Generally, two theories are considered to be at the heart of economic thinking on discrimination: taste-based discrimination and statistical discrimination. Taste-based discrimination assumes that some people want to avoid interaction with members of minority groups. This type of discrimination results from animus. Statistical discrimination is based on rational behaviour with decision making based on limited information. Bertrand et al. (2005) argue that in addition to conscious tasted-based and rational statistical discrimination there may be unconscious discrimination to which they refer as implicit discrimination. In their overview of the main differences between statistical discrimination and taste-based discrimination Guryan and Charles (2013) emphasise that much of the empirical work has been focused on establishing that there is discrimination rather than on the cause of this discrimination. Some empirical studies address the question on the origin of the discrimination. Bertrand and Duflo (2017) provide a general overview of field experiments on discrimination while Neumark (2018) presents an overview of experimental research on labour market discrimination. Lang and Kahn-Lang Spitzer (2020) discuss recent economic research on racial discrimination.
Sports data are suitable to analyse the existence and sometimes the nature of discrimination whereby the focus is on racial bias. In these studies there is often a distinction between Black and non-Black players based on photographs from websites or book publications. Szymanski (2000), for example, uses wage data from English professional football to study racial discrimination. He finds that conditional on their wage bill, clubs with a higher share of Black players perform better which suggests that the Black players were underpaid. Our study contributes to the discrimination literature by investigating the relationship between race and performance evaluations of professional football players provided by mass media (Principe and Van Ours 2021). We study this issue in the context of Serie A in Italy, a country where episodes of racism are often reported both on and off the sports contexts.
In Italy there are three printed sports newspapers - La Gazzetta dello Sport, Il Corriere dello Sport and TuttoSports – with average circulations of about 250,000, 170,000 and 120,000, respectively. However, most readers access these sports newspapers through their websites. For example, La Gazzetta's website has a daily traffic of about 1.5 million unique users. All sports newspapers rate the performance of football players from the Italian top league Serie A, generally on Monday. The newspaper ratings – in Italian ‘pagelle’ (‘report cards’) – evaluate the performance in the spirit of the school's report cards, ranging from one (very poor performance) to ten (excellent performance). These are assigned by professional sports journalists employed by the newspaper, but they are reported anonymously. We assembled a unique data set recording ratings relating to 409 outfield players, covering a nine-season period from 2009-2010 to 2017-2018. This provides a longitudinal dataset of 1,835 player-season observations.
We use several sources of data. First, information about individual player's characteristics (i.e. birth year, position on the pitch and international appearances) and performance are extracted from the website whoscored.com. Second, we collected information about the ratings given by the three Italian sports newspapers: La Gazzetta dello Sport, Il Corriere dello Sport and Tuttosport. We use the end-of-season overall average performance provided by the newspapers. Third, data on players yearly wages – recorded net of taxes and excluding any performance-related bonus – are taken from an annual report, published at the beginning of each football season by La Gazzetta dello Sport. For this reason, our dataset has 1,627 player-season observations for wages, as players who join the league in the following transfer windows (i.e. in January) are not recorded.
One of the recurrent issues is how to establish whether an individual is part of a minority group. Our variable of interest referring to the race of the player is coded through a visual inspection of players’ photographs on the website transfermarkt.com. This is an established method in the economic research on discrimination in the sports labour market, since the discriminators prejudge an individual based on appearance. We recruited four students to look at each player’s photograph and asked them to label the player as either Black or non-Black. We focus our analysis on the dichotomous situation in which the player is labelled Black by at least one of the students or non-Black, i.e. labelled Black by none of the students.
Newspaper ratings and wages
Figure 1a presents kernel densities of the average ratings of players. There is a clear difference between the two densities whereby Black players receive more low ratings than non-Black players, especially between 4.5 and 5. Only a few non-Black players receive such a low rating. Figures 1b shows kernel densities of log wages. The density is broader for Black players in particular because there are higher wages for Black players, although in the very high wage group non-Black players are over-represented.
Figure 1 Densities ratings and log wages for Black and non-Black football players
Non-Black players on average receive a higher newspaper rating. Whereas Black players on average receive 5.67, non-Black players receive 5.78 – a difference of 0.09. The largest difference in average rating is for Corriere and TuttoSport; the smallest difference is for Gazzetta.
These differences are highlighted in the left side of Figure 2, which shows the results of linear regressions with a dummy variable for being black as explanatory variable. Also, for wages there is a difference between Black and non-Black players. However, as shown in Figure 2 this difference is not significantly different from zero. To find out whether there is a racial bias in ratings and log wages we performed OLS-regressions where is addition to the dummy variable for being black, we added a wide range of player performance characteristics, club fixed effects and seasonal fixed effects. The right side of Figure 2 shows that this hardly affects the main outcomes, confirming that there is a racial bias in newspaper ratings but not in wages.
Figure 2 Effect of being a Black football player on newspaper ratings and log wages
Note: Unconditional = only including dummy variable for being black; conditional = including a wide range of performance characteristics, seasonal fixed effects and club fixed effects; presented are point estimates and 95% confidence intervals.
We find no evidence of racial bias in players’ wages. We speculate that for clubs there is sufficient competition to remove racial wage discrimination, i.e. clubs simply want value for money and are willing to pay market wages for the players they want. We do, however, find evidence of a racial bias in newspaper ratings of professional football players in Italy. Evaluation provided by sports newspapers do not only reflect on-field performance. With our data we are not able to identify the racial bias mechanisms at work. We hypothesise that this is unconscious discrimination related to stereotyping. Racial bias seems to be present only at the low end of the skill distribution. Therefore, it may be that the discrimination is unintentional without the discriminator being aware. This suggests that exposure to research outcomes establishing a racial bias may reduce discrimination. There is some evidence of this. When an academic study on racial bias among professional basketball referees was published, nothing happened to the bias. However, when the study received media coverage and the awareness of the racial bias was raised, the bias disappeared (Pope et al. 2018).
In other words, although unintentional discrimination is as harmful to the discriminated as intentional discrimination, there is an easy cure. Creating awareness by exposure might be able to wipe out racial bias in newspaper ratings and perhaps also in the behaviour of others involved in football or sports in general. To the extent that there are spillover effects from newspaper reports to opinions in society, making clear that there is a racial bias will be helpful in reducing discrimination across the board.
Bertrand, M, D Chugh and S Mullainathan (2005), “Implicit discrimination”, American Economic Review 95(2): 94-98.
Bertrand, M and E Duflo (2017), “Field experiments on discrimination”, in A V Banerjee and E Duflo (Eds.), Handbook of Field Experiments, Volume 1, pp. 309-393, North-Holland.
Guryan, J and K K Charles (2013), “Taste-based or statistical discrimination: The economics of discrimination returns to its roots”, Economic Journal, 123(572), F417-F432.
Lang, K and A Kahn-Lang Spitzer (2020), “Race discrimination: An economic perspective”, Journal of Economic Perspectives, 34(2), 68-89.
Neumark, D (2018), “Experimental research on labor market discrimination”, Journal of Economic Literature, 56(3), 799-866.
Pope, D G, J Price and J Wolfers (2018), “Awareness reduces racial bias”, Management Science, 64(11), 4988-4995.
Principe, F and J C van Ours (2021), “Racial Bias in Newspaper Ratings of Professional Football Players”, CEPR Discussion Paper No 16419.
Szymanski, S (2000), “A market test for discrimination in the English professional soccer leagues”, Journal of Political Economy, 108(3), 590-603.