Yet another rejection: Aygül Muhiddin finds it frustratingly hard to get a job. If she could only be invited for a job interview so that they can learn how committed and hard-working she is! Aygül Muhiddin has a suspicion that she is discriminated against because of her name; after all, didn’t she go to the same schools as everyone else? Academics are interested in discrimination just as much as Aygül Muhiddin – a fictitious job seeker as academics sometimes use them to study the phenomenon. They send out equivalent CVs from minority and majority candidates to measure discrimination. In a meta-analysis of such field experiments, we synthesized 738 experiments to examine patterns of discrimination.
Field Experiments Are Used to Detect Discrimination
It may take Aygül Muhiddin a couple of rejections and a cursory comparison to her friends from the majority population to conclude that she has been discriminated against. Academics are often more careful – not because they find discrimination any more acceptable, but because empirically it is very difficult to be sure discrimination has really taken place. The crucial intention to treat minority candidates differently is tough to measure, and few employers would admit to discriminating against minority candidates; social (and legal) norms clearly discourage anyone from admitting to acts of discrimination.
In the UK, racial discrimination was outlawed in 1965, but few academics believed that this meant discrimination against minority candidates would cease completely. To determine discrimination in a context where few employers will admit to discrimination, it is common to use field experiments. Two fictitious CVs are sent to the same employer, differing substantively only in the name of the applicant: one has a minority name – like Aygül Muhiddin – the other one a majority name. By comparing the rate at which applicants with a minority name are invited for an interview with that of applicants with a majority name, the degree of discrimination is enumerated. This is a controlled experiment called a “correspondence test” and draws on real decisions by real human resources personnel. As a result, even hardened critics among academics find it plausible that discrimination can be captured using this method.
This basic experiment can be varied to examine discrimination against different minority groups, against men and women, or specific to an occupation. Using a systematic search of such experiments in OECD countries between 1990 and 2015, we could identify 738 experiments, published in 43 separate studies. We used meta-analysis to examine patterns of discrimination, and draw inferences about the likely mechanisms behind the observed discrimination.
Taste-based and Statistical Discrimination
A basic distinction is between “taste-based discrimination” and “statistical discrimination.” Taste-based discrimination means that the employer dislikes members of the ethnic minority. He or she is willing to hire a less qualified person to avoid having an ethnic minority employee. Statistical discrimination, by contrast, means that the employer does not hire ethnic minorities because he or she believes them to be less qualified. The ethnicity of the candidate is used to draw inferences about unobserved characteristics, and the employer may draw on past experience, stereotypes, hearsay, or simple gut feelings in making this evaluation. While for Aygül Muhiddin both forms of discrimination have the same consequence, the policy interventions necessary to combat discrimination may be quite different.
Pervasive Discrimination Against Ethnic Minority Candidates
The meta-analysis identifies relatively consistent patterns of discrimination against minority candidates across countries and time. On average, ethnic minority candidates need to send twice as many applications to be invited for an interview as majority candidates do. There are large differences between specific minority groups. For instance, on average Arabs need to send nearly twice as many applications than Turks, or in a particular study sales assistants might be invited less often than nurses. The national economic context does not seem to have a direct impact on call-back rates, nor are there systematic gender differences. Moreover, call-back rates were similar before and after the introduction of EU directives 2000/43/EC and 2000/78/EC, both of which were intended to to reduce discrimination.
Call-back rates are indistinguishable between immigrants and their children. Children of immigrants – the so-called second generation – tend to have local qualifications. This means that employers have no reason to use the ethnicity of the applicant as a proxy to estimate the quality of a diploma, for example. The kind of discrimination that remains is likely taste-based. By contrast, in countries where long application packs are the norm, the difference in call-back rates for minority candidates is greatly reduced. This fact suggests that with additional information on the candidates, statistical discrimination can probably be reduced.
That might spare the real Aygül Muhiddins out there a few rejections.
Eva Zschirnt and Didier Ruedin (2016). Ethnic Discrimination in Hiring Decisions: A Meta-Analysis of Correspondence Tests 1990–2015. Journal of Ethnic and Migration Studies. doi: 10.1080/1369183X.2015.1133279.