ArticlePDF Available

Salary Determination in the German “Bundesliga”: A Panel Study

Authors:

Abstract and Figures

Abstract Using a hitherto unavailable panel data set from the top-tier in German pro- fessional football (n>6,000 player-year-observations) we investigate the de- terminants of player ,remuneration. We find ,that not only past and recent performance, but also region of birth and leadership skills influence individ- ual salaries. We also,document,how,the impact of these,characteristics va- ries across the salary distribution. Keywords: Salaries; Football; Pay Determination JEL-Code: L83, J31
Content may be subject to copyright.
Salary Determination in the German “Bundesliga”:
A Panel Study
Marcel Battré1, Christian Deutscher2 and Bernd Frick3
Abstract
Using a hitherto unavailable panel data set from the top-tier in German pro-
fessional football (n>6,000 player-year-observations) we investigate the de-
terminants of player remuneration. We find that not only past and recent
performance, but also region of birth and leadership skills influence individ-
ual salaries. We also document how the impact of these characteristics va-
ries across the salary distribution.
Keywords: Salaries; Football; Pay Determination
JEL-Code: L83, J31

1 University of Paderborn, Department of Management, Warburger Strasse 100, D-33098 Paderborn, E-
mail: marcel.battre@wiwi.uni-paderborn.de
2 University of Paderborn, Department of Management, Warburger Strasse 100, D-33098 Paderborn, E-
mail: christian.deutscher@wiwi.uni-paderborn.de
3 Corresponding author. University of Paderborn, Department of Management, Warburger Strasse 100,
D-33098 Paderborn, E-mail: bernd.frick@notes.upb.de and Institute of Labor Law and Industrial Re-
lations in the European Community, Campus II, University of Trier, D-54286 Trier
1. Research Question and Motivation
The “escalating” and/or “skyrocketing” salaries of professional football players have
recently become a highly controversial issue in Germany. Perhaps surprisingly, this has
not always been the case: When in the summer of 1954 the members of the German
national team after their glorious victory in the World Cup final against Hungary re-
turned home, each player received a gratification of 2,000 DM – about six months’ pay
of a male full-time employee (Müller-Jentzsch 1989: 148). By that time, the “enorm-
ous” amount was considered by most people a well-deserved recognition for an out-
standing performance.
The public opinion, however, changed gradually. On July 28th, 1962 when the represen-
tatives of the 21 different regional football associations in Germany agreed to introduce
a single first division, they also introduced a minimum and a maximum salary (the for-
mer being 250 DM per month and the latter 1,200 DM per month)4. Moreover the
maximum transfer fee was set at 50,000 DM, of which a maximum of 5,000 DM could
be paid to the player (all these caps were finally abandoned in 1972). The salaries of the
top players soon started to rise: In 1966, Uwe Seeler – at the time captain of the national
team – earned 50,000 DM, while Günther Netzer in 1972 was paid 100,000 DM al-
ready. Another five years later, top-scorer Gerd Müller earned 500,000 DM per season.
In 1987, Rudi Völler was paid 1.1 Mio. DM and in 1992 Andreas Möller made 1.7 Mio.
DM. Upon his return from the Italian “Serie A” to the Bundesliga in 1995, Lothar Mat-
thäus was paid 2.5 Mio. DM; an amount that he more than tripled until 19985. In 2001,
Stefan Effenberg as well as Oliver Kahn were paid 9.5 Mio. DM (Sonnenberg 2002: 24-
29).
This development – that can mainly be attributed to the development of the TV reve-
nues generated by the clubs – has, for most of the time, been accompanied by public
discussion about the “adequacy” of player salaries and has recently even attracted the

4 In the same year the average salary of a full-time blue-collar worker amounted to 7,775 DM. i.e. about
60% of a football player’s annual income. Today, the average player salary is about 45 times the aver-
age salary in Germany.
5 In the latter year, Oliver Bierhoff – later Matthäus’ teammate in Munich – earned more than 12 Mio.
DM in Italy.
attention of a number of politicians. Shortly before Christmas 2007, Norbert Lammert,
the president of the German “Bundestag” (the nation’s parliament) gave the following
statement:
„I am particularly annoyed by the salary explosion that we have recently expe-
rienced in professional sport in general and in soccer in particular. (…) This is
something I cannot understand at all“ (Source: Onabrücker Zeitung, 23. Dec.
2007).
This caused the president of the German Football Association, Theo Zwanziger, to re-
spond as follows:
„From a “moral” point of view, the salaries of many professional soccer players are
too high – as are the incomes of most actors and some top managers.“ (source:
Südddeutsche Zeitung, 9. Jan. 2008).
Given the steadily increasing ticket sales and merchandising revenues it is hardly sur-
prising that football fans seem to be quite relaxed with regard to the level and the devel-
opment of player salaries. In an online opinion poll that was started shortly after the
interviews were published, the daily newspaper “Die Welt” asked its readers the follow-
ing question: Should politicians be concerned about the development of player salaries
in professional football? The results were as follows:
Table 1
Are Fans Envious?
Possible Responses %
Yes, because politicians are obliged to intervene if certain developments in
society are causing discontent.
19
Yes, because the salaries in football are simply too high. 0
No, because politicians should in principle abstain from intervening in private
businesses.
36
No, because the salaries are the result of market forces. 45
Source: welt.de (last access on 8. Jan. 2009)
The fans position is nicely summarized in the following quote by sports journalist Oskar
Beck:
„We football fans are a rather strange species. We complain when our heroes earn
enormous amounts of money, but at the same time we readily accept increasing
higher ticket prices if this enables our favorite club to sign yet another top-scorer.
Moreover, we are prepared to pay 19.90€ for the memoirs of Stefan Effenberg and
the diaries of Lothar Matthäus as if it were the most recent works of nobel lau-
reates Heinrich Böll and Günter Grass” (source: Die Welt, 30. December 2007).
Summarizing, it appears that fans have fewer problems with the “escalating” and “sky-
rocketing” salaries than politicians and journalists seem to expect (or perhaps even hope
for). From an economic point of view, however, the question is not whether the salaries
reported above are “adequate” or “excessive”, but whether the observable variation in
player remuneration can be explained by differences in individual performance and the
clubs’ ability to pay (which, in turn, is a function of past and recent sporting success,
market size, and tradition).
The reminder of the paper is organized as follows: Section 2 refers to the neoclassical
model of the labor market to derive testable hypothesis concerning the remuneration of
professional football players and provides a short review of the available (very limited)
empirical evidence. In section 3, we summarize our hypotheses. Section 4 includes a
description of the unique data set we use, presents some descriptive statistics and the
estimation results. We conclude with a brief summary, some policy implications and a
number of suggestions for further research.
2. The Remuneration of Professional Football Players: Theory and Previous Evi-
dence
2.1. Theory
„Professional sport offers a unique opportunity for labor market research. There is no
other research setting 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, and we have a com-
plete data set of worker-employer matches over the career of each production worker
and supervisor in the industry. … Moreover, professional sports leagues have expe-
rienced major changes in labor market rules and structure … creating interesting natural
experiments that offer opportunities for analysis“ (Kahn 2000: 75).
In the absence of labor market restrictions (such as salary caps, reserve clauses and/or
draft rules) players will be paid according to their marginal product, i.e. the wage an
individual player receives is a function of his talent and his experience. However, since
the clubs differ with respect to their drawing potential – there are “small market” and
“large market” teams – they also differ with respect to their “ability to pay”.
Since it rests on a number of critical assumptions (such as player mobility, complete
information, and risk neutrality) the neoclassical model of wage determination has often
been rejected not only by sports fans, but also by some highly respected economists:
“… the elementary classical model presents a very poor description of employ-
ment relations in advanced economies” (Milgrom and Roberts 1992: 329).
However, the problems that are characteristic for most – if not all – “real life” labor
contracts (information asymmetries, incompleteness, importance of implicit elements)
are clearly less important in professional team sports. Here, an individual player‘s per-
formance can easily be measured, „shirking“ can be detected at low cost, effort and tal-
ent can be evaluated not only by a player‘s current club but also by other teams. It is,
therefore, plausible to assume that in the German “Bundesliga“ – as in other profession-
al team sports leagues with an unregulated labor market – players are paid mainly ac-
cording to their performance.
Moreover, professions in which talent is highly valued by consumers are usually charac-
terized by a highly skewed distribution of earnings: Small differences in talent translate
into large differences in pay (Rosen 1981). Player reputation not only attracts (addition-
al) spectators, but advances in technology facilitate the reproduction of matches at low
cost. Together, these two effects lead to an expansion of the market. Thus, players are
neither completely homogenous nor completely specialized. This creates a situation of
bilateral monopoly in which players and teams share a surplus or economic rent. Only a
few players who are sufficiently differentiated can shift surpluses (rents) completely
into salaries; these players will tend to be the ‘superstars’ of their sports.
2.2. Previous Evidence
To the best of our knowledge, only two studies have been published in English so far
that seek to identify the determinants of player salaries. Lucifora and Simmons (2003)
use information on 533 outfield players from the Italian “Serie A” and “Serie B” at the
beginning of the 1995/96 season (i.e. a cross section). They find that individual perfor-
mance (measured primarily by the number of games played and goals scored) has a sta-
tistically significant and economically relevant influence on salaries. Moreover, earn-
ings are highly convex in the individual’s career goal-scoring rate and the assist rate,
suggesting the existence of a considerable “superstar effect”. Lehmann and Schulze
(2008) use 651 player-year-observations from the German “Bundesliga” in the seasons
1998/99 and 1999/2000. Their performance measures also have the expected and statis-
tically significant influence on salaries. Surprisingly, however, media presence has a
positive, but declining influence, suggesting decreasing returns to popularity – a finding
that is difficult to reconcile with the concept of “superstardom”6.
Summarizing, these papers show that salaries of professional athletes are not just ran-
dom, that systematic factors determine these salaries to a large extent and that these
systematic factors such as age, experience and performance are very similar to those
found in other occupations. Where sports teams differ in structure of earnings is that the
distribution of salaries is even more highly skewed than in standard occupations and
also that sports teams apply more stringent selection procedures into occupations. For
example, poor performance by a player results in being dropped from team squad and
very quickly being discarded; there are high levels of mobility within the industry (be-
tween teams) and into and out of the industry, with shorter careers than in most occupa-
tions (Bryson, Frick and Simmons 2009).

6 Publications in German include Lehmann and Weigand (1999), Lehmann (2000), Huebl and Swieter
(2002), and Frick and Deutscher (2009). With the exception of the latter, all these papers use much
smaller samples from short sub-periods since the early 1990s.
3. Testable Hypotheses
The observable variance in player salaries is primarily due to the variance in talent and
performance:
1. Player salaries will increase with performance (league appearances, goals), expe-
rience (age) and popularity (appearances in the national team).
2. The most recent performance – i.e. in the last season – will have a greater impact on
player salaries than (previous) career performance.
Moreover, the clubs’ different ability to pay (which, in turn, is a function of the size of
the respective market, the club’s history and its sporting performance) will also affect
player salaries significantly.
4. Structure and Development of Player Salaries in the “Bundesliga”
4.1. Available Data
Our primary source of information is “Kicker”, a highly respected soccer magazine that
offers market valuations of players assessed at the beginning of a season for 13 consecu-
tive years (1995/96-2007/08) as a proxy for undisclosed salary, which remains private
and confidential not only in Germany, but in the rest of in Europe too. We can be confi-
dent of the reliability of these proxies for several reasons. First, the correlation between
Kicker salary figures and the ones from another reliable source (www.transfermarkt.de)
is high, at 0.75 (Torgler et al. 2006). Second, the player valuations in Kicker magazine
have been compiled by a stable team of experts who have established consistent practice
over a long period. We therefore interpret the players’ market values as published by
Kicker as particularly reliable. Aggregating the individual market values across teams
and dividing these by a constant factor of 1.5 results in the aggregated wage bill of each
team in the Bundesliga as published in the annual reports of the German Football Asso-
ciation over the period 1996-2007. Furthermore, the correlation between Kicker player
valuations and a subset of actual salary data obtained from the Bundesliga has been
found to be high, at 0.80 (Frick, 2003).
The size of our sample is quite large: We have 6,147 „player-year-observations“ for
1,993 different players to which we add player characteristics (such as age, number of
career games played, number of games played last season, number of career goals
scored, number of goals scored last season, number of career international appearances,
number of international appearances last season, team captain (dummy), position (a set
of three dummies), region of birth (six dummies), and previous league) as well as team
characteristics (win percentage, average attendance) that are also available from pre-
and post-season special issues of “Kicker”.
4.2. Descriptive Evidence
It appears from Figure 1 that average player salaries have increased from 550,000€ in
1995/96 to about 1.3 Mio. € in 2007/08. Interestingly, the standard deviation constantly
oscillates around the mean, suggesting that the dispersion of player salaries has re-
mained more or less constant over time7. The decline in player salaries in the seasons
2003/04 and 2004/05 has to be attributed to the insolvency of the Kirch group, the com-
pany that had bought the TV rights for a record amount of 695 Mio. DM per year start-
ing with the 2000/01 season. Moreover, player salaries differ considerably by position:
In 2007/08 goalkeepers on average earn about 900,000 € while forwards are paid an
average of 1.45 Mio. € (see Figure 2). The salaries of defenders and midfielders are
higher than those of goalkeepers, but lower than those of forwards. Although statisti-
cally significant (2007/08: F=3.08; p<.05), these averages hide considerable variation
within the different groups of players. Particularly in the case of goalkeepers, the stan-
dard deviation of individual salaries – and, therefore, the corresponding coefficient of
variation – is rather high.

7 This is interestingly insofar, as Theo Zanziger in the interview quoted above also argued that many
politicians by supporting the developments that has been induced by the Bosman-ruling of the Euro-
pean Court of Justice in December 1995 “have made few particularly gifted players richer and richer
and the clubs poorer and poorer”. He then went on to argue that “UEFA and the national associations
will do their very best to introduce an individual salary cap and to reach a more egalitarian wage struc-
ture in professional football.“
Figure 1
The Development of Player Salaries in the Bundesliga (in 1,000€)
1995 1998 2000 2002 2004 2006
500
600
700
800
900
1000
1100
1200
1300
____: Mean; ----: Standard Deviation
Figure 2
Player Salaries by Position (in 1,000€)
1995 1998 2000 2002 2004 2006
200
400
600
800
1000
1200
1400
1600
____: Forward; ----: Midfielder; …..: Defender; -.-.-: Goalkeeper
Perhaps also surprising is the fact that the wage premium of forwards seems to decline
over the yars. Whether this is due to changes in the supply of forwards (relative to other
positions) or to changes in the (again, relative) quality of all players under contract, re-
mains to be seen.
4.3. Econometric Findings
Estimation Methods
We start with the estimation of an OLS model (with robust standard errors), a Random-
Effects model as well as a Median Regression model8. We then present the findings of
various quantile regressions (.10, .25, .75, .90) with and without bootstrapped standard
errors (200 repetitions). The results are comparable to those obtained from OLS as well
as RE- and MR-estimation. However, few of the coefficients remain constant over the
percentiles.
The model to be estimated is of the following general form:
lnPAY = α0 + α1 AGE + α2 AGE2 + α3 GPL + α4 CGP + α5 CGP2 + α6 CGP3 + α7 IAL+
α8 IAL2 + α9 IAL3 + α10 IAP + α11 IAP2 + α12 IAP3 + α13 GSL + α14 CGS +
α15 CGS2 + α16 CGS3 + α17 TEN + α18 CAP + α19 FDD + α20 PD + α21 RD +
α22 TD + α23 YD + ε
where AGE: Player Age
GPL: number of appearances in Bundesliga in last season
CGP: number of career appearances in Bundesliga
IAL: international appearances last season
IAP: international appearances in career
GLS: goals scored last season in Bundesliga
CGS: career goals scored in Bundesliga
CAP: captain of team (0=no; 1=yes)

8 Although the Hausman-Test suggests using the results from the fixed effects estimation, we report the
findings of the random effects estimation. The problem is that region of birth is a constant for each
player and cannot be used in a fixed effects estimation. However, the differences between the remain-
ing coefficients in the RE- and the FE- estimations are negligible.
FDD: previous team in first division abroad (0=no; 1=yes)
PD: position dummies (ref.: goalkeeper)
RD: region of birth dummies (ref.: Germany)
TD: team dummies (ref.: Borussia Moenchengladbach)
YD: year dummies (ref.: 2001/02)
Thus, our models distinguish between a player’s career performance and his most recent
(i.e. last season) performance. The most recent performance (measured by, inter alia, the
number of games played, the number of international appearances and the number of
goals scored) is, of course not included in the career performance (the results of our
OLS, RE and MR estimations are displayed in Table 1 below)9.
Figure 3
Kernel Density Estimate of Player Salaries
0.1 .2 .3 .4
Density
10 12 14 16
lnpay
Kernel density estimate
Normal density
Kernel density estimate

9 Contrary to the situation in most American team sports leagues with their abundance of performance
figures, measurement of individual player performance in (European) football can be problematic es-
pecially for defenders whose task it is to prevent the opposing team’s forwards to score goals. While
counting the number of goals scored, shots on goal and assists is straightforward, it is far more diffi-
cult to assess the performance of defensive players. We therefore present our estimation results sepa-
rately for the different groups of players (see Table A2 in the Appendix).
Most studies of pay determination in football rely on the standard conditional expecta-
tions model. However, the focus on the conditional mean is likely to misrepresent the
relationship between pay and performance if there are differences in the returns to per-
formance along the conditional distribution. Several studies of salary determination in
other professional (North American) team sports use quantile regression estimation
since log salary measures tend to have even greater kurtosis values than standard occu-
pations (Hamilton 1997, Reilly and Witt 2007, Berri and Simmons 2009, Simmons and
Berri 2009, Leeds and Kowalewski 2001 Vincent and Eastman 2009). OLS salary re-
gressions are sensitive to the presence of outliers and can be inefficient if the log salary
measure has a highly non-normal distribution as is often the case in professional team
sports. In contrast, quantile regression estimates are more robust. Presence of non-
normality is indicated by a large kurtosis value and the D’Agostino et al. (1990) test is
performed by the sktest command in Stata 10.1. In our panel, the p-value for the test
statistic of the null hypothesis that kurtosis does not depart from the value associated
with a normal distribution is 0.000 and hence our log salary data depart from normali-
ty10, a result that is similar to those found in some studies of North American sports
(e.g. Berri and Simmons, 2009 on NFL). One further advantage of quantile regression is
that it facilitates examination of salary returns to characteristics at different points in the
salary distribution. That is, we can investigate the impacts of the available performance
measures at any quantile of the salary distribution, not just the conditional mean. More-
over, the quantile regression approach is semi-parametric in that it avoids assumptions
about the parametric distribution of the regression error term, an especially suitable fea-
ture where the data are heteroskedastic as in our case. To ensure robustness of standard
errors, we bootstrap with 200 replications. We report quantile regression estimates in
Table 2.
Results
Our main findings can be summarized as follows (see Tables 1 and 2 as well as Figures
4-15 below):

10 This is surprising insofar as the Kernel density estimate of log of player salaries (see Figure 3 above)
seem to suggest that our dependent variable shows a normal distribution.
Table 1
Estimation Results I: Various Methods
Variable Random Effects Robust OLS Median Regression
B T B T B T
AGE .5121 22.43*** .4559 18.99*** .4361 23.71***
AGE2 -.0092 -21.48*** -.0083 -18.69*** -.0079 -23.12***
GPL .0191 25.66*** .0240 31.95*** .0226 33.12***
CGP .0042 7.48*** .0056 11.27*** .0057 12.46***
CGP2 *100 -.0021 -5.97*** -.0028 -9.18*** -.0030 -10.26***
CGP3 * 10000 .0033 5.46*** .0043 8.07*** .0046 9.06***
IAL .0848 6.86*** .0903 6.04*** .0909 8.02***
IAL2 -.0071 -3.56*** -.0081 -2.79*** -.0094 -5.01***
IAL3 .0002 2.19 ** .0002 1.74 * .0003 4.09***
IAP .0118 4.19*** .0125 5.36*** .0131 5.94***
IAP2 -.0003 -3.40*** -.0003 4.17*** -.0003 -4.48***
IAP3 *1000 .0017 2.99*** .0016 3.67*** .0016 3.67***
GSL .0444 14.24*** .0465 16.28*** .0513 18.26***
CGS -.0129 -4.71*** -.0114 -4.69*** -.0077 -3.56***
CGS2 .0002 4.13*** .0002 4.38*** .0001 3.31***
CGS3 * 1000 -.0011 -3.68*** -.0011 -4.15*** -.0007 -3.11***
TEN -.0142 -4.43*** -.0187 -6.46*** -.0153 -6.53***
CAP .2692 6.60*** .3406 10.17*** .3718 10.50***
FDD .5910 12.46*** .6159 11.41*** .6346 15.11***
DEF .2113 5.17*** .0990 3.20*** .0539 2.24 **
MID .2677 6.65*** .1667 5.34*** .0965 4.04 **
FOR .3157 7.14*** .2167 5.97*** .1020 3.68 **
S_AM .4494 8.23*** .3778 9.87*** .3824 11.91***
N_AM -.0822 -0.73 + -.1785 -1.92 * -.1510 -2.10 **
W_EU .2442 6.62*** .1848 7.00*** .1969 8.53***
E_EU .0774 2.23 ** .0329 1.36 * .0200 0.95 +
AFR .0654 1.24 + -.0117 -0.30 + -.0166 -0.52 +
AS_AU .0928 1.28 + .0099 0.20 + .0185 0.42 +
CONST 5.8725 19.30*** 6.8245 21.14*** 7.1631 29.21***
Team Dummies included
Season Dummies included
N of Obs. 6,147 6,147 6,147
Obs. per Player 1-13 --- ---
N of Players 1,993 --- ---
R2*100 61,7 62,7 40,5
F-value --- 164.5*** ---
Wald Chi2 6,672.0*** --- ---
LM-Test 392.0*** --- ---
Raw Sum of Dev. --- --- 4,656.6
Min Sum of Dev. --- --- 2,772.6
+ not significant; * p < .10; ** p < .05; *** p < .01
Table 2
Estimation Results II: Quantile Regressions
Variable .1 Quantile .25 Quantile .75 Quantile .9 Quantile
AGE .5415*** .5485*** .3660*** .2829***
AGE2 -.0097*** -.0099*** -.0068*** -.0055***
GPL .0347*** .0271*** .0173*** .0124***
CGP .0050*** .0058*** .0047*** .0030***
CGP2 *100 -.0027*** -.0034*** -.0021*** -.0001 **
CGP3 * 10000 .0042*** .0057*** .0030*** .0013 +
IAL .0340 ** .0568*** .1241*** .1129***
IAL2 -.0003 + -.0034 * -.0149*** -.0114***
IAL3 .0000 + .0000 + .0006*** .0004***
IAP .0108*** .0119*** .0126*** .0122***
IAP2 -.0002 ** -.0003*** -.0002*** -.0002 *
IAP3 *1000 .0014 ** .0019*** .0013*** .0009 +
GSL .0453*** .0511*** .0486*** .0425***
CGS -.0094 ** -.0038 + -.0132*** -.0077 *
CGS2 .0002*** .0000 + .0003*** .0002 **
CGS3 * 1000 -.0014*** .0000 + -.0001*** -.0009 **
TEN -.0134*** -.0181*** -.0201*** -.0177***
CAP .3662*** .3742*** .3114*** .3296***
FDD .7485*** .6895*** .5848*** .4772***
DEF .2154*** .1049*** -.0002 + -.1560***
MID .2414*** .1458*** .0756*** -.0537 +
FOR .2832*** .1634*** .1111*** -.0170 +
S_AM .3010*** .3086*** .3863*** .4230***
N_AM -.1989 + -.0509 + -.2002*** -.2519 *
W_EU .1999*** .1992*** .1637*** .1627***
E_EU .0635 * .0690*** -.0344 + .0085 +
AFR -.0153 + .0538 + -.0389 + -.0320 +
AS_AU .1296 + .1042 ** -.2022*** -.1494 *
CONST 4.6571*** 5.1341*** 8.6862*** 10.4911***
Team Dummies included
Season Du
m
mies included
N of Cases 6,147 6,147 6,147 6,147
Pseudo R2*100 43.6 42.4 39.2 39.2
Raw Sum of Dev. 2,196.5 3,891.5 3,577.2 1,934.0
Min Sum of Dev. 1,239.1 2,240.8 2,139.6 1,175.5
+ not significant; * p < .10; ** p < .05; *** p < .01
First, age, career games played, international appearances over career and international
appearances last season have a statistically significant non-linear influence on salaries.
The statistically significant coefficient of the cubic term suggests existence of “superstar
effects” (Lucifora and Simmons 2003).
A strange result is obtained for career goals scored: The coefficient of the linear and the
cubic term are significant and negative, while the coefficient of the squared term is posi-
tive and significant – a finding that is not that easy to explain …
Second, goals scored last season as well as games played last season have a significant-
ly positive and strictly linear influence on annual income, i.e. there are no decreasing
returns to either goals scored or games played.
Comparing the returns to career performance and to performance in the last season, it
appears that “historical merits” do not count very much, i.e. recent performance is – as
expected – far more important than past performance.
Third, defenders, midfielders and forwards earn significantly higher salaries than goal-
keepers. The premiums for these positions, however, differ considerably across estima-
tions: The effect is most pronounced in the RE-estimation and weakest in the MR mod-
el.
Fourth, region of birth is also important: Players from South America and Western Eu-
rope receive a considerable pay premium while players from the “rest of the world” are
neither favored nor “discriminated” against. The pay premium for South Americans and
West Europeans is not surprising: Other things equal, players from these regions attract
larger crowds (Wilson and Ying 2003) and contribute more to merchandising revenues
than other players (Kalter 1999).
Finally, team captains and players who moved from a first division club abroad to Ger-
many are paid a significant premium, too. In the former case this is obviously due to
“leadership skills” that are required for the job and that are, therefore, particularly re-
warded in the market (Kuhn and Weinberger 2005).
Figure 4
Player Age and Player Salaries (I)
Figure 5
Number of Games Played Last Season and Player Salaries (I)
Figure 6
Number of Career Games Played and Player Salaries (I)
Figure 7
International Appearances in Last Season and Player Salaries (I)
Figure 8
Career International Appearances and Player Salaries (I)
Figure 9
Goals Scored Last Season and Player Salaries (I)
Figure 10
Player Age and Player Salaries (II)
Figure 11
Number of Games Played Last Season and Player Salaries (II)
Figure 12
Number of Career Games Played and Player Salaries (II)
Figure 13
International Appearances in Last Season and Player Salaries (II)
Figure 14
Career International Appearances and Player Salaries (II)
Figure 15
Goals Scored Last Season and Player Salaries (II)
Few of the coefficients retain their magnitude across the different quantiles of the salary
distribution:
5. Summary and Implications
Players are remunerated by the market according to their innate talent and their perfor-
mance with the most recent performance being far more important than the performance
delivered years ago.
The OLS- and the RE-models explain more than 60% of the observable variance in
player salaries. This is quite high and indicates that the available performance measures
– although far from ideal – are indeed well suited for the empirical analysis. The quan-
tile regressions, in turn, demonstrate that restricting the analysis to the standard models
is problematic insofar as the focus on the conditional mean is likely to misrepresent the
relationship between pay and performance because there are considerable differences in
the returns to performance along the conditional distribution.
The analysis can – and will be – extended in different directions:
• A first example for an extension: Controlling for player age, height, position and
national league, Bryson, Frick and Simmons (2009) find that both-feet players enjoy
a pay premium of more than 50% while left-footed players receive a statistically sig-
nificant premium of 15%.
Another possible step is to estimate the different models separately for goalkeepers,
defenders, midfielders and forwards (this has been done already, discussion of the re-
sults, however, is beyond the scope of this presentation)
Moreover, the number of (previous as well as recent) international appearances can
and should be weighted by the “quality” of the respective national team, i.e. its posi-
tion in the annual ranking if FIFA. This has also been done in the meantime, but the
results are not reported here due to space constraints.
Finally, estimating the models for different sub-periods will possibly yield informa-
tion about the wage determination process over time.
Appendix
Table A1
Means and Standard Deviations
Variable Mean Std. Dev. Min. Max.
PAY 909,014 889,577 17,043 10,000,000
lnPAY 13.31 0.96 9.74 16.12
GPL 13.27 12.62 0 34
GSL 1.63 3.14 0 28
IAL 1.43 3.08 0 25
CGP 55.81 80.61 0 540
CGS 6.34 14.93 0 171
IAP 7.54 16.56 0 130
TEN 2.67 3.12 0 21
CAP 0.04 - 0 1
FDD 0.04 - 0 1
GK 0.11 - 0 1
DEF 0.28 - 0 1
MID 0.39 - 0 1
FOR 0.22 - 0 1
GER 0.58 - 0 1
S_AM 0.05 - 0 1
N_AM 0.01 - 0 1
W_EU 0.13 - 0 1
E_EU 0.16 - 0 1
AFR 0.05 - 0 1
AS_AU 0.02 - 0 1
Table A2
The Determinants of Player Salaries by Position
Literature
Becker, G.S. (1993): The Economic Way of Looking at Behavior. Journal of Political
Economy, 101, pp. 385-409.
Berri, D.J. and R. Simmons (2009): Race and the Evaluation of Signal Callers in the
National Football League. Journal of Sports Economics, 10, pp. 23-43.
Bryson, A., B. Frick and R. Simmons (2009): The Returns to Scarce Talent: Footedness
and Player Remuneration in European Football, unpublished manuscript.
Buchinsky, M. (1998): Recent Advances in Quantile Regression Models: A Practical
Guideline for Empirical Research. Journal of Human Resources, 33, pp. 88-126.
D’Agostino, R.B., A. Balanger and R.B. Jr. D’Agostino (1990): A Suggestion for Using
Powerful and Informative Tests of Normality, The American Statistician, 44, pp.
316-327.
Deutsche Fussball-Liga (2002-2009): Bundesliga-Report, Frankfurt.
Frick, B. (2007): The Football Players‘ Labor Market: Empirical Evidence from the
Major European Leagues. Scottish Journal of Political Economy, 54, pp. 422-446.
Hamilton, B.H. (1997): Racial Discrimination and Professional Basketball Salaries in
the 1990s. Applied Economics, 29, pp. 287-296.
Huebl, L. and D. Swieter (2002): Der Spielermarkt in der Fußball-Bundesliga.
Zeitschrift für Betriebswirtschaft, Ergänzungsheft 4, Sportökonomie, 72, pp. 105-
125.
Kahn, L.M. (2000): The Sports Business as a Labor Market Laboratory. Journal of Eco-
nomic Perspectives, 14, pp. 75-94.
Kalter, F. (1999): Ethische Kundenpräferenzen im professionellen Sport? Der Fall der
Fußballbundesliga. Zeitschrift für Soziologie, 28, pp. 219-234.
Kicker (1995-2008): Special Issues „Bundesliga“, Nürnberg.
Koenker, R. and G. Bassett (1978): Regression Quantiles. Econometrica, 46, pp. 33-50.
Kuhn, P. and C. Weinberger (2005): Leadership Skills and Wages. Journal of Labor
Economics, 23, pp. 395-436.
Leeds, M. and S. Kowalewski (2001): Winner Take All in the NFL. The Effect of the
Salary Cap and Free Agency on the Compensation of Skill Position Players. Journal
of Sports Economics, 2, pp. 244-256.
Lehmann, E. (2000): Verdienen Fußballspieler was sie verdienen? in: Schellhaaß, H.-M.
(ed.): Sportveranstaltungen zwischen Liga- und Medieninteressen, Karl Hofmann:
Schorndorf, pp. 97-121.
Lehmann, E.E. and G.G. Schulze (2008): What Does it Take to be a Star? The Role of
Performance and the Media for German Soccer Players. Applied Economics Quarter-
ly, 54, pp. 59-70.
Lehmann, E. and J. Weigand (1999): Determinanten der Entlohnung von
Profifußballspielern – Eine empirische Analyse für die deutsche Bundesliga.
Betriebswirtschaftliche Forschung und Praxis, 51, pp. 124-135.
Lucifora, C. and R. Simmons (2003): Superstar Effects in Sports: Evidence from Italian
Soccer. Journal of Sports Economics, 4, pp. 35-55.
Milgrom, P. and J. Roberts (1992): Economics, Management, and Organization, Engle-
wood Cliffs, NJ: Prentice-Hall.
Reilly, B. and R. Witt (2007): Determinants of Base Pay and the Role of Race in Major
League Soccer: Evidence from the 2007 League Season, Department of Economics,
University of Surrey, Discussion Paper D19/07.
Rosen, S. (1981): The Economics of Superstars. American Economic Review, 71, pp.
845-858.
Simmons, R. and D.J. Berri (2009): Gains from Specialization and Free Agency: The
Story from the Gridiron, Review of Industrial Organization, 34, pp. 81-98.
Sonnenberg, M. (2002): Als die Gehälter laufen lernten. SportBild, 2, pp. 24-29.
Vincent, C. and B. Eastman (2009): Determinants of Pay in the NFL. A Quantile Re-
gression Approach. Journal of Sports Economics, 10, pp. 256-277.
Wilson, D. and Y.-H. Ying (2003): Nationality Preferences for Labour in the Interna-
tional Football Industry. Applied Economics, 35, pp. 1551-1559.
... For example, racial discrimination in European football leagues has been highlighted repeatedly, both in terms of fan behaviour and institutionalized barriers (Bradbury, 2013;Kilvington, 2013;Cleland and Cashmore, 2014;Dixon et al., 2016). However, over the years, many scholars have investigated player discrimination on multiple fronts using various indicators, and have obtained results both affirming and negating the existence of discrimination (Reilly and Witt, 1995;Szymanski, 2000, Pedace, 2008Frick, 2008;Gallo et al., 2013;Bullough and Mills, 2014;Posso et al., 2016). Yet, most studies have not focused on salary and the relationship between pay and performance as the crux for evaluating disparities. ...
... Goddard and Wilson (2008) reported an advantage for black players in career progression and retention in English professional football. Frick (2008) found that in professional football in Germany, the players from South America enjoyed a higher premium than their colleagues did; yet, Gallo et al. (2013) found evidence of discrimination by referees against non-white players in the English Premier League. Bullough and Mills (2014) reported a decline over a twenty-year period in the appearance of English players in the English Premier League pointing to diminishing opportunities for indigenous players. ...
Article
This study contributes to research on migrant pay disparities by analysing the impact of players' domestic/foreign status on performance-based pay offered to professional footballers, to understand if foreign players benefit from a preferential labour market. We used information from publicly available data of 275 footballers who played for two consecutive seasons in the Italian league Serie A. We found that the relationship between previous and current performance was partially mediated by the current salary. This result reinforced earlier findings on the pay-performance relationship, where seasonal performance is particularly relevant. Moreover, our results show that pay discrimination does not indicate a straightforward (dis)advantage for one group, but presents a more complex picture. We have examined possible underlying reasons for these disparities and offered suggestions for further research. We conclude by discussing how clubs and managers could consider incentives to strengthen pay-performance relationships by being sensitive to the complex influence of players' origins.
... This popularity has also attracted researchers' interest. Much work in this field studies European leagues, mostly the German Bundesliga, the English Premier League, the Spanish Primera Division La Liga, and Italian Serie A and Serie B. Feess et al., 9 Frick,10 Lehman and Schulze 11 ,and Battre et al. 12 are just a few of many studies that explore the determinants of differences in players' salaries in the German Bundesliga using a cross-section or panel data covering various shorttime spans. Garcia-del-Barrio and Pujol 13,14 study Spanish La Liga's 2001-2002 season, while Lucifora and Simmons 15 work on Italian Serie A and Serie B during the 1995-1996 season. ...
... Otherwise, it will cause a multicollinearity problem, as players who play more games will have higher statistics for the number of goals scored, assists, offsides, fouls committed, and yellow and red cards received. Battre et al. 12 found that in German Bundesliga the most recent performance, i.e. in the last season, has a greater impact on players' salaries than a previous career performance. Hence, in the first model here, the number of games started in the first eleven in the previous season and the number of games as a substitute in the last season are used as performance indicators rather than the current year. ...
Article
Professional soccer is the world’s most popular sport; a number of National Leagues are under the control of National Associations. The economic theory behind soccer is the continuing competition to earn much more than other sports do in the sports market. Since the supply of talent is limited, teams’ demand for certain professionals is so strong that it leads to salary differences between players. Therefore, in this study, attention is given to the determinants of the differences in workers’ salaries in the Major League Soccer labor market using Generalized Least Squares (GLS) estimation on panel data from 2007 to 2016. Birth place is the most influential determinant of a player’s salary, along with a player’s position, a player’s age, whether the player has a national team duty, and the number of games in which the player started in the first eleven. Conversely, moving from one Major League Soccer team to another and the number of games played as a substitute have a negative effect on players’ salaries.
... This layoff also results in considerable financial costs to the club. Based on the average salary of strikers in the Bundesliga in the 2006/07 season, a club incurs 42,647 euros in wages per game for every player unable to compete (12). Secondly, professional football rules allow each team to make three substitutions every game. ...
Article
Full-text available
This study aimed to explore the effect of DST implementation on key performance variables and match events in football games played in the top three football leagues in Germany.
... The results estimate the additional salary for team captains to be 20 to 30 per-cent, depending on the specification of the model. A study by Battré et al. (2009) follows up on this approach, identifying the wage premium for captains of German Bundesliga teams to be between 25 and 35 percent, again including a variety of control variables. ...
Article
Full-text available
Football players’ wages are, in most clubs, the largest share of expenses and one of the most important incentives to keep players motivated. How are football players encouraged to constantly improve their performance and what is the effect of good performances on wages? Are wage spreads within teams being set to motivate players, as in a rank-order tournament? Using accurate data from Portuguese First League football player wages, we study the relationship between wages and players’ performance. Using minutes, number of games and performance statistics during the season as a proxy for players’ performance, we show that players who play more minutes earn a wage prize and those at the top face a convex relationship suggesting the setup of a tournament within football teams.
Thesis
Full-text available
The world of football is followed by billions of people around the world. The financial stake of this sport goes far beyond the understanding of many fans. Many criticisms occurred in the last few years due to some tremendous salaries given to the football players. To understand this up-to-date subject, this paper will investigate what are the determinants of these high wages. We found that the wages are mainly affected by the income of the clubs but also the position on the pitch, the image a player brings, his age, and the grade he receives (which includes goals, assists etc).
Article
The collection and analysis of observational and experimental data represent the main tools for assessing the presence, the extent, the nature, and the trend of discrimination phenomena. Data analysis techniques have been proposed in the last 50 years in the economic, legal, statistical, and, recently, in the data mining literature. This is not surprising, since discrimination analysis is a multidisciplinary problem, involving sociological causes, legal argumentations, economic models, statistical techniques, and computational issues. The objective of this survey is to provide a guidance and a glue for researchers and anti-discrimination data analysts on concepts, problems, application areas, datasets, methods, and approaches from a multidisciplinary perspective. We organize the approaches according to their method of data collection as observational, quasi-experimental, and experimental studies. A fourth line of recently blooming research on knowledge discovery based methods is also covered. Observational methods are further categorized on the basis of their application context: labor economics, social profiling, consumer markets, and others.
Article
Discrimination data analysis has been investigated for the last fifty years in a large body of social, legal, and economic studies. Recently, discrimination discov-ery and prevention has become a blooming research topic in the knowledge discov-ery community. This chapter provides a multi-disciplinary annotated bibliography of the literature on discrimination data analysis, with the intended objective to pro-vide a common basis to researchers from a multi-disciplinary perspective. We cover legal, sociological, economic and computer science references.
Article
Within society, there are many publications regarding expectations of firm performance. These include analysis, industry insiders and ranking service. Within sports, bookmakers provide betting odds that reflect the expected outcome of matches and are determined by several factors. The purpose of this study is to analyse the determinants of expected match outcome (i.e. betting odds) and compare them to those of actual match outcome. Data from the 2010/2011 season of the German Bundesliga (soccer) are made available that are combined with betting odds for each match. Regression results show that several factors, such as difference in standing between team and opponent, difference in competitions played, difference in points earned over the last five matches, differences in standing last season and home advantage significantly influence expected match outcome. On the contrary, actual match outcome is only determined by the latter two factors.
Article
Full-text available
In an earlier paper, Kowalewski and Leeds showed that free agency and the salary cap brought profound changes to the level and nature of players' salaries in the National Football League (NFL). Their study is limited, however, by the fact that--unlike most other professional athletes--football players are evaluated by position-specific statistics. The authors improve on their earlier work by performing quantile regressions on data for specific positions to show how free agency and the salary cap affected compensation. They show that the new bargaining regime greatly increased the reward to performance.
Article
Full-text available
We investigate the salary returns to the ability to play football with both feet. The majority of footballers are predominantly right footed. Using two data sets, a cross-section of footballers in the five main European leagues and a panel of players in the German Bundesliga, we find robust evidence of a substantial salary premium for two-footed ability, even after controlling for available player performance measures. We assess how this premium varies across the salary distribution and by player position.
Article
For testing that an underlying population is normally distributed the skewness and kurtosis statistics, √b1and b2, and the D’Agostino-Pearson K2 statistic that combines these two statistics have been shown to be powerful and informative tests. Their use, however, has not been as prevalent as their usefulness. We review these tests and show how readily available and popular statistical software can be used to implement them. Their relationship to deviations from linearity in normal probability plotting is also presented.
Article
Using data from the pre-2004-2005 lockout period, we use quantile regression to estimate the earnings function of forwards and defensemen in the National Hockey League (NHL). We find that the explanatory power of Mincer's earnings equation is smaller for low-paid players than for high-paid stars. More importantly, we find significant differences in the returns to measures of performance and other variables across the conditional earnings distribution. Our estimation results suggest that the conditional expectation model used in previous studies misses some of the subtleties of the earnings determination process in professional hockey.
Article
This article investigates wage determination among professional soccer players appearing in the Italian league. Given the popularity of "top" soccer players, the relationship between individual productivity and pay can lead to "superstar" effects. In that context, the marginal revenue product of a soccer player is related to the extra price that a spectator is willing to pay to see him play (live or on television) times the number of spectators who are attracted. The authors use rare data on individual earnings and other personal characteristics of a set of soccer players in the 1995-1996 Italian league season to estimate human capital earnings equations and test for superstar effects in wage determination via convexity of earnings in performance. Earnings are found to be highly convex in two performance measures after controlling for a set of personal characteristics and team fixed effects.
Article
This paper provides a guideline for the practical use of the semi-parametric technique of quantile regression, concentrating on cross-section applications. It summarizes the most important issues in quantile regression applications and fills some gaps in the literature. The paper (a) presents several alternative estimators for the covariance matrix of the quantile regression estimates; (b) reviews the results for a sequence of quantile regression estimates; and (c) discusses testing procedures for homoskedasticity and symmetry of the error distribution. The various results in the literature are incorporated into the generalized method of moments frame-work. The paper also provides an empirical example using data from the Current Population Survey, raising several important issues relevant to empirical applications of quantile regression. The paper concludes with an extension to the censored quantile regression model.
Article
We investigate the salary returns to the ability to play football with both feet. The majority of footballers are predominantly right footed. Using two data sets, a cross-section of footballers in the five main European leagues and a panel of players in the German Bundesliga, we find robust evidence of a substantial salary premium for two-footed ability, even after controlling for available player performance measures. We assess how this premium varies across the salary distribution and by player position.