Mechanisms of Superstar Formation in German Soccer: Empirical Evidence

Abstract

Based on the competing theories of superstar formation proposed by Rosen (1981) and Adler (1985) it is controversial if first hand observable talent or other factors like past consumption and popularity influence stardom. This article investigates the emergence of superstars in German soccer. We use data on market values and individual player performance and publicity data to differentiate between Rosen's and Adler's theory of superstar formation. Running quantile regressions we find evidence that Adler's theory applies to German soccer stars. Therefore, not only investments in physical talent but also the cultivation of popularity is an adequate strategy for becoming a superstar.

Acknowledgements

We are grateful to Leif Brandes, Rainer Winkelmann and to the seminar participants at the Eastern Economic Association conference 2006 in Philadelphia and at the International Association of Sports Economics (IASE) conference 2006 in Bochum, to the editor and two anonymous referees for helpful comments. Remaining errors are, of course, our own.

Notes

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

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. Scott, Long and Somppi (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 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.

5. Average match attendance was calculated by the Kicker soccer magazine.

6. Unfortunately, we were not able to include further seasons because popularity data on previous seasons was only partially available.

7. “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.

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

9. 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).

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

11. Unfortunately data relating to salaries, signing fees, bonuses, or transfer fees are not available on a grand scale. In the 1999–2000 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 did not include any bonuses, signing fees, or transfer fees.

12. In addition, the Pearson χ² independence test rejects independency at the 0% level of significance. Thus, the two institutions seem to deliver corresponding data. However, having only estimations we cannot prove that the data used were totally free from arbitrariness.

13. Of course, the evaluations of sports journalists do not necessarily depend on the same criteria as those employed by coaches and club managers.

14. We 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 our results in any significant way. Due to the ease of interpretation we use the unweighted dummies.

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

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

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

18. The interview was held on 18th August 2005.

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

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

21. The database used contained 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).

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

23. 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).

24. 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.

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