There is widespread evidence that being called to a job interview depends, at least in part, on one's race. There are also racial gaps in wages (see Guryan and Charles 2013 for a review of the debate and evidence). However, data limitations often make it difficult to be sure whether racial gaps emanate from discrimination and, if so, what the source of that discrimination might be. The issue is of concern not only because discrimination –whatever its source – is to the detriment of the discriminated group, but also because it leads to an economically inefficient allocation of resources.
Two models of discrimination
Economists have two non-excludable models to explain discrimination.
- Employers, employees, and customers may have a taste for discrimination, whereby hiring, working with, or being served by an individual from the discriminated group has a psychic cost.
Employers may indulge this taste by refusing to hire, say, black workers or, if they do, they may pay them less than other employees. Employees may need to be compensated for working with members of the discriminated group. Customers served by those from the discriminated group may seek lower prices to compensate them for having to do so. Whatever the source of the discrimination, the discriminated group ends up worse-off.
- An alternative model states that the origin of the gap in employment and wages between groups arises due to employers lacking information on individual productivity at the point of hiring.
To reduce that uncertainty, they use information on the average productivity of the group of origin. This is often termed ‘statistical discrimination’. It can arise when economic actors perceive one group to be less productive than another.
Unfortunately, as noted by Levitt (2004: 433) “in general, empirical tests have a difficult time distinguishing between taste-based and information-based models of discrimination”. Efforts to identify discrimination in hiring face the additional problem that discrimination is unlawful. Thus, the literature focuses on the intention to hire. Typically, fake curriculum vitas, differing only by the implied race of the applicant, are sent to recruiting employers. Differences in call back rates may imply discrimination, either on the grounds of taste or perhaps for statistical discrimination reasons. Furthermore, these studies can only observe variation in call back rate and not in hiring decisions.
New test for a taste-based discrimination
In a recent paper (Bryson and Chevalier 2014), we propose a new test to isolate the contribution of taste-based discrimination. We do so by examining the hiring decisions in an on-line game based on the English Professional Football League, the Fantasy Premier League (FPL). The aim of the game is to hire 15 footballers playing in the English Premier League and to accumulate points according to their performance on the pitch. As such, FPL participants are employers making actual hiring decisions in a virtual labour market. Furthermore, these employers are ‘free’ to discriminate because there is no legal impediment to them doing so.
Unlike previous studies, we are able to isolate the effects of taste-based discrimination. This is because in the virtual labour market of the FPL, workers do not interact, they do not even know they have been hired by a FPL participant. Firms in the FPL do not have customers – they produce points for the fantasy league, which only have value for the manager of the firm. Finally, full information on the productivity of each potential employee is fully disclosed, as such there is no statistical discrimination. Thus, the only potential source of discrimination in this study is taste-based discrimination, which would be easy, since photographs of all potential employees are available, and discrimination is not unlawful.
- However, analysing data from three seasons (2009/10 to 2011/12), we find that the FPL employers do not discriminate on grounds of race when choosing their team, either in initial hiring or through the season.
Figure 1 reports the net change in demand for footballers depending on their performance (measured in points) in the previous game. More productive workers see a surge in their demand and there is no difference between white and non-white players (the divergences towards the tails of the graph are not statistically significant). This result is robust to controlling for other potentially relevant factors and to a variety of sensitivity tests.
Figure 1. Productivity and labour demand in the FPL
There is no evidence of racial discrimination, either in initial hiring or through the season, in a context where employers are fully aware of current and prospective workers’ productivity. So what are we to make of this result in relation to the existing literature? It would be wrong simply to dismiss it on the grounds that it is hard to extrapolate from a virtual labour market in footballers to a real-world setting. Instead, we would emphasise the advantages that our setting has in isolating the potential for taste-based discrimination in hiring decisions. The fact that we find no such discriminatory behaviour against non-whites is consistent with other studies which suggest that racial differences in market outcomes are not driven by taste considerations. Instead, other factors are at play. For example, a number of studies point to the importance of statistical discrimination (List 2004, Zussman 2013, Doleac and Stein 2013) while others, such as Plug et al. (2014), emphasise the importance of worker sorting based on perceptions of discrimination in certain occupations or among particular types of employer. Since those channels are closed in our setting, it is perhaps unsurprising that we do not find any evidence of discrimination.
Our findings are also consistent with studies which point to a diminution in the extent of racial discrimination in sports on the part of employers and fans. The difference in discriminatory behaviour between the sport context and the general labour market is likely to be driven by the availability of good productivity measures for all possible employees in the sports labour market. Although such measures are not usually available in the general labour market, any policies that can minimise opportunities for statistical discrimination should be piloted and evaluated.
Bryson, A and A Chevalier (2014), “What Happens When Employers are Free to Discriminate? Evidence from the English Barclays Premier Fantasy Football League”, CEP Discusssion Paper No. 1283.
Doleac, J L and L C D Stein (2013), “The Visible Hand: Race and Online Market Outcomes”, The Economic Journal, 123, F469-F492.
Guryan, J and K K Charles (2013), “Taste-based or Statistical Discrimination: The Economics of Discrimination Returns to its Roots”, The Economic Journal, 123, 572: F417-432.
Levitt, S (2004), “Testing Theories of Discrimination: Evidence from The Weakest Link”, Journal of Law and Economics, 47(2): 431–52.
Plug, E, D Webbink, and N Martin (2014), “Sexual Orientation, Prejudice and Segregation”, Journal of Labor Economics, 32, 1: 123-159.
Zussman, A (2013), “Ethnic Discrimination: Lessons from the Israeli Online Market for Used Cars”, The Economic Journal, 123, F433-F468.