars The existence of rents is not sufficient for a media firm to earn above average returns. If the resources which generate the rents are not owned by the firm, the suppliers (in our case the superstars) may bid up the price of their resources to the point where they capture the differential value won from customers (Bowman & Ambrosini, 2000). The question how much value the media firm is able to retain is answered by the relative bargaining powers of the resource supplier and the firm. Resource suppliers with a powerful bargaining position are able to capture a large proportion of the created value, whereas resource suppliers with weak bargaining power will find themselves obtaining far less value. How powerful are superstars as external resource suppliers of media content?

In Rosen’s theory superstars have a certain degree of monopolistic power due to their exceptional talent. Since consumers strongly prefer to watch the best performers, superstars are not replaceable without significant quality losses. Superstars cannot be separated from the activity in which they excel. Therefore, they display high bargaining power, which enables them to capture large parts of the generated rents (Borghans & Groot, 1998).

In Adler’s (1985) star theory, superstars enjoy high bargaining power due to the star specific consumption capital. Since consumption capital is irreversible and not transferable, it creates lock-in-effects and significant switching costs. Thus Adler stars have high bargaining power. For example, a person who has become a connoisseur of the actress Meg Ryan is not willing to substitute a movie with Meg Ryan for one without her. Ravid (1999) shows that movie stars capture most of the value added they create. There exists broad casual evidence indicating that movie stars very quickly adjust their fees to reflect their value. John Travolta for instance multiplied his fee almost with 100 after the success of his film Pulp Fiction. Weinstein (1998) who analyzes the evolution of profit-sharing contracts in the Hollywood movie sector illustrates how proven stars are more likely to sign contracts with gross-profit shares. Superstars have more assertiveness and require higher remuneration. They take this

112

Different Star Strategies in the Media

compensation in the form of a profit or revenue share. Weinstein (1998) argues that sharing contracts are not primarily intended to align the incentives of actors with those of the studios. They are rather a sign of high bargaining power. Superstars like Julia Roberts, Jim Carrey or Tom Hanks do not only earn guaranteed $ 20 million but also 20% back end (Franck & Opitz, 2003). Hence, in spite of revenues in excess of half a billion dollars, the film Forrest Gump for example failed to make a profit (Ravid, 1999). The marginal production costs of films or television programs partly do not decrease but increase with (expected) audience size. Bourreau, Gensollen, and Perani (2002) explain what seems to be an atypical production cost function for the media sector with the fact that superstars are able to negotiate remuneration based on the expected mean audiences they draw as a result of their rare talent.

I conclude that “self-made” superstars in the sense of Rosen and Adler are excellent in value creation and value capture. Media firms, therefore, have clear incentives to find substitutes with comparable value creation potential but less bargaining power. In the following I argue that celebrities draw large audiences without having substantial bargaining power to adopt the created value.

6.4 „Manufactured“ Celebrities Despite the differences in the emergence of Rosen and Adler superstars, there is also a unifying element in both theories. They presuppose that superstars have exceptional talent and provide services of perceived superior quality. This assumption is obvious for Rosen superstars. However, it is also required for Adler stars because the notion of consumption capital stipulates that there is hidden talent and/or quality which need to be discovered through a learning process. If there was nothing to discover, learning would be superfluous and consumption capital inexistent. Because their stardom is based on their own capabilities Rosen and Adler stars are “self-made” to a significant degree.

This is not necessarily the case for celebrities. The most widely quoted definition of celebrity was given by Boorstin (1961, p. 57): “The celebrity is a person who is known for his well-knowness.” According to Boorstin (1961) celebrities’ appearances are

Different Star Strategies in the Media

113

pseudo-events14; they appear to be meaningful but are in fact insubstantial. He explains the distinction between “self-made” stars – which he calls heroes – and celebrities as follows: “We can fabricate fame, we can at will (…) make a man or woman well known; but we cannot make him great. We can make a celebrity, but we can never make a hero. (…) The hero created himself; the celebrity is created by the media. The hero was a big man; the celebrity is a big name” (Boorstin, 1961, p. 48+61). Celebrity status is artificially producible by media publicity. In the democracy of pseudo-events anyone may become a celebrity if only he or she manages to get into the news and to stay there (Boorstin, 1961). Superstars all share admirable qualities – qualities that somehow set them apart form the rest of us – whereas celebrities need not do anything special (Gamson, 1994). Concerning superstars, fame and popularity are related to an exceptional talent. Celebrities though are famous because they have been made to be. For example, reality television turns ordinary people into well-known celebrities just by providing a publicity platform.

Marshall’s assumption that “(…) no celebrity possesses any meaning of consequence” (Marshall, 1997, p. 11) is a heroic simplification, of course. In reality, the boundary between “self-made” superstars and “manufactured” celebrities is more blurred. Most celebrities may also have a moderate level of talent and superstars have also profited from publicity platforms. But the fact that the well-knowness of celebrities has become a viable commodity all by itself is intrinsic to their meaning. Fame may stand independent of accomplishment, heroics, or talent (Rein, Kotler, Hamlin, & Stoller, 2006). However, given the postulated triviality of celebrities, how can their attraction be explained?

14

Turner (2004) describes a “pseudo-event” as an event planned and staged entirely for the media, which accrues significance through the scale of its media coverage rather than through any more disinterested assessment of its importance.

Different Star Strategies in the Media

114

6.4.1 Viewer Drawing Capability of “Manufactured” Celebrities I argue that celebrities generate gossip externalities. They attract audiences by providing a projection screen for all kind of rumors, judgments, or interpretations. The pleasure of gossip lies in the exchange of news, in the development of new story lines, in discussing and sharing evaluations. It is not necessary for gossip that the information is demonstrably true; in fact, too much truth can stop the gossip game (Gamson, 1994). Celebrities are in many ways better objects for this game than acquaintances like e.g. neighbors. “Celebrities are like neighbors whom nearly everyone knows, in nearly every social setting, and “stuff” about them is easier to find and share than information about friends or colleagues. More important, celebrity gossip is a much freer realm, much more game-like than acquaintance gossip: there are no repercussions and there is no accountability” (Gamson, 1994, p. 176). Joke Hermes who wrote a book about “Reading Women’s Magazines” observed that most women find talking about their favorite celebrities a comfortable way of spending their time with other people: “Gossip draws speakers together in their sharing and evaluation of ‘news’ about ‘third persons who are not present’” (Hermes, 1995, p. 131). Jane Johnson, a reporter of the successful British celebrity magazine Closer, even believes that: “Celebrity gossip is a national obsession and a unifying experience across all social groups” (Johnson, 2004, p. 55).

The interaction benefit of gossip increases with the number of people knowing the tidings of a particular celebrity. The activity of discussion, story telling, interpretation, or judgment is typically subject to network externalities. The more popular a celebrity is, the easier gossip circulation becomes. The interaction benefit is, therefore, an increasing function of the celebrity’s popularity. This creates a self-energizing virtuous cycle: a celebrity with a large audience becomes more and more valuable to each viewer, as he or she attracts ever more viewers. Leibenstein (1950) named the observation that people often follow the crowd as “bandwagon effect”. The bandwagon effect emerges if people’s valuations of a commodity (and thus demand for this good) increase when they observe others consuming the same commodity. Banerjee (1992, p.

Different Star Strategies in the Media

115

798) defines this herd behavior as “everyone doing what everyone else is doing.” Individuals decide whether or not to follow a rising celebrity depending on the number of people currently paying attention to this person. Popular support for an individual by the media may thus suddenly gain momentum and escalate. Of course, bandwagon effects might also happen randomly. But in the majority of cases media corporations consciously set agenda15 and promote new celebrities.

In contrast to Adler’s conception of superstars, star attraction of celebrities is not linked to the consumption benefit of the performance but rather to the subsequent interaction benefit. Therefore, the star attraction of celebrities is no longer necessarily based on talent. Of course, positive network externalities of popularity also exist in Adler’s superstar theory. According to Adler (1985) popularity indirectly increases star attraction by simplifying the accumulation of consumption capital. Consumption capital, however, has no value of its own; it only generates a benefit by enhancing the valuation of the star’s performance. But the underlying quality of the performance still depends on the star’s talent. Referring to Adler stars even enormous popularity cannot replace missing talent.

6.4.2 Bargaining Power of “Manufactured” Celebrities The participants of the Pop Idol series are ordinary people who – due to the high-profile associated with the show – are becoming well-known celebrities. Nobody would know them if they had not been in the media. Since celebrity gossip does neither rely on extraordinary talent nor on specific consumption capital, celebrities are easy to replace and, therefore, have low bargaining power.16 Pop Idol candidates who are not willing to sign the very restrictive contracts are promptly exchanged by other applicants. Hence, media corporations obtain the lion’s share of the generated revenues and routinely seek

15

See e.g. Maxwell, & Shaw (1993) for a review on the research about “agenda setting”, which denotes the media’s ability to influence the public’s opinion.

16

“Accidental celebrities” (Turner, Bonner & Marshall, 2000), who are individuals getting into the focus of attention initially through an uncontrollable incident, however, can sell their stories for large sums. Monica Lewinski, Diana’s former butler Paul Burrell, or kidnap victim Natascha Kampusch are examples of this category. “Accidental celebrities” are not interchangeable.

116

Different Star Strategies in the Media

to find unspoiled fresh prospects they can “discover” and groom for stardom. In reality television programs such as Pop Idol the media producers have incorporated celebrity emergence into the format. Therefore, these celebrities are particularly dependent upon the program that made them visible (Turner, 2004).

6.4.3 Market Segmentation Even though “manufacturing” celebrities is more lucrative for the media than employing “self-made” superstars, it is obvious that a total substitution of “self-made” superstars in the media has not occurred. Why? Briefly, the market potential of “manufactured” celebrities is limited to certain kinds of entertainment like for example game shows, docu-soaps and pop music to some extent. Celebrity “creation” is feasible wherever an activity does not have clearly measurable quality indicators and whenever its consumption does not require specific knowledge. In sports or classical music, for example, celebrity status cannot be created from scratch. A set of well-established tournaments relying either on objective quality indicators like e.g. time performance and/or on institutionalized voting procedures by proven expert judges determine winners in these fields. The Olympic sprint finals or the Olympic gymnastics finals are obviously not decided by public voting. Here audiences are interested in the discovery of superior talent, and factors like the speed and the dexterity of the contestants cannot be substituted by sympathy or publicity. As long as audiences are interested in the talent superiority of performers, “self-made” superstars will not disappear from the media despite their ability to capture large parts of the generated economic value.

6.5 Conclusion Broadcasting stars is a common strategy to increase audience size. However, catching high attention and reaching high media ratings does not suffice to capture a rent in a competitive environment. The latter strongly depends on the relative bargaining powers of the stars as external resource suppliers and the media firm. I analyzed the viewer drawing capability and the bargaining power of both superstars and celebrities. Whereas “self-made” superstars attract audiences based on the perceived excellence of the

Different Star Strategies in the Media

117

provided service, celebrities draw viewers by offering a source of gossip. However, while superstars exert strong bargaining power due to the singularity of their performances and/or the consumers’ accumulation of specific consumption capital, “manufactured” celebrities are interchangeable and thus have low market power to capture value. No wonder that the creation and exploitation of celebrities has become a large business in the media sector. But the market potential of “manufactured” celebrities is limited because they typically prevail only in “talent free” entertainment.

Summary and Outlook

119

7 Summary and Outlook

Both superstars and celebrities are individuals of high profile. Even though superstars are defined by their enormous salaries and celebrities by their remarkable fame, they often combine both. This thesis attempts to shed light on the economic mechanisms that explain the emergence and the impact of superstars and celebrities. While the economic star literature looks back on more than 20 years of research, celebrities have been less addressed in the economic literature and if so, mostly in consumer and marketing studies.

In chapter 2 and 3, the competing but not mutually exclusive theories of Rosen (1981) and Adler (1985) are empirically analyzed in German soccer. While the first study deals with the question what characteristics determine the emergence of superstars, the second study investigates the marginal revenue product of star players for their team. These two studies belong together, since in the long run teams are only able to pay high salaries to their stars if they attract additional viewers and generate extra merchandizing, gate or television revenue in return. Running quantile regressions, I find evidence that the market values of German soccer stars are driven by hidden talent characteristics, proxies of the consumers’ past consumption activities and popularity variables. Firsthand observable talent measures like goals and assists, however, do not significantly increase the market values of superstars. Hence, Rosen’s theory of superstar formation stressing the importance of firsthand observable talent is not supported for German soccer stars. Outstanding soccer players fit to Adler’s conception. Since a soccer game is a team product, the assessment and evaluation of a particular player require specific knowledge. The longer a player already performed in the first German Bundesliga and the more popularity he enjoys, the easier the accumulation of player specific knowledge

120

Summary and Outlook

and the better the evaluation of his performance on the pitch becomes. Hence, players cannot only intensify their investments in physical talent but they can also make higher popularity investments in order to increase the likelihood of becoming a superstar.

In the future the study in chapter 2 may be extended by including more detailed performance data. The Opta Index for example provides unique data about the individual number of shots, successful passes, dribbles, tackles and clearances, or saved shots by the goalkeeper. Since soccer is a highly interactive game based on the combination of complementary player skills, it is very difficult to properly assess a player’s talent. Especially for defenders, the number of goals and assists alone does not adequately measure talent. More precise individual performance data enables a more accurate analysis of a player’s talent and field performance in the future. A second critical point of the present study is the distinction between firsthand observable and hidden talent. According to the study, goals and assists are identifiable and measurable by the spectators without requiring significant specialized knowledge, whereas a player’s “true” talent is hidden. Only experts are able to detect the latter. However, it is not clear how difficult talent assessment really is. Therefore, it might be better to drop the fuzzy distinction between firsthand observable and hidden talent characteristics and to just rely on the ceteris paribus assumption of multiple regressions that given a certain talent level, popularity increases a star’s earnings according to Adler (1985), but does not according to Rosen (1981).

Chapter 3 analyzes how stars facilitate fan support using longitudinal match attendance data of all clubs in the first German soccer league during a nine year period. The results indicate that the exact channel of generating star attraction (by field performance or popularity) largely differs depending on firstly whether a player is a nationwide superstar or a local hero and secondly whether attendance at home or on the road is investigated. While superstars enhance attendance both at home and on the road by outstanding field performances, local heroes attract home fans by mere popularity. Robustness checks reveal that the results are sensitive to the chosen star definition. Performance related star attraction of superstars is no longer observed in the data, if the superstar category is extended to account for the 5% or 8% most expensive players of

Summary and Outlook

121

the league. The economic star literature clearly postulates that superstardom is a “small number” phenomenon (see Rosen, 1981). Therefore, I defined the superstars as players whose market values are in the top 2% quantile of the league’s market value distribution and a local hero as the most valued player of a particular team without superstars. Since superstars increase attendance on the road, they display a positive externality on other teams in the same league. Based on the results in chapter 3, the value of the average superstar externality sums up to approximately € 430’000 per season. At home the superstars’ scores increase attendance by 8.4% which corresponds to 57’703 tickets sold additionally. In total, German soccer superstars are expected to generate approximately € 1’440’0001 gate revenues. However, since we have no information about merchandizing and television revenues, we are not able to pass a sentence whether the German soccer stars are over- or underpaid.

Superstar salaries are controversial. The enormous wages of a view on the top are despised in most societies. From an economic perspective the efficiency of superstar earnings is crucial. Inefficient allocations would allow making one person better off without making others worse off. Inefficiency may have several causes: e.g. market power, information asymmetries, or externalities. I argue in chapter 4 that the player’s earnings in soccer are inefficiently high due to positional externalities based on the league tournament. Any action – for example engaging a superstar player by paying higher wages – that increases one club’s chances of winning must necessarily reduce the chances of others. These positional externalities lead to “rat races” or “positional arm races” in which the clubs tend to overinvest in player talent. The overinvestment hypothesis in chapter 4 stands in contrast to the positive externality of star attraction in section 3. While the latter theoretically leads to underpaid stars, the former causes overpaid players. Since annual growth rates of salaries exceeded annual growth rates of revenues in most European soccer leagues during the last decade, the overpaid hypothesis seems to dominate in reality. In 2002 the leading European soccer clubs reacted to the highly increasing salaries by signing a voluntary salary cap agreement to

1

The star effect at home corresponds to € 1’009’800, given an average admission prize of € 17.50. In combination with the star effect on the road, this sums up to approximately € 1’440’000.

122

Summary and Outlook

limit annual salaries to 70% of annual revenues. The simple model of two identical profit maximizing clubs in section 4 though suggests that this voluntary salary cap is very unlikely to be honored.

The concept of celebrities is introduced in chapter 5. In line with Boorstin (1961), a celebrity is defined as an individual who is known for his or her well-knowness. Unlike superstars, the fame of celebrities does not necessarily depend on extraordinary talent or special achievements but may rely on pure publicity. Style rules over substance regarding celebrities. What matters is no longer the consumption benefit of a certain service, but rather the subsequent interaction benefit of being able to relate to others by gossiping about celebrities. Gossip draws people together. Celebrities are well-suited for gossip because stories about celebrities are widely available and there is no danger of accountability or repercussions. The fame of celebrities originates from a bandwagon effect: the more popular a celebrity is, the easier gossip circulation becomes, which then fuels further popularity in a self-enforcing virtuous cycle. It is efficient to concentrate on the most famous celebrities, because they provide the highest interaction benefit.

In this thesis, I treat celebrities as trivial individuals who do not possess any special talent at all. This is a strong and often false assumption, of course. However, due to character of the interaction benefit, attention drawing ability does no longer necessarily depend on talent. The rise of numerous shallow reality television formats offers enough casual evidence that even insubstantial celebrities are able to generate high ratings. But who snowballs into a celebrity and who does not? I suggest that the media plays a critical role for celebrity emergence because anyone can become a celebrity, if only he or she gets into the media. The high publicity of media platforms manages to trigger the self-energizing bandwagon effects that “create” most famous celebrities.

Chapter 6 integrates the “creation” of celebrities by the media into a strategic context. Celebrities not only need the “oxygen” of media publicity, but the media also highly profits from “manufacturing” celebrities. Media firms are mainly in the business of producing audiences. And the audience attracting capability of superstars and celebrities is one of the traditional instruments employed by the media to increase the number of

Summary and Outlook

123

viewers. Analyzing both the viewer drawing potential and the bargaining power, I derive that celebrities are more lucrative for the media than superstars. Whereas superstars attract audiences based on the perceived excellence of the provided service, celebrities draw viewers by offering a source of gossip. However, while superstars exert strong bargaining power due to the singularity of their performances and/or the consumers’ accumulation of specific consumption capital, “manufactured” celebrities are interchangeable and thus have low market power to capture value. This may explain why the “creation” and “exploitation” of celebrities has become such a large business in the media sector. It is obvious though that a total substitution of superstars in the media has not occurred. It seems that the market potential of “manufactured” celebrities is limited to certain kinds of entertainment like game shows, docu-soaps or pop music to some extent. Celebrities typically prevail only in “talent free” entertainment.

The economics of superstars and celebrities is an ongoing topic worthwhile being followed in the future. A crucial question which certainly deserves further investigations is the issue of talent determination. Scholars face great difficulties measuring talent. Throsby (1994) writes: “While it is quite plausible to take estimated earnings functions and to attribute at least some of the (often large) unexplained residual to differences in talent, such a hypothesis remains untestable when no independent measure of talent is forthcoming” (p. 19). The lack of adequate talent indicators might be one reason why empirical superstar studies which confront Rosen’s with Adler’s theory mostly provide support for the Adler version of superstar effects (see e.g. Hamlen, 1991 & 1994 or Chung & Cox, 1994). The same applies to the present study in chapter 2. Since Rosen’s superstar theory requires that “small differences in talent become magnified in large earnings differences” (p. 846), identifiable and quantifiable talent measures are needed for an empirical validation. And even when quantifiable talent measures are available, the question about the validity often remains unanswered. For example, Hamlen (1991 & 1994) investigated the competing superstar theories in the music industry. In doing so he used the harmonic content of the voice as a singer’s talent indicator. “Voice quality” is a clearly quantifiable variable which measures the „richness“ and „depth“ of the voice (see Hamlen, 1991, p. 731). Hamlen’s results show that differences in talent improve record sales less than proportional. He fails to find a

124

Summary and Outlook

magnification effect. But does voice quality really matter? In classical music or opera, presumably yes. Concerning pop or rock music, however, I have strong doubts about the relevance of voice quality. More important factors for the success of such singers are probably charm, sex-appeal, or the show on stage (Schulze, 2003). We need both valid and quantifiable talent measures not only to empirically test superstar theories but also to draw a clear distinction between talented superstars and trivial celebrities. Such measures are hard to find. I agree with Connoly and Krueger (2005, p. 37) that “a major limitation of tests of superstar models is the absence of natural units with which to measure talent.” A second difficulty of superstar analysis is the diversity of tastes. Both Rosen (1981) and Adler (1985) assume identical consumers who demand an equal unspecified artistic activity. Different tastes are handled by saying that the diversity of tastes just confines a seller’s market without changing the basic mechanisms for superstar formation (see Adler, 1985, p. 211). Consumers of similar tastes compose different categories. Each category (e.g. opera, rock, or pop music) thereby constitutes a market with its own stars. However, Adler (1985) clearly states that stars do not emerge unless at least a group of people have similar tastes. Thus, even though Stigler and Becker (1976) argued in favor of stable tastes, the formation of generally accepted tastes is a very important and relevant issue regarding superstars and celebrities. It is hard to deny the fact that firstly, tastes are diverse and that secondly, such preferences might also change over time. Interdisciplinary approaches to the phenomenon of superstars and celebrities, therefore, provide promising future research topics.

References

125

8 References

Adler, M. (1985). Stardom and talent. The American Economic Review, 75, 208-212. Adler, M. (2006). The economics of superstars: theories and their application. V. Ginsburgh & D. Throsby (eds.). Handbook of Economics of Art and Culture. Elsevier. Agrawal, J. & Kamakura, W. A. (1995). The economic worth of celebrity endorsers: an event study analysis. Journal of Marketing, 59, 56-63. Albert, S. (1998). Movie stars and the distribution of financially successful films in the motion picture industry. Journal of Cultural Economics, 22, 249-270. Akerlof, G. A. (1976). The Economics of Caste and of the Rat Race and Other Woeful Tales. The Quarterly Journal of Economics, 90, 599-617. Alchian, A. & Demsetz, H. (1972). Production, information costs, and economic organization. The American Economic Review, 62, 777-795. Banerjee, A. V. (1992). A simple model of herd behavior. The Quarterly Journal of Economics, 107, 798-817. Bell, J. (2007). David Beckham is coming to America. The New York Times, January 11, 2007. Beresford, P., & Boyd, S. (2006). Millionaires in film and TV. Sunday Times, April 23. Berri, D. J., Schmidt, M. B., & Brook, S. L. (2004). Stars at the gate. The impact of star power on NBA gate revenues. Journal of Sports Economics, 5, 33-50. Berri, D. J. & Schmidt, M. B. (2006). On the road with the National Basketball Association’s superstar externality. Journal of Sports Economics, 7, 347-358. Boorstin, D. J. (1961). The image or what happened to the American dream. New York: Harper and Row.

References

126

Borghans, L. & Groot, L. (1998). Superstardom and monopolistic power: why media stars earn more than their marginal contribution to welfare. Journal of Institutional and Theoretical Economics, 154, 546-572. Borland, J. & MacDonald, R. (2003). Demand for sport. Oxford Review of Economic Policy, 19, 478-502. Bourreau, M., Gensollen, M., & Perani, J. (2002). Economies of scale in the media industry. Working Paper, ENST. Bowbrick, P. (1983). The economics of superstars: comment. The American Economic Review, 73, 459. Bowman, C. & Ambrosini, V. (2000). Value creation versus value capture: towards a coherent definition of value in strategy. British Journal of Management, 11, 1-15. Brandes, L., Franck, E., & Nüesch, S. (2007). Local heroes and superstars – An empirical analysis of star attraction in German Soccer, Journal of Sports Economics, Prepublished June 12, 2007; DOI: 10.1177/1527002507302026. Brauer, D. A. & Hickok, S. (1995). Explaining the growing inequality in wages across skill levels. Federal Reserve Bank of New York Economic Policy Review, January 1995. Brown, E., Spiro, R., & Keenan, D. (1991). Wage and non-wage discrimination in professional basketball: do fans affect it? American Journal of Economics and Sociology, 50, 333-345. Buchinski, M. (1998). Recent advances in quantile regression models: a practical guide for empirical research. Journal of Human Resources, 33, 88-126. Bundesliga (2006). Bundesliga Report 2006. Press conference on 14th March 2006 in Frankfurt/Main. Bull, C. (1987). The Existence of Self-Enforcing Implicit Contracts. The Quarterly Journal of Economics, 102, 147-159. Buraimo, B., Forrest, D., & Simmons, R. (2006). Freedom of entry, market size and competitive outcome: evidence from English soccer. Working Paper, Lancaster University. Burdekin, R. C. & Idson, T. L. (1991). Customer preferences, attendances and the racial structure of professional basketball teams. Applied Economics, 23, 179-186.

References

127

Burtless, G. (1995). International trade and the rise in earnings inequality. Journal of Economic Literature, 33, 800-816. Byers, J. D., Peel, D. A., & Thomas, D. A. (2001). Habit and long memory in attendance demand: the case of football support. Research Paper, University of Wales. Carmichael, F., Forrest, D., & Simmons, R. (1999). The labour market in association football: who gets transferred and for how much? Bulletin of Economic Research, 51, 125-150. Carmichael, F., Thomas, D., & Ward, R. (2000). Team performance: the case of English premiership football. Managerial and Decision Economics, 21, 31-45. Carmichael, F., Thomas, D., & Ward, R. (2001). Production and efficiency in association football. Journal of Sports Economics, 2, 228-243. Chung, K. H. & Cox, R. A. K. (1994). A stochastic model of superstardom: an application of the Yule distribution. Review of Economics and Statistics, 76, 771775. Clark, A. E. & Oswald, A. J. (1996). Satisfaction and comparison income. Journal of Public Economics, 61, 359-381. Connolly, M. & Krueger, A. B. (2005). Rockonomics. The economics of popular music, NBER Working Paper 11282, National Bureau of Economic Research. Cowen, T. (2000). What price fame? Cambridge/London: Harvard University Press. Czarnitzki, D. & Stadtmann, G. (2002). Uncertainty of outcome versus reputation: empirical evidence for the first German football division. Empirical Economics, 27, 101-112. Daly, S. (2006). Turns out, the show is really the “Idol”. St. Petersburg Times. January 15. Dann, G. (2004). American Idol: from selling a dream to selling a nation. Mediations, 1, 15-21. Dean, D. H. (2004). Consumer reaction to negative publicity. Journal of Business Communication, 41, 192-210. Dietl, H., Franck, E., & Roy, P. (2003). Überinvestitionsprobleme in einer Sportliga (Overinvestment in a Sports League). Betriebswirtschaftliche Forschung und Praxis, 5, 528-540.

References

128

Dietl, H., Franck, E., & Lang, M. (2005). Overinvestment in European Football leagues. Working Paper, University of Zürich. Dietl, H., Franck, E., & Nüesch, S. (2006). Are voluntary salary caps agreements selfenforcing?. European Sport Management Quarterly, 6, 23-34. Dobbson, S. & Goddard, J. (2001). The Economics of Football. Cambridge: Cambridge University Press. Doyle, G. (2002). Understanding media economics. London: Sage Publications. Ehlers, M. & Writer, S. (2004). Pot of gold within reach; ‘heart and soul’ seen as springboard to Barrino’s career. The News & Observer, May 27. Elberse, A. (2006). The power of stars in creative industries: do stars drive the success of movies? Working Paper, Harvard Business School. Erdogan, B. Z. (1999). Celebrity endorsement. A literature review. Journal of Marketing Management, 15, 291-315. Eschweiler, M. & Vieth, M. (2004). Preisdeterminanten bei Spielertransfers in der Fußball-Bundesliga (Prize determinants of player transfers in the German soccer league). Zeitschrift für Betriebswirtschaft, 64, 671-692. Evans, A. & Wilson, G. (1999). Fame: the psychology of stardom. London: Vision. Falter, J. & Pérignon, C. (2000). Demand for football and intramatch winning probability: an essay on the glorious uncertainty of sports. Applied Economics, 32, 1757-1765. Farhi, P. (2003). TV’s new reality: hit shows are here today, gone tomorrow. Washington Post, February 17. Farrell, J. & Saloner, G. (1985). Standardization, compatibility, and innovation. Rand Journal of Economics, 16, 70-83. Feess, E. & Muehlheusser, G. (2003). Optimal Revenue Sharing in Professional Sports Leagues. Working Paper, University of Berne. Ferrer-i-Carbonell, A. (2005). Income and well-being: an empirical analysis of the comparison income effect. Journal of Public Economics, 89, 997-1019. Forrest, D., Simmons, R., & Feehan, P. (2002). A spatial cross-sectional analysis of the elasticity of demand for soccer. Scottish Journal of Political Economy, 49, 236255.

References

129

Fort, R. & Quirk, J. (1995). Cross-Subsidization, Incentives, and Outcomes in Professional Team Sports Leagues. Journal of Economic Literature, 33, 12651299. Fort, R. (2000). European and North American Sports Differences (?). Scottish Journal of Political Economy, 47, 431-455 Franck, E. (2001). Warum gibt es Stars? – Drei Erklärungsansätze (Why do stars exist? – Three explanations). Wirtschaftsdienst, 81, 59-64. Franck, E. & Opitz, C. (2003). Julia Roberts, Tom Hanks & Co. Wie Stars zur effizienten Zuordnung von Filmen auf Filmkonsumenten beitragen (How stars provide

an

efficient

matching

of

films

and

film-consumers).

Wirtschaftswissenschaftliches Studium, 32, 203-208. Franck, E. & Nüesch, S. (2007). Avoiding ‘star wars’ – Celebrity creation in the media, Kyklos, 60, 211-230. Franck, G. (1998). Ökonomie der Aufmerksamkeit (The economics of attention). Wien: Hanser. Frank, R. H. & Cook, P. J. (1995). The winner-take-all society. New York: The Free Press. Frank, R. H. (2003). Microeconomics and behavior. Boston: McGraw-Hill. Frey, B. S. (1998). Superstar museums: an economic analysis. Journal of Cultural Economics, 22, 113-125. Frick, B. (2001). Die Einkommen von “Superstars” und “Wasserträgern” im professionellen Teamsport’ (The incomes of „superstars“ and „benchwarmers“ in professional team sports). Zeitschrift für Betriebswirtschaft, 71, 701-720. Gamson, J. (1994). Claims of fame. celebrity in contemporary America. Berkeley/Los Angeles/London: University of California Press. Garcia, J. & Rodriguez, P. (2002). Match attendance revisited. Empirical evidence from the Spanish football league. Journal of Sports Economics, 3, 18-38. Gibbons, R. (1992). Game theory for applied economists. Princeton, NJ: Princeton University Press. Hamlen, W. (1991). Superstardom in popular music: empirical evidence. Review of Economics and Statistics, 73, 729-733.

References

130

Hamlen, W. (1994). Variety in superstardom in popular music. Economic Inquiry, 32, 395-406. Hardy, P. (2004). Pop Idol brings soundcarrier, broadcasting, format and publishing success to Fuller and Bertelsmann. Music&Copyright, February 18. Hart, O. D. (1988). Incomplete contracts and the theory of the firm. Journal of Law, Economics, & Organization, 4, 119-139. Hausman, J. A. & Leonard, G. K. (1997). Superstars in the National Basketball Association: economic value and policy. Journal of Labor Economics, 15, 586624. Hausman, J. A. (1978). Specification tests in econometrics. Econometrica, 46, 12511271. Heermann, P. W. (2004). Salary Caps – Kartellrechtliche Grenzen (Salary caps – boundaries of the antitrust law). K. Zieschang & C. Klimmer (eds.), Unternehmensführung im Profifussball (Management in professional football). Berlin: Erich Schmid Verlag. Hermes, J. (1995). Reading women’s magazines. Cambridge: Polity Press. Holmes, S. (2004a). ‘All you’ve got to worry about is the task, having a cup of tea, and doing a bit of sunbathing.’ Approaching celebrity in Big Brother. S. Holmes & D. Jermyn (eds.). Understanding reality television. London: Routledge. Holmes, S. (2004b). Reality goes pop. Reality TV, popular music, and narratives of stardom in Pop Idol. Television & New Media, 5, 147-172. Hübl, L. & Swieter, D. (2002). Der Spielermarkt in der Fußball-Bundesliga (Player market in the German soccer league). Zeitschrift für Betriebswirtschaft, Special Issue 72, 105-123. Idson, T. L. & Kahane, L. H. (2000). Team effects on compensation. An application to salary determination in the National Hockey League. Economic Inquiry, 38, 345357. James, J. (2002). Reinventing reality. Time International, November 4, 64. James, J. (2003). The Joy of Becks. Online-Time, visited on October third 2005. Johnson, G. E. (1997). Changes in earnings inequality. The role of demand shifts. Journal of Economic Perspectives, 11, 41-54. Johnson, J. (2004). The secret: sex and celebs. British Journalism Review, 15, 51-56.

References

131

Jones, D. & Boon, G. (2003). Annual Review of Football Finance. Manchester: Deloitte. Jones, D. & Boon, G. (2004). Annual Review of Football Finance. Manchester: Deloitte. Jones, D. & Boon, G. (2005). Annual Review of Football Finance. Manchester: Deloitte. Jones, G. (2005). Annual Review of Football Finance. Düsseldorf: Deloitte. Kahn, L. M. (1993). Free agency, long-term contracts and compensation in Major League Baseball: estimates from panel data. Review of Economics and Statistics, 75, 157-164. Kahn, M. L. (2000). The sports business as a labor market laboratory. Journal of Economic Perspectives, 14, 75-94. Katz, M. L. & Shapiro, C. (1985). Network externalities, competition, and compatibility. The American Economic Review, 75, 424-440. Kern, M. & Süssmuth, B. (2005). Managerial efficiency in German top league soccer: an econometric analysis of club performances on and off the pitch. German Economic Review, 6, 485-506. Késenne, S. (2000). The Impact of Salary Caps in Professional Team Sports. Scottish Journal of Political Economy, 47, 422-430. Késenne, S. (2003). The Salary Cap Proposal of the G-14 in European Football. European Sports Management Quarterly, 3, 120-128. Koenker, R. & Bassett, G. (1978). Quantile regression. Econometrica, 46, 33-50. Koenker, R. & Hallock, K. F. (2000). Quantile regression. An introduction. Working Paper, http://www.econ.uiuc.edu/~roger/research/intro/rq.pdf, visited on May 24, 2006. Kremer, M. (1993). The O-ring theory of economic development. Quarterly Journal of Economics, 108, 551-575. Krueger, A. B. (2005). The economics of real superstars: the market for rock concerts in the material world. Journal of Labor Economics, 23, 1-30. Kuan, C. (2004). An introduction to quantile regression. http://www.sinica.edu.tw/econ/ activities/conpdf/chung-ming-kuan.pdf, visited January 19, 2007.

References

132

Kubat, R (1998). Der Markt für Spitzensport (The market for professional sports). Peter Lang. Kuczynski, A. (2002). ‘American Idol’ star wants to withdraw from Sept. 11 event. The New York Times, September 6. Learmonth, M. (2006). The show’s big winners. Variety, May 8. Lehmann, E. & Weigand, J. (1999). Determinanten der Entlohnung von Profifussballspielern – Eine empirische Analyse für die deutsche Bundesliga (Wage determination among professional soccer players – evidence from the German league). Betriebswirtschaftliche Forschung und Praxis, 51, 124-135. Lehmann, E. (2000). Verdienen Fussballspieler was sie verdienen? (Do football players earn what they deserve?). H. M. Schnellhaaȕ (eds.). Sportveranstaltungen zwischen Liga- und Medieninteressen (Sporting events between the interests of the league and the media). Schorndorf: Verlag Karl Hofmann. Lehmann, E. & Schulze, G. G. (2005). What does it take to be a star? – the role of performance and the media for German soccer players. Working Paper, MaxPlanck-Institute for Research into Economic Systems. Lehn, K. (1982). Property rights, risk sharing, and player disability in Major League Baseball. Journal of Law and Economics, 25, 347-366. Leibenstein, H. (1950). Bandwagon, snob, and veblen effect in the theory of consumers’ demand. The Quarterly Journal of Economics, 64, 183-207. Lieberman, D. (2005). ‘American Idol’ zooms from hit show to massive business. USA Today. March 30. Lucifora, C. & Simmons, R. (2003). Superstar effects in sport. Evidence from Italian soccer. Journal of Sports Economics, 4, 35-55. MacDonald, G. (1988). The economics of rising stars. The American Economic Review, 78, 155-166. Marshall, A. (1923). Principles of Economics. 8th edition, London. Marshall, D. P. (1997). Celebrity and power. Fame in contemporary culture. Minneapolis/London: University of Minnesota Presss. Mas-Colell, A., Whinston, M. D., & Green, J. R. (1995). Microeconomic Theory. Oxford: Oxford University Press.

References

133

Mathur, L. K., Mathur, I., & Rangan, N. (1997). The wealth effects associated with celebrity endorsers: the Michael Jordan phenomenon. Journal of Advertising Research, 37, 67-74. Maxcy, J. (2004). Motivating long-term employment contracts: risk management in Major League Baseball. Managerial Decision and Economics, 25, 109-120. Maxwell. E. M. & Shaw, D. L. (1993). The evolution of agenda setting research: twenty-five years in the marketplace of ideas. Journal of Communication, 43, 5867. McCracken, G. (1989). Who is the celebrity endorser? Cultural foundations of the endorsement process. Journal of Consumer Research, 16, 310-322. Merton, R. K. (1968). The Matthew effect in science. Science, 159, 56-63. Michie, J. & Oughton, C. (2004). Competitive balance in football. Trends and effects. Working Paper, University of London. Monaco, J. (1978). Celebrity. The media as image makers. New York: Delta. Mortimer, R. (2004). A brand that’s never idol. Brand Strategy, 188, 20-23. Mullin, C. J. & Dunn, L. F. (2002). Using baseball card prices to measure star quality and Monopsony. Economic Inquiry, 40, 620-632. Noll, R. G. (1974). The economics of discrimination in professional sports: the case of baseball. A. H. Pascal (eds.). Government and the Sports Business, Washington. Olsen, E. (2002). Slaves of celebrity: did American Idol contestants sign their career away? Calgary Herald, October 5. Owen, B. M. & Wildman, S. S. (1992). Video Economics. Cambridge: Harvard University Press. Peteraf, M. A. & Barney, J. B. (2003). Unraveling the resource-based tangle. Managerial and Decision Economics, 24, 309-323. Piccoli, S. (2006). What price, Idol?, Entertainment News, August 2. Piketty, T. & Quian, N. (2006). Income inequality and progressive income taxation in China and India, 1986-2015. Working Paper, Centre for Economic Policy Research. Piketty, T. & Saez, W. (2003). Income inequality in the United States. The Quarterly Journal of Economics, 118, 1-39.

References

134

Prag, J. & Cassavant, J. (1994). An empirical study of determinants of revenues and marketing expenditures in the motion picture industry. Journal of Cultural Economics, 18, 217-235. Ravid, S. A. (1999). Information, blockbusters, and stars: a study of the film industry. Journal of Business, 72, 463-492. Rein, I., Kotler, P., Hamlin, M., & Stoller, M. (2006). High visibility. Transforming your personal and professional brand. New York: McGraw-Hill. Reinstein, D. A. & Snyder, C. (2005). The influence of expert reviews on consumer demand for experience goods: a case of study of movie critics. The Journal of Industrial Economics, 53, 27-51. Rindova, V. P., Pollock, T. G. and Hayward, M. L. A. (2006). Celebrity firms: the social construction of market popularity. Academy of Management Review, 31, 5071. Rosen, S. (1981). The economics of superstars. The American Economic Review, 71, 845-858. Rosen, S. (1982). Authority, control and the distribution of earnings. Bell Journal of Economics, 13, 311-323. Rosen, S. (1983). The economics of superstars: reply. The American Economic Review, 73, 460-462. Rosen, S. (1986). Prizes and incentives in elimination tournaments. The American Economic Review, 76, 701-715. Rosen, S. & Sanderson, A. (2001). Labour markets in professional sports. The Economic Journal, 111, 47-68. Rottenberg, S. (1956). The Baseball Players’ Labor Market. Journal of Political Economy, 64, 242-258. Rojek, C. (2001). Celebrity. London: Reaktion. Roy, P. (2004). Die Zuschauernachfrage im professionellen Teamsport (Match attendance in professional team sports). Aachen: Shaker. Salganik, M. J., Dodds, P. S., & Watts, D. J. (2006). Experimental study of inequality and unpredictability in an artificial cultural market. Science, 311, 854-856. Saloner, G. A., Shepard, A., & Podolny, J. (2005). Strategic management. New York: John Wiley and Sons.

References

135

Sanghera, S. (2002). Where there’s a Will. Financial Times, April 2. Schmidt, M. B. & Berri, D. J. (2001). Competitive balance and attendance. The case of Major League Baseball. Journal of Sports Economics, 2, 145-167. Schulze, G. G. (2003). Superstars. R. Towse (eds.). The handbook of cultural economics. Cheltenham: Edward Elgar. Scoggings, J. F. (1993). Shirking or stochastic productivity in Major League Baseball: comment. Southern Economic Journal, 60, 239-240. Scott, F., Long, J., & Scompii, K. (1985). Salary vs. marginal revenue product under monopsony and competition: the case of professional basketball. Atlantic Economic Journal, 13, 50-59. Scully , G. W. (1974). Pay and performance in Major League Baseball. The American Economic Review, 64, 915-930. Scully, G. W. (1995). The market structure of sports. Chicago: The University of Chicago Press. Seaman, B. A. (2003). Cultural and sports economics: conceptual twins? Journal of Cultural Economics, 27, 81-126. Serwer, A. (2005). The man who bought Elvis; investor Robert Sillerman is combining the kind, American Idol, and other entertainment assets to build his next media conglomerate. Fortune, December 12. Shapiro, C. and Varian, H. R. (1999). Informational rules. A strategic guide to the network economy. Boston/Massachusetts: Harvard Business School Press. Simmons, R. (1996). The demand for English league football: a club-level analysis. Applied Economics, 28, 139-155. Sloane, P. J. (1971). The economics of professional football: the football club as a utility maximizer. Scottish Journal of Political Economy, 17, 121-145. Smart, B. (2005). The sport star. Modern sport and the cultural economy of sporting celebrity. London: Sage. Sportfive (2004). European Football 2004. Hamburg. Staudohar, P. D. (1998). Salary caps in professional team sports. Compensation and Working Conditions, Spring, 3-11.

References

136

Staudohar, P. D. (1999). Salary caps in professional team sports. C. Jeanrenaud & S. Késenne (eds.). Competition policy in professional sports. Antwerpen: Standaard Edition. Stigler, G. & Becker, G. (1977). De gustibus non est disputandum. The American Economic Review, 67, 76-90. Stollenwerk, H. J. (1996). Sport – Zuschauer – Medien (Sport – audiences – media). Band 4 der Reihe “Edition Sport & Freizeit“. Aachen: Meyer und Meyer. Szymanski, S. (2003a). The economic design of sporting contests. Journal of Economic Literature, 16, 1137-1187. Szymanski, S. (2003b). The assessment: the economics of sport. Oxford Review of Economic Policy, 19, 467-477. Szymanski, S. (2004). Professional team sports are only a game. The Walrasian fixedsupply conjecture model, contest-Nash equilibrium, and the invariance principle. Journal of Sports Economics, 5, 111-126. Szymanski, S. & Késenne, S. (2004). Competitive balance and gate revenue sharing in team sports. Journal of Industrial Economics, 52, 165-177. Telser, L. (1980). A theory of self-enforcing agreements. Journal of Business, 53, 2744. Throsby, D. (1994). The production and consumption of the arts: a view of cultural economics. Journal of Economic Literature, 32, 1-29. Tirole, J. (1988). The theory of industrial organization. Cambridge, MA: The MIT Press. Torgler, B., Schmidt, S. L., & Frey, B. S. (2006). Relative income position and performance: an empirical panel analysis. Working paper: University of Zurich. Towse, R. (1997). The earnings of singers: an economic analysis. R. Towse (ed.). Cultural economics: the arts, the heritage and the media industries. Cheltenham: Edward Elgar Publishing Limited. Turner, G., Bonner, F., & Marshall, P.D. (2000). Fame games: the production of celebrity in Australia. Melbourne: Cambridge University Press. Turner, G. (2004). Understanding celebrity. London: Sage. Turner, G. (2006). The mass production of celebrity. ‘Celetoids’, reality TV and the ‘demiotic turn’. International Journal of Cultural Studies, 9, 153-165.

References

137

Vrooman, J. (1995). A general theory of professional sports leagues. Southern Economic Journal, 61, 971-990. Vrooman, J. (2000). The economics of American sports leagues. Scottish Journal of Political Economy, 47, 364-398. Wallace, W. T., Seigerman, A., & Holbrook, M. B. (1993). The role of actors and actresses in the success of films: how much is a movie star worth?, Journal of Cultural Economics, 17, 1-27. Weinstein, M. (1998). Profit sharing contracts in Hollywood: evolution and analysis, Journal of Legal Studies, 27, 67-112. Williams, R. (2006). Multicollinearity. www.nd.edu/~rwilliam/stats2/l11.pdf, visited on December 21, 2006. Whitney, J. D. (1993). Bidding till bankrupt: destructive competition in professional team sports. Economic Inquiry, 31, 100-115. Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. Cambridge/London: The MIT Press. Wooldridge, J. M. (2003). Introductory econometrics: a modern approach. Second edition. South-Western. Young, S. M. & Pinsky, D. (2006). Narcissism and celebrity. Journal of Research in Personality, 40, 463-471. Zeitchik, S. & LaPorte, N. (2006). The money tree. Variety, May 8.

Index

139

9 Index

Addiction 19, 86, 110 Attention 12, 14, 27, 87, 90, 92-95, 97,

Gossip 2, 10, 15, 84, 91, 93-95, 114, 115, 117, 122, 123

99, 106, 107, 115, 116, 122 Herd behavior 92, 115 Bargaining 68, 99, 104, 108, 11, 112, 115-117, 123 Bandwagon effect 10, 15, 90, 91, 92, 94, 95, 114, 115, 122

Imperfect substitutability 4 Inefficiency 2, 7, 8, 14, 65-67, 89, 121 Inequality 1, 25 Incentives 9, 72, 73, 112

Competitive advantage 107 Competitive balance 50, 51, 63, 69, 76, 77 Consumption capital 5, 19, 20, 30, 38, 48, 84, 86, 88, 90, 91, 95, 110-112, 115, 117

Labor market 2, 3, 65, 67 Learning 4, 5, 19, 23, 28, 110, 112 Local hero 14, 45-47, 51, 54, 56, 60, 62, 120-121 Luck 5, 20, 87

Economies of scale 3, 4, 18, 85

Mass media 83, 97

Exogeneity 53, 60

Matthew effect 4 Media 1-3, 10-15, 18, 22, 31, 36, 61, 62,

Fame 2, 9-12, 83, 84, 90, 92, 97, 99,

83, 84, 92-95, 97-104, 106-108,

102, 107, 113, 122

111-117, 122, 123

Fixed-effects model 52-54

Merchandizing 119, 121 Monopolistic power 8, 65, 111

Gate attendance 46, 50, 61

Multicollinearity 32, 37

140

Nash equilibrium 78 Network externalities 4, 5, 9, 19, 88, 89, 91, 110, 114

Index

Star attraction 41, 43, 46-63, 65, 86, 107, 115, 120-121 Star quality 43, 48

Ordinary least squares (OLS) 33, 34

Random-effects model 52

O-ring theory 6

Reality television 2, 12, 98, 99, 113,

Overinvestment 67, 73, 79, 80, 121

116, 122 Rent-seeking 8

Pareto efficiency 13, 65

Revenue sharing 63, 69, 71-72

Pop Idol 12, 15, 98-105, 115-116 Popularity 5, 10-15, 17, 19, 22, 27, 30,

Talent 2, 4-6, 8-10, 13, 15, 17-19, 21-

36, 38, 39, 42, 43, 48, 49, 56, 83,

24, 27-31, 33-39, 43, 46, 47, 58-60,

84, 86, 87, 90, 92, 95, 105, 107,

70, 73, 76, 84-90, 95, 105, 109, 112,

113, 115, 119, 120, 122

113, 115-117, 119, 120, 122-124

Property right 8

Taste 7, 85, 86, 104, 105, 109, 110, 124

Pseudo-event 9, 83, 99, 113

Team product 23, 38, 119

Public goods 65

Technological change 1

Publicity 2, 9, 11-13, 20, 22, 31, 33, 38,

Television 2, 12, 18, 21, 28, 33, 83, 86,

48, 62, 93, 95, 97, 99, 102, 105-107,

95, 97-101, 104, 106-108, 112, 113,

113, 116, 122

116, 119, 121, 122 Tournament 8, 65, 116, 121

Quality 2-9, 21, 22, 28, 43, 46-48, 83, 85-87, 90, 92, 95, 109, 111, 112,

Trigger strategy 73, 74 Triviality 9, 113

115, 116, 123, 124 Quantile regression 13, 17, 34-36, 119

Value creation 88, 89, 107, 112 Value capture 107, 112

Salary cap 26, 67-69, 75-81, 122

Virtuous circle 91, 92, 95, 114, 122

Self-enforcement 67, 68, 77 Sex-appeal 124

Wages 65, 66, 121

Soccer 17, 19, 22-29, 31, 36, 38, 39, 41-

Walrasian model 69

44, 47-51, 55, 56, 60-65, 119-121

Winner-takes-all 1, 7, 8, 71