Abstract
The immigration of professional athletes hides several realities of discrimination. Often, the high salaries associated with these athletes help hide many discriminatory practices that still occur even at the highest competitive level. We looked at the reality of professional goalkeepers in the main OECD football leagues (specifically, the six main leagues in Europe), over more than 10 seasons. Using Mincerian equations and Blinder–Oaxaca decomposition, we found that discrimination is mainly based on the nationality of the athletes, with discrimination based on skin color taking a secondary role. Complementary evidence based on the athlete's performance, experience or biological factors such as age was also discussed.
Keywords
Introduction
In addition to the migration flows that make the news due to the volume of displaced people and the media-covered fragility of the conditions of movement of migrants and their families, other migrations are exposed to significant difficulties, including discrimination, prejudice, and the deterioration of contractual conditions, but which, however, do not receive significant media attention. One migration flow that can be classified as such involves professional sportspeople, especially soccer players. Thousands of players leave their clubs every year to start contracts with other clubs in different countries. However, while the escalation in transfer fees and the records broken in these amounts season after season by some of the players leads to the crystallized image of these migration flows appearing to be one of a flow surrounded by very favorable conditions for those involved, the reality that we discuss here, focusing on the main professional championships of this sport in the OECD, is quite different. As we have already mentioned, the leading destination for this migratory movement is the championships of many OECD countries, especially the European championships, which are still seen as the most demanding in professional terms, but also those where the expected salaries, performance bonuses, and media exposure are the most notable.
Within this evidence, however, there is a discussion about the discriminatory practices that football players of certain nationalities suffer, particularly players who play as goalkeepers. As we will highlight throughout the paper, it is clear that many goalkeepers, unlike players in other football positions, are white or easily characterized as Caucasian. Although there has been an effort to contradict this interpretation, current figures show several discriminatory patterns in selecting top-level soccer players who arrive in Europe.
These discriminatory patterns were observed over more than 10 sports seasons and considered skin color, nationality, physical distance between championships, and other controlled characteristics of the athletes and the respective championships. Our analysis with Mincerian quantile equations and Blinder–Oaxaca decomposition revealed important determinants of salaries’ differences related to experience, including the number of games without conceding goals and team players’ market value. A complimentary and detailed analysis of skin color, nationality, and differences in professional championships generates important implications in sports policies. These policies must focus on several wage discrimination patterns, especially the nationality-related ones, which we identify as particularly determining.
The remaining sections of this work are as follows. Section 2 reviews the literature on wage differentiation. Section 3 presents our empirical analysis, supported by detailed analyses involving Mincerian equations and the decomposition between explained and unexplained effects of the discrimination found. Last, Section 4 concludes the article.
Wage differences in sports and soccer: Classic explanations
One of the universal values of sports is reducing differences between athletes in favor of the quality of competition. The 10th Sustainable Development Goal (SDG) also explicitly reduces labor inequalities and fights against hidden discriminatory practices in professional environments (namely, SDG 10.3 and SDG 10.6). However, professional sports allow different contracts and salaries, marked by different levels of prestige that shape different careers for different athletes.
Within the world of professional sports, soccer is a media sport that involves considerable financial dimensions. However, given the different teams, seasons, and championships, differences between athletes, including in their wages and career longevity, are obvious. 1
In the most competitive type of soccer, played in European professional championships, the European Union (EU) regulations place players from EU member countries on equal employment terms, thereby creating an asymmetry of opportunities that affects players from nations outside the EU. 2 Recently, several authors also noted that few black goalkeepers play in the premier leagues of the most valued championships on the European continent. 3 That observation follows from other empirical work that has revealed sources of discrimination in professional sports, including athlete's nationality and other identifying characteristics.4–6
Against that backdrop, in our study, we sought to determine whether any empirical evidence supports the claim that wage discrimination exists in the world of goalkeepers playing in European soccer championships. In an analysis of five championships (i.e., championships of OECD countries, namely the first leagues in England, Spain, Italy, France, and Portugal) since the 2010–2011 season, we tested whether wage differences among goalkeepers could be explained by classic reasons associated with recent performance and accumulated experience or by discrimination, including skin color and/or nationality. These leagues tend to be identified as the wealthiest leagues in Europe, having the most competitive soccer teams. 7
Although the study of wage differentiation is even richer in other sectors of the economy a range of work has been intended to address explanations for the existence of wage differences in the world of sports.5,6,8 To date, it has pointed to rather classic explanations subsequently considered in sports economics:
Different capacities for productivity and contribution to the payer's objectives; Differentiated accumulation of human capital (i.e., training, schooling, and/or experience1,9 Theories linked to credentialism and dualism, which explain the tendency of social groups to replicate in younger generations the differences in remuneration derived from access to training and universities, which in turn also differ; Segmentation theories, which explain how characteristics observed in workers condition the expectation of remuneration destined to them
10
; Theory of the cost of preparing the worker and/or looking for a suitable worker available in the market, developed by Adam Smith and also revisited recently,2,11 and The rigidity of the labor supply, including the number of workers available for the function and the nature of the function, which, from the community's perspective, may require more significant personal effort costs.
Those explanations show how the salaries paid differ among athletes for reasons ranging from performance achieved in recent seasons to professional maturity and including expectations regarding the athlete's characteristics, availability on the market (e.g., in transfers or training schools), capacity to negotiate their salary, legal representation, and interested clubs.
Other works addressing those themes, whether for athletes or sports agents, have been conducted.12–14 Investigations in line with credentialist theory showed that sports practice is associated with higher-income workers. 13 At a more detailed level, the set of conditions made possible by higher incomes ends up motivating workers and, thus, athletes to engage in practice more regularly, which in sports improves performance.
That line of investigation therefore places special emphasis on the initial conditions of athletes, as well as their ability to position themselves in favorable negotiation conditions, beginning with their initial contracts.15,16
However, the difference observed among teams that manage to gain public support compared with ones that do not have also been detailed.14,17 Teams with public support or subsidies gain financial reinforcements that allow the hiring of players with better expected performance, which contributes to an acceleration of the imbalances observed in the salaries subsequently paid to athletes or in contracts’ length (Frick, 2011).17,18
The theory of superstars additionally suggests that certain athletes have a public image that generates significant advertising revenue, which particularly influences the associated contracts (Soebbing et al., 2023; Bernardo et al., 2022; Scarfe et al., 2021; Lucifora and Simmons, 2003).19–22 That differentiating media capacity also contributes to accentuating the wage heterogeneity observed.
Last, recent investigations have demonstrated how sports influence wages in other sectors beyond sports. 23 Therefore, an increased understanding of the forces and dynamics of salaries in various professional sports can also illuminate dimensions that are significant for economic sectors on a global level.
Discrimination in sports and soccer
Along with the classic dimensions already mentioned, complementary work has explored other dimensions of discrimination in sports, including gender, skin color, nationality, and sexual orientation.24–27 Such work shows that wage inequality exists according to gender, as can be explained by the maturity of professional competitions—generally, older competitions award higher salaries to athletes—to the different standards of audiences or cultural habits that tend to discriminate against women professionally.26,28 Discrimination by sexual orientation was also recently examined along with discrimination by skin color and nationality, in salary differences that could not be attributed to classic explanations, including performance or experience with the competition.25,29
There have also been other works that broaden the approach to include social and labor inequality in spectators and how such dimensions influence the demand for sporting events and the inequality of the related athletes’ earnings, observing how women and minorities occupy less-central roles in sports organizations. 30
The case of goalkeepers
Among goalkeepers in the professional reality of European soccer, only a few are not Caucasian (referred to as “white.”) Whereas finding non-white players in other positions (e.g., defenders and attackers) is rather easy, there remain few black goalkeepers. 3 Usually, goalkeepers’ salaries are determined year-to-year within the contractual period, as happens for most of professional soccer players. At the end of a season (or a mid-season) and following the performance of the player conditioned on the sporting competence of the team, there is space for slight changes of the contractual terms, namely salary's updates. 10
In such a differentiated position in relation to the team strategy, goalkeepers might be expected to have equivalent salaries when it comes to the same team, the same professional championship, or professional championships regulated by the same framework (e.g., championships regulated by the Union of European Football Associations). However, the similarity of goalkeepers’ salaries is not evident, whether on the same team or in the same championship, and during the same season. Searching for explanations, models based on perfect competition cannot offer convincing enlightenment because, even in an extreme situation, soccer players cannot be randomly placed in any position, given the career and specific training of each player. A player defined as a defender or midfielder can serve as a goalkeeper in limited periods during a soccer game (e.g., when the usual goalkeeper receives a red card and there is no available goalkeeper for the substitution); however, in practice, players trained as goalkeepers tend to indeed be the goalkeepers of the teams. Therefore, models supported by imperfect competition would always be theoretically more appropriate to discuss wages across different soccer players on a team.
From the literature, we have identified the following classic explanations for the lack of homogeneity among goalkeepers’ salaries:
In addition to those explanations, 27 we want to advance the possibility of discrimination found for sectors of the economy in general due to innate characteristics, including skin color and nationality. 32 On that topic, it has been observed the difficulty of finding non-white goalkeepers in the most competitive European football leagues. 3 However, that observation may not necessarily translate into latent discrimination or prejudice if, for example, we empirically observe that wage differences are not significantly explained by skin color or nationality. We also have to differentiate European leagues.2,11 Although the action of the Union of European Football Associations has tended to create a normative set regulating European football as a homogeneous universe, it is convenient to endogenously differentiate each championship.
What has not yet been said about goalkeepers’ discrimination
After reviewing the previous literature, we verified that, despite the development of several authors who explain the wage differentiation in certain industrial or service realities, there is still a long way to explain the reasons for the discrimination observed in professional goalkeepers, in male soccer.12–14
On the one hand, the scarcity of black or Asian goalkeepers in European professional soccer has been highlighted in the sports press, just as the issue is beginning to be identified in other sports (namely, in MotoGP, as Patterson 2021 identified). 33 As registered, the singular position that a goalkeeper has in a football team can contribute to this salience of the phenomenon. However, the demographic increase of immigrant communities in Europe, as well as the sporting development of second and third generation communities, has been increasing significant pressure for this signalled rarity of non-Caucasian goalkeepers to be reduced. 10
However, more than the signalled rarity, it is important to look at wage differentiation as well as to identify the dimensions responsible for the differentiation in question. In a primary insight, it could be argued that the reported rarity of goalkeepers of non-European origin in European professional football could be associated with different team market values and higher salaries for professionals of non-European origin.2,11 For this, we will have to use proper techniques in which we attribute the remaining characteristics of a professional to other professionals in order to understand whether this professional is undervalued or overvalued. The ‘Mincerian quantile equations’ and the ‘Blinder-Oaxaca decomposition’ that we will develop in section 3.1 will help in this task.
Additionally, we will have to verify whether, more than ‘wage gaps’, there are ‘competence gaps’ or ‘experience gaps’ that can help explain the salary penalties that some goalkeepers have when compared to other professionals. In this verified possibility, the discussion around discrimination in professional sport would become more complex. Actually, in addition to identifying the sources of wage differentiation, we would have to search for more primary sources of inequality, namely inequality in access to development channels, professional sport, which would make some more ‘competent’ to the detriment of other athletes.
Empirical analysis
Mincerian quantile equations and Blinder–Oaxaca decomposition
This section describes the research methods used to analyze goalkeeper salaries. The main goal was to understand if salary differences are due to performance-related factors (like experience and games played) or discriminatory factors, such as nationality or skin color.
The study used two main techniques:
− Mincerian Quantile Equations: This method analyzes how different factors influence salaries not just for the average player, but for players at different salary levels (e.g., the lowest paid, the average, and the highest paid). It is a more advanced version of a standard salary analysis and helps to avoid some common statistical errors.32,34 − Blinder–Oaxaca Decomposition: This technique separates the salary differences between two groups (for example, white and non-white goalkeepers) into two components35,36:
Explained Effects: The part of the salary difference that can be explained by factors like a player's experience, number of games played, and their team's market value. Unexplained Effects: The part of the salary difference that cannot be explained by the available data. This is used as a proxy for the discrimination effect.
To address some limitations of the Blinder-Oaxaca method, 37 we used a refined approach called the Recentered Influence Function (RIF) decomposition, which allows for a more detailed analysis of the unexplained differences across various salary levels.38,39 By applying these methods, we aimed to provide a clearer picture of whether goalkeepers are paid based on their professional attributes or if their salaries are also affected by discrimination related to their skin color or nationality.
Data sources and descriptive analysis
To build our database, we used official information from the various clubs that, between the 2010–2011 and 2019–2020 seasons, formed each major men's league of the five observed championships (i.e., English, Spanish, Italian, French, and Portuguese championships). The source of salaries associated with each athlete was www.capology.com since the 2013–2014 season and, before that, https://salarysport.com/. In turn, the website www.transfermarkt.com made it possible to observe the following variables for each athlete and for each season: experience (i.e., years as a professional), height (cm), skin color as observed from Player Image (just differentiating between white and non-white), matches per season, goals conceded, and team's market value (a biannual estimation of Transfermarkt's team considering dimensions like players’ performance, market rumors or injuries). Let us clarify that our classification of a goalkeeper as a white/non-white player followed the majority of responses from a survey responded by 65 students of Master and PhD Degrees, inquired in the period of this work's data collection, also following Principe and van Ours (2022). 40 All our data is available upon request.
Table 1 presents the central values for annual salary and the respective standard deviations. It distinguishes goalkeepers by country league and athletes’ skin color. In general, the league that best remunerated goalkeepers among the six leagues studied was the Spanish league, whereas the one that remunerated goalkeepers the worst was the Portuguese league. Regarding skin color, with the exception of France, in the other leagues, non-white goalkeepers received lower wages than white ones. The biggest salary gap was found in the Italian league, in which a non-white goalkeeper earned, on average, just more than 4% of the income of their white counterparts.
Annual salary of goalkeepers by men's league and skin color from the 2010–2011 to 2019–2020 seasons (euros, at current prices).
Note. Standard deviations appear in parentheses. 1557 observations considering 698 individuals.
Table 2 shows the average salaries among the goalkeepers by continent of nationality. Goalkeepers of European nationality were paid the most, followed by goalkeepers from North or South America. By contrast, goalkeepers of Asian nationality playing in Europe had some of the worst salaries. Regarding differences in skin color, the largest wage gap emerged among goalkeepers of American nationality, such that a non-white goalkeeper receives on average 46.26% of the wages of whites. As for goalkeepers of African nationality, the position was reversed; a white goalkeeper received on average 49.07% of the wage of a goalkeeper who is also African but not white.
Salary by nationality and skin color (euros, at current prices).
Note. Standard deviations appear in parentheses. 1557 observations considering 698 individuals.
As shown in Table 3, our sample consisted of 1554 observations. The average logarithm of the athletes’ annual salary was 12.50, and, in absolute values, the annual average was €840,357.60. The average experience for the observed goalkeepers was approximately 10 years. The goalkeepers had an average height of 1.89 m, and there was a predominance of white goalkeepers (i.e., > 92%) in the six leagues. On average, goalkeepers played 37 games in the previous season, with an average of 45 goals conceded. By nationality, more than 80% were European, while approximately 13% were from North or South America.
Descriptive statistics.
Results of the models
Table 4 allows discussing the determinants of the salaries of goalkeepers. As can be observed, variables traditionally related to human capital were significant for the remuneration of those athletes; on average, one more year of experience increased the salary by approximately 16%. Such influence decreases as the athlete acquires more experience, however, and for the highest quantile of wages, the effect is lower. The number of matches in the previous season also contributed positively to a higher salary; for each additional game played, the athlete would observe an increase of 0.6% in their expected income. However, for each goal conceded, the reduction was nearly in the same proportion (i.e., 0.58%), thereby reinforcing the importance of a goalkeeper's having played more games while conceding fewer goals. Regarding the variable related to team market value, for each 1% increase in the team's market value, the goalkeeper's salary would increase by approximately 0.76%.
Wage determinants in ordinary least squares (OLS) and quantiles.
Note. Robust standard errors have been estimated. *** p < .01. ** p < .05. * p < .1.
Regarding the innate variables, height proved to be significant only for the lowest quantile (i.e., Q25), which may indicate that remuneration is more related to the athlete's impulsion (i.e., jumping skills) than height. By skin color, being non-white contributed positively to higher wages, which we interpret as a possible form of positive discrimination in the soccer market for that characteristic of goalkeepers that, contrary to expectations, anticipated negative discrimination. In the highest quantile (i.e., Q75), being non-white contributed to a little more than 20% higher wages.
Regarding variables in the theory of the segmented labor market, we studied the possible wage returns of the leagues of the various countries compared to the English league, considered to be a reference league in the software (Stata V.16.0). As shown in our results, a goalkeeper who played in Spain earned an average of 63.47% more than a goalkeeper working in the English league. In contrast to the Spanish league, the Portuguese league pays its goalkeepers the worst; in fact, a goalkeeper who worked in Portugal earned 73.61% less on average than a goalkeeper who worked in England.
By nationality, goalkeepers from North or South America earned 35.34% more on average than ones from Europe. There was also a negative effect on goalkeepers from Asia, who earned the lowest salaries. Compared with a goalkeeper of European nationality, one of Asian nationality received approximately 50% less when observing the median quantile (i.e., at Q50 regression).
Table 5 presents the results of the Blinder–Oaxaca decomposition between the white and non-white groups. Combining the reading of the explained and unexplained effects clarified why a given professional earned less or more. If their group tended to present inferior characteristics identified in the model, then their group had a negative estimated coefficient in the explained effects. When the goalkeeper earned less because the discriminatory characteristic was especially influential—for example, skin color, as shown in Table 5—then the group had a negative estimated coefficient in the unexplained effects.
White × non-white.
Note. Robust standard errors have been estimated.
*** p < .01. ** p < .05. * p < .1.
The difference in salary between white and non-white goalkeepers was 27.44%. Put differently, for the salaries of non-whites to be equal to the salaries of whites, they would have to increase by more than 27%. Part of that difference can be explained by individual characteristics related to the revisited theories, including the theory of human capital (i.e., experience and accumulated games without conceding goals) and segmentation (i.e., specificity of each championship). In a counterfactual analysis, if non-white goalkeepers had the same productive attributes as white goalkeepers, then their salaries would increase by more than 90% on average (i.e., explained effects). The unexplained effects, meanwhile, can be understood as a proxy for wage discrimination based on skin color, because the comparison is between whites and non-whites. The results point to possible positive discrimination in favor of non-whites in the market for goalkeepers. If the goalkeeper belonged to the white group, then their salary would reduce by an average of 33.05%. The positive discrimination is more pronounced at lower salary quantiles, reaching 65.02% at the 25th percentile (Q25) and 40.65% at the median (Q50). In sum, white goalkeepers tend to earn more in the observed championships (€ 371,475.51 against € 291,481.89) mostly due to the explained effects—that is, due to having more experience, more games played, and a higher team market value.
We agree that this result is a challenging and counterintuitive finding and requires further discussion. The manuscript already indicates that, on average, non-white goalkeepers tend to earn less than white goalkeepers (€291,481.89 versus €371,475.51). However, this wage gap is primarily explained by productive and market attributes, such as more experience, more games played, and the higher market value of white goalkeepers’ team. The counterfactual analysis (explained effects) showed that if non-white goalkeepers had the same productive attributes as white goalkeepers, their wages would increase by over 90% on average.
What this study identified as a “positive contribution to higher wages” for non-whites, or “possible positive discrimination,” refers to the “unexplained effects” in the Blinder-Oaxaca decomposition. This means that, after controlling for productive characteristics, if a goalkeeper belonged to the White group, their salary would decrease by an average of 33.05%, indicating that the characteristics of non-Whites (when present) are differentially valued, or that, for the same set of attributes, non-Whites are relatively better paid. This “positive discrimination” was even more pronounced at the lowest salary quantiles (65.02% in Q25 and 40.65% at the median, Q50).
The rarity of non-White goalkeepers in elite European leagues, corroborated by our study (>92% are White), 3 may be explained by the fact that those who reach this level are considered “exceptional” and, consequently, valued above average in terms of their inherent attributes, resulting in an “unexplained” salary premium. This is not necessarily a case of self-selection to higher-paying clubs, but rather that, given the barrier to entry, those who overcome it may be perceived as having a “something extra.”
Table 6 shows the results of the Blinder–Oaxaca decomposition of salaries comparing each league in relation to the other five studied leagues. Analyzing the salaries of goalkeepers who work in Spain compared with athletes who play in the leagues of the other respective countries—France, England, Italy, and Portugal—revealed a negative difference of 77.87%. That is, a goalkeeper who did not play in Spain received, on average, slightly less than 22% of the salary of a goalkeeper who worked in the Spanish league. That difference can be explained by productive characteristics (i.e., performance) as well as market characteristics. If the difference between the characteristics of Spanish and other goalkeepers were valued by the standards of goalkeepers in other leagues (such as experience or market value), Spanish goalkeepers’ salaries would suffer a reduction in their salaries by approximately 56%. The unexplained effects can be understood as a proxy for the league's effect, which also underscores that the Spanish league tends to value goalkeepers more, because if those goalkeepers were not working in Spain, their salaries would drop by approximately 50%.
Country league.
Note. Robust standard errors have been estimated.
*** p < .01. ** p < .05. * p < .1.
Analyzing the effect of the other leagues in light of the same reasoning mentioned for the Spanish league revealed that the English league has shown similar behavior, both for the explained and unexplained effects. For the other countries, the explained effects indicate that if the goalkeepers of those leagues had the same performance-related characteristics as the goalkeepers of the other leagues observed, then their salaries would suffer an increase of 56.62%, 180.67%, and 27.78% for France, Italy, and Portugal, respectively. The unexplained effects (i.e., the league's effect) were estimated to be negative for France and Italy, thereby indicating that those leagues overvalue their athletes, because if goalkeepers from those countries played in other leagues, then their salaries would fall by 16.97% and 64.52%, respectively. 41 By contrast, the Portuguese league tends to undervalue goalkeepers, for if the same athletes who work in Portugal were defending teams in other leagues, then their salaries would increase by more than 2.68 times—that is, 268.44%.
Table 7 presents the results of the salary decomposition comparing the nationality of the goalkeepers. Once again, to clarify the interpretation, we referred to the meaning of the estimated coefficients for the explained and unexplained effects. An estimated coefficient of positive direction for the explained effects meant the impulse given by the variables included in the model, beyond nationality. Therefore, an estimated coefficient in the positive direction for the explained effects of a given nationality meant that the other groups of nationalities tended to present higher values in the variables other than those of nationality. If the estimated coefficient was negative for the explained effects, then it would mean that other nationality groups tended to have lower values in variables other than nationality. Thus, goalkeepers of European nationality had more salary-enhancing attributes than goalkeepers of American, African, Asian, and Oceanian nationalities. If European goalkeepers had the same characteristics as the other goalkeepers of other nationalities, then their wages would fall by an average of 21.79%. That difference means that, in expected value, European goalkeepers tend to have more experience, play more games per season, and have greater team market value. The opposite emerged for players from the American and Asian continents, whose characteristics had lower values than in the comparison group (i.e., less experience, fewer games played per season, and lower team market value). Therefore, if goalkeepers born in the Americas or in Asia had the same characteristics as goalkeepers from other continents, then their salaries would increase by 32.22% and 165.85%, respectively.
Continent of nationality.
Note. Robust standard errors have been estimated.
*** p < .01. ** p < .05. * p < .1.
Regarding the unexplained effects, one interpretation is that an estimated positive coefficient in the unexplained effects means that if goalkeepers had a nationality other than their actual nationality, then they would earn more, which indicates a perverse or discriminatory effect of nationality. That dynamic affects Asian goalkeepers, for example, because if they had a nationality from a different continent, then their income would increase by 56.96% on average. By contrast, for players born in North or South America, who even exhibited a premium for nationality, the estimated coefficient was statistically significant at 1%; therefore, if those goalkeepers were not American, then their salaries would fall by 28.71% on average.
Robustness checks
For assessing the robustness of the previous findings, we ran additional regressions, controlling with variables like the logarithmic value of the team of each goalkeeper (Source: transfermakt.de, Mean: 4.8451, Standard deviation: 1.0309, Minimum: 2.4578, Maximum: 7.0900) or the overall rating (Source: whoscored.com, Mean: 6.6260, Standard deviation: 0.3730, Minimum: 4.95, Maximum: 8.17).
The rationale for these variables follows literature.27,29,42 The value of the team has been observed as an exogenous dimension influencing the individual value attributed to a player. Teams with more significant values tend to update the values of their players, also in a rising move. The overall rating is an additional indicator of the sporting performance of the athlete. Although being a construction based on a limited number of variables observed along the season, this indicator allows an additional assessment regarding the goalkeeper's expected competence. Therefore, we wanted to assess the consistency of the previous outcomes by controlling with these innovations.
As observed in Table 8, it is confirmed that experience has a positive impact on salary - for each year of experience the athlete has an increase of 9.94% in their salaries (quantile 50). This impact reduces as wages increase. The variable “skin color” was associated, again, with significant and positive estimated coefficients for the last quantile. For each goal conceded, goalkeepers’ wages drop by an average of 0.22%. In relation to the market value, the impact is inelastic: for each 1% increase in the goalkeeper's market value, the increase in their salaries was 0.47% on average; this impact is estimated to be smaller for higher wage quantiles.
Wage determinants in ordinary least squares (OLS) and quantiles.
Note. Robust standard errors have been estimated.
*** p < .01. ** p < .05. * p < .1.
Regarding the different leagues, it is possible to notice that a goalkeeper who works in Spain earns, on average, more than the goalkeeper who plays in the Premier League - the goalkeeper who plays in Spain earns, on average, 32.7% more than the goalkeeper playing in the English league. In contrast to the Spanish league, the Portuguese league is the league that pays its goalkeepers the worst: compared to the Premier League, a goalkeeper who works in Portugal earns less (−51.03%) than a goalkeeper who works in England. Regarding the athletes’ nationality continent, comparatively to European nationality, a goalkeeper from Asia earns 35% less than a European goalkeeper, when observing the first quantile (25).
Regarding the logarithmic value of the team, the estimated coefficient of this variable proved to be highly significant and higher for the higher quantiles, that is, the higher the team value, the higher the goalkeepers’ salaries - for a 1% increase in the quantile (75), the return on goalkeepers’ salaries is 43.31% higher too. We have also included an indicator of Player rating in this stage (fbref.com). This indicator identifies the percentage of goals saved of each Goalkeeper in the analyzed period. 27 However, this indicator of Player rating did not have any significant estimated coefficient.
Table 9 updates Table 6, considering the introduction of the control variables identified in Table 8. It thus shows the salary decomposition comparing the League of the country in question in relation to the other nations studied. In general, the observations follow those of Table 6.
Country league (robustness checks).
Note. Robust standard errors have been estimated.
*** p < .01. ** p < .05. * p < .1.
When analyzing the salaries of goalkeepers who work in Spain compared to athletes who play in the leagues of the other respective countries, France, England, Italy and Portugal, a negative difference of −77.76% is observed. A goalkeeper who does not play in Spain receives, on average, just over 22% of the salary of a goalkeeper who works in the Spanish league. Part of this difference is explained by performance and market characteristics. If the difference between the characteristics of Spanish and other goalkeepers were valued by the standards (such as experience or market value) of goalkeepers in other leagues, the salaries of Spanish goalkeepers would suffer a reduction in their salaries by approximately 64.63%. The unexplained part can be understood as a proxy for the League effect, which also reinforces that the Spanish league tends to value goalkeepers more because if these goalkeepers were not working in Spain, their salaries would be reduced by approximately 37%.
When analyzing the effect of the other leagues on the others, following the same reasoning mentioned above for the Spanish league, it is noted that the English league has a similar behavior to the explained part. As for Italy, the explained part was of a lower magnitude (but in the same direction), showing that if Italian goalkeepers had the same performance and personal characteristics as goalkeepers in other leagues, their salaries would be reduced by 31.55%. For the other countries, the explained part points out that if the goalkeepers of these countries had the same performance characteristics as the goalkeepers of the other leagues observed, their salaries would suffer an increase of 87.64% and 95.58% for France and Portugal, respectively.
The unexplained part (“League effect”) was negative for France, indicating that this league tends to overvalue its athletes because if goalkeepers of France played in other leagues, their salaries would reduce by 31.53. In contrast, the Portuguese league tends to undervalue goalkeepers because if these same athletes who play in Portugal were defending teams from other leagues, their salaries would increase by more than 1.42 times, that is 142.38%.
Table 10 now presents the salary breakdown results comparing the goalkeepers’ nationality. Once again, for the purpose of clarifying the reading, we recall the meaning of the estimated coefficients for ‘Explained’ and for ‘Unexplained’.
Continent of nationality (robustness checks).
Note. Robust standard errors have been estimated.
*** p < .01. ** p < .05. * p < .1.
An estimated coefficient of positive direction for ‘Explained’ means, in this case, the impulse given by the variables included in the model beyond the nationality and observed in the other nationalities. Therefore, an estimated coefficient of positive direction for ‘Explained’, in a given nationality, means that the other groups of nationalities tend to present higher values in the other variables (other than the nationality identification). If the estimated coefficient is negative for ‘Explained’, then it means that other nationality groups tend to have lower values in other variables (other than nationality identification). Thus, nationality goalkeepers inserted in the European area have more salary enhancement attributes than American, African, Asian and Oceanian nationality goalkeepers. If European goalkeepers had the same characteristics as other goalkeepers of other nationalities, their salaries reduced by an average of 18.69%. This means that, in expected value, European goalkeepers have more experience, more games per season and greater market value.
The opposite is seen in the nationalities of the continents of America, Africa and Asia, for example, because goalkeepers with nationalities from these continents have characteristics with lower values than the comparison group (that is, less experience, fewer games per season and lower value of Marketplace). Therefore, if goalkeepers born in America, Africa and Asia had the same characteristics as goalkeepers from other continents, their income would increase by 23.02%, 110.57% and 136.64% respectively.
Regarding the ‘Unexplained’ part, we also suggest the following interpretation. An estimated coefficient of positive direction in the ‘unexplained’ part means that if goalkeepers had a different nationality than they have, then they would earn more (indicating a perverse or discriminatory nationality effect). This is what happens with Asian goalkeepers, for example, who have a wage penalty, because if these athletes had another nationality, their income would increase by 35.31% on average. Contrarily, those born in America observe a premium for nationality - the estimated coefficient is statistically significant at 1%, meaning that if these goalkeepers were not American, their wages would reduce, on average, 25.19%.
Once again, the update performed with the inclusion of control variables revalidated the insights from Table 7.
Discussion
In our investigation, we took as a starting point the possibility of discrimination in the world of professional goalkeepers in soccer. Our hypothesis was that there would be discrimination beyond conventional explanations, which are based on the accumulation of human capital (i.e., experience) but also on the signaling of competence (e.g., by games in which the goalkeeper conceded few goals). However, we also wanted to assess whether skin color or nationality would contribute to justifying the observed differences in the salaries attributed to goalkeepers in six of the primary European leagues.
Using Mincerian equations and Blinder–Oaxaca decomposition, we found results that add evidence to the literature on discrimination in professional sports. Specifically, in the case of goalkeepers, although few non-whites played in the observed championships, they tended to earn above-average wages. Thus, being non-white ultimately seems to be a valued characteristic. However, nationality matters as a source of decreased wages. In particular, being a goalkeeper of non-European nationality was associated with a salary reduction. Following other works 43 we can highlight the effect of the Bosman ruling that brought the status of community player to players with a nationality associated with an EU member country, which in practice allows a player with the attribute to be considered within the national contingent in the championship of another EU member country. Obviously, that possibility can restrict access to those championships for players with a citizenship outside the EU. 44
Our interpretation of the explained and unexplained effects affords additional evidence. For example, we can refute the idea that non-white goalkeepers earned less than white goalkeepers due to skin color. The results in Table 5 show that though non-white goalkeepers indeed earned less on average, it was due to characteristics such as professional experience or market value. Table 6 additionally shows marked heterogeneity in the wages paid over the observed championships, showing that there are championships such as the Portuguese one that tend to pay goalkeepers who play there less than expected.
We acknowledge that our results are in line with recent findings. 40 We also know that previous researches showed that team owners exhibited racial discrimination in pay setting in the 1970s and 1980s. 45 Currently, such discriminatory practices are not as evident in salary relations, especially in professional sports. However, this does not prevent alternative discriminatory practices from emerging in the world of professional sports. As we had already anticipated, increased difficulties from scouting departments (particularly affecting undervalued athletes), negative media exposure (both in newspapers and television commentary), or manipulation of fan groups focused on devaluing certain athletes are current strategies that resemble, in terms of professional and sporting consequences, the discriminatory practices of the past. In fact, several studies show that fans’ racist behaviour has still been assessed as present on several occasions. 46
Last, Table 7 reveals that the player's nationality significantly influences the observed patterns in salary. According to our results, European goalkeepers tended to earn more than goalkeepers from other continents. Obviously, we have an additional source of discussion here. Our results ask for a further discussion on European championships – these leagues tend to enable European goalkeepers to acquire the skills specially valued by European championships. However, we have also observed that goalkeepers of American nationality with the same performance characteristics tend to experience the benefits of favorable discrimination. Such evidence may relate to the role of players’ transfer agents, which also merits further investigation. 47
Additionally, we also run other regressions controlling for several variables related to the performance of each keeper, like (among many others) Post-Shot Expected Goals minus Goals Allowed, Post-Shot Expected Goals per shot on target, Penalty save percentage, Touches, Crosses Stopped (%), from the aforementioned source fbref.com. However, these regressions (available upon request) did not change previous outcomes.
Conclusion and future challenges
This article addresses the possibility of wage discrimination among goalkeepers playing in the European soccer championships. Specifically, in addition to the classic hypotheses related to human capital and market segmentation, we tested the possibility that skin color and nationality interfere with the evidence.
Using Mincerian equations and Blinder–Oaxaca decomposition, we identified a set of challenging results. This empirical strategy allowed to study the wage distribution considering different quantiles and controlling for several dimensions. We also extended the original effort by deepening it with a robustness section.
First, contrary to a certain idea propagated in some media channels, being non-white does not lead to lower wages. Our evidence suggests that the difference in wages observed primarily stems from the difference in values observed in dimensions related to performance or the athlete's team market value.
However, our results also show the existence of discrimination associated with nationality. In that case, it becomes simpler for goalkeepers from European countries to earn higher salaries. Even so, goalkeepers competing in the Portuguese championship face a salary penalty compared with colleagues who play in the English, Spanish, Italian, or French championships.
These results raise the opportunity of discussing some implications for the soccer industry. We highlight the two most prominent. First, as we have concluded, there is a higher chance of discriminating against players because of nationality than because of skin color. This can happen because of significantly different conditions of sports formation and training. Therefore, reinforcing policies for reducing these starting inequalities can ensure more opportunities for all young sportsmen. Second, top leagues pay better for specific positions, like the positions studied here, the goalkeepers. Besides the common implications already debated over the seasons – namely, the increasing financial and sportive gaps across European leagues – this can also generate a certain kind of discrimination against good goalkeepers not able to move to the top leagues for exogenous reasons (like the agencies managing their contracts).
Our results raise three lines of investigation. Firstly, the sample should be extended to other professional championships, including those from other continents, and involving other sources for data. We consider it promising to expand our database, namely, by incorporating data from complementary sources such as “Who Scored”. Secondly, the observations made about soccer should be detailed by expanding the sample beyond goalkeepers. Last, we recommend exploring other measures of discrimination, namely the minutes of play for each player, the influence of different coaches, and the coincidence of the language spoken by the player in relation to the official language of the championship where they play.
Footnotes
Data availability statement
Data referred in the work is publicly available from the sources and the Authors.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Fundação para a Ciência e a Tecnologia, (grant number UIDB/03182/2020).
