On the influence of foreign players on the success of football1 clubs

On the influence of foreign players on the success of football

clubs

Vicente Royuela

Department of Econometrics and Statistics, University of Barcelona.

Avda Diagonal 690. 08034 Barcelona (Spain) Fax +34 934021821

[email protected]

Roberto Gásquez

Department of International Economics, University of Barcelona

Avda Diagonal 690. 08034 Barcelona (Spain)

[email protected]

ABSTRACT

This paper analyzes whether having more foreign players may influence the performance

of football teams. We do so by comparing a cross section of close to one thousand football

clubs all over the world. We find that, on average, those teams in leagues with more

foreign players display better results in the World classification. Nevertheless, within

every league, where all teams have the same regulations, having more foreign players has

no effect. In the end, when all teams have the same possibilities for importing better

players from abroad, what matters is the financial power to choose the better ones.

JEL codes: C33, C53, L83

Key words: Elo rating, football success, FIFA ranking, international migration

Acknowledgements

V. Royuela thanks the support of ECO2016-75805-R.

That is soccer in North America.

1. INTRODUCTION

Football is the most globalized sport in the world

, and it is an important part of the global

economy

. The importance of football is such that it can even be used as an indicator of

social development (Gásquez and Royuela, 2014).

Nowadays international migration is a very important phenomenon, and football is not

alien to this reality. Football clubs try to hire the best players, no matter where they come

from, while football players aim to join the best clubs to enjoy better salaries and

professional prospects. In a globalized sport such as football, talent can be anywhere and

what results is an international dimension, probably larger than in any other profession.

This article analyzes the impact of such migration flows on performance, i.e. if having a

larger proportion of foreign football players significantly influences the success of

football clubs. Most papers analyzing the impact of foreign football players are addressed

at the national team level. Our contribution expands current knowledge by considering a

comprehensive data set of international clubs all over the world that allows for conducting

both national comparisons and a detailed analysis at the club level.

Our results confirm that having more foreign football players favors the performance of

clubs at the international level, although such influence vanishes within each individual

national league, where every club faces the same level of restrictions in hiring foreign

talent. Having more foreign players only has a positive effect for clubs in football

confederations where a learning process can ultimately benefit home clubs. On the other

hand, in better-ranked leagues we do not observe any benefit once we account for local

football norms, as all clubs have the same possibilities for hiring better players which

makes it, in the end, a financial issue.

Next, section II reviews several facts and the existing literature on the topic. Section III

introduces the theoretical-analytical framework used in this research. Section IV presents

the data sources. Section V sets out the empirical model and presents the estimation

results, several additions to the model and a sensitivity analysis and robustness checks.

Finally, section VI offers some conclusions.

2. LITERATURE REVIEW AND STYLISED FACTS

Several works have studied the phenomenon of migration in football. Specifically, they

have focused on analyzing the effect of footballer migration on the performance of

national teams.

According to FIFA, 2014 Brazil’s World Cup reached 3.2 billion people, and one billion watched the

final.

Dimitrov et al. (2006), cited by the European Commission’s White Paper on Sport, estimates that the

sports industry in the European Union accounts for a global impact of 3.7 percent of total GDP and 5.4

percent of total employment.

A few recent studies (Baur and Lehman, 2007; Gelage and Dobson, 2007; Berlinschi et

al., 2013; Yakamura, 2009; Allan and Moffat, 2014) have investigated the benefits of

having national association football players playing in clubs outside their domestic

league. Competing in higher quality leagues allows them access to better training and

tactical methods, and players who play abroad improve the performance of the national

team. As opposed to these authors and contrary to conventional economical wisdom,

Frick (2009) finds that the migration of players to the financially rewarding leagues in

Western Europe does not improve national team performance.

Baur and Lehman (2007) examine the effect of having a large proportion of foreign

players on the performance of the national team. Contrary to public opinion, they defend

that having more foreigners in your league may result in the sporting success of your

national team. These authors conclude that imports in a football league improve the

performance of the national team, because players benefit from knowledge-spillovers.

Imported players have some skills or qualities from which other players can learn and

benefit. They suggest as a future research agenda extending the study of the effect of

imported players on football clubs, research that, to our knowledge, had not been

addressed yet.

Along the same lines, Alvarez et al. (2011) look at whether there is an impact on the

performance of a national basketball team from having non-domestic players within the

national leagues. When skilled labor is imported, skill levels of local workers may be

raised by contact with new techniques and practices. With the study of European

basketball, the authors demonstrate that an increase in the number of foreigners in a

domestic league tends to generate an improvement in the performance of the national

team.

Migration and labor mobility, is the human side of the agglomeration story.

Consequently, we can see the positive effects of these flows in terms of the three sources

of agglomeration economies reported by Duranton and Puga (2004). Among them, one

can expect that the matching effect dominates: stronger and more successful clubs,

usually with higher financial resources, are the ones expected to hire the best players,

regardless of their origin. Still, as reported in many of the papers studying the impact on

national teams’ performance, the learning effect can be substantial, through knowledge-

spillovers, which can take place both at club level and at national level. Finally, sharing

common legal and administrative frameworks within a national league or international

environment (such as UEFA’s Champions League) can help to exploit fully the market

potential of foreign players, by having a larger global audience worldwide or by

improving a club’s merchandising sales.

Agglomeration economies also have several other consequences, for example

distributional (Behrens and Robert-Nicoud, 2014). On this theme, Milanovic (2005)

focuses on the impact of football players’ international migration on inequality between

clubs. He develops a theoretical model predicting that opening of football markets reduces

inequality between national teams due to skills spillover between players. Binder and

Findlay (2012) study the effects on competitive balance of the Bosman Ruling on

National and Club Football in Europe.

According to their results, the competitive balance

in domestic leagues has not decreased over time. That is, imported players have gone to

a variety of clubs, not just the top clubs. In another area of research, Kleven et al. (2013)

analyze the effects of tax on international migration. The authors found evidence that

football migration is conditioned by taxes.

In fact, legal barriers and conditionings for migration are one of the key aspects to be

considered. According to data from the Football Observatory, since the Bosman ruling

the percentage of foreign players recruited by clubs in the “Big Five” European football

leagues

increased from around 19% in the 1995/1996, to around 46% in the 2014/2015.

In recent years we see that some clubs have come to have more than 90% international

migrant players (e.g. Swansea F.C. in the Premier League).

This reality has turned into a debate in the media. Attitudes towards migration of

footballers raise several issues related to the political economy of high-level sport, but

also raise broader questions about national identity, citizenship, freedom of work and the

inclusion or exclusion of foreigners in local labor markets (Taylor, 2006).

Both UEFA and FIFA have tried to, and in fact have partially been able to, limit the

number of foreigners in order to preserve the national identity of clubs. Critics argue that

excessive mobility threatens the configuration of local identities and worsens national

football team performance: former FIFA president, Sepp Blatter, defended that having

more foreigners is neither good for the development of football, nor for the education of

young players, and supported FIFA in opening the door to foreign players but not so much

that this identity is lost. Other examples of this attitude can be found in the words of the

former Italian prime minister and AC Milan president, Silvio Berlusconi, who said he

dreamed of seeing his club without foreigners. Regarding this, Giulianotti and Robertson

(2004) note that this process of globalization in football has as its counterpart a growing

sense of dispossession among fans, and stress the importance of maintaining the balance

between globalization and identity.

Another line of critique is the negative impact for national teams of excessive volumes of

foreign players: despite the Premier League being considered one of the best leagues in

the world, and English clubs among the strongest in Europe, the English team does not

achieve similar success. Several voices blame the increase of foreign players in the

Premier League clubs for such weak National team results. In this context, several football

The Bosman ruling established freedom of movement football players, as workers, within the European

Union. In December 1995, the European Court of Justice ruled that the provision, whereby out-of-contract

players could only move between two clubs in different European (EU) countries if a transfer fee was

agreed between the clubs, was incompatible with Article 48 of the ‘‘Treaty of Rome’’ which relates to

freedom of movement of labor. Moreover, Article 48 was also ruled as incompatible with restrictions on

the number of foreign players permitted in a team.

The Bosman jurisprudence was later extended to citizens of European countries that were not European

Union member states by the Malaja, Kolpak and Simutenkov cases and to citizens of African, Caribbean

and Pacific countries by the 2000 Cotonou agreement.

The “Big Five” are England, Spain, Germany, Italy and France.

federations restrict the entry of foreign players, aiming to ensure the success of national

football teams. With the intention of restricting the number of foreign players, in 2000,

FIFA and UEFA sought support from the European Parliament to amend the Amsterdam

Treaty, and recognize football as a cultural activity to stop the effects of the Bosman

ruling.

In 2008, FIFA approved the application of the “6 + 5 rule” to force clubs to field six

players eligible for the national team to protect the identity of national teams. However,

the European Commissioner for Employment, Valdmiri Špidla, challenged this idea

because "players are workers and the principles of free movement must be respected. The

rule of '6 + 5' constitutes direct discrimination"

and that the European Commission would

take legal action against any country that approved the controversial proposal by FIFA to

limit the number of foreigners in football clubs.

Finally, a different rule was created for the "protection of young players": since 2008/09,

clubs in the UEFA Champions League and UEFA Europa League are required to include

a minimum of eight homegrown players of the country in a squad limited to 25 players.

UEFA defines "homegrown players in a country" as those who, regardless of their

nationality, have been trained by such club or by another club in the same national

association for at least three years when the player was between 15 and 21 years old.

UEFA regulations have no conditions of nationality, since those conditions would be

illegal in the European Union (the Bosman ruling).

In line with this European policy, many countries have restrictions on foreign players,

varying extensively within the same area. As for Latin America, Argentina allows only

four foreigners in the club, whereas in Brazil the number is three, in Chile seven, five for

Mexico and six for Peruvian clubs. European countries face a huge diversity of rules for

players from EU-countries and non-EU countries: no quota for non-EU-players (Austria,

Belgium, England

, Germany, Poland, Portugal, Scotland, Serbia, Wales, The

Netherlands), no quota for non-EU-players but only a certain amount can be brought to

the games (Czech Republic, Croatia, Denmark, Finland, Hungary, Iceland, Russia,

Sweden, Slovakia) and a limited amount of foreigners/non EU players (Belarus, Bulgaria,

France, Greece, Israel, Italy, Norway, Romania, Spain, Turkey). For example, Spain only

allows for three non-EU players. The differences between EU and non-EU players have

As established by the EU treaties, the principle of free movement may not apply to cultural activities,

since culture is one of the areas, along with defense, which are not subject to these rules, considering them

outside the economic space and responsibility of each State. For this reason, some governments of EU

countries have requested the declaration of sport as a cultural activity, with the aim of maintaining this area

outside the regulations imposed by the court.

Statement of Commissioner Vladimir Špidla regarding FIFA’s “6+5” rule, accessible at

http://ec.europa.eu/social/main.jsp?langId=en&catId=89&newsId=424&furtherNews=yes

Clubs have no obligation to play a certain number of homegrown players in a match in the national league.

England only allows the entry of foreign players if they play regularly in some of the best 70 national

football teams of the world.

resulted in seeking an EU passport for players as a way to avoid restrictions and to be part

of the more competitive leagues in the world.

As reported above, most academic literature addressing this phenomenon is concentrated

on analyzing the impact of foreign players on a national team’s performance. To the best

of our knowledge, to date only Karaka (2008) has focused on the study of the impact of

international migration on club performance in a small data set and with no information

at the club level. Our work expands current knowledge by considering a comprehensive

and wide database at the international level of about one thousand clubs worldwide.

3. FROM THEORY TO EMPIRICS

Bernard and Busse (2004) develop a theoretical framework to analyze the determinants

of success in sport, specifically at the Olympic Games. On the other hand, many other

works have studied the determinants of national football team success.

Gásquez and

Royuela (2016) apply the model developed by Bernard and Busse (2004) to a national

football team’s performance. We follow this theoretical framework, where the empirical

expression is as follows:





= 



+ 







+ 







+ 







+ 







+ 



(1)

Where 



refers to football success, 



to population, 



to economic resources, 



warm temperature, 



to football-related institutions at country level, and 



is error term.

This model assumes that the talent of players is randomly distributed and consequently

we would expect success to be proportional to population. Still, despite having a large

population, wealthy countries are more likely to have public and/or private organizations

willing to invest in improving good players. The variable for temperature follows

Hoffman et al. (2002), who claim that the optimal mean annual temperature for sporting

practice is 14º C and that shifts from this temperature can hinder success. Football rather

than political institutions (Leeds and Leeds, 2009) may be connected with football

performance (as in Gásquez and Royuela, 2016).

These variables are suitable for studies at national team level, but they may not be enough

for analyzing performance of clubs, as they do not capture differences within national

leagues. Consequently, we adapt these variables to analyze the determinants of

Some of the most striking cases of passport forgeryto obtain dual citizenship are: the Argentinian Veron,

convicted of falsifying his Italian passport for an alleged great grandfather, the Brazilian Dida and

Uruguayan Alvaro Recoba who received penalties for passport fraud. Many other players have found a way

to obtain dual nationality. For instance, Spain allows for citizenship after several years of legal and

continuous residence (just two years for nationals of Latin American countries, Andorra, Philippines,

Equatorial Guinea, Portugal or persons of Sephardic origin).

Hoffman et al. (2002), Houston and Wilson (2002), Torgler (2004a), Torgler (2004b), Torgler (2004c),

Hoffman et al. (2006), Macmillan and Smith (2007), Gelade and Dobson (2007), Leeds and Leeds (2009),

Yamamura (2009), Binder and Findlay (2012), Berlinschi et al. (2013), Allan and Moffat (2014) and Jacobs

(2014) have studied the determinants of the success of national football teams.

See Additional Material number 1 for model development.

performance at the club level. Castellanos et al. (2007) analyze the determinants of

success taking cities instead of countries as units of analysis. They consider that success

at club level is a function of the size and wealth of its city.

They consider these factors

using population and local GDP per capita. These authors end their work by arguing that

future studies should address the importance of inherently non-economic factors of the

city/club, such as culture, weather conditions, institutions or historical excellence

(tradition) in the context of football performance.

In addition to considering GDP per capita as a proxy of socioeconomic conditions, we

prefer to consider the economic ‘power’ of every club, proxying the capacity to hire more

and better (foreign) players. Szymanski (2003) and Osso and Szymansky (1991) reports

a positive relationship between expenditure on player’s wages,transfers and position for

twelve English Clubs. Likewise, Fløtnes (2011) argues that the more important factors

for clubs’ success are player wages and financial resources through operating income.

When competing in the elite division or at international level, access to financial resources

partly determines how successful a football club can be. The Economist (2014) illustrates

a strong link between the amount spent on wages and the points won by 34 English clubs

that played in the top division between 1996 and 2014.

For these reasons, we focus the explanatory variables of the equation (1) in our study at

the club level by considering 



, the economic resources of the club, 



, the population of

the city where each club is located, 



, an index that takes into account the average

temperature of each city, and 



, an indicator of local football institutions, that we proxy

with confederations and country dummies.

In addition to these variables, we consider the proportion of foreign football players, 



which allows us to capture the impact of migration on football clubs´ success. We cannot

distinguish with our approach between learning, sharing and matching. Rather, we simply

account for a global impact of this variable. We finally add more controls by considering

the social engagement of clubs through the capacity of the Stadium, 



, as a proxy of

attendance.

Finally, then, our equation becomes:





= 



+ 







+ 







+ 







+ 







+ 







+ 



 



+ 



(2)

4. DATA

Our empirical strategy relies on the use of a worldwide dataset. It is difficult, though, to

find official comparable statistics for capturing success of football clubs in an

Walker (1986), Burger and Walters (2003), Troelsen (2005) and Fløtnes (2011) argue that the most

populous cities offer a greater internal potential market for their football teams.

As anecdotal evidence, Sam Allardyce, former manager of West Ham United and English football

manager, came up with a straightforward explanation for footballing performance: “Where you finish in

the league depends on the money you’ve spent. It’s a statistical fact, that.”

international environment. Other works in football literature have worked with points in

domestic leagues, rankings at the European level, etc., which makes it difficult to make

worldwide comparisons. Our final database considers 971 clubs from the First Divisions

of 71 Leagues.

We finally rely on the Elo rating score published by

www.footballdatabase.com. The data set referred to is February 2016. This ranking

follows the methodology Elo Rating, also followed in Gasquez and Royuela (2016), who

show its advantages over official scores, such as FIFA ranking.

The top ten clubs

according to this ranking in February 2016 are displayed in Table 1, while Figure 1 shows

the boxplot of the clubs according to every football confederation

Table 1. World top 10 clubs according to

Elo Ratig score. February 2016.

League Club Ranking

Elo

Points

Spain

FC Barcelona

2082

Germany

Bayern Munich

1989

Spain

Real Madrid

1967

France

Paris Saint-Germain

1958

Italy

Juventus FC

1942

Spain

Atlético Madrid

1908

Italy

SSC Napoli

1855

England

Arsenal FC

1822

England

Tottenham Hotspur

1818

Germany

Borussia Dortmund

1810

Figure 1. Box plot of Elo rating

score by Football Confederation

Table 2. Top clubs and national leagues by share of foreign players

Club (National League)

(%) Foreign

players

National League

(%) Foreign

players

AS Monaco (France)

100%

Canada

77.98%

Watford FC (England)

93.10%

England

69.34%

Swansea City (England)

91.30%

Cyprus

58.31%

Inverness Caledonian Thistle FC (Scotland)

88.00%

Belgium

56.36%

NK Zavrc (Slovenia)

85.19%

Portugal

55.65%

FC Vaduz (Switzerland)

84.62%

Italy

55.48%

Chelsea FC (England)

84.00%

Luxembourg

55.43%

Stoke City (England)

84.00%

Switzerland

52.41%

Manchester City (England)

83.33%

Germany

49.54%

Inter Milan (Italy)

82.61%

France

49.30%

The explanatory variables refer to 2015. To find the proportion of foreign players, we

used information at www.transfermarkt.es. Table 2 displays the top ten clubs and top ten

leagues by share of foreign football players, while Figure 2 displays the box-plot of this

variable by confederation. On average, the confederation with the largest share of foreign-

Appendix 1 lists the considered football leagues.

A detailed analysis of the methodology for calculating the Elo rating is shown in Additional Material 2.

UEFA, CONCACAF, CONMEBOL, AFC, CAF, and OFC. Supporting Information Additional Material

3 provides details of the countries in each confederation.

1,200 1,400 1,600 1,800 2,000

Points

UEFA

CONCACAF

CONMEBOL

CAF

AFC & OFC

born football players is UEFA (35%), followed by CONCACAF (34%), AFC and OFC

(19%), CONMEBOL (12%) and CAF (11%). It is important to compare these figures

against global ones: about 200 million people in the world, around 3% of total world

population, live outside their country of birth. As in football, the proportion of

international immigrants is greater in developed countries: they represent more than 12

percent of the total population in OECD countries.

Figure 2. Box plot of the share of foreign players, by Football Confederation

The proxy we consider of the economic power of football clubs is the market value

published by www.transfermarkt.es. This variable is highly correlated with the budget of

clubs and consequently can be used to proxy the total wage bill.

To know which city

the football club is from, we use www.soccerway.com information. City population data

comes from Wikipedia, and refers to the administrative definition of the city rather than

corresponding to the metropolitan of functional urban area. We follow Hoffman et al.

(2002) and consider as weather indicator an index considering the deviation from the

optimal mean annual temperature for sporting performance, which is settled at 14º C.

Thus, our weather variables is computed as 



= (14)



. The Average

city temperature is extracted from http://www.weatherbase.com. We will consider a list

of country (league) dummies to capture differentiated football-related institutions. The

capacity of the stadium is obtained mostly from www.soccerway.com, being

complemented using Wikipedia. The same sources are used to find the year of foundation

of the club, a variable that we will used later for identification issues, together with the

average player age at each club, also obtained from www.transfermarkt.es. Appendix 2

summarizes the description and sources of all considered variables, and Table 3 reports

the descriptive statistics of the considered variables.

Additional Material 5 displays the correlation between market value and to see why the market value of

the Squad is a good Proxy of the budget of the clubs.

0 .2 .4 .6 .8 1

% Foreign Players

UEFA

CONCACAF

CONMEBOL

CAF

AFC & OFC

At this stage, we look at the correlation between the Elo ranking score and the share of

foreign players. Figure 3 displays the scatter plot between these two variables for the 971

considered clubs. The left panel considers all information, the central panel the average

for every national league, and the right panel plots the information of all clubs once

national averages are removed (within transformation). In the first case the correlation is

0.36, which increases to 0.44 between countries and shrinks to 0.28 when national

averages are discounted (correlations for the log of the Elo rating are very similar).

Consequently, we have a first insight into a positive relationship between these two

variables, which is much stronger at the national level than within each national league.

The next step will try to find out if this correlation holds once we control for other factors.

Table 3. Descriptive statistics

Variable Obs. Mean Std. Dev. Min. Max.

Corr(Xi,

ln Points)

Corr(Xi, %

For. Pl.)

Elo Points

971

1414.9

133.77

1159

2082

ln Elo Points

971

7.251

0.091

7.055

7.641

% Foreign Players

971

0.277

0.196

0.359

Market Value

971

26.14

63.01

0.025

704.8

0.623

0.434

Population

970

1,396,643

2,965,882

25,333

24,152,700

0.125

-0.079

Weather

971

33,59

43,42

.0001

289,00

-0.125

-0.139

Stadium Capacity

971

22,714.7

18,805.9

368

105,064

0.538

0.134

Year Founded

968

1,943

36.6

1,863

2,014

-0.378

-0.171

Age

971

25.6

1.66

18.5

33.6

0.082

0.191

Figure 3. Scatter plots between Elo ranking score and the Share of Foreign Players

5. RESULTS

1200 1400 1600 1800 2000

Elo ranking score

0.0000 0.2000

0.4000 0.6000 0.8000 1.0000

Overall information

1200 1300 1400

1500 1600 1700

0 .2 .4 .6 .8

Between countries

-200 0 200 400

-.4

-.2 0 .2 .4

Within countries

% Foreign Players

The estimation strategy tries to avoid endogeneity problems because of a possible causal

link between the percentage of foreigners and performance of football clubs. We have

obtained the data of foreign players playing at a club since September 2015 and the

ranking of the club in February 2016.

Thus, the causality is from the variable number

of foreign players to success in football clubs, and not vice versa. Nevertheless, we admit

that such variables may have some time series persistence, and consequently some reverse

causality can exist in the data, together with the omitted variables problem. In order to

solvethese problems, at least partially, we incorporate an instrumental variable approach

based on a two-step procedure following Brückner (2012, 2013) and Castells-Quintana

(2016). We use the year of foundation of each club to build instruments for football

success in an equation explaining the share of international football migrants. Later we

use the residual of this equation as an instrument for the share of football migrants in our

main equation, together with the average age of players in every club. We explain this

identification strategy in appendix 3.

Table 4 displays the results of the main model where we use the variables in logs. We

introduce variables sequentially and columns 1 to 7 display OLS estimates, while

columns 8 to 11 show IV estimates. We can see how in a first stage the share of foreign

players is positively associated with clubs’ success, even if external components are

controlled for (columns 1 and 2). Nevertheless, when financial and market potential

variables are introduced (columns 3 and 4), model adjustment improves dramatically and

the parameter for the share of foreign players turns significant and negative, which we

interpret as a clear sign of the strength of financial aspects in sport success. When we

introduce institutions in columns 5 (confederations dummies) and 6 (country dummies),

the share of foreign players becomes insignificant. One can interpret this result in terms

of the importance of financial variables and national regulations as main drivers on

international football rankings. When national regulations are controlled, competition for

foreign players is balanced within every country and this factor becomes negligible.

Clubs’ performance basically depends on their financial health. To be clear: while those

leagues with a higher share of foreign players have better ratings than leagues with less

imported players, within each league where all clubs face the same type of restrictions in

hiring foreign players, this factor finally becomes negligible.

Nevertheless, at this stage we have not accounted for reverse causality. Columns 7 to 10

report IV estimates using as instruments the residual of the two-step procedure based on

instrumenting the 2010 Elo rating when explaining the September 2015 share of foreign

players. This generated instrument is used together with the average age of all football

players in each club (and its square) in columns 8 to 10 to report the over identification

statistics. This set of regressions report again a non-significant parameter, which we

understand as a sort of robustness check of the previous results. Still, the insignificance

of the parameter of the interest variable might be the result of non-linearities or an omitted

This indicator considers the results of recent months. For example, on 14.02.2016 (date of obtaining the

data of this ranking) Leicester City was Leader of the Premier League and consequently was ranked 15th,

while the same week of the previous year it ranked 416th.

variables problem. We have performed a number of regressions, available upon request,

excluding variables with non-significant parameters (Population and Weather), including

the square of the log of the Stadium Capacity and even the ratio between the Market value

and the Stadium Capacity. The share of foreign players was not significant in any of these

regressions. We finally included the square of the share of foreign players in a list of

regressions with permutations of the other control variables. These regressions never

reported a significant parameter for the share of foreign players. We admit, though, that

as in any empirical work, some relevant variables, such as quality of foreign players or

the leagues, are lacking. Narrowing the scope of analysis, for instance looking at a specific

territory or confederation, would allow to include new information, an aspect that we

leave for further research.

We investigate next if our result is a global outcome or if it is specific to some World

regions or Football confederations. We perform additional regressions separated by

Football confederations. Both OLS and IV estimates are presented in table 5, including

in all cases national dummies. We see there that, in general, the share of international

football players is not significantly associated with clubs’ success. This is an additional

proof of the small impact of this variable compared with the economic ones, including

the market value and the capacity of the stadium. We also included the square of the share

of football players (results not reported), with similar results.

We finally check for robustness by using the ranking of clubs rather than Elo points. We

also use the Elo rating rather than the log of the index. The results (not reported here for

brevity but displayed in Additional Material 8) basically replicate former results, with the

exception of the marginally significant parameter for the CONCACAF subsample for the

model using the ranking rather than the Elo points. Overall, then, the benefits of having a

larger share of foreign players does not exist either in the global sample or in any

confederation.

Table 4. Estimation results

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

OLS

2SLS

% Foreign players

0.1658***

0.1780***

-0.0127

-0.0435**

-0.0159

0.0174

-0.0120

-0.0122

(0.039)

(0.034)

(0.030)

(0.022)

(0.019)

(0.015)

(0.014)

(0.013)

ln Population

0.0152***

0.0002

-0.0000

0.0004

0.0014

0.0012

(0.002)

(0.001)

Weather Index

-0.0002*

0.0001

-0.0000

(0.000)

ln Market value

0.0510***

0.0212***

0.0196**

0.0563***

0.0577***

(0.008)

(0.009)

(0.008)

(0.006)

ln Market value

0.0068***

0.0067***

0.0047***

0.0048***

(0.001)

ln Stadium Capacity

0.0097**

0.0092**

0.0092***

0.0099***

0.0098***

(0.004)

(0.003)

(0.002)

Confederation Dummies

YES

Country Dummies

YES

Kleibergen-Paap rk LM statistic

170.8 (0.000)

170.9 (0.000)

Hansen J-Statistic (p-val)

1.797 (0.180)

1.878 (0.171)

Observations

971

970

R-squared

0.126

0.246

0.604

0.658

0.680

0.807

0.806

Note: Robust standard errors clustered by country in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Equations (8) to (10) report not-clustered robust standard errors to get the

over-identification statistics. Columns (7) and (8) use the generated residual resulting from the two-step Brückner strategy described in appendix 3. Columns (9) and (10) add

the average age of players and its square respectively, what allows for computing the over-identification statistic. KP refers to the under-identification Kleibergen-Paap LM

statistic, while the J statistic corresponds to the over-identification Hansen J-Statistic.

Table 5. Estimation results by confederation.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

UEFA

CONMEBOL

CONCACAF

CAF

AFC - OFC

UEFA

CONMEBOL

CONCACAF

CAF

AFC - OFC

OLS

% Foreign players

0.00643

0.00138

0.0852

0.169*

0.0189

-0.0120

-0.0003

0.0465

0.0232

-0.0631

(0.0153)

(0.0886)

(0.0630)

(0.0869)

(0.0393)

(0.015)

(0.089)

(0.060)

(0.083)

(0.046)

ln Population

0.00201

0.00104

0.00370

-0.000964

0.00318

0.0018

0.0010

0.0038

-0.0008

0.0019

(0.00135)

(0.00221)

(0.00568)

(0.00209)

(0.00362)

(0.001)

(0.002)

(0.005)

(0.002)

(0.003)

Weather Index

4.13e-05

3.67e-05

-0.000275

7.29e-05

-1.10e-05

0.0000

-0.0003*

0.0000

-0.0000

(9.65e-05)

(0.000109)

(0.000164)

(0.000159)

(0.000140)

(0.000)

ln Market value

0.0727***

0.0910***

0.0348

0.0518***

-0.00958

0.0742***

0.0910***

0.0409

0.0535***

-0.0038

(0.00812)

(0.0292)

(0.0760)

(0.00558)

(0.0136)

(0.008)

(0.028)

(0.069)

(0.006)

(0.014)

ln Market value

0.00298***

-0.00781

0.00313

0.0134***

0.0295***

0.0030***

-0.0078

0.0025

0.0138***

0.0285***

(0.00109)

(0.00510)

(0.0148)

(0.00120)

(0.00463)

(0.001)

(0.005)

(0.013)

(0.001)

(0.005)

ln Stadium Capacity

0.00565*

0.0196***

0.0171

0.000219

0.00907

0.0057*

0.0196***

0.0169

0.0007

0.0087

(0.00321)

(0.00682)

(0.0124)

(0.00401)

(0.00915)

(0.003)

(0.006)

(0.011)

(0.004)

(0.009)

KP statistic (p-val)

131.2 (0.000)

36.56 (0.000)

9.800 (0.020)

30.72 (0.000)

13.52 (0.004)

J-Statistic (p-val)

5.517 (0.063)

3.242 (0.198)

0.427 (0.808)

2.680 (0.262)

1.639 (0.441)

Country dummies

YES

Observations

571

129

124

571

129

125

R-squared

0.715

0.603

0.427

0.596

0.556

0.858

0.630

0.589

0.706

0.688

Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All models use the generated residual resulting from the two-step Brückner strategy together with

the average Age (and its square) of players in every club. KP refers to the under-identification Kleibergen-Paap LM statistic, while the J statistic corresponds to the

overidentification Hansen J-Statistic.

6. CONCLUSIONS

This work analyses the socio-economic determinants of sporting success of football clubs

in a worldwide cross-section sample, and especially inspects the effect of the proportion

of foreign players when other factors are controlled for. We use the Elo rating as indicator

for the world ranking of close to one thousand clubs in 71 leagues. We use a two-step

procedure as an identification strategy: in a regression explaining the proportion of

foreign football players in 2015 we instrument the Elo rating in 2010 by means of the

year of foundation of every club together with the local conditions for playing football.

The residual of this equation is later used as an instrument of the share of foreign players

in a regression explaining the Elo rating in 2016.

We observe that, on average, leagues with higher proportions of foreign players are the

ones with better positioned clubs. On the contrary, within each league, having more

foreign players has a negligible impact on clubs´ performance.

As expected, we find that the fundamental explanatory variable of clubs´ success is

money. Our key finding is that having more foreign football players favors the

performance of clubs at an international level: national regulations allowing more foreign

players will result in better performance of these clubs in an international framework.

Nevertheless, such influence vanishes within each national league, where each club faces

the same level of restrictions in hiring foreign talent. In the end having better players will

be the result of financial constraints. We believe that our work can be improved by

including relevant variables such as the quality of the players and even dealing with the

quality of the leagues. Working with a panel data set would also allow for controlling for

non-observable variables at the club level.

REFERENCES

Allan, G. J. and Moffat, J. (2014). “Muscle drain versus gain in association football:

technology transfer through player migrations and manager immigration.” Applied

Economic Letters, 21(7), 490-493.

Alvarez, J., Forrest, D. and Sanz, I. (2011). “Impact of importing foreign talent on

performance levels of local co-workers.” Labour Economics, 18, 287–96.

Baur, D. G. and Lehmann, S. (2007). “Does the mobility of football players influence the

success of the national team?” IIIS Discussion Paper No. 217, Institute for International

Integration Studies, Dublin.

Behrens, K., and Robert-Nicoud, F. (2014). “Survival of the fittest in cities: Urbanisation

and inequality.” The Economic Journal, 124(581): 1371-1400.

Berlinschi, R., Schokkaert, J. and Swinnen, J. (2013). “When drains and gains coincide:

migration and international football performance.” Labour Economics, 21, 1–14.

Bernard, A. B. and Busse, M. R. (2004). “Who wins the Olympic Games: economic

resources and medal total.” The Review of Economic and Statistics, 86(1), 413-417.

Binder, J. J. and Findlay, M. (2012). “The effects of the Bosman ruling on national and

club teams in Europe.” Journal of Sports Economics, 13, 107–29.

Brückner, M. (2012) “Economic growth, size of the agricultural sector, and urbanization

in Africa.” Journal of Urban Economics, 71, 26–36.

Brückner, M. (2013) “On the simultaneity problem in the aid and growth debate.” Journal

of Applied Econometrics, 28, 126–150.

Burger, D. and Walters, S. (2003). “Market Size, Pay, and Performance.” Journal of Sport

Economics, 4(2), 108-125.

Castellanos, P., Dopico, J. and Sánchez, J.M. (2007). “The Economic Geography of

football success: empirical evidence from European Cities.” Rivista di Diritto et

Economia dello Sport, 3(2), 67-88.

Castells-Quintana, D. (2016) “Malthus living in a slum: Urban concentration,

infrastructure and economic growth. ” Journal of Urban Economics, 92, 31-47.

Dimitrov, C., C. Helmenstein, A. Kleissner, B. Moser, and J. Schindler. (2006). “Die

Makroökonomischen Effekte des Sports in Europa.” Studie im Auftrag des

Bundeskanzlermats, Sektion Sport, März. Vienna: Sports Econ Austria.

Duranton, G. and Pugba, D. (2004). “Micro-Foundations of Urban Agglomeration

Economies.” Handbook of Urban and Regional Economics, Edition 1, Vol. 4: 2063-2117,

J.V. Henderson and J.F. Thisse (eds.)

FIFA. (2014). “FIFA World Cup Brazil. Television Audience Report. Available at

http://resources.fifa.com/mm/document/affederation/tv/02/74/55/57/2014fwcbraziltvaud

iencereport(draft5)(issuedate14.12.15)_neutral.pdf

Fløtnes, T. (2011). “Factors of success for Norwegian top football clubs - And why profit

making is difficult for European football clubs.” Copenhagen Business School. M. Sc.

Applied Economics and Finance.

Frick, B. (2009). “Globalization and factor mobility: the impact of the ‘Bosman-Ruling’

on player migration in professional soccer.” Journal of Sports Economics, 10, 88–106.

Gásquez, R. and Royuela, V. (2014). “Is football an indicator of development at the

international level?” Social Indicator Research, 117(3), 827-848.

Gásquez, R. and Royuela, V. (2016). “The Determinants of International Football

Success: A Panel Data Analysis of the Elo Rating.” Social Science Quarterly, 97(2), 125-

141.

Gelade, G. A. and Dobson, P. (2007). “Predicting the comparative strengths of national

football teams.” Social Science Quarterly, 88, 244–58.

Giulianotti, R. and Robertson R. (2004). “The globalization of football: a study in the

glocalization of the 'serious life'.” The British Journal of Sociology, 55(4), 545-68.

Hoffman, R., Ging, L.C. and Ramasamy. B. (2002). “The socio-economic determinants

of the international soccer performance.” Journal of Applied Economics, 5(2), 253-272.

Hoffman, R., Ging, L.C., Matheson, V. and Ramasamy. B. (2006). “International

women’s football and gender inequality.” Applied Economic Letters, 13, 999-1001.

Houston, R.G. and Wilson, D. (2002). “Income, leisure and proficiency: an economic

study of football performance.” Applied Economics Letters, 9, 939-943.

Jacobs, J. (2014). “Programme-level determinants of women's international football

performance.” European Sport Management Quarterly, 14(5), 521-537.

Karaca, O. (2008). “The impact of foreign players on international football performance.”

Munich Personal RePEc Archive, MPRA Paper No. 11064. Online at http://mpra.ub.uni-

muenchen.de/11064/

Kleven, H.J., Landais. C. and Saez, E. (2013). “Taxation and International Migration of

Superstars: Evidence from the European Football Market.” American Economic Review,

103(5), 1892-1924.

Leeds, M.A. and Leeds, E.M. (2009). “International Soccer Success and National

Institutions.” Journal of Sports Economics, 10(4), 369-390.

Macmillan, T. and Smith, I. (2007). “Explaining International Soccer Rankings.” Journal

of Sports Economics, 8(2), 202-213.

Milanovic, B. (2005). “Globalization and goals: does soccer show the way?” Review of

International Political Economy, 12(5), 829-850.

Osso and Szymansky (1991). “Who are the Champions? (An analysis of football and

architecture).” Business Strategy Review Summer, 113-130.

Taylor, M. (2006). “Global Players? Football, Migration and Globalization, c. 1930-

2000.” Historical Social Research, 31(1), 7-30.

Torgler B. (2004a). “The Economics of the FIFA Football Worldcup.” KYKLOS, 57(2),

287-300.

Torgler B. (2004b). “The determinants of women’s international soccer performances.”

Center for Research in Economics, Managements and Arts, working paper no. 2004-19.

Torgler B. (2004c). “Historical Excellence in football world cup tournaments: empirical

evidence with data from 1930 to 2002.” Center for Research in Economics, Managements

and Arts, working paper no. 2004-18.

Troelsen, T. (2005). “Centralisering af dansk fodbold: En statistisk analyse med bud på

årsagsforklaringer. Kommunal støtte til professionel fodbold.” I J. Magnussen, & R. K.

Storm (red.), Professionel fodbold. Klim, Århus, s. 69-86.

Szymanski, S. (2003). “The Economic Design of Sporting Contest.” Journal of Economic

Literature, 41 (4), 1137-1187.

The Economist. (2014) “Everything to play for. Good managers matter, but not as much

as money does.” Available at http://www.economist.com/news/britain/21601873-good-

managers-matter-not-much-money-does-everything-play.

Walker, B. (1986). “The demand for professional league football and the success of

football league teams: some city size effects.” Urban Studies, 23(3), 209-219.

Yamamura, E. (2009). “Technology transfer and convergence of performance: an

economic study of FIFA football ranking.” Applied Economics Letters, 16, 261–6.

Appendices of the paper “On the influence of foreign players on

the success of football clubs”

Appendix 1. Included football leagues

Albania

25.

Finland

49.

Norway

Algeria

26.

France

50.

Peru

Argentina

27.

Georgia

51.

Poland

Australia

28.

Germany

52.

Portugal

Austria

29.

Ghana

53.

Qatar

Azerbaijan

30.

Greece

54.

Qatar

Belarus

31.

Hungary

55.

Romania

Belgium

32.

Iceland

56.

Russia

Bosnia-

Herzegovina

33.

India

57.

Saudi Arabia

10.

Brazil

34.

Iran

58.

Scotland

11.

Bulgaria

35.

Israel

59.

Serbia

12.

Canada

36.

Italy

60.

Slovakia

13.

Chile

37.

Japan

61.

Slovenia

14.

China

38.

Kazakhstan

62.

South Africa

15.

Colombia

39.

Korea, South

63.

Spain

16.

Costa Rica

40.

Lebanon

64.

Sweden

17.

Croatia

41.

Luxembourg

65.

Switzerland

18.

Cyprus

42.

Macedonia

66.

Tunisia

19.

Czech

Republic

43.

Mexico

67.

Turkey

20.

Denmark

44.

Moldova

68.

Ukraine

21.

Ecuador

45.

Montenegro

69.

United

States

22.

Egypt

46.

Morocco

70.

Uruguay

23.

England

47.

Netherlands

71.

Wales

24.

Estonia

48.

New

Zealand

Appendix 2. Variables: definition and sources

Variable

Description

Source

Points

Elo rating points of World Football Club

classification

http://footballdatabase.com/

Ranking

Worls ranking based on Elo rating points of

World Football Club classification

http://footballdatabase.com/

Pop

Population of the city where every club is located

Wikipedia

Market value

Market value of total players of the Team

http://www.transfermarkt.com/

Share Foreign players

Share of foreign players of each team (%)

http://www.transfermarkt.com/

Age

Average age of players of each team

http://www.transfermarkt.com/

Weather Index

(TEMP-14) squared, where TEMP refers the

weather of the city (in log)

http://www.weatherbase.com/

Foundation

Years of Foundation of the team

www.soccerway.com

complemented with Wikipedia

Capacity_Stadium

Capacity of the Stadium

www.soccerway.com

complemented with Wikipedia

CONCAFAF

Confederation of North, Central American and

Caribbean Association Football

CONCAFAF

CONMEBOL

South American Football Confederation

CONMEBOL

AFC

Asian Football Confederation

AFC

CAF

Confederation of African Football

CAF

OFC

Oceania Football Confederation

OFC

UEFA

European Union of Association Football

UEFA

Appendix 3. Identification strategy

As it is hard to find an instrument for the share of foreign players in every football team, we

follow a two-step procedure following Brückner (2012, 2013) and Castells-Quintana (2016) to

adjust for simultaneity bias. By using an instrument for the Elo ranking in 2010 we are able to

build a valid instrument for the share of foreign players in our substantive estimation. The starting

point is a simultaneous equation model where football success (



) and the share of foreign

football players (



) are mutually related:





= (



) + 



(A1)





= (



) + 



(A2)

We are interested in estimating parameter , but if  is not zero, OLS estimates in A1 will be

biased and inconsistent. To overcome this problem we propose using instrumental variables for

football migrants. If we can consistently estimate  in A2, we can build an instrument to be used

in A1 by capturing the residual: 



= (



) = 











. Using this generated variable as

instrument, the IV estimate of A1 will be free of simultaneity bias:







(



(





)

,



)



(



(





)

,



)

= +



(



(





)

,



)



(



(





)

,



)

= +



(





,



)



(



(





)

,



)

(A3)

Still, as far as 

(





, 



)

0 the omitted variable bias will exist. In order to avoid that bias, we

include in our estimate country fixed effects in A2 together with the population size of every city.

In this strategy, timing is an important aspect. Thus, our first step consists on explaining the share

of foreign football players as a function of past football success: we regress the share of foreign

footballers in September 2015 against the Elo ranking dated in January 2010. We include then a

set of country dummies (





) that proxy national institutions, such as legal barriers.







= 





+ 



+ 



(A4)

As an instrument for the Elo ranking in 2010 (





), we use information based on the history of

every club. In particular we account for the year of foundation of the club and we compute its

seniority (





). With this variable we build the following set of instruments: the rank of seniority

within every league (1), and its square (2), plus the rank of the ratio between the capacity of the

stadium and the seniority of every team (3), and its square (4). Finally, we add the weather

indicator differentiated by country. These instruments are expected to be correlated at some stage

with the success of every team, but not to affect the share of foreign players. The first stage of

equation A4 becomes:







= 



_



+ 



_





+ 



_/



+ 



_/





+ 



+ 



+ 



Table A7.1 displays the results of the first (A5) and second stage (A4) regressions. We find that

our instruments are correlated with the classification of every team within its league (remember

that we include league’s dummies) and that in the second stage we find a non-significant Sargan

statistic for over identification, what is a signal of the good performance of our instruments. Still,

we have also performed additional checks for the exclusion restriction. First, we have computed

the correlation between the residuals of equations A4 and A4, to check if 

(





, 



)

= 0. We

obtain a correlation coefficient of -0.0378, which is not significant even at 10% (p-val = 0.24).

Figure A7.1 displays the scatterplot between both residuals. Finally, figure A7.2 shows the

scatterplots of the generated residuals of equation A4, which are correlated with the share of

migrants but not with the Elo points.

Table A7.1. Identification strategy

ln Elo Points

2010

OLS (1st stage)

% Foreign

Players

2SLS

Rank Old

-0.0044***

(0.001)

Rank Old

0.000036

(0.000061)

Rank (Capacity/Old)

-0.0077***

(0.001)

Rank (Capacity/Old)

0.0001**

(0.000)

ln Elo points 2010

0.4218***

(0.112)

Interaction Weather -Index # Country Dummies

YES

Country Dummies

YES

Observations

968

R-squared

0.587

0.712

Anderson canon. corr. LM statistic for Underidentification

Chi-sq(73)=232 P-val=0.000

Sanderson-Windmeijer multivariate F test of excluded

instruments:

F(73, 825) = 3.56 P-val= 0.0000

Sargan Statistic test for Overidentification

Chi-sq(72) P-val = 0.159

Figure A7.1. Scatter plots - generated residuals eq. A4 and A5

Figure A7.2. Scatter plots for the generated residuals

-.2 -.1 0 .1

Residual ln Elo points 2010

-.4 -.2

0 .2 .4 .6

Resid % For Players

0.0000 0.2000 0.4000 0.6000 0.8000 1.0000

% Foreign Players

-.4 -.2

0 .2 .4 .6

Resid % For Players

7 7.2 7.4 7.6

ln Elo Points 2010

-.4 -.2 0

.2 .4 .6

Resid % For Players

7 7.2 7.4

7.6

ln Elo Points 2016

-.4 -.2 0 .2

.4 .6

Resid % For Players

Additional Materials of the paper “On the influence of foreign

players on the success of football clubs”

Additional Material 1. Theoretical foundation of the empirical

model, based on Bernard and Busse (2004)

The production function of talent (



) of the football teams in country  requires a population (



economic resources (



), a warm temperature (



), a number of football-related institutions (



)

and some organizational skills (







= (



, 



, 



, 



, 



)

The relative football success, 





, obtained by the country is a function of the talent in that

particular country:





 





 





= 





= (



)

A Cobb-Douglas talent production function is assumed:





= 



























This characterization leads to the following specification for a country’s relative success at

football:







= ln















= 



+ 



+ 



+ 



+ 







As the socioeconomic variable, can be expressed as the product of population and per capita

income, the specification to be estimated is:





= 



+ 







+ 













+ 







+ 







+ 



where 



the error term that is distributed normally.

Additional Material 2. The World Football Elo Rating System

The World Football Elo Ratings are based on the Elo rating system, developed by Dr. Arpad Elo.

This system is used by FIDE, the international chess federation, to rate chess players. In 1997

Bob Runyan adapted the Elo rating system to international football and posted the results on the

Internet. He was also the first maintainer of the World Football Elo Ratings web site. The system

was adapted to football by adding a weighting for the kind of match, an adjustment for the home

team advantage, and an adjustment for goal difference in the match result.

These ratings take into account all international matches for which results could be found. Ratings

tend to converge on a team's true strength relative to its competitors after about 30 matches.

Ratings for teams with fewer than 30 matches should be considered provisional. Match data are

primarily from International Football 1872 - Present

The ratings are based on the following formulas:

= R

+ K × (W - W

)

is the new rating; R

is the old (pre-match) rating.

K is the weight constant for the tournament played:

• 60 for World Cup finals;

• 50 for continental championship finals and major intercontinental tournaments;

• 40 for World Cup and continental qualifiers and major tournaments;

• 30 for all other tournaments;

• 20 for friendly matches.

K is then adjusted for the goal difference in the game. It is increased by half if a game is won by

two goals, by 3/4 if a game is won by three goals, and by 3/4 + (N-3)/8 if the game is won by four

or more goals, where N is the goal difference.

W is the result of the game (1 for a win, 0.5 for a draw, and 0 for a loss).

is the expected result (win expectancy), either from the chart or the following formula:

= 1 / (10

(-dr/400)

+ 1)

dr equals the difference in ratings plus 100 points for a team playing at home.

Sample Winning Expectancies

Difference in rating

Higher related

Lower related

0.500

0.500000

0.514

0.486

0.529

0.471

0.543

0.457

0.557

0.443

0.571

0.429

0.585

0.415

0.599

0.401

0.613

0.387

0.627

0.373

100

0.640

0.360

110

0.653

0.347

120

0.666

0.334

130

0.679

0.321

140

0.691

0.309

150

0.703

0.297

160

0.715

0.285

170

0.727

0.273

180

0.738

0.262

190

0.749

0.251

200

0.760

0.240

210

0.770

0.230

220

0.780

0.220

230

0.790

0.210

240

0.799

0.201

250

0.808

0.192

260

0.817

0.183

270

0.826

0.174

280

0.834

0.166

290

0.841

0.159

300

0.849

0.151

325

0.867

0.133

350

0.882

0.118

375

0.896

0.104

400

0.909

0.091

425

0.920

0.080

450

0.930

0.070

475

0.939

0.061

500

0.947

0.053

525

0.954

0.046

550

0.960

0.040

575

0.965

0.035

600

0.969

0.031

625

0.973

0.027

650

0.977

0.023

675

0.980

0.020

700

0.983

0.017

725

0.985

0.015

750

0.987

0.013

775

0.989

0.011

800

0.990

0.010

Additional Material 3. Football Confederations

a. The Asian Football Confederation (AFC) is the governing body of association football in Asia.

It has 47 member countries, located in the main on the Asian continent. All the transcontinental

countries with territory straddling both Europe and Asia are members of UEFA (Azerbaijan,

Armenia, Georgia, Kazakhstan, Russia and Turkey). Israel, although it lies entirely in Asia, is

also a UEFA member. Australia, formerly in the OFC, has been in the AFC since 2006, and the

Oceanian island of Guam, a territory of the United States, is also a member of the AFC.

b. The Confederation of African Football (CAF) represents the national football associations of

Africa.

c. The Confederation of North, Central American and Caribbean Association Football

(CONCACAF) is the continental governing body for association football in North America,

Central America and the Caribbean.

d. The South American Football Confederation (CONMEBOL) is the continental governing

body for association football in South America.

e. The Oceania Football Confederation (OFC) is one of the six continental confederations of

international association football, consisting of New Zealand and island nations such as Tonga,

Fiji and other Pacific Island countries. In 2006, the OFC’s largest and most successful nation,

Australia, left to join the Asian Football Confederation.

f. The Union of European Football Associations (UEFA) is the administrative body for

association football in Europe and, partially, Asia. UEFA membership coincides with sovereign

countries in Europe, although some UEFA members are transcontinental states (e.g. Turkey).

Several Asian countries have also been admitted to the European football association: Azerbaijan,

Armenia, Georgia, Kazakhstan, Israel, Russia and Turkey, which had previously been members

of the Asian football association.

Additional Material 4. Using market value as a proxy for budget

for salaries

The top ten clubs with higher Market Value are displayed in table A5.1.

Table A5.1. World top ten clubs by Market Value (M €)

League

Club

Market value

Spain

Real Madrid

704.8

Spain

FC Barcelona

689.5

Germany

Bayern Munich

578.55

England

Manchester City

501.75

England

Chelsea FC

490

England

Arsenal FC

431

France

Paris Saint-Germain

423.75

England

Manchester United

411.25

Italy

Juventus FC

379.8

England

Liverpool FC

367.1

In order to exemplify how Market Value is a good proxy of the Budget of the Club, we

use data of the Spanish League, for which we record the Budget linked to the maximum

amount that can be devoted to wages by every club, according to the rules dictated by the

Professional Football League. Table A5.2 and Figure A5.1 report this data, and shows a

strong correlation, close to 99%.

Table A5.2. Market Value – Budget devoted

to salaries. Spanish League 2015-06.

Club(s)

Market

Value

Budget

Athletic Bilbao

134

53.6

Atlético Madrid

325

159.6

Celta de Vigo

75.7

22.6

Deportivo de La Coruña

53.95

17.8

FC Barcelona

689.5

421.7

Getafe CF

51.1

20.7

Granada CF

61.9

25.2

Levante UD

58.75

25.9

Málaga CF

57.9

28.7

Rayo Vallecano

41.45

20.9

RCD Espanyol Barcelona

60.6

30.6

Real Betis Balompié

39.1

Real Madrid

704.8

431.3

Real Sociedad

115.4

56.6

SD Eibar

40.8

19.11

Sevilla FC

186.2

105.13

Sporting Gijón

39.05

14.6

UD Las Palmas

29.25

18.4

Valencia CF

282

122.8

Villarreal CF

135.8

61.5

Figure A5.1. Scatter plot Market

Value – Budget devoted to salaries

R² = 0,988

100

150

200

250

300

350

400

450

0 200 400 600 800

Additional Material 5. Robustness checks. Negative binomial and

log-log estimates

Tabla A8.1. Negative binomial regressions of World Ranking

(1)

(2)

(3)

(4)

(5)

(6)

World

Sample

World

Sample

World

Sample

World

Sample

UEFA

CONMEBOL

CONC

Share Foreign players

0.728***

-0.149

0.813***

0.198

0.261

0.302

-0.7

(0.225)

(0.169)

(0.210)

(0.152)

(0.168)

(1.232)

(0.4

ln Population

0.00926

-0.00782

0.0108

-0.00381

-0.0144

-0.000909

-0.0

(0.0198)

(0.0126)

(0.0179)

(0.0109)

(0.0133)

(0.0335)

(0.06

Weather Index

-0.000299

0.000772

-0.000384

0.000665

0.000151

0.000146

0.00

(0.000601)

(0.000693)

(0.000634)

(0.000630)

(0.000831)

(0.000980)

(0.00

ln Market value

-0.140***

-0.493***

-0.156***

-0.520***

-0.612***

-0.868**

-0.2

(0.0395)

(0.0545)

(0.0461)

(0.0406)

(0.0682)

(0.409)

(1.4

ln Market value

-0.108***

-0.105***

-0.103***

-0.0989***

0.0443

-0.1

(0.00905)

(0.00908)

(0.00987)

(0.00662)

(0.00886)

(0.0742)

(0.2

ln Capacity Stadium

-0.0962**

-0.0891***

-0.0987***

-0.0959***

-0.0524

-0.213**

-0.13

(0.0382)

(0.0296)

(0.0375)

(0.0250)

(0.0363)

(0.105)

(0.04

SFP_res

54.24***

60.76***

52.26***

-3.413

73.

(8.340)

(7.463)

(9.273)

(26.76)

(50.

Constant

8.037***

9.653***

8.106***

9.700***

9.568***

9.749***

11.35

(0.275)

(0.266)

(0.299)

(0.224)

(0.318)

(0.998)

(2.7

Country Dummies

YES

ln alpha

-1.107***

-1.589***

-1.181***

-1.729***

-1.768***

-1.358***

-1.89

(0.0815)

(0.0860)

(0.0856)

(0.101)

(0.123)

(0.139)

(0.1

Observations

970

968

571

129

Note: Robust standard errors clustered by country in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Models

including as regressor SFP_res are controlling for endogeneity by means of the Control Function Approach,

as we include the generated residual of an equation where the Share of Foreign Players depend on all control

variables plus a list of excluding instruments, namely the generated instrument reported in appendix 7 plus

the average age of players in every team.

Tabla A8.2. Elo points regressions. IV estimates.

(1)

(2)

(3)

(4)

(5)

(6)

World

Sample

UEFA

CONMEBOL

CONCACAF

CAF

AFC - OFC

Share Foreign players

-18.0137

-18.5044

-0.8221

67.0160

31.4755

-90.2161

(18.526)

(21.141)

(130.435)

(84.807)

(114.631)

(63.521)

ln Population

1.3537

2.1962

1.4402

5.5291

-1.4372

2.5610

(1.466)

(1.837)

(3.024)

(7.157)

(3.179)

(4.552)

Weather Index

-0.0268

0.0635

0.0533

-0.3807*

0.0550

-0.0342

(0.086)

(0.129)

(0.149)

(0.197)

(0.225)

(0.181)

ln Market value

76.2571***

97.1647***

126.0772***

44.9985

74.1756***

-6.7516

(7.772)

(10.320)

(40.356)

(97.329)

(7.622)

(18.640)

ln Market value

9.1983***

7.2045***

-9.6598

6.4672

19.3227***

40.0886***

(1.182)

(1.427)

(7.135)

(18.929)

(1.765)

(6.347)

ln Capacity Stadium

13.3003***

7.0710

27.9330***

24.2508

1.6509

12.5905

(3.366)

(4.429)

(9.311)

(15.219)

(5.385)

(11.719)

Country Dummies

YES

KP statistic (p-val)

171.1

(0.000)

131.2

(0.000)

36.56 (0.000)

9.800 (0.020)

30.72

(0.000)

13.52

(0.004)

J-Statistic (p-val)

2.524

(0.283)

6.454

(0.039)

3.224 (0.200)

0.429 (0.807)

2.669

(0.263)

1.638

(0.441)

Observations

970

571

129

125

R-squared

0.818

0.867

0.632

0.597

0.713

0.689

Note: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. All models use the generated

residual resulting from the two-steps Brückner strategy together with the average Age (and its square) of

players in every team. KP refers to the under-identification Kleibergen-Paap LM statistic, while the J

statistic corresponds to the over-identification Hansen J-Statistic.